1 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
5 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
7 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
9 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
10 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
14 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
15 Disable -fdelete-null-pointer-checks for ELF toolchain.
17 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
18 Kito Cheng <kito.cheng@gmail.com>
20 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
21 (nds32le-*-*, nds32be-*-*): Integrate checking process.
22 (nds32*-*-*): Add glibc and uclibc conditions.
23 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
24 (TARGET_EXCEPT_UNWIND_INFO): Define.
25 * config/nds32/elf.h: New file.
26 * config/nds32/linux.h: New file.
27 * config/nds32/nds32-elf.opt: New file.
28 * config/nds32/nds32-linux.opt: New file.
29 * config/nds32/nds32-fp-as-gp.c
30 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
31 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
33 (nds32_asm_file_end): Ditto.
34 (nds32_print_operand): Ditto.
35 (nds32_insert_attributes): Ditto.
36 (nds32_init_libfuncs): New function.
37 (TARGET_HAVE_TLS): Define.
38 (TARGET_INIT_LIBFUNCS): Define.
39 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
41 (TARGET_ELF): Apply different mcmodel setting.
42 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
43 been migrated into elf.h and linux.h files.
44 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
45 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
46 (mcmodel): The content has been migrated into nds32-elf.opt and
47 nds32-linux.opt files.
48 * config/nds32/t-elf: New file.
49 * config/nds32/t-linux: New file.
51 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
52 Shiva Chen <shiva0217@gmail.com>
54 * config/nds32/constants.md (unspec_volatile_element): Add
55 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
56 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
58 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
59 (make_pass_nds32_fp_as_gp): Declare.
60 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
62 (nds32_asm_function_end_prologue): Remove unused asm output.
63 (nds32_asm_function_begin_epilogue): Remove unused asm output.
64 (nds32_asm_file_start): Output necessary fp_as_gp information.
65 (nds32_option_override): Adjust register usage.
66 (nds32_expand_prologue): Consider fp_as_gp situation.
67 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
68 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
71 (simple_return): Ditto.
72 (omit_fp_begin): Output special directive for fp_as_gp.
73 (omit_fp_end): Output special directive for fp_as_gp.
74 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
75 mforbid-fp-as-gp): New options.
77 2018-06-01 Mark Wielaard <mark@klomp.org>
79 * dwarf2out.c (dwarf2out_finish): Remove generation of
82 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
84 * gimple-ssa-store-merging.c: Include gimple-fold.h.
85 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
86 (struct merged_store_group): Add bit_insertion field.
87 (dump_char_array): Use standard hexadecimal format.
88 (merged_store_group::merged_store_group): Set bit_insertion to false.
89 (merged_store_group::apply_stores): Use optimal buffer size. Deal
90 with BIT_INSERT_EXPR stores. Move up code updating the mask and
91 also print the mask in the dump file.
92 (pass_store_merging::gate): Minor tweak.
93 (imm_store_chain_info::coalesce_immediate): Fix wrong association
94 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
95 stores with INTEGER_CST stores.
96 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
97 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
98 and use it throughout. Generate bit insertion sequences if need be.
99 (pass_store_merging::process_store): Remove redundant condition.
100 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
102 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
104 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
105 the 128-bit floating point types. Fix function comment.
107 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
109 * config/aarch64/aarch64-simd.md
110 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
112 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
115 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
117 PR tree-optimization/85989
118 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
120 (backprop::intersect_uses): Check it when deciding whether this
121 is a backedge reference.
122 (backprop::process_block): Add each phi to m_visited_phis
123 after visiting it, then clear it at the end.
125 2018-06-01 Richard Biener <rguenther@suse.de>
127 * tree-vectorizer.h (vect_dr_stmt): New function.
128 (vect_get_load_cost): Adjust.
129 (vect_get_store_cost): Likewise.
130 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
131 Use vect_dr_stmt instead of DR_SMTT.
132 (vect_record_base_alignments): Likewise.
133 (vect_calculate_target_alignment): Likewise.
134 (vect_compute_data_ref_alignment): Likewise and make static.
135 (vect_update_misalignment_for_peel): Likewise.
136 (vect_verify_datarefs_alignment): Likewise.
137 (vector_alignment_reachable_p): Likewise.
138 (vect_get_data_access_cost): Likewise. Pass down
139 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
140 (vect_get_peeling_costs_all_drs): Likewise.
141 (vect_peeling_hash_get_lowest_cost): Likewise.
142 (vect_enhance_data_refs_alignment): Likewise.
143 (vect_find_same_alignment_drs): Likewise.
144 (vect_analyze_data_refs_alignment): Likewise.
145 (vect_analyze_group_access_1): Likewise.
146 (vect_analyze_group_access): Likewise.
147 (vect_analyze_data_ref_access): Likewise.
148 (vect_analyze_data_ref_accesses): Likewise.
149 (vect_vfa_segment_size): Likewise.
150 (vect_small_gap_p): Likewise.
151 (vectorizable_with_step_bound_p): Likewise.
152 (vect_prune_runtime_alias_test_list): Likewise.
153 (vect_analyze_data_refs): Likewise.
154 (vect_supportable_dr_alignment): Likewise.
155 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
156 (vect_gen_prolog_loop_niters): Likewise.
157 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
158 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
160 (vect_recog_mask_conversion_pattern): Likewise.
161 (vect_try_gather_scatter_pattern): Likewise.
162 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
163 to vect_get_store_cost.
164 (vect_get_store_cost): Get stmt_info instead of DR.
165 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
166 (vect_get_load_cost): Get stmt_info instead of DR.
168 2018-06-01 Richard Biener <rguenther@suse.de>
171 * gimple-fold.c (var_decl_component_p): Also allow offsetted
172 vars wrapped in MEM_REFs.
174 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
176 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
177 Fix subreg tests so that we only return a choice between
178 GENERAL_REGS and FP_REGS if the original classes included both.
180 2018-06-01 Richard Biener <rguenther@suse.de>
183 * tree-ssa-structalias.c (get_function_part_constraint):
184 Handle NULL fi->decl.
185 (find_func_aliases_for_call): Properly handle indirect
187 (find_func_clobbers): Likewise.
188 (ipa_pta_execute): Likewise.
189 (create_variable_info_for): For functions that are ifunc_resolver
190 resolve to a varinfo that contains the result of the resolver
192 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
195 2018-05-31 Michael Collison <michael.collison@arm.com>
197 * config/aarch64/aarch64.md:
198 (*fix_to_zero_extenddfdi2): New pattern.
199 * gcc.target/aarch64/fix_extend1.c: New testcase.
201 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
205 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
206 and BUILT_IN_STRNCMP_EQ.
207 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
209 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
210 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
211 (gimple_fold_builtin): Likewise.
212 * tree-ssa-strlen.c (compute_string_length): New function.
213 (determine_min_obsize): New function.
214 (handle_builtin_string_cmp): New function to handle calls to
215 string compare functions.
216 (strlen_optimize_stmt): Add handling to builtin string compare
218 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
219 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
220 * tree.c (build_common_builtin_nodes): Add new defines of
221 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
223 2018-05-31 Jakub Jelinek <jakub@redhat.com>
226 * bb-reorder.c (pass_partition_blocks::gate): Return false for
227 functions with naked attribute.
229 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
231 * config/i386/sse.md (avx_vec_concat<mode>):
232 Substitute concat_tg_mode mode attribute with xtg_mode.
233 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
234 (concat_tg_mode): Remove mode attribute.
236 2018-05-31 Martin Sebor <msebor@redhat.com>
239 * calls.c (alloc_max_size): Correct a logic error/typo.
240 Treat excessive arguments as infinite. Warn for invalid arguments.
241 * doc/invoke.texi (-Walloc-size-larger-than): Update.
243 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
246 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
247 and movx for Haswell.
249 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
250 Cesar Philippidis <cesar@codesourcery.com>
253 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
254 when emitting error on private/firstprivate reductions.
255 * omp-low.c (lower_omp_target): Avoid reference-type processing
256 on pointers for firstprivate clause.
258 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
260 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
263 * config/aarch64/aarch64-simd.md
264 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
265 (aarch64_ld1_x3_<mode>): Likewise
266 (aarch64_st1x2<VALLDIF:mode>): Likewise
267 (aarch64_st1_x2_<mode>): Likewise
268 (aarch64_st1x3<VALLDIF:mode>): Likewise
269 (aarch64_st1_x3_<mode>): Likewise
270 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
271 (vld1_s8_x3): Likewise.
272 (vld1_u16_x3): Likewise.
273 (vld1_s16_x3): Likewise.
274 (vld1_u32_x3): Likewise.
275 (vld1_s32_x3): Likewise.
276 (vld1_u64_x3): Likewise.
277 (vld1_s64_x3): Likewise.
278 (vld1_f16_x3): Likewise.
279 (vld1_f32_x3): Likewise.
280 (vld1_f64_x3): Likewise.
281 (vld1_p8_x3): Likewise.
282 (vld1_p16_x3): Likewise.
283 (vld1_p64_x3): Likewise.
284 (vld1q_u8_x3): Likewise.
285 (vld1q_s8_x3): Likewise.
286 (vld1q_u16_x3): Likewise.
287 (vld1q_s16_x3): Likewise.
288 (vld1q_u32_x3): Likewise.
289 (vld1q_s32_x3): Likewise.
290 (vld1q_u64_x3): Likewise.
291 (vld1q_s64_x3): Likewise.
292 (vld1q_f16_x3): Likewise.
293 (vld1q_f32_x3): Likewise.
294 (vld1q_f64_x3): Likewise.
295 (vld1q_p8_x3): Likewise.
296 (vld1q_p16_x3): Likewise.
297 (vld1q_p64_x3): Likewise.
298 (vst1_s64_x2): Likewise.
299 (vst1_u64_x2): Likewise.
300 (vst1_f64_x2): Likewise.
301 (vst1_s8_x2): Likewise.
302 (vst1_p8_x2): Likewise.
303 (vst1_s16_x2): Likewise.
304 (vst1_p16_x2): Likewise.
305 (vst1_s32_x2): Likewise.
306 (vst1_u8_x2): Likewise.
307 (vst1_u16_x2): Likewise.
308 (vst1_u32_x2): Likewise.
309 (vst1_f16_x2): Likewise.
310 (vst1_f32_x2): Likewise.
311 (vst1_p64_x2): Likewise.
312 (vst1q_s8_x2): Likewise.
313 (vst1q_p8_x2): Likewise.
314 (vst1q_s16_x2): Likewise.
315 (vst1q_p16_x2): Likewise.
316 (vst1q_s32_x2): Likewise.
317 (vst1q_s64_x2): Likewise.
318 (vst1q_u8_x2): Likewise.
319 (vst1q_u16_x2): Likewise.
320 (vst1q_u32_x2): Likewise.
321 (vst1q_u64_x2): Likewise.
322 (vst1q_f16_x2): Likewise.
323 (vst1q_f32_x2): Likewise.
324 (vst1q_f64_x2): Likewise.
325 (vst1q_p64_x2): Likewise.
326 (vst1_s64_x3): Likewise.
327 (vst1_u64_x3): Likewise.
328 (vst1_f64_x3): Likewise.
329 (vst1_s8_x3): Likewise.
330 (vst1_p8_x3): Likewise.
331 (vst1_s16_x3): Likewise.
332 (vst1_p16_x3): Likewise.
333 (vst1_s32_x3): Likewise.
334 (vst1_u8_x3): Likewise.
335 (vst1_u16_x3): Likewise.
336 (vst1_u32_x3): Likewise.
337 (vst1_f16_x3): Likewise.
338 (vst1_f32_x3): Likewise.
339 (vst1_p64_x3): Likewise.
340 (vst1q_s8_x3): Likewise.
341 (vst1q_p8_x3): Likewise.
342 (vst1q_s16_x3): Likewise.
343 (vst1q_p16_x3): Likewise.
344 (vst1q_s32_x3): Likewise.
345 (vst1q_s64_x3): Likewise.
346 (vst1q_u8_x3): Likewise.
347 (vst1q_u16_x3): Likewise.
348 (vst1q_u32_x3): Likewise.
349 (vst1q_u64_x3): Likewise.
350 (vst1q_f16_x3): Likewise.
351 (vst1q_f32_x3): Likewise.
352 (vst1q_f64_x3): Likewise.
353 (vst1q_p64_x3): Likewise.
355 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
357 * config/msp430/msp430.c (msp430_output_labelref): Prepend
358 user_label_prefix to name.
360 * tree-core.h: Update comment about the format of NAME string
361 passed to handler in attribute_spec.
363 * config/msp430/msp430.md: Remove erroneous subreg expression from
364 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
365 zero_extend{q,h}isi2.
367 2018-05-30 Borislav Petkov <bp@suse.de>
369 * doc/extend.texi: Document some architecture specific
370 constraints and sort entries.
372 2018-05-30 Martin Sebor <msebor@redhat.com>
375 * builtins.c (expand_builtin_stpcpy_1): New function.
376 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
377 only if the former succeeds.
379 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
381 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
384 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
386 * doc/invoke.texi (-flinker-output): Document
388 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
390 * passes.c (ipa_write_summaries): Only modify statements if body
392 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
393 incrementally linking.
394 (ipa_passes): Likewise.
395 * lto-cgraph.c (lto_output_node): When incrementally linking do not
396 pass down resolution info.
397 * common.opt (flag_incremental_link): Update info.
398 * gcc.c (plugin specs): Turn flinker-output=* to
399 -plugin-opt=-linker-output-known
400 * toplev.c (compile_file): Also cut compilation when doing incremental
402 * flag-types. (enum lto_partition_model): Add
403 LTO_LINKER_OUTPUT_NOLTOREL.
404 (invoke.texi): Add -flinker-output docs.
405 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
406 link same way as WPA; do not stream in dead initializers.
408 * dwarf2out.c (dwarf2out_die_ref_for_decl,
409 darf2out_register_external_decl): Support incremental link.
411 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
413 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
415 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
417 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
418 it down to simple_object_copy_lto_debug_sections.
419 (run_gcc): Determine incremental LTO link time and configure
420 lto1 into non-wpa mode, disable renaming of debug sections.
422 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
424 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
425 descriptions of various incorrectly documented functions.
427 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
430 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
431 address check not strict.
433 2018-05-30 Richard Biener <rguenther@suse.de>
435 PR tree-optimization/85964
436 * tracer.c (better_p): Drop initialized count check, we only
437 call the function with initialized counts now.
438 (find_best_successor): Do find a best edge if one
439 has uninitialized count.
440 (find_best_predecessor): Likewise. Do BB frequency check only
441 if count is initialized.
443 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
445 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
446 (aarch64_ldrstr_offset_compare): New.
447 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
448 load/store orderings.
449 (aarch64_gen_adjusted_ldpstp): Likewise.
451 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
453 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
454 Check for subset of GENERAL_REGS and FP_REGS.
455 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
458 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
460 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
461 and wi::to_poly_offset. Add the current offset and then check
462 whether the sum fits, rather than using an unchecked addition of
463 a checked term. Check for a shwi rather than a uhwi.
464 * expr.c (get_bit_range): Use tree_to_poly_uint64.
465 (store_constructor): Use poly_int_tree_p.
466 (expand_expr_real_1): Likewise.
467 * function.c (assign_temp): Likewise.
468 * fold-const.c (const_binop): Use poly_int_tree_p and
470 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
472 * ipa-icf-gimple.c (func_checker::compare_operand): Use
473 to_poly_offset for MEM offsets.
474 * ipa-icf.c (sem_variable::equals): Likewise.
475 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
476 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
477 wi::to_poly_offset for BIT_FIELD_REF offsets.
478 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
480 * var-tracking.c (emit_note_insn_var_location): Use
483 2018-05-29 Jim Wilson <jimw@sifive.com>
485 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
487 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
490 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
491 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
493 (sse4_1_round<mode>2): Use nonimmediate_operand
494 for operand 1 predicate.
496 2018-05-29 Martin Sebor <msebor@redhat.com>
497 Richard Biener <rguenther@suse.de>
500 * calls.c (get_size_range): Call determine_value_range instead
502 * tree-vrp.h (determine_value_range): Declared new function.
503 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
505 2018-05-29 Richard Biener <rguenther@suse.de>
507 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
508 sure to use non-pattern stmts for get_earlier_stmt arguments.
509 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
510 called on pattern stmts.
511 (get_later_stmt): Likewise.
513 2018-05-29 Martin Liska <mliska@suse.cz>
515 PR gcov-profile/85759
516 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
519 2018-05-29 Jakub Jelinek <jakub@redhat.com>
521 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
523 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
527 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
528 VEC_PACK_FLOAT_EXPR): New tree codes.
529 * tree-pretty-print.c (op_code_prio): Handle
530 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
531 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
532 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
533 * tree-inline.c (estimate_operator_cost): Likewise.
534 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
535 * fold-const.c (const_binop): Likewise.
536 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
537 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
538 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
539 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
540 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
541 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
542 * expr.c (expand_expr_real_2): Likewise.
543 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
544 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
545 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
547 * optabs.c (expand_widen_pattern_expr): For
548 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
549 sign from result type rather than operand's type.
550 (expand_binop_directly): For vec_packu_float_optab and
551 vec_packs_float_optab allow result type to be different from operand's
553 * optabs-tree.c (optab_for_tree_code): Handle
554 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
555 VEC_PACK_FLOAT_EXPR. Formatting fixes.
556 * tree-vect-generic.c (expand_vector_operations_1): Handle
557 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
559 * tree-vect-stmts.c (supportable_widening_operation): Handle
561 (supportable_narrowing_operation): Handle FLOAT_EXPR.
562 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
563 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
564 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
565 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
567 (vec_pack<floatprefix>_float_<mode>): New expander.
568 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
570 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
571 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
572 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
573 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
574 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
576 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
577 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
578 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
579 VEC_PACK_FLOAT_EXPR): Document.
581 2018-05-29 Richard Biener <rguenther@suse.de>
583 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
585 (stmt_vec_info_vec): Make pointer.
586 (init_stmt_vec_info_vec): Remove.
587 (free_stmt_vec_info_vec): Likewise.
588 (set_stmt_vec_info_vec): New function.
589 (free_stmt_vec_infos): Likewise.
590 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
591 (set_vinfo_for_stmt): Likewise.
592 (get_earlier_stmt): Likewise.
593 (get_later_stmt): Likewise.
594 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
595 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
596 (vec_info::~vec_info): Free stmt_vec_infos.
597 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
598 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
599 (pass_slp_vectorize::execute): Likewise.
600 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
601 (free_stmt_vec_info_vec): Likewise.
602 (set_stmt_vec_info_vec): New function.
603 (free_stmt_vec_infos): Likewise.
604 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
605 the global stmt_vec_info_vec.
606 * tree-parloops.c (gather_scalar_reductions): Use
607 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
610 2018-05-29 Richard Biener <rguenther@suse.de>
612 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
614 2018-05-29 Martin Liska <mliska@suse.cz>
615 David Malcolm <dmalcolm@redhat.com>
617 * vec.c (test_reverse): New.
618 (vec_c_tests): Add new test.
619 * vec.h (vl_ptr>::reverse): New function.
621 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
623 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
625 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
628 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
630 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
632 2018-05-28 Richard Biener <rguenther@suse.de>
634 PR tree-optimization/85933
635 * tree-vect-data-refs.c (vect_record_base_alignments): Only
636 look at stmts marked as vectorizable.
638 2018-05-28 Richard Biener <rguenther@suse.de>
640 PR tree-optimization/85934
641 * tree-vect-generic.c (expand_vector_operations_1): Hoist
642 vector boolean check before scalar optimization.
644 2018-05-28 Jakub Jelinek <jakub@redhat.com>
646 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
649 2018-05-28 Mark Wielaard <mark@klomp.org>
651 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
652 if it is an expression containing a minus sign.
654 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
656 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
658 2018-05-27 Paul Koning <ni1d@arrl.net>
660 * config/pdp11/pdp11.md (truncsihi2): Remove.
662 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
663 Chung-Ju Wu <jasonwucj@gmail.com>
665 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
667 (unaligned_store_dw): Ditto.
668 * config/nds32/nds32-memory-manipulation.c
669 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
670 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
671 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
672 (emit_setmem_word_loop): Rename to ...
673 (emit_setmem_doubleword_loop): ... this.
674 (nds32_gen_dup_4_byte_to_word_value): New function.
675 (nds32_gen_dup_8_byte_to_double_word_value): New function.
676 (nds32_expand_setmem_loop): Refine implementation.
677 (nds32_expand_setmem_loop_v3m): Ditto.
678 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
681 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
683 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
685 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
687 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
688 (nds32_init_machine_status): Initialize machine->attr_naked_p and
689 machine->attr_no_prologue_p.
690 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
691 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
692 (nds32_expand_epilogue): Consider attr_naked_p.
693 (nds32_expand_epilogue_v3pop): Likewise.
694 (nds32_can_use_return_insn): Likewise.
695 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
696 attr_no_prologue_p fields.
697 * config/nds32/nds32.opt (mret-in-naked-func): New option.
699 2018-05-27 Jakub Jelinek <jakub@redhat.com>
702 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
705 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
707 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
709 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
711 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
713 (*<floatsuffix>floatv2div2sf2): Rename to ...
714 (*float<floatunssuffix>v2div2sf2): ... this.
715 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
716 (float<floatunssuffix>v2div2sf2_mask): ... this.
717 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
718 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
719 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
721 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
723 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
725 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
727 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
729 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
731 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
732 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
733 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
734 gen_ufix_truncv8dfv8si2.
735 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
736 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
737 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
738 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
739 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
740 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
741 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
742 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
744 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
748 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
750 2018-05-25 Jim Wilson <jimw@sifive.com>
752 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
753 * config/riscv/riscv.c (struct machine_function): Add
754 interrupt_handler_p and attribute_checked_p fields.
755 (riscv_attribute_table): Add interrupt.
756 (riscv_interrupt_type_p): New.
757 (riscv_save_reg_p): Save extra regs for interrupt handler.
758 (riscv_use_save_libcall): Return false for interrupt handler.
759 (riscv_first_stack_step): Add forward declaration.
760 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
761 for interrupt handler with large frame. Use it for saved reg list.
762 (riscv_expand_prologue): Move flag_stack_usage_info support to
763 eliminate duplication.
764 (riscv_expand_epilogue): Generate mret for interrupt handler.
765 (riscv_epilogue_uses): New.
766 (riscv_can_use_return_insn): Return false for interrupt handler.
767 (riscv_function_ok_for_sibcall): Likewise.
768 (riscv_set_current_function): Add interrupt handler support.
769 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
770 * config/riscv/riscv.md (UNSPECV_MRET): New.
772 (riscv_frflags, riscv_fsflags): Use tab after opcode.
774 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
776 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
778 PR tree-optimization/85712
779 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
780 this candidate has already been replaced in-situ by a copy.
782 2018-05-25 Jason Merrill <jason@redhat.com>
784 PR c++/80485 - inline function non-zero address.
785 * symtab.c (nonzero_address): Check DECL_COMDAT.
787 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
790 * config/alpha/alpha.md (ashlsi3): New insn pattern.
791 (*ashlsi_se): Rename from *ashldi_se. Define as sign
792 extension of SImode operation. Use const123_operand predicate.
794 (*saddl_se_1): Ditto.
796 (*ssubl_se_1): Ditto.
797 * config/alpha/predicates.md (const123_operand): New predicate.
798 * config/alpha/constraints.md (P): Use IN_RANGE.
800 2018-05-25 Richard Biener <rguenther@suse.de>
802 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
804 (ref_maybe_used_by_stmt_p): Likewise.
805 (stmt_may_clobber_ref_p): Likewise.
806 (stmt_may_clobber_ref_p_1): Likewise.
807 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
809 (ref_maybe_used_by_stmt_p): Likewise.
810 (stmt_may_clobber_ref_p): Likewise.
811 (stmt_may_clobber_ref_p_1): Likewise.
812 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
813 the alias oracle to disambiguate DRs with stmts DR analysis
815 (vect_analyze_data_refs): Do not give up on not analyzable
816 DRs for BB vectorization. Remove code truncating the dataref
819 2018-05-25 Jakub Jelinek <jakub@redhat.com>
822 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
823 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
824 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
826 2018-05-25 Richard Biener <rguenther@suse.de>
828 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
829 function, combining stmt data ref gathering and fatal analysis
831 (vect_analyze_data_refs): Remove now redudnant code and simplify.
832 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
833 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
834 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
835 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
837 2018-05-25 Bin Cheng <bin.cheng@arm.com>
839 PR tree-optimization/85720
840 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
841 SCC if all partitions are builtins.
842 (version_loop_by_alias_check): New parameter. Generate cancelable
843 runtime alias check if all partitions are builtins.
844 (distribute_loop): Update call to above function.
846 2018-05-25 Bin Cheng <bin.cheng@arm.com>
848 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
849 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
850 (parm_default_def_partition_arg): Ditto.
851 (set_parm_default_def_partition): Ditto.
852 (get_parm_default_def_partitions): Ditto and make it static.
853 (get_undefined_value_partitions): Ditto and make it static.
854 (remove_ssa_form): Refactor call to init_var_map here.
855 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
856 computation for loop region.
857 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
858 (register_default_def): Delete.
859 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
860 (parm_default_def_partition_arg): Ditto.
861 (set_parm_default_def_partition): Ditto.
862 (get_parm_default_def_partitions): Ditto and make it static.
863 (get_undefined_value_partitions): Ditto and make it static.
864 (coalesce_with_default, coalesce_with_default): Update comment.
865 (create_coalesce_list_for_region): New func factored out from
866 create_outofssa_var_map.
867 (populate_coalesce_list_for_outofssa): New func factored out from
868 create_outofssa_var_map and coalesce_ssa_name.
869 (create_outofssa_var_map): Delete.
870 (coalesce_ssa_name): Refactor to support live range computation.
871 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
872 (get_parm_default_def_partitions): Delete.
873 (get_undefined_value_partitions): Ditto.
874 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
875 computation for loop region.
876 (new_tree_live_info, loe_visit_block): Ditto.
877 (live_worklist, set_var_live_on_entry): Ditto.
878 (calculate_live_on_exit, verify_live_on_entry): Ditto.
879 * tree-ssa-live.h (struct _var_map): New fields.
880 (init_var_map): Change decl.
881 (region_contains_p): New.
883 2018-05-25 Bin Cheng <bin.cheng@arm.com>
885 * tree-ssa-live.h (live_merge_and_clear): Delete.
887 2018-05-25 Richard Biener <rguenther@suse.de>
890 * tree-dump.c (dequeue_and_dump): Remove access to removed
891 operand 2 of a SWITCH_EXPR.
893 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
895 * doc/sourcebuild.texi (vect_double_cond_arith): Include
896 multiplication and division.
897 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
898 (cond_udiv@var{m}, cond_umod@var{m}): Document.
899 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
900 (cond_udiv_optab, cond_umod_optab): New optabs.
901 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
902 (IFN_COND_RDIV): New internal functions.
903 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
904 TRUNC_MOD_EXPR and RDIV_EXPR.
905 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
906 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
908 (SVE_INT_BINARY): Include mult.
909 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
910 (optab, sve_int_op): Handle mult.
911 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
913 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
914 for SVE_INT_BINARY_SD.
916 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
918 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
919 (optab, sve_int_op): Handle div and udiv.
920 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
921 for SVE_INT_BINARY_SD.
922 (*<optab><mode>3): New insn for the same.
924 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
926 * tree-vect-patterns.c: Include predict.h.
927 (vect_recog_divmod_pattern): Restrict check for division support
928 to when optimizing for size.
930 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
932 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
933 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
934 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
935 (gimple_match_op::set_op): Likewise.
936 (gimple_resimplify4): Declare.
937 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
938 (expr::gen_transform): Likewise.
939 (decision_tree::gen): Generate a simplification routine for 4 operands.
940 * gimple-match-head.c (gimple_simplify): Add an overload for
941 4 operands. In the top-level function, handle up to 4 call
942 arguments and call gimple_resimplify4.
943 (gimple_resimplify4): New function.
944 (build_call_internal): Pass a fourth operand.
945 (maybe_push_to_seq): Likewise.
946 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
947 Fold VEC_COND_EXPRs of an operation and a default value into
948 an IFN_COND_* function if possible.
949 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
951 (SVE_COND_FP_BINARY): Include them.
952 (optab, sve_fp_op): Handle them.
953 (SVE_INT_BINARY_REV): New code iterator.
954 (SVE_COND_FP_BINARY_REV): New int iterator.
955 (commutative): New int attribute.
956 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
958 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
960 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
961 (*cond_<optab><mode>): New patterns for reversed operands.
963 2018-05-25 Richard Biener <rguenther@suse.de>
965 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
966 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
967 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
968 (STMT_VINFO_GROUPED_ACCESS): Adjust.
969 * tree-vect-data-refs.c (everywhere): Adjust users.
970 * tree-vect-loop.c (everywhere): Likewise.
971 * tree-vect-slp.c (everywhere): Likewise.
972 * tree-vect-stmts.c (everywhere): Likewise.
973 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
975 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
977 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
979 (gcc_cv_as_section_exclude): ... this.
980 Try Solaris as #exclude syntax.
981 * configure: Regenerate.
982 * config.in: Regenerate.
983 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
985 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
986 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
988 * varasm.c (default_elf_asm_named_section): Don't check if
989 HAVE_GAS_SECTION_EXCLUDE is defined.
991 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
993 * doc/md.texi: Update the documentation of the cond_* optabs
994 to mention the new final operand. Fix GET_MODE_NUNITS call.
995 Describe the scalar case too.
996 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
997 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
999 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1000 (get_conditional_internal_fn): Update comment.
1001 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1002 accumulator value as a final argument to conditional functions.
1003 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1004 a define_expand and add an "else" operand. Assert for now that
1005 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1006 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1007 (*cond_<optab><mode>): New patterns.
1008 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1009 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1010 (UNSPEC_COND_EOR): Delete.
1011 (optab): Remove associated mappings.
1012 (SVE_INT_BINARY): New code iterator.
1013 (sve_int_op): Remove int attribute and add "minus" to the code
1015 (SVE_COND_INT_OP): Delete.
1016 (SVE_COND_FP_OP): Rename to...
1017 (SVE_COND_FP_BINARY): ...this.
1019 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1021 * optabs.c (can_reuse_operands_p): New function.
1022 (maybe_legitimize_operands): Try to reuse the results for
1025 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1027 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1028 Add {q} suffix to insn mnemonic.
1030 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1032 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1033 (msp430_warn_func_return): New.
1035 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1037 * fold-const.c (tree_nonzero_bits): New function.
1038 * fold-const.h (tree_nonzero_bits): Likewise.
1039 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1040 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1042 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1046 * varasm.c (assemble_alias): Check ifunc_resolver only on
1049 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1052 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1053 when memory input operand is handled.
1055 2018-05-24 Luis Machado <luis.machado@linaro.org>
1057 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1059 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1061 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1063 * match.pd: Delay FMA folds until after vectorization.
1065 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1068 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1069 address check not strict.
1071 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1073 * gimple-match.h (gimple_match_op): New class.
1074 (mprts_hook): Replace parameters with a gimple_match_op *.
1075 (maybe_build_generic_op): Likewise.
1076 (gimple_simplified_result_is_gimple_val): Replace parameters with
1077 a const gimple_match_op *.
1078 (gimple_simplify): Replace code_helper * and tree * parameters with
1079 a gimple_match_op * parameter.
1080 (gimple_resimplify1): Replace code_helper *, tree and tree *
1081 parameters with a gimple_match_op * parameter.
1082 (gimple_resimplify2): Likewise.
1083 (gimple_resimplify3): Likewise.
1084 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1085 parameters with a gimple_match_op * parameter.
1086 * gimple-match-head.c (gimple_simplify): Change prototypes of
1087 auto-generated functions to take a gimple_match_op * instead of
1088 separate code_helper * and tree * parameters. Make the same
1089 change in the top-level overload and update calls to the
1090 gimple_resimplify routines. Update calls to the auto-generated
1091 functions and to maybe_push_res_to_seq in the publicly-facing
1092 operation-specific gimple_simplify overloads.
1093 (gimple_match_op::MAX_NUM_OPS): Define.
1094 (gimple_resimplify1): Replace rcode and ops with a single res_op
1095 parameter. Update call to gimple_simplify.
1096 (gimple_resimplify2): Likewise.
1097 (gimple_resimplify3): Likewise.
1098 (mprts_hook): Replace parameters with a gimple_match_op *.
1099 (maybe_build_generic_op): Likewise.
1100 (build_call_internal): Replace type, nargs and ops with
1101 a gimple_match_op *.
1102 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1103 with a single gimple_match_op *. Update calls to mprts_hook,
1104 build_call_internal and gimple_simplified_result_is_gimple_val.
1105 Factor out code that is common to the tree_code and combined_fn cases.
1106 * genmatch.c (expr::gen_transform): Replace tem_code and
1107 tem_ops with a gimple_match_op called tem_op. Update calls
1108 to the gimple_resimplify functions and maybe_push_res_to_seq.
1109 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1110 res_ops. Update call to the gimple_resimplify functions.
1111 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1112 (decision_tree::gen): Make the functions take a gimple_match_op *
1113 called res_op instead of separate res_code and res_ops parameters.
1114 Update call accordingly.
1115 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1116 and ops with a single res_op parameter. Update calls to
1117 maybe_build_generic_op and maybe_push_res_to_seq.
1118 (fold_stmt_1): Update calls to gimple_simplify and
1119 replace_stmt_with_simplification.
1120 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1121 and gimple_simplified_result_is_gimple_val.
1122 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1124 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1125 with a gimple_match_op *.
1126 (vn_nary_build_or_lookup): Likewise. Update call to
1127 vn_nary_build_or_lookup_1.
1128 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1129 gimple_match_op *. Update calls to the gimple_resimplify routines
1130 and to gimple_simplified_result_is_gimple_val.
1131 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1132 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1133 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1134 (visit_nary_op): Likewise.
1135 (visit_reference_op_load): Likewise.
1137 2018-05-23 Luis Machado <luis.machado@linaro.org>
1139 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1140 modifier for printing the step amount.
1142 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1145 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1148 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1150 * doc/sourcebuild.texi (Endianness): New subsubsection.
1152 2018-05-23 Luis Machado <luis.machado@linaro.org>
1154 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1155 <prefetch_dynamic_strides>: New const bool field.
1156 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1157 prefetch_dynamic_strides.
1158 (exynosm1_prefetch_tune): Likewise.
1159 (thunderxt88_prefetch_tune): Likewise.
1160 (thunderx_prefetch_tune): Likewise.
1161 (thunderx2t99_prefetch_tune): Likewise.
1162 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1164 (aarch64_override_options_internal): Update to set
1165 PARAM_PREFETCH_DYNAMIC_STRIDES.
1166 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1167 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1168 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1169 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1170 prefetch-dynamic-strides setting.
1172 2018-05-23 Luis Machado <luis.machado@linaro.org>
1174 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1175 <minimum_stride>: New const int field.
1176 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1177 minimum_stride field defaulting to -1.
1178 (exynosm1_prefetch_tune): Likewise.
1179 (thunderxt88_prefetch_tune): Likewise.
1180 (thunderx_prefetch_tune): Likewise.
1181 (thunderx2t99_prefetch_tune): Likewise.
1182 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1183 <default_opt_level>: Set to 3.
1184 (aarch64_override_options_internal): Update to set
1185 PARAM_PREFETCH_MINIMUM_STRIDE.
1186 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1187 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1188 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1189 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1190 stride is constant and is below the minimum stride threshold.
1192 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1194 * config/arm/arm-cpus.in (mode26): Delete.
1195 (armv4): Delete mode26 reference.
1196 * config/arm/arm.c (arm_configure_build_target): Delete use of
1199 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1201 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1203 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1204 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1206 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1207 (floatunsdidf2): Ditto.
1209 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1211 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1212 (fixuns_trunc<mode>si2_avx512f): Ditto.
1213 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1214 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1215 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1217 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1219 PR rtl-optimization/79985
1220 * df-scan.c (df_insn_refs_collect): Remove special case for
1221 global registers and asm statements.
1223 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1225 * extend.texi (Global Register Variables): Rewrite the bullet list.
1226 Note that the register is available for allocation. Note that access
1227 via inline asm must use constraints. Add note about async-signal
1228 handlers. Remove paragraph about automagic register selection.
1230 2018-05-23 Richard Biener <rguenther@suse.de>
1232 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1233 of fixed offset from memset VN.
1235 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1237 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1239 (alloc_cand_and_find_basis): Initialize first_interp field.
1240 (slsr_process_mul): Modify first_interp field.
1241 (slsr_process_add): Likewise.
1242 (slsr_process_cast): Modify first_interp field for each new
1244 (slsr_process_copy): Likewise.
1245 (dump_candidate): Dump first_interp field.
1246 (replace_mult_candidate): Process all interpretations, not just
1248 (replace_rhs_if_not_dup): Likewise.
1249 (replace_one_candidate): Likewise.
1251 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1253 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1255 (aarch64_needs_frame_chain): New function.
1256 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1258 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1261 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1262 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1263 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1264 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1265 as true for strict-align.
1266 (aarch64_can_inline_p): Perform checks even when callee has no
1267 attributes to check for strict alignment.
1268 * doc/extend.texi (AArch64 Function Attributes): Document
1270 * doc/invoke.texi: (AArch64 Options): Likewise.
1272 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1274 PR tree-optimization/85853
1275 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1276 the handling of the root of the node to...
1277 (vect_slp_analyze_node_operations_1): ...this new function,
1278 and run the whole thing with the child nodes' def types
1279 set according to their SLP node's def type.
1281 2018-05-23 Richard Biener <rguenther@suse.de>
1284 * tree-data-ref.c (create_runtime_alias_checks): Defer
1285 and ignore overflow warnings.
1287 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1289 PR tree-optimization/85822
1290 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1293 2018-05-23 Richard Biener <rguenther@suse.de>
1295 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1296 memset constants via native_interpret_expr.
1298 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1301 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1303 (cgraph_node::create_alias): Likewise.
1304 (cgraph_node::get_availability): Check ifunc_resolver instead
1305 of looking up ifunc attribute.
1306 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1307 * varasm.c (do_assemble_alias): Likewise.
1308 (assemble_alias): Likewise.
1309 (default_binds_local_p_3): Likewise.
1310 * cgraph.h (cgraph_node): Add ifunc_resolver.
1311 (cgraph_node::only_called_directly_or_aliased_p): Return false
1313 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1315 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1316 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1317 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1318 instead of looking up ifunc attribute.
1320 2018-05-22 Luis Machado <luis.machado@linaro.org>
1322 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1324 2018-05-22 Martin Sebor <msebor@redhat.com>
1327 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1328 only when expasion succeeds.
1329 (expand_builtin_strcmp): Same.
1330 (expand_builtin_strncmp): Same.
1332 2018-05-22 Martin Sebor <msebor@redhat.com>
1334 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1336 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1337 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1339 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1340 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1342 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1343 Add check for legitimate_address.
1344 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1345 (aarch64_swap_ldrstr_operands): New.
1346 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1349 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1350 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1352 * config/aarch64/aarch64.md: New patterns to generate stp
1354 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1355 single words and double words.
1356 (load_pair_sw, load_pair_dw): Likewise.
1357 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1359 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1361 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1362 for different mode ldpstp and add peephole for merged zero stores.
1364 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1366 (aarch64_gen_store_pair): Rename calls to match new patterns.
1367 (aarch64_gen_load_pair): Rename calls to match new patterns.
1368 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1369 (load_pair<DREG:mode><DREG2:mode>): ... This.
1370 (store_pair<mode>): Rename to...
1371 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1372 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1374 (V_INT_EQUIV): Handle SImode.
1375 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1378 2018-05-22 Martin Sebor <msebor@redhat.com>
1381 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1382 or ajust the presumed bound on an operation to avoid unnecessary
1385 2018-05-22 Martin Sebor <msebor@redhat.com>
1387 PR tree-optimization/85826
1388 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1389 assuming that a DECL necesarily has a constant size.
1391 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1394 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1396 2018-05-22 Richard Biener <rguenther@suse.de>
1398 PR tree-optimization/85834
1399 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1400 non-constant and non-zero memset arguments.
1402 2018-05-22 Martin Liska <mliska@suse.cz>
1405 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1407 2018-05-22 Richard Biener <rguenther@suse.de>
1409 PR tree-optimization/85863
1410 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1411 comparisons when vectype is specified.
1412 (vectorizable_condition): Do not specify vectype for
1413 vect_is_simple_cond when SLP vectorizing.
1415 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1418 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1419 define __ibm128 as long double.
1420 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1421 as a distinct type when IEEE 128-bit support is enabled.
1422 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1423 128-bit types to use the correct functions.
1424 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1425 convert between 128-bit floating point types that have different
1426 modes but the same representation, instead of using gen_lowpart to
1428 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1430 (IFKF_reg): New attributes to give the register constraints for
1432 (extend<mode>tf2_internal): New insns to mark an explicit
1433 conversion between 128-bit floating point types that have a
1434 different mode but share the same representation.
1436 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1438 PR tree-optimization/85814
1439 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1440 a null return from get_strinfo when unsharing the next
1441 strinfo in the chain.
1443 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1446 * varasm.c (weak_finish): Clean up weak_decls.
1448 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1450 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1451 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1452 UNSPEC_UADALP values.
1453 * config/aarch64/iterators.md (ABAL): New int iterator.
1456 (sur): Add mappings for the above.
1457 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1459 (aarch64_<sur>abal<mode>_4): Likewise.
1460 (aarch64_<sur>adalp<mode>_3): Likewise.
1461 (<sur>sadv16qi): New define_expand.
1463 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1465 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1466 (*movdf_internal): Ditto.
1467 (*rcpsf2_sse): Ditto.
1468 (*rsqrtsf2_sse): Ditto.
1469 (*sqrt<mode>2_sse): Ditto.
1471 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1473 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1475 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1476 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1477 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1478 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1480 * config/aarch64/arm_neon.h: Likewise.
1481 * config/aarch64/iterators.md (VQ_I): New.
1483 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1485 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1487 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1490 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1492 * config/nds32/constraints.md (S): New constraint.
1493 * config/nds32/nds32.md (call_internal): Use constraint S.
1494 (call_value_internal): Likewise.
1495 (sibcall_internal): Likewise.
1496 (sibcall_value_internal): Likewise.
1498 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1499 Chung-Ju Wu <jasonwucj@gmail.com>
1501 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1504 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1505 Chung-Ju Wu <jasonwucj@gmail.com>
1507 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1508 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1509 (nds32_rtx_costs_impl): Simplify.
1510 (nds32_address_cost_impl): Simplify.
1511 (nds32_init_rtx_costs): New function.
1512 (nds32_rtx_costs_speed_prefer): Likewise.
1513 (nds32_rtx_costs_size_prefer): Likewise.
1514 (nds32_address_cost_speed_prefer): Likewise.
1515 (nds32_address_cost_speed_fwprop): Likewise.
1516 (nds32_address_cost_size_prefer): Likewise.
1517 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1518 * config/nds32/nds32.c (nds32_option_override): Use
1519 nds32_init_rtx_costs function.
1521 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1523 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1524 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1525 (TARGET_PIPELINE_N8): Likewise.
1526 (TARGET_PIPELINE_N10): Likewise.
1527 (TARGET_PIPELINE_N13): Likewise.
1528 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1530 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1532 * config/nds32/nds32-fpu.md: Update copyright year.
1534 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1536 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1538 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1540 * config/nds32/nds32.c
1541 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1542 * config/nds32/nds32.opt (minline-asm-r15): New option.
1544 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1546 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1548 * config/nds32/nds32.md (abssi2): New pattern.
1550 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1552 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1553 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1554 Merge insn pattern from sse_cvtsi2ss<round_name> and
1555 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1556 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1557 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1558 using SWI48 mode iterator.
1559 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1560 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1561 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1562 pattern from sse_cvttss2si<round_saeonly_name>
1563 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1564 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1565 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1566 using SWI48 mode iterator.
1567 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1568 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1569 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1570 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1571 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1572 using SWI48 mode iterator.
1573 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1574 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1575 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1576 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1577 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1578 SWI48 mode iterator.
1579 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1580 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1581 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1582 pattern from sse_cvttsd2si<round_saeonly_name>
1583 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1585 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1587 * config/nds32/nds32-md-auxiliary.c
1588 (nds32_valid_smw_lwm_base_p): Refine.
1589 (nds32_output_smw_single_word): Refine.
1590 (nds32_output_smw_double_word): New.
1591 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1593 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1595 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1596 (nds32_output_stack_pop): Refine.
1597 (nds32_expand_unaligned_load): Refine.
1598 (nds32_expand_unaligned_store): Refine.
1600 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1601 Chung-Ju Wu <jasonwucj@gmail.com>
1603 * config/nds32/constants.md: Add TP_REGNUM constant.
1604 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1605 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1607 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1608 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1609 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1610 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1611 and PIC code generation.
1612 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1614 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1616 * config/nds32/nds32.md: Support TLS and PIC.
1617 * config/nds32/nds32.c: Support TLS and PIC.
1618 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1619 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1622 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1624 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1625 mode with E_ prefix.
1627 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1628 Chung-Ju Wu <jasonwucj@gmail.com>
1630 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1631 * config/nds32/nds32-md-auxiliary.c
1632 (symbolic_reference_mentioned_p): New.
1633 (nds32_legitimize_ict_address): New.
1634 (nds32_expand_ict_move): New.
1635 (nds32_indirect_call_referenced_p): New.
1636 (nds32_symbol_binds_local_p): Delete.
1637 (nds32_long_call_p): Modify.
1638 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1639 * config/nds32/nds32-protos.h
1640 (symbolic_reference_mentioned_p): Declare.
1641 (nds32_legitimize_ict_address): Declare.
1642 (nds32_expand_ict_move): Declare.
1643 (nds32_indirect_call_referenced_p): Declare.
1644 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1645 (nds32_relax_group): Use nds32_ict_const_p as condition.
1646 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1647 (nds32_asm_file_start): Output ict_model directive in asm code.
1648 (nds32_legitimate_address_p): Consider indirect call.
1649 (nds32_print_operand): Consider indirect call.
1650 (nds32_print_operand_address): Consider indirect call.
1651 (nds32_insert_attributes): Handle "indirect_call" attribute.
1652 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1653 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1654 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1655 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1656 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1657 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1658 (TARGET_ICT_MODEL_SMALL): Define.
1659 (TARGET_ICT_MODEL_LARGE): Define.
1660 * config/nds32/nds32.md (movsi): Consider ict model.
1661 (call, call_value): Consider ict model.
1662 (sibcall, sibcall_value): Consider ict model.
1663 * config/nds32/nds32.opt (mict-model): New option.
1664 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1667 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1668 Monk Chiang <sh.chiang04@gmail.com>
1669 Jim Wilson <jimw@sifive.com>
1671 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1672 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1673 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1674 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1675 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1676 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1677 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1678 compute save_libcall_adjustment properly.
1679 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1680 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1681 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1682 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1683 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1684 (ABI_SPEC): Handle mabi=ilp32e.
1685 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1686 (RVE): Add RVE mask.
1687 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1688 <-march>: Add rv32e as an example.
1690 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1693 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1694 (intra_create_variable_infos): Handle C++ constructors.
1696 2018-05-18 Martin Liska <mliska@suse.cz>
1698 * passes.def: Remove a redundant pass.
1700 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1703 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1705 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1707 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1709 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1711 (armv2, armv2a, armv3, armv3m): Delete architectures.
1712 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1713 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1714 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1716 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1717 (*mulsidi3adddi): Likewise.
1718 (mulsidi3): Likewise.
1719 (*mulsidi3_nov6): Likewise.
1720 (umulsidi3): Likewise.
1721 (umulsidi3_nov6): Likewise.
1722 (umaddsidi4): Likewise.
1723 (*umulsidi3adddi): Likewise.
1724 (smulsi3_highpart): Likewise.
1725 (*smulsi3_highpart_nov6): Likewise.
1726 (umulsi3_highpart): Likewise.
1727 (*umulsi3_highpart_nov6): Likewise.
1728 * config/arm/arm.h (arm_arch3m): Delete.
1729 * config/arm/arm.c (arm_arch3m): Delete.
1730 (arm_option_override_internal): Update armv3-related comment.
1731 (arm_configure_build_target): Delete use of isa_bit_mode32.
1732 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1733 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1734 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1736 (mulusa3): Likewise.
1737 * config/arm/arm-protos.h (arm_arch3m): Delete.
1738 * config/arm/arm-tables.opt: Regenerate.
1739 * config/arm/arm-tune.md: Likewise.
1740 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1741 deleted architectures.
1743 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1745 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1746 (armv5t, armv5te): New features.
1747 (ARMv5, ARMv5e): Delete fgroups.
1748 (ARMv5t, ARMv5te): Adjust for above changes.
1750 (armv5, armv5e): Delete arches.
1751 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1753 (*call_reg_arm): Likewise.
1754 (*call_value_reg_armv5): Likewise.
1755 (*call_value_reg_arm): Likewise.
1756 (*call_symbol): Likewise.
1757 (*call_value_symbol): Likewise.
1758 (*sibcall_insn): Likewise.
1759 (*sibcall_value_insn): Likewise.
1761 (prefetch): Likewise.
1762 (define_split and define_peephole2 dependent on arm_arch5):
1764 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1766 (TARGET_ARM_QBIT): Likewise.
1767 (TARGET_DSP_MULTIPLY): Likewise.
1768 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1769 (arm_arch5, arm_arch5e): Delete.
1770 (arm_arch5t, arm_arch5te): Declare.
1771 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1772 (arm_arch5t): Declare.
1773 (arm_option_reconfigure_globals): Update for the above.
1774 (arm_options_perform_arch_sanity_checks): Update comment, replace
1775 use of arm_arch5 with arm_arch5t.
1776 (use_return_insn): Likewise.
1777 (arm_emit_call_insn): Likewise.
1778 (output_return_instruction): Likewise.
1779 (arm_final_prescan_insn): Likewise.
1780 (arm_coproc_builtin_available): Likewise.
1781 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1782 arm_arch5e with arm_arch5t and arm_arch5te.
1783 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1784 (arm_arch5t, arm_arch5te): Declare.
1785 * config/arm/arm-tables.opt: Regenerate.
1786 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1787 * config/arm/t-multilib: Likewise.
1788 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1789 instead of arm_arch5.
1790 (*call_reg_thumb1): Likewise.
1791 (*call_value_reg_thumb1_v5): Likewise.
1792 (*call_value_reg_thumb1): Likewise.
1793 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1795 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1797 2018-05-18 Martin Liska <mliska@suse.cz>
1799 PR gcov-profile/84846
1800 * doc/gcov.texi: Document -t option of gcov tool.
1802 2018-05-18 Martin Liska <mliska@suse.cz>
1804 PR gcov-profile/84846
1805 * gcov.c (print_usage): Add new -t option.
1806 (process_args): Handle the option.
1807 (generate_results): Use stdout as output when requested by
1810 2018-05-18 Martin Liska <mliska@suse.cz>
1812 PR gcov-profile/84846
1813 * coverage.c (coverage_init): Write PWD to .gcno file.
1814 * doc/gcov.texi: Document how working directory is printed.
1815 * gcov-dump.c (dump_gcov_file): Print PWD.
1816 * gcov.c (output_intermediate_file): Likewise.
1817 (read_graph_file): Read PWD string.
1818 (output_lines): Print PWD.
1820 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1823 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1824 for retval and return false if all args to phi are zero.
1826 2018-05-18 Richard Biener <rguenther@suse.de>
1828 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1830 (evrp_dom_walker::before_dom_children): Call it.
1832 2018-05-18 Richard Biener <rguenther@suse.de>
1834 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1835 results when processing array refs with variable index.
1837 2018-05-18 Toon Moene <toon@moene.org>
1839 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1840 directly after that of -floop-interchange. Indicate that both
1841 options are enabled by default when specifying -O3.
1843 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1845 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1846 iterator. Delete separate integer-mode vec_set<mode> expander.
1847 (aarch64_simd_vec_setv2di): Delete.
1848 (vec_setv2di): Delete.
1849 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1850 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1851 the "w, r" alternative.
1853 2018-05-18 Martin Liska <mliska@suse.cz>
1855 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1856 * tree-pass.h (make_pass_lower_switch_O0): New function.
1857 * tree-switch-conversion.c (node_has_low_bound): Remove.
1858 (node_has_high_bound): Likewise.
1859 (node_is_bounded): Likewise.
1860 (class pass_lower_switch): Make it a template type and create
1862 (pass_lower_switch::execute): Add template argument.
1863 (make_pass_lower_switch): New function.
1864 (make_pass_lower_switch_O0): New function.
1865 (do_jump_if_equal): Remove.
1866 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1867 all the hard work to tree optimization passes.
1869 2018-05-18 Martin Liska <mliska@suse.cz>
1871 * dbgcnt.c (limit_low): Renamed from limit.
1872 (limit_high): New variable.
1873 (dbg_cnt_is_enabled): Check for upper limit.
1874 (dbg_cnt): Adjust dumping.
1875 (dbg_cnt_set_limit_by_index): Add new argument for high
1877 (dbg_cnt_set_limit_by_name): Likewise.
1878 (dbg_cnt_process_single_pair): Parse new format.
1879 (dbg_cnt_process_opt): Use strtok.
1880 (dbg_cnt_list_all_counters): Remove 'value' and add
1882 * doc/invoke.texi: Document changes.
1884 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1886 * doc/sourcebuild.texi (scalar_all_fma): Document.
1887 * tree.def (FMA_EXPR): Delete.
1888 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1889 * internal-fn.c (ternary_direct): New macro.
1890 (expand_ternary_optab_fn): Likewise.
1891 (direct_ternary_optab_supported_p): Likewise.
1892 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1893 * builtins.c (fold_builtin_fma): Delete.
1894 (fold_builtin_3): Don't call it.
1895 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1896 * expr.c (expand_expr_real_2): Likewise.
1897 * fold-const.c (operand_equal_p): Likewise.
1898 (fold_ternary_loc): Likewise.
1899 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1900 * gimple.c (DEFTREECODE): Likewise.
1901 * gimplify.c (gimplify_expr): Likewise.
1902 * optabs-tree.c (optab_for_tree_code): Likewise.
1903 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1904 * tree-eh.c (operation_could_trap_p): Likewise.
1905 (stmt_could_throw_1_p): Likewise.
1906 * tree-inline.c (estimate_operator_cost): Likewise.
1907 * tree-pretty-print.c (dump_generic_node): Likewise.
1908 (op_code_prio): Likewise.
1909 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1910 * tree-ssa-operands.c (get_expr_operands): Likewise.
1911 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1912 * fold-const-call.h (fold_fma): Delete.
1913 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1914 CFN_FNMA and CFN_FNMS.
1916 * genmatch.c (combined_fn): New enum.
1917 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1918 (commutative_op): New function.
1919 (commutate): Use it. Handle more than 2 operands.
1920 (dt_operand::gen_gimple_expr): Use commutative_op.
1921 (parser::parse_expr): Allow :c to be used with non-binary
1922 operators if the commutative operand is known.
1923 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1924 CFN_FMS, CFN_FNMA and CFN_FNMS.
1925 (backprop::process_assign_use): Remove FMA_EXPR handling.
1926 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1927 (gen_hsa_fma): New function.
1928 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1929 IFN_FNMA and IFN_FNMS.
1930 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1931 * gimple-fold.h (follow_all_ssa_edges): Declare.
1932 * gimple-fold.c (follow_all_ssa_edges): New function.
1933 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1934 gimple_build interface and use follow_all_ssa_edges to fold the result.
1935 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1936 instead of checking for optabs directly.
1937 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1938 rather than FMA_EXPRs.
1939 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1940 call to IFN_FMA instead of an FMA_EXPR.
1942 2018-05-17 Jim Wilson <jimw@sifive.com>
1944 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1945 sign extended promoted subreg, and the range does not have the sign bit
1946 set, then do a sign extend.
1948 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1949 test, check for sign extended subreg and/or constant operands, and
1950 do a sign extend in that case.
1952 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1954 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1955 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1957 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1958 Change logics_shift_reg to logics_shift_imm.
1959 (thunderx2t99_fp_loadpair_basic): Delete.
1960 (thunderx2t99_fp_storepair_basic): Delete.
1961 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1962 (thunderx2t99_asimd_polynomial): Delete.
1963 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1964 and neon_fp_mul_d_scalar_q.
1965 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1966 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1967 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1968 (thunderx2t99_asimd_lut): Add missing tbl types.
1969 (thunderx2t99_asimd_ext): Delete.
1970 (thunderx2t99_asimd_load1_1_mult): Delete.
1971 (thunderx2t99_asimd_load1_2_mult): Delete.
1972 (thunderx2t99_asimd_load1_ldp): New.
1973 (thunderx2t99_asimd_load1): New.
1974 (thunderx2t99_asimd_load2): Add missing *load2* types.
1975 (thunderx2t99_asimd_load3): New.
1976 (thunderx2t99_asimd_load4): New.
1977 (thunderx2t99_asimd_store1_1_mult): Delete.
1978 (thunderx2t99_asimd_store1_2_mult): Delete.
1979 (thunderx2t99_asimd_store2_mult): Delete.
1980 (thunderx2t99_asimd_store2_onelane): Delete.
1981 (thunderx2t99_asimd_store_stp): New.
1982 (thunderx2t99_asimd_store1): New.
1983 (thunderx2t99_asimd_store2): New.
1984 (thunderx2t99_asimd_store3): New.
1985 (thunderx2t99_asimd_store4): New.
1987 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1989 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1990 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1992 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1993 Segher Boessenkool <segher@kernel.crashing.org>
1996 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1999 2018-05-17 Richard Biener <rguenther@suse.de>
2001 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2002 for pruning loop and prune defs feeding only already visited PHIs.
2004 2018-05-17 Richard Biener <rguenther@suse.de>
2006 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2008 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2009 Richard Biener <rguenther@suse.de>
2011 PR tree-optimization/85793
2012 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2013 for VMAT_ELEMENTWISE.
2015 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2017 * internal-fn.h (lookup_internal_fn): Declare
2018 * internal-fn.c (lookup_internal_fn): New function.
2019 * gimple.c (gimple_build_call_from_tree): Handle calls to
2021 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2022 internal function names.
2023 * tree-pretty-print.c (dump_generic_node): Likewise.
2024 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2026 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2028 * gimple-fold.h (gimple_build): Make the function forms take
2029 combined_fn rather than built_in_function.
2030 (gimple_simplify): Likewise.
2031 * gimple-match-head.c (gimple_simplify): Likewise.
2032 * gimple-fold.c (gimple_build): Likewise.
2033 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2034 rather than gimple_build_call_internal.
2035 (get_initial_defs_for_reduction): Likewise.
2036 (vect_create_epilog_for_reduction): Likewise.
2037 (vectorizable_live_operation): Likewise.
2039 2018-05-17 Martin Liska <mliska@suse.cz>
2041 * gimple-ssa-sprintf.c (format_directive): Do not use
2042 space in between 'G_' and '('.
2044 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2047 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2048 even if the mask is not all ones.
2051 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2053 (ix86_gimple_fold_builtin): Likewise.
2056 * config/i386/i386.c: Include tree-vector-builder.h.
2057 (ix86_vector_shift_count): New function.
2058 (ix86_fold_builtin): Fold shift builtins by scalar count.
2059 (ix86_gimple_fold_builtin): Likewise.
2061 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2062 _mm512_setzero): New intrinsics.
2064 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2065 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2067 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2068 code generation for cases where splatting a value is not useful.
2069 * simplify-rtx.c (simplify_ternary_operation): Simplify
2070 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2071 mode to a vec_concat.
2073 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2075 * config.gcc: Support "goldmont-plus".
2076 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2078 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2079 PROCESSOR_GOLDMONT_PLUS.
2080 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2081 (processor_target_table): Add "goldmont-plus".
2082 (PTA_GOLDMONT_PLUS): Define.
2083 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2084 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2085 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2086 (fold_builtin_cpu): Add "goldmont-plus".
2087 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2088 (ix86_option_override_internal): Add "goldmont-plus".
2089 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2090 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2091 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2092 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2094 2018-05-17 Richard Biener <rguenther@suse.de>
2096 PR tree-optimization/85757
2097 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2098 remove defs that only feed that PHI from further processing.
2100 2018-05-16 Jim Wilson <jimw@sifive.com>
2102 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2104 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2106 2018-05-16 Mark Wielaard <mark@klomp.org>
2108 * dwarf2out.c (count_index_strings): New function.
2109 (output_indirect_strings): Call count_index_strings and generate
2110 header for dwarf_version >= 5.
2112 2018-05-16 Mark Wielaard <mark@klomp.org>
2114 * dwarf2out.c (dwarf_FORM): New function.
2115 (set_indirect_string): Use dwarf_FORM.
2116 (reset_indirect_string): Likewise.
2117 (size_of_die): Likewise.
2118 (value_format): Likewise.
2119 (output_die): Likewise.
2120 (add_skeleton_AT_string): Likewise.
2121 (output_macinfo_op): Likewise.
2122 (index_string): Likewise.
2123 (output_index_string_offset): Likewise.
2124 (output_index_string): Likewise.
2125 (count_index_strings): Likewise.
2127 2018-05-16 Carl Love <cel@us.ibm.com>
2129 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2130 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2132 2018-05-16 Martin Jambor <mjambor@suse.cz>
2134 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2135 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2137 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2139 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2140 (fnma<mode>4): Likewise.
2141 (fms<mode>4): Likewise.
2142 (fnms<mode>4): Likewise.
2143 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2144 (aarch64_fnma<mode>4): Likewise.
2145 (aarch64_fms<mode>4): Likewise.
2146 (aarch64_fnms<mode>4): Likewise.
2147 (aarch64_fnmadd<mode>4): Likewise.
2149 2018-05-16 Jason Merrill <jason@redhat.com>
2151 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2153 2018-05-16 Richard Biener <rguenther@suse.de>
2155 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2156 (dump_stmt_cost): Declare.
2157 (add_stmt_cost): Dump cost we add.
2158 (add_stmt_costs): New function.
2159 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2161 (vect_analyze_stmt): Adjust prototype.
2162 (vectorizable_condition): Likewise.
2163 (vectorizable_live_operation): Likewise.
2164 (vectorizable_reduction): Likewise.
2165 (vectorizable_induction): Likewise.
2166 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2167 cost vector to pass to vectorizable_ and record afterwards.
2168 (vect_model_reduction_cost): Take cost vector argument and adjust.
2169 (vect_model_induction_cost): Likewise.
2170 (vectorizable_reduction): Likewise.
2171 (vectorizable_induction): Likewise.
2172 (vectorizable_live_operation): Likewise.
2173 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2174 SLP_TREE_NUMBER_OF_VEC_STMTS.
2175 (vect_analyze_slp_cost_1): Remove.
2176 (vect_analyze_slp_cost): Likewise.
2177 (vect_slp_analyze_node_operations): Take visited args and
2178 a target cost vector. Avoid processing already visited stmt sets.
2179 (vect_slp_analyze_operations): Use a local cost vector to gather
2180 costs and register those of non-discarded instances.
2181 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2182 (vect_schedule_slp_instance): Remove copying of
2183 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2185 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2186 adding cost. Record cost entry location.
2187 (vect_prologue_cost_for_slp_op): Function to compute cost of
2188 a constant or invariant generated for SLP vect in the prologue,
2189 split out from vect_analyze_slp_cost_1.
2190 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2191 (vect_model_promotion_demotion_cost): Likewise.
2192 (vect_model_store_cost): Likewise, make static.
2193 (vect_model_load_cost): Likewise.
2194 (vectorizable_bswap): Add cost vector arg and adjust.
2195 (vectorizable_call): Likewise.
2196 (vectorizable_simd_clone_call): Likewise.
2197 (vectorizable_conversion): Likewise.
2198 (vectorizable_assignment): Likewise.
2199 (vectorizable_shift): Likewise.
2200 (vectorizable_operation): Likewise.
2201 (vectorizable_store): Likewise.
2202 (vectorizable_load): Likewise.
2203 (vectorizable_condition): Likewise.
2204 (vectorizable_comparison): Likewise.
2205 (can_vectorize_live_stmts): Likewise.
2206 (vect_analyze_stmt): Likewise.
2207 (vect_transform_stmt): Adjust calls to vectorizable_*.
2208 * tree-vectorizer.c: Include gimple-pretty-print.h.
2209 (dump_stmt_cost): New function.
2211 2018-05-16 Richard Biener <rguenther@suse.de>
2213 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2214 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2215 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2216 (check_name): New callback.
2217 (dse_classify_store): Track cycles via a visited bitmap of PHI
2218 defs and simplify handling of in-loop and across loop dead stores
2219 and properly fail for loop-variant refs. Handle byte-tracking with
2220 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2223 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2225 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2226 (vect_get_mask_type_for_stmt): Likewise.
2227 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2229 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2230 to determine the statement's vector type and the vector type that
2231 should be used for calculating nunits. Deal with cases in which
2232 the type has to be deferred.
2233 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2234 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2235 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2236 (vect_determine_vf_for_stmt): New functions, split out from...
2237 (vect_determine_vectorization_factor): ...here.
2238 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2239 (vect_get_mask_type_for_stmt): New functions, split out from
2240 vect_determine_vectorization_factor.
2242 2018-05-16 Richard Biener <rguenther@suse.de>
2244 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2245 verify the [VEC_]COND_EXPR embedded comparison.
2247 2018-05-15 Martin Sebor <msebor@redhat.com>
2249 PR tree-optimization/85753
2250 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2251 RECORD_TYPE in addition to ARRAY_TYPE.
2253 2018-05-15 Martin Sebor <msebor@redhat.com>
2256 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2258 2018-05-15 Richard Biener <rguenther@suse.de>
2260 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2261 add by_clobber_p one. Change algorithm to collect all defs
2262 representing uses we need to walk and try reducing them to
2263 a single one before failing.
2264 (dse_dom_walker::dse_optimize_stmt): Adjust.
2266 2018-05-13 Mark Wielaard <mark@klomp.org>
2268 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2269 (size_of_loc_descr): Likewise.
2270 (output_loc_operands): Likewise.
2271 (output_loc_operands_raw): Likewise.
2272 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2273 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2274 (hash_loc_operands): Likewise.
2275 (compare_loc_operands): Likewise.
2277 2018-05-14 Mark Wielaard <mark@klomp.org>
2279 * dwarf2out.c (count_index_addrs): New function.
2280 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2282 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2284 PR tree-optimization/83648
2285 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2286 return value as malloc candidate.
2288 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2291 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2292 as true in call to suggest_attribute.
2294 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2296 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2297 -mreadonly-in-sdata.
2299 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2301 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2303 (aarch64_crypto_aesd_fused): Likewise.
2305 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2307 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2308 (movsi_aarch64): Likewise.
2309 (load_pairsi): Likewise.
2310 (load_pairdi): Likewise.
2311 (store_pairsi): Likewise.
2312 (store_pairdi): Likewise.
2313 (load_pairsf): Likewise.
2314 (load_pairdf): Likewise.
2315 (store_pairsf): Likewise.
2316 (store_pairdf): Likewise.
2317 (zero_extend): Likewise.
2318 (trunc): Swap alternatives.
2319 (fcvt_target): Add '?' to prefer w over r.
2321 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2324 * config/i386/i386.md: Disallow non-commutative arithmetics in
2325 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2326 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2327 in the peephole2 before it.
2329 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2331 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2332 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2333 (ix86_handle_option): Handle -mcldemote.
2334 * config.gcc: New header.
2335 * config/i386/cldemoteintrin.h: New file.
2336 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2337 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2339 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2340 OPTION_MASK_ISA_CLDEMOTE.
2341 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2342 (ix86_valid_target_attribute_inner_p): Ditto.
2343 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2344 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2345 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2346 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2347 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2349 * config/i386/i386.opt: Add -mcldemote.
2350 * config/i386/x86intrin.h: New header.
2351 * doc/invoke.texi: Add -mcldemote.
2353 2018-05-14 Richard Biener <rguenther@suse.de>
2355 * doc/match-and-simplify.texi: Adjust :s documentation.
2357 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2359 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2360 intended memcpy size.
2361 (REORDER_45): Likewise.
2363 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2365 * sort.cc: New file.
2366 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2367 * vec.c (qsort_chk): Use gcc_qsort.
2368 * Makefile.in (OBJS-libcommon): Add sort.o.
2369 (build/sort.o): New target. Use it...
2370 (BUILD_RTL): ... here, and...
2371 (build/gencfn-macros): ... here, and...
2372 (build/genmatch): ... here.
2374 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2375 Chung-Ju Wu <jasonwucj@gmail.com>
2377 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2378 * config/nds32/nds32-graywolf.md: New file.
2379 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2380 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2382 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2383 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2384 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2385 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2386 * config/nds32/pipelines.md: Include n15 settings.
2388 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2389 Chung-Ju Wu <jasonwucj@gmail.com>
2391 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2392 * config/nds32/nds32-n13.md: New file.
2393 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2394 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2396 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2397 * config/nds32/nds32.md (pipeline_model): Add n13.
2398 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2399 * config/nds32/pipelines.md: Include n13 settings.
2401 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2402 Chung-Ju Wu <jasonwucj@gmail.com>
2404 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2405 * config/nds32/nds32-n10.md: New file.
2406 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2407 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2409 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2410 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2411 * config/nds32/nds32.md (pipeline_model): Add n10.
2412 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2413 * config/nds32/pipelines.md: Include n10 settings.
2415 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2416 Kito Cheng <kito.cheng@gmail.com>
2417 Chung-Ju Wu <jasonwucj@gmail.com>
2419 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2420 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2421 Add enum values for DSP extension instructions.
2422 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2424 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2425 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2427 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2428 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2429 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2430 * config/nds32/nds32-intrinsic.md: Likewise.
2431 * config/nds32/nds32_intrinsic.h: Likewise.
2432 * config/nds32/nds32-md-auxiliary.c: Likewise.
2433 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2434 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2435 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2436 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2437 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2438 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2439 TYPE_DMAC in switch statement.
2440 * config/nds32/nds32.c: New checking and implementation for DSP
2441 extension instructions.
2442 * config/nds32/nds32.h: Likewise.
2443 * config/nds32/nds32.md: Likewise.
2444 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2445 * config/nds32/predicates.md: Implement new predicates for DSP
2448 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2450 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2451 Reformat alternatives and attributes so it is easier to identify
2452 which constraints/attributes go with which instruction.
2453 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2454 (mov<mode>_softfloat32, FMOVE64): Likewise.
2455 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2456 (mov<mode>_softfloat64, FMOVE64): Likewise.
2458 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2460 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2462 (Basic PowerPC Built-in Functions): The new name of the
2463 subsection previously known as "PowerPC Built-in Functions".
2464 (Basic PowerPC Built-in Functions Available on all Configurations):
2466 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2467 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2468 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2469 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2471 2018-05-11 Martin Jambor <mjambor@suse.cz>
2474 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2477 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2480 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2482 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2484 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2485 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2486 (ix86_handle_option): Handle -mwaitpkg.
2487 * config.gcc: New header.
2488 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2489 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2490 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2492 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2493 OPTION_MASK_ISA_WAITPKG.
2494 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2495 (ix86_option_override_internal): Add PTA_WAITPKG.
2496 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2497 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2498 IX86_BUILTIN_TPAUSE.
2499 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2500 __builtin_ia32_umwait and __builtin_ia32_tpause.
2501 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2502 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2503 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2504 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2505 UNSPECV_TPAUSE): New.
2506 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2507 * config/i386/i386.opt: Add -mwaitpkg.
2508 * config/i386/waitpkgintrin.h: New file.
2509 * config/i386/x86intrin.h: New header.
2510 * doc/invoke.texi: Add -mwaitpkg.
2512 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2515 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2517 (cortex-m0): Use armv6s-m isa.
2518 (cortex-m0plus): Likewise.
2519 (cortex-m1): Likewise.
2520 (cortex-m0.small-multiply): Likewise.
2521 (cortex-m0plus.small-multiply): Likewise.
2522 (cortex-m1.small-multiply): Likewise.
2524 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2525 Jakub Jelinek <jakub@redhat.com>
2527 PR tree-optimization/85692
2528 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2529 source permute as well.
2531 2018-05-11 Martin Liska <mliska@suse.cz>
2534 * doc/extend.texi: Document LLVM style format for no_sanitize
2537 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2539 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2540 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2541 (mode_supports_vsx_dform_quad): Likewise.
2542 (mode_supports_vmx_dform): Move these functions to be next to the
2543 other mode_supports functions.
2544 (mode_supports_dq_form): Likewise.
2545 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2546 mode_supports_dq_form.
2547 (reg_offset_addressing_ok_p): Likewise.
2548 (offsettable_ok_by_alignment): Likewise.
2549 (rs6000_legitimate_offset_address_p): Likewise.
2550 (legitimate_lo_sum_address_p): Likewise.
2551 (rs6000_legitimize_address): Likewise.
2552 (rs6000_legitimize_reload_address): Likewise.
2553 (rs6000_secondary_reload_inner): Likewise.
2554 (rs6000_preferred_reload_class): Likewise.
2555 (rs6000_output_move_128bit): Likewise.
2557 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2559 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2560 Generate SImode target register for null target.
2561 <case IX86_BUILTIN_XGETBV>: Ditto.
2562 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2563 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2565 2018-05-10 Carl Love <cel@us.ibm.com>
2567 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2568 dcbtt and dcbtstt if operands[2] is 0.
2570 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2573 * config/i386/sse.md (usadv64qi): New expander.
2575 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2577 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2578 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2579 -maltivec=be support.
2580 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2581 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2582 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2583 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2584 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2585 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2586 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2587 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2588 altivec_vsumsws): Adjust.
2589 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2590 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2591 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2593 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2594 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2595 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2596 (altivec_lve<VI_char>x): Delete expand.
2597 (*altivec_lve<VI_char>x_internal): Rename to...
2598 (altivec_lve<VI_char>x): ... this.
2599 (altivec_lvxl_<mode>): Delete expand.
2600 (*altivec_lvxl_<mode>_internal): Rename to ...
2601 (altivec_lvxl_<mode>): ... this.
2602 (altivec_stvxl_<mode>): Delete expand.
2603 (*altivec_stvxl_<mode>_internal): Rename to ...
2604 (altivec_stvxl_<mode>): ... this.
2605 (altivec_stve<VI_char>x): Delete expand.
2606 (*altivec_stve<VI_char>x_internal): Rename to ...
2607 (altivec_stve<VI_char>x): ... this.
2608 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2609 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2610 reduc_plus_scal_<mode>): Adjust.
2611 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2613 (rs6000_cpu_cpp_builtins): Adjust.
2614 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2615 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2616 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2617 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2618 -maltivec=be support.
2619 (rs6000_split_vec_extract_var): Adjust.
2620 (rs6000_split_v4si_init): Adjust.
2621 (swap_selector_for_mode): Delete.
2622 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2623 altivec_expand_stvex_be): Delete.
2624 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2625 -maltivec=be support.
2626 (rs6000_gimple_fold_builtin): Ditto.
2627 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2629 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2630 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2631 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2632 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2633 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2634 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2635 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2636 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2637 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2638 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2639 anonymous split): Adjust.
2640 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2641 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2643 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2645 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2646 when --with-gxx-include-dir is also specified.
2647 * configure: Regenerate.
2649 2018-05-09 Jim Wilson <jimw@sifive.com>
2652 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2653 * config/riscv/t-withmultilib: New.
2654 * config/riscv/withmultilib.h: New.
2655 * doc/install.texi: Document RISC-V --with-multilib-list support.
2657 2018-05-09 Richard Biener <rguenther@suse.de>
2659 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2661 (vect_bb_vectorization_profitable_p): Adjust. Compute
2662 actual scalar cost using the cost vector and the add_stmt_cost
2665 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2667 PR rtl-optimization/85645
2668 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2669 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2671 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2673 PR rtl-optimization/85645
2674 * shrink-wrap.c (spread_components): Return a boolean saying if
2675 anything was changed.
2676 (try_shrink_wrapping_separate): Iterate spread_components until
2677 nothing changes anymore.
2679 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2681 PR rtl-optimization/85645
2682 * regrename.c (build_def_use): Also kill the chains that include the
2683 destination of a REG_CFA_REGISTER note.
2685 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2687 PR rtl-optimization/85645
2688 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2689 insn that has a REG_CFA_REGISTER note.
2691 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2693 * cfgexpand.c (expand_clobber): New function.
2694 (expand_gimple_stmt_1): Use it.
2695 * tree-vect-stmts.c (vect_clobber_variable): New function,
2697 (vectorizable_simd_clone_call): ...here.
2698 (vectorizable_store): Emit a clobber either side of an
2699 IFN_STORE_LANES sequence.
2700 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2702 2018-05-09 Tom de Vries <tom@codesourcery.com>
2705 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2706 (define_insn "trap_if_false"): Add exit after trap.
2708 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2710 PR rtl-optimization/85638
2711 * bb-reorder.c: Include common/common-target.h.
2712 (create_forwarder_block): New function extracted from...
2713 (fix_up_crossing_landing_pad): ...here. Rename into...
2714 (dw2_fix_up_crossing_landing_pad): ...this.
2715 (sjlj_fix_up_crossing_landing_pad): New function.
2716 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2717 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2718 from both partitions and exit the loop after one iteration.
2720 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2723 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2725 (Basic PowerPC Built-in Functions): The new name of the
2726 subsection previously known as "PowerPC Built-in Functions".
2727 (Basic PowerPC Built-in Functions Available on all Configurations):
2729 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2731 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2733 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2735 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2738 2018-05-08 Jim Wilson <jimw@sifive.com>
2740 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2741 (LD_EMUL_SUFFIX): New.
2742 (LINK_SPEC): Use it.
2744 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2746 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2748 (Basic PowerPC Built-in Functions): The new name of the
2749 subsection previously known as "PowerPC Built-in Functions".
2750 (Basic PowerPC Built-in Functions Available on all Configurations):
2752 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2754 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2756 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2758 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2761 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2764 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2765 after cmpelim optimization.
2767 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2769 * config.gcc: Support "goldmont".
2770 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2771 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2773 * config/i386/i386.c (m_GOLDMONT): Define.
2774 (processor_target_table): Add "goldmont".
2775 (PTA_GOLDMONT): Define.
2776 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2777 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2778 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2779 (fold_builtin_cpu): Add "goldmont".
2780 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2781 (ix86_option_override_internal): Add "goldmont".
2782 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2783 (processor_type): Add PROCESSOR_GOLDMONT.
2784 * config/i386/i386.md: Add CPU "glm".
2785 * config/i386/glm.md: New file.
2786 * config/i386/x86-tune.def: Add m_GOLDMONT.
2787 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2789 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2792 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2794 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2795 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2796 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2799 * config/i386/i386.c (ix86_fold_builtin): Handle
2800 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2803 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2804 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2806 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2809 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2810 (check_arch): Likewise.
2811 (check_fpu): Return the result rather than printing it.
2812 (end arch): Fix operator precedence.
2813 (end cpu): Likewise.
2814 (END): Print the result from check_fpu.
2816 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2817 Alan Hayward <alan.hayward@arm.com>
2818 David Sherwood <david.sherwood@arm.com>
2820 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2821 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2822 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2823 (*fcmuo<mode>_and): New patterns.
2825 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2827 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2828 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2829 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2830 (cmp_op, sve_imm_con): New code attributes.
2831 (SVE_COND_INT_CMP, imm_con): Delete.
2832 (cmp_op): Remove above unspecs from int attribute.
2833 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2835 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2836 comparison-specific unspecs.
2837 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2838 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2839 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2840 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2841 (*vec_fcm<cmp_op><mode>): Rename to...
2842 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2843 (*vec_fcmuo<mode>): Rename to...
2844 (*fcmuo<mode>): ...this and adjust likewise.
2845 (*pred_fcm<cmp_op><mode>): New pattern.
2846 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2847 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2849 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2851 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2852 (aarch64_emit_sve_predicated_cond): New function.
2853 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2854 (aarch64_emit_unspec_cond_or): Replace with...
2855 (aarch64_emit_sve_or_conds): ...this new function. Use
2856 aarch64_emit_sve_ptrue_op for the individual comparisons and
2857 aarch64_emit_binop to OR them together.
2858 (aarch64_emit_inverted_unspec_cond): Replace with...
2859 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2860 aarch64_emit_sve_ptrue_op for the comparison and
2861 aarch64_emit_unop to invert the result.
2862 (aarch64_expand_sve_vec_cmp_float): Update after the above
2863 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2865 2018-05-07 Nathan Sidwell <nathan@acm.org>
2867 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2868 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2869 (Backwards Compatibility): Likewise.
2871 2018-05-07 Luis Machado <luis.machado@linaro.org>
2875 2018-05-07 Luis Machado <luis.machado@linaro.org>
2877 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2878 <prefetch_dynamic_strides>: New const bool field.
2879 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2880 prefetch_dynamic_strides.
2881 (exynosm1_prefetch_tune): Likewise.
2882 (thunderxt88_prefetch_tune): Likewise.
2883 (thunderx_prefetch_tune): Likewise.
2884 (thunderx2t99_prefetch_tune): Likewise.
2885 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2886 (aarch64_override_options_internal): Update to set
2887 PARAM_PREFETCH_DYNAMIC_STRIDES.
2888 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2889 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2890 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2891 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2892 prefetch-dynamic-strides setting.
2894 2018-05-07 Luis Machado <luis.machado@linaro.org>
2896 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2897 <minimum_stride>: New const int field.
2898 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2899 minimum_stride field.
2900 (exynosm1_prefetch_tune): Likewise.
2901 (thunderxt88_prefetch_tune): Likewise.
2902 (thunderx_prefetch_tune): Likewise.
2903 (thunderx2t99_prefetch_tune): Likewise.
2904 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2905 (aarch64_override_options_internal): Update to set
2906 PARAM_PREFETCH_MINIMUM_STRIDE.
2907 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2908 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2909 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2910 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2911 stride is constant and is below the minimum stride threshold.
2913 2018-05-07 Luis Machado <luis.machado@linaro.org>
2915 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2918 2018-05-07 Luis Machado <luis.machado@linaro.org>
2920 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2921 <prefetch_dynamic_strides>: New const bool field.
2922 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2923 prefetch_dynamic_strides.
2924 (exynosm1_prefetch_tune): Likewise.
2925 (thunderxt88_prefetch_tune): Likewise.
2926 (thunderx_prefetch_tune): Likewise.
2927 (thunderx2t99_prefetch_tune): Likewise.
2928 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2929 (aarch64_override_options_internal): Update to set
2930 PARAM_PREFETCH_DYNAMIC_STRIDES.
2931 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2932 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2933 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2934 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2935 prefetch-dynamic-strides setting.
2937 2018-05-07 Luis Machado <luis.machado@linaro.org>
2939 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2940 <minimum_stride>: New const int field.
2941 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2942 minimum_stride field.
2943 (exynosm1_prefetch_tune): Likewise.
2944 (thunderxt88_prefetch_tune): Likewise.
2945 (thunderx_prefetch_tune): Likewise.
2946 (thunderx2t99_prefetch_tune): Likewise.
2947 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2948 (aarch64_override_options_internal): Update to set
2949 PARAM_PREFETCH_MINIMUM_STRIDE.
2950 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2951 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2952 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2953 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2954 stride is constant and is below the minimum stride threshold.
2956 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2959 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2960 the type is addressable. Don't force op into register if it has
2963 2018-05-05 Roland McGrath <mcgrathr@google.com>
2966 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2967 any section for which we don't know a specific type it should have,
2968 regardless of name. Previously this was done only for the exact
2969 names ".init_array", ".fini_array", and ".preinit_array".
2970 (default_elf_asm_named_section): Add comment about
2971 relationship with default_section_type_flags and SECTION_NOTYPE.
2972 (get_section): Don't consider it a type conflict if one side has
2973 SECTION_NOTYPE and the other doesn't, as long as neither has the
2974 SECTION_BSS et al used in the default_section_type_flags logic.
2976 2018-05-05 Tom de Vries <tom@codesourcery.com>
2979 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2980 (workaround_barsyncs): New function.
2981 (nvptx_reorg): Use workaround_barsyncs.
2982 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2983 (define_expand "nvptx_membar_cta"): New define_expand.
2984 (define_insn "*nvptx_membar_cta"): New insn.
2986 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2988 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2989 To improve optimization opportunities.
2990 * builtin-types.def: The new needed builtin types for the above.
2992 2018-05-04 Richard Biener <rguenther@suse.de>
2994 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2995 * gimple-ssa-store-merging.c
2996 (imm_store_chain_info::output_merged_store): Remove redundant create,
2997 release split_store vector contents on failure.
2998 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2999 scalar stmt vector on cache hit.
3001 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3003 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3005 * config.gcc (powerpc-xilinx-eabi*): Remove.
3006 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3008 (fusion_addis_mem_combo_load): Ditto.
3009 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3011 (rs6000_cpu_cpp_builtins): Ditto.
3012 * config/rs6000/rs6000-linux.c
3013 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3014 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3015 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3017 (rs6000_setup_reg_addr_masks): Ditto.
3018 (rs6000_init_hard_regno_mode_ok): Ditto.
3019 (rs6000_option_override_internal): Ditto.
3020 (legitimate_lo_sum_address_p): Ditto.
3021 (rs6000_legitimize_address): Ditto.
3022 (rs6000_legitimize_reload_address): Ditto.
3023 (rs6000_legitimate_address_p): Ditto.
3024 (abi_v4_pass_in_fpr): Ditto.
3025 (setup_incoming_varargs): Ditto.
3026 (rs6000_gimplify_va_arg): Ditto.
3027 (rs6000_split_multireg_move): Ditto.
3028 (rs6000_savres_strategy): Ditto.
3029 (rs6000_emit_prologue_components): Ditto.
3030 (rs6000_emit_epilogue_components): Ditto.
3031 (rs6000_emit_prologue): Ditto.
3032 (rs6000_emit_epilogue): Ditto.
3033 (rs6000_elf_file_end): Ditto.
3034 (rs6000_function_value): Ditto.
3035 (rs6000_libcall_value): Ditto.
3036 * config/rs6000/rs6000.h: Ditto.
3037 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3038 (TARGET_MINMAX): ... this. New.
3039 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3040 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3041 (*movsi_internal1_single): Delete.
3042 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3043 mfpu=, mxilinx-fpu): Delete.
3044 * config/rs6000/singlefp.h: Delete.
3045 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3046 * config/rs6000/t-rs6000: Ditto.
3047 * config/rs6000/t-xilinx: Delete.
3048 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3049 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3050 (VStype_simple): Delete.
3051 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3052 * config/rs6000/xfpu.h: Delete.
3053 * config/rs6000/xfpu.md: Delete.
3054 * config/rs6000/xilinx.h: Delete.
3055 * config/rs6000/xilinx.opt: Delete.
3056 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3057 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3059 2018-05-04 Tom de Vries <tom@codesourcery.com>
3062 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3065 2018-05-04 Richard Biener <rguenther@suse.de>
3068 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3069 (expand_complex_div_wide): Likewise.
3070 (expand_complex_operations_1): Likewise.
3071 (expand_complex_libcall): Preserve EH info of the original stmt.
3072 (tree_lower_complex): Handle removed blocks.
3073 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3074 on complex multiplication and division libcall builtins.
3076 2018-05-04 Richard Biener <rguenther@suse.de>
3079 * fold-const.c (negate_expr_p): Restrict negation of operand
3080 zero of a division to when we know that can happen without
3082 (fold_negate_expr_1): Likewise.
3084 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3087 * real.h (real_nextafter): Declare.
3088 * real.c (real_nextafter): New function.
3089 * fold-const-call.c (fold_const_nextafter): New function.
3090 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3091 CASE_CFN_NEXTTOWARD.
3092 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3093 even when arg1_mode is different from arg0_mode.
3095 2018-05-03 Nathan Sidwell <nathan@acm.org>
3097 * doc/extend.texi (Deprecated Features): Remove
3099 (Backwards Compatibility): Likewise.
3100 * doc/invoke.texi (C++ Language Options): Likewise.
3101 (C++ Dialect Options): Likewise.
3103 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3106 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3107 _mm512_mask_mullox_epi64): New intrinsics.
3109 2018-05-03 Tom de Vries <tom@codesourcery.com>
3112 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3113 dump files): Add offload-tree.
3115 2018-05-03 Richard Biener <rguenther@suse.de>
3117 PR tree-optimization/85615
3118 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3119 to loops not nested in BBs loop father to avoid creating multi-entry
3122 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3124 PR tree-optimization/70291
3125 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3126 arguments. Change return type to tree. Emit libcall as a new
3127 statement rather than replacing existing one when inplace_p is true.
3128 (expand_complex_multiplication_components): New function.
3129 (expand_complex_multiplication): Expand floating-point complex
3130 multiplication using the above.
3131 (expand_complex_division): Rename inner_type parameter to type.
3132 Update expand_complex_libcall call-site.
3133 (expand_complex_operations_1): Update expand_complex_multiplication
3134 and expand_complex_division call-sites.
3136 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3139 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3140 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3141 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3142 the highest significant bit of the shift count mask is clear. In
3143 check whether and[sq]i3 is needed verify that all significant bits
3144 of the shift count other than the highest are set.
3146 2018-05-02 Tom de Vries <tom@codesourcery.com>
3149 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3150 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3151 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3152 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3153 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3154 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3155 __builtin_goacc_parlevel_size.
3157 2018-05-02 Richard Biener <rguenther@suse.de>
3159 PR tree-optimization/85597
3160 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3161 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3164 2018-05-02 Tom de Vries <tom@codesourcery.com>
3167 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3168 dump files): Add ltrans-tree.
3170 2018-05-02 Tom de Vries <tom@codesourcery.com>
3173 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3174 dump files): Add wpa-ipa.
3176 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3178 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3179 powerpc*-*-linux*paired* target.
3180 * config/rs6000/750cl.h: Delete.
3181 * config/rs6000/paired.h: Delete.
3182 * config/rs6000/paired.md: Delete.
3183 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3185 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3186 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3187 comment. Remove paired float support.
3188 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3189 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3191 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3192 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3194 * config/rs6000/rs6000.c: Remove paired float support.
3195 (paired_expand_vector_init, paired_expand_vector_move,
3196 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3197 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3198 paired_expand_builtin, paired_expand_predicate_builtin,
3199 paired_init_builtins): Delete.
3200 * config/rs6000/rs6000.h: Remove paired float support.
3201 * config/rs6000/rs6000.md: Remove paired float support.
3202 (move_from_CR_ov_bit): Delete.
3203 * config/rs6000/rs6000.opt (mpaired): Delete.
3204 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3205 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3207 2018-05-02 Richard Biener <rguenther@suse.de>
3210 * gimplify.c (gimplify_save_expr): When in SSA form allow
3211 SAVE_EXPRs to compute to SSA vars.
3213 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3216 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3217 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3218 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3219 clobber operands[2], instead use a new pseudo. Formatting fixes.
3221 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3223 PR tree-optimization/85586
3224 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3225 exit early for statements in the same group if the accesses are
3228 2018-05-02 Tom de Vries <tom@codesourcery.com>
3231 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3234 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3236 PR tree-optimization/85143
3237 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3239 2018-05-01 Tom de Vries <tom@codesourcery.com>
3242 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3243 not found" error message.
3245 2018-05-01 Tom de Vries <tom@codesourcery.com>
3248 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3249 * vec.c (test_ordered_remove_if): New function.
3250 (vec_c_tests): Call test_ordered_remove_if.
3251 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3252 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3253 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3254 VEC_ORDERED_REMOVE_IF.
3256 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3258 PR tree-optimization/82665
3259 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3260 pointer subtraction where arguments come from a memchr call.
3262 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3264 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3265 --push-state --as-needed and --pop-state instead of --as-needed and
3266 --no-as-needed if ld supports it.
3267 * configure: Regenerated.
3270 * doc/install.texi2html: Replace _002d with - and _002a with * in
3271 generated html files using sed.
3273 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3276 * gcc-rich-location.c (blank_line_before_p): New function.
3277 (use_new_line): New function.
3278 (gcc_rich_location::add_fixit_insert_formatted): New function.
3279 * gcc-rich-location.h
3280 (gcc_rich_location::add_fixit_insert_formatted): New function.
3282 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3284 * selftest.c (assert_streq): Rename "expected" and "actual" to
3285 "val1" and "val2". Extend NULL-handling to cover both inputs
3286 symmetrically, while still requiring both to be non-NULL for a pass.
3287 * selftest.h (assert_streq): Rename "expected" and "actual" to
3289 (ASSERT_EQ): Likewise.
3290 (ASSERT_EQ_AT): Likewise.
3291 (ASSERT_KNOWN_EQ): Likewise.
3292 (ASSERT_KNOWN_EQ_AT): Likewise.
3293 (ASSERT_NE): Likewise.
3294 (ASSERT_MAYBE_NE): Likewise.
3295 (ASSERT_MAYBE_NE_AT): Likewise.
3296 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3297 the assertion to pass.
3298 (ASSERT_STREQ_AT): Likewise.
3300 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3302 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3303 interaction with -pie.
3305 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3307 * selftest.h: Fix alphabetization of per-source-file selftest
3310 2018-04-30 Jason Merrill <jason@redhat.com>
3312 PR c++/61982 - dead stores to destroyed objects.
3313 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3316 2018-04-30 Jason Merrill <jason@redhat.com>
3318 * tree.c (build_clobber): New.
3319 * tree.h: Declare it.
3320 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3322 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3324 * diagnostic-show-locus.c (layout::layout): Update for
3325 location_get_source_line returning a char_span.
3326 (struct char_span): Move to input.h.
3327 (struct correction): Update for fields in char_span becoming
3329 (struct source_line): Update for location_get_source_line
3330 returning a char_span.
3331 (layout::print_line): Likewise.
3332 * edit-context.c (edited_file::print_content): Likewise.
3333 (edited_file::print_diff_hunk): Likewise.
3334 (edited_file::print_run_of_changed_lines): Likewise.
3335 (edited_file::get_num_lines): Likewise.
3336 (edited_line::edited_line): Likewise.
3337 * final.c (asm_show_source): Likewise.
3338 * input.c (location_get_source_line): Convert return type
3339 from const char * to char_span, losing the final "line_len"
3341 (dump_location_info): Update for the above.
3342 (get_substring_ranges_for_loc): Likewise. Use a char_span
3343 when handling the literal within the line.
3344 (test_reading_source_line): Update for location_get_source_line
3345 returning a char_span.
3346 * input.h (class char_span): Move here from
3347 diagnostic-show-locus.c, converting from a struct to a class.
3348 Make data members private.
3349 (char_span::operator bool): New.
3350 (char_span::length): New.
3351 (char_span::get_buffer): New.
3352 (char_span::operator[]): New.
3353 (char_span::subspan): Make const.
3354 (char_span::xstrdup): New.
3355 (location_get_source_line): Convert return type from const char *
3356 to char_span, losing the final "line_size" param.
3358 2018-04-30 Jan Hubicka <jh@suse.cz>
3360 * lto-wrapper.c (ltrans_priorities): New static var.
3361 (cmp_priority): New.
3362 (run_gcc): Read priorities and if doing parallel build order
3363 the Makefile by them.
3365 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3367 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3369 2018-04-30 Richard Biener <rguenther@suse.de>
3371 * tree-cfg.c (verify_address): Remove base argument, add
3372 flag whether to check TREE_ADDRESSABLE and do that.
3373 (verify_expr): Remove.
3374 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3375 (verify_gimple_assign_single): Likewise.
3376 (verify_gimple_switch): Likewise.
3377 (verify_expr_location_1): Dereference tp once. Add (disabled)
3378 piece from verify_expr.
3379 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3381 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3383 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3385 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3387 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3388 (small_data_pattern): Likewise.
3389 (arc_rewrite_small_data): Likewise.
3390 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3391 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3392 (get_symbol_alignment): New function.
3393 (legitimate_small_data_address_p): Likewise.
3394 (legitimate_scaled_address): Update, call
3395 legitimate_small_data_address_p.
3396 (output_sdata): New static variable.
3397 (arc_print_operand): Update how we handle small data operands.
3398 (arc_print_operand_address): Likewise.
3399 (arc_legitimate_address_p): Update, use
3400 legitimate_small_data_address_p.
3401 (arc_rewrite_small_data_p): Remove.
3402 (arc_rewrite_small_data_1): Likewise.
3403 (arc_rewrite_small_data): Likewise.
3404 (small_data_pattern): Likewise.
3405 (compact_sda_memory_operand): Update to use
3406 legitimate_small_data_address_p and get_symbol_alignment.
3407 (prepare_move_operands): Don't rewite sdata pattern.
3408 (prepare_extend_operands): Remove.
3409 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3411 (zero_extendqisi2): Likewise.
3412 (zero_extendhisi2): Likewise.
3413 (extendqihi2): Likewise.
3414 (extendqisi2): Likewise.
3415 (extendhisi2): Likewise.
3419 * config/arc/constraints.md (Usd): Change it to memory constraint.
3421 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3423 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3424 as source of std instructions.
3425 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3426 allow 6-bit constants as source for store instructions.
3427 (movdi_insn): Update instruction pattern to allow 6-bit constants
3428 as source for store instructions.
3430 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3432 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3434 2018-04-30 Nathan Sidwell <nathan@acm.org>
3435 Sandra Loosemore <sandra@codesourcery.com>
3437 * dumpfile.c (dump_open): Allow '-' for stdout.
3438 * doc/invoke.texi (Developer Options): Document dump filename
3439 determination early. Document stdin/stdout selection.
3441 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3443 Microblaze Target: PIC data text relative
3445 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3446 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3448 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3449 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3450 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3451 New addressing mode for data-text relative position indepenedent code.
3452 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3453 'ADDRESS_SYMBOLIC_TXT_REL'.
3454 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3455 (microblaze_legitimate_pic_operand): Exclude function calls from
3456 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3457 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3459 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3460 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3461 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3462 for 'address + offset'.
3463 (microblaze_expand_prologue): Add new function prologue call for
3465 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3466 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3467 table in case of TARGET_PIC_DATA_TEXT_REL.
3468 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3469 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3470 Add new macros 'UNSPEC_TEXT',
3471 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3472 + exclude function calls from 'UNSPEC_PLT' in case of data text
3474 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3475 new target hook for generating address diff vector tables in case of
3477 * doc/tm.texi : Regenerate.
3478 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3479 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3480 of addr diff vector generation.
3481 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3482 target hook definition.
3483 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3484 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3485 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3486 option for data text relative.
3488 2018-04-30 Richard Biener <rguenther@suse.de>
3490 * tree-chrec.h (evolution_function_is_constant_p): Remove
3492 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3494 2018-04-30 Richard Biener <rguenther@suse.de>
3497 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3499 2018-04-30 Richard Biener <rguenther@suse.de>
3501 PR tree-optimization/28364
3502 PR tree-optimization/85275
3503 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3504 copying first exit test.
3506 2018-04-28 Mark Wielaard <mark@klomp.org>
3508 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3510 (dwarf_AT): Handle DW_AT_addr_base.
3511 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3513 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3516 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3517 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3518 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3519 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3521 2018-04-28 Richard Biener <rguenther@suse.de>
3523 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3524 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3525 to reflect use. Only add interesting stmts.
3527 2018-04-27 Martin Jambor <mjambor@suse.cz>
3530 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3531 the jump function allows for passing through aggregate values.
3533 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3535 * input.h (in_system_header_at): Convert from macro to inline
3537 (from_macro_expansion_at): Likewise.
3538 (from_macro_definition_at): Likewise.
3540 2018-04-27 Jeff Law <law@redhat.com>
3542 * config.gcc: Mark tile* targets as deprecated/obsolete.
3544 2018-04-27 Richard Biener <rguenther@suse.de>
3546 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3549 2018-04-27 Richard Biener <rguenther@suse.de>
3551 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3553 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3555 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3556 with Yd constraint. Set "preferred_for_speed" attribute from
3557 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3559 (*movdi_internal): Ditto.
3560 (movti_interunit splitters): Remove
3561 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3562 (movdi_interunit splitters): Ditto.
3563 * config/i386/constraints.md (Ye): Remove.
3564 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3566 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3569 * config/aarch64/constraints.md (Usg): Limit to 31.
3572 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3574 PR tree-optimization/85529
3575 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3576 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3577 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3578 zero extension or masking of the MSB bit.
3579 (optimize_range_tests): Add FIRST_BB argument, pass it through
3580 to optimize_range_tests_var_bound.
3581 (maybe_optimize_range_tests, reassociate_bb): Adjust
3582 optimize_range_tests callers.
3584 2018-04-26 Richard Biener <rguenther@suse.de>
3585 Jakub Jelinek <jakub@redhat.com>
3587 * cgraph.h (symbol_table): Just declare debug method here.
3588 * symtab.c (symbol_table::debug): Define.
3590 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3592 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3594 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3596 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3597 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3598 (*movdi_internal): Substitute Yi and Yj constraint with x
3599 and Ym and Yn constraint with y constraint. Update "isa"
3600 attribute and set "preferred_for_speed" attribute from
3601 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3602 (*movsi_internal): Ditto.
3603 (*movdf_internal): Ditto.
3604 (*movsf_internal): Ditto.
3605 (*zero_extendsidi2): Ditto.
3606 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3607 (sse2_loadld): Ditto.
3608 (*vec_extract<ssevecmodelower>_0): Ditto.
3609 (*vec_extractv4si_0_zext_sse4): Ditto.
3610 (vec_concatv2di): Ditto.
3611 (*vec_dup<mode>): Ditto.
3612 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3613 * config/i386/constraints.md (Yi): Remove.
3618 2018-04-26 Nathan Sidwell <nathan@acm.org>
3620 * dumpfile.c (dump_open): New.
3621 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3622 (dump_finish): Detect stdio/stderr by value not name.
3624 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3626 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3628 2018-04-26 Tom de Vries <tom@codesourcery.com>
3631 * config/nvptx/nvptx.c (verify_neutering_jumps)
3632 (verify_neutering_labels): New function
3633 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3635 2018-04-26 Tom de Vries <tom@codesourcery.com>
3638 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3639 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3642 2018-04-26 Richard Biener <rguenther@suse.de>
3643 Tom de Vries <tom@codesourcery.com>
3646 * lto-streamer-out.c (output_function): Fixup loops if required to match
3647 discovery done in the reader.
3649 2018-04-26 Richard Biener <rguenther@suse.de>
3651 PR tree-optimization/85116
3652 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3653 have a loop exit from the single latch predecessor. Remove
3654 case of header with just condition.
3655 (ch_base::copy_headers): Exclude infinite loops from any
3657 (pass_ch::execute): Record exits.
3659 2018-04-26 Richard Biener <rguenther@suse.de>
3661 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3662 prologue cost vector and pass it to vect_get_load_cost.
3663 (vect_get_peeling_costs_all_drs): Likewise.
3664 (vect_peeling_hash_get_lowest_cost): Likewise.
3665 (vect_enhance_data_refs_alignment): Likewise.
3667 2018-04-26 Richard Biener <rguenther@suse.de>
3670 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3671 checking of integer<->pointer conversions.
3672 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3673 sign-/zero-extending pointer types.
3674 (expand_omp_for_static_chunk): Likewise.
3676 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3677 Jean Lee <xiaoyur347@gmail.com>
3679 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3680 (TARGET_ASAN_SHADOW_OFFSET): Define.
3681 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3682 true for -fsanitize=address.
3684 2018-04-25 Mark Wielaard <mark@klomp.org>
3686 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3689 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3691 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3692 than "alu", remove explicit "memory" and "imm_disp" attributes.
3693 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3696 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3697 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3698 gen_lowpart_no_emit.
3700 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3703 * config/i386/i386.c (ix86_expand_builtin): Change memory
3704 operand to XI, extend p0 to Pmode.
3705 * config/i386/i386.md: Change unspec volatile and operand
3706 1 mode to XI, change operand 0 mode to P.
3708 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3710 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3711 GET_MODE_MASK before any checking.
3712 (nds32_can_use_bset_p): Likewise.
3713 (nds32_can_use_btgl_p): Likewise.
3715 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3717 * config/nds32/nds32-doubleword.md: New define_split pattern for
3718 illegal register number.
3720 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3722 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3724 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3726 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3728 2018-04-25 Richard Biener <rguenther@suse.de>
3730 * lto-streamer.h (LTO_major_version): Bump to 8.
3732 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3734 * BASE-VER: Set to 9.0.0.
3736 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3738 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3739 in __abskf2 and __powikf2.
3741 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3744 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3745 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3746 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3747 Use the above on operand 2. Reindent.
3748 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3750 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3753 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3755 * config/i386/i386.opt (mcet): Removed.
3756 * doc/install.texi: Remove -mcet documentation.
3757 * doc/invoke.texi: Likewise.
3759 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3762 * doc/install.texi: Remove -mcet from bootstrap-cet.
3764 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3767 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3768 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3772 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3773 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3774 containing a CONST_VECTOR.
3776 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3778 * doc/install.texi: Update newlib dependency for nvptx.
3780 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3783 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3784 instead of INTVAL when shifting x left.
3786 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3788 PR tree-optimization/85478
3789 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3790 vect_grouped_store_supported for single element vectors.
3792 2018-04-24 Richard Biener <rguenther@suse.de>
3795 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3796 load cost increase to the case of non-constant step.
3798 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3801 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3802 destination if any_malformed_asm.
3804 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3807 * expr.c (store_field): In the bitfield case, if the value comes from
3808 a function call and is returned in registers by means of a PARALLEL,
3809 do not change the mode of the temporary unless BLKmode and VOIDmode.
3811 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3813 PR rtl-optimization/85423
3814 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3815 dependencies to debug insns when the previous insn is non-debug.
3817 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3819 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3820 enums into a single definition.
3821 (fls): Fix predicates and printing.
3824 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3826 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3827 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3828 and short u6 immediate.
3829 (check_if_valid_sleep_operand): Remove.
3830 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3832 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3834 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3835 flag_always_save_lp condition.
3836 * config/nds32/nds32.opt (malways-save-lp): New option.
3838 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3840 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3841 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3842 * config/nds32/nds32.h
3843 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3844 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3846 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3848 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3849 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3851 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3852 Chung-Ju Wu <jasonwucj@gmail.com>
3854 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3856 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3857 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3859 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3861 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3863 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3865 * config/nds32/nds32-protos.h (nds32_data_alignment,
3866 nds32_local_alignment): Declare.
3867 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3868 nds32_local_alignment): New functions.
3869 (TARGET_CONSTANT_ALIGNMENT): Define.
3870 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3872 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3874 * config/nds32/nds32.c
3875 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3876 (TARGET_MODES_TIEABLE_P): Likewise.
3878 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3880 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3883 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3884 Chung-Ju Wu <jasonwucj@gmail.com>
3886 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3887 for unaligned access.
3888 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3890 * config/nds32/nds32-intrinsic.md: Likewise.
3891 * config/nds32/nds32_intrinsic.h: Likewise.
3892 * config/nds32/nds32.h (nds32_builtins): Likewise.
3893 * config/nds32/nds32.opt (munaligned-access): New option.
3894 * config/nds32/nds32.c (nds32_asm_file_start): Display
3895 flag_unaligned_access status.
3897 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3899 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3900 -mno-relax is present.
3901 * config/riscv/linux.h (LINK_SPEC): Ditto.
3903 2018-04-20 Martin Sebor <msebor@redhat.com>
3906 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3908 (gimple_fold_builtin_stxcpy_chk): Same.
3909 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3911 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3914 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3915 __powikf2 when long double is IEEE 128-bit.
3917 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3919 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3920 step to make sure stack always aligned.
3922 2018-04-20 Carl Love <cel@us.ibm.com>
3925 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3926 size check for arg0.
3928 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3929 Tom de Vries <tom@codesourcery.com>
3932 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3933 Emit insns for calls too.
3934 (nvptx_find_par): Always look for worker-level predecessor insn.
3935 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3937 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3938 (nvptx_process_pars): Propagate frames for calls.
3940 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3943 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3945 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3946 (ix86_handle_option): Don't handle OPT_mibt.
3947 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3949 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3951 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3952 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3953 (ix86_target_macros): Define __CET__ with flag_cf_protection
3954 for -fcf-protection.
3955 * config/i386/i386.c (isa2_opts): Remove -mibt.
3956 * config/i386/i386.h (TARGET_IBT): Removed.
3957 (TARGET_IBT_P): Likewise.
3958 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3959 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3960 * config/i386/i386.opt (mcet): Update help message.
3963 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3964 -mcet as an alias for -mshstk.
3966 2018-04-20 Richard Biener <rguenther@suse.de>
3969 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3970 complexity by forcing a single use of the multiply operand.
3972 2018-04-20 Martin Jambor <mjambor@suse.cz>
3975 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3976 recursion dependency to only apply to non-clones.
3978 2018-04-20 Martin Jambor <mjambor@suse.cz>
3981 * ipa-cp.c (create_specialized_node): Check that clones of
3982 self-recursive edges exist during IPA-CP.
3984 2018-04-19 Toon Moene <toon@moene.org>
3986 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3989 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3991 PR tree-optimization/85467
3992 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3993 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3994 VECTOR_CST element to type.
3996 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3999 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4000 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4001 (builtin_longjmp): Likewise.
4002 (save_stack_nonlocal): New pattern.
4003 (restore_stack_nonlocal): Likewise.
4005 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4008 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4009 Replace ASM_OUTPUT_LABEL with fprintf.
4011 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4014 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4015 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4016 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4017 define __IBT__ and __SHSTK__ for -fcf-protection.
4018 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4020 (ix86_trampoline_init): Likewise.
4021 (x86_output_mi_thunk): Likewise.
4022 (ix86_notrack_prefixed_insn_p): Likewise.
4023 (ix86_option_override_internal): Don't disallow -fcf-protection.
4024 * config/i386/i386.md (rdssp<mode>): Also enable for
4026 (incssp<mode>): Likewise.
4027 (nop_endbr): Likewise.
4028 * config/i386/i386.opt (mcet): Change help message to built-in
4032 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4033 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4034 enable CET built-in functions.
4036 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4038 * common/config/i386/i386-common.c
4039 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4040 OPTION_MASK_ISA_MOVDIRI_UNSET,
4041 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4042 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4043 * config.gcc (movdirintrin.h): New header.
4044 * config/i386/cpuid.h (bit_MOVDIRI,
4045 bit_MOVDIR64B): New bits.
4046 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4048 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4049 (VOID, PVOID, PCVOID)): New function types.
4050 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4051 __builtin_ia32_directstoreu_u64,
4052 __builtin_ia32_movdir64b): New builtins.
4053 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4054 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4056 (ix86_valid_target_attribute_inner_p): Ditto.
4057 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4058 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4059 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4060 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4061 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4062 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4063 (movdiri<mode>, movdir64b_<mode>): New.
4064 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4065 * config/i386/immintrin.h: Include movdirintrin.h.
4066 * config/i386/movdirintrin.h: New file.
4067 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4069 2018-04-19 Richard Biener <rguenther@suse.de>
4072 * cfg.c (clear_bb_flags): When loop state says we have
4073 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4075 2018-04-19 Richard Biener <rguenther@suse.de>
4077 PR tree-optimization/84737
4078 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4079 copying restrict info.
4080 (vect_setup_realignment): Use it.
4081 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4082 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4083 the first DR to all generated stores.
4084 (vectorizable_load): Likewise for loads.
4086 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4088 PR tree-optimization/85446
4089 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4090 the integral and pointer types to have the same precision.
4092 * doc/install.texi: Document --disable-cet being the default and
4095 2018-04-18 Martin Liska <mliska@suse.cz>
4097 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4100 2018-04-18 Martin Liska <mliska@suse.cz>
4103 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4106 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4107 arguments if they are comparable.
4109 2018-04-18 Martin Liska <mliska@suse.cz>
4112 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4115 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4118 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4121 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4122 ENDBR after calling __morestack.
4124 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4127 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4128 by using gcc_base_ver to generate a gcc_driver_version, and use
4129 it when generating GCC_DRIVER_NAME.
4130 * configure: Regenerate.
4132 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4135 * config.gcc: Obsolete powerpc*-*-*spe*.
4137 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4140 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4141 (stabstr_D): Change type of unum from unsigned int to
4142 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4145 2018-04-17 Jim Wilson <jimw@sifive.com>
4148 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4149 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4150 Set arg_pointer_offset after using pretend_args_size.
4152 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4154 PR rtl-optimization/85431
4155 * dse.c (record_store): Ignore zero width stores.
4158 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4159 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4160 __builtin_stack_restore rather than after it.
4161 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4162 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4163 argument instead of virtual_dynamic_stack_rtx.
4165 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4167 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4169 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4170 Add note to error message to explain internal mapping of overloaded
4171 built-in function name to non-overloaded built-in function name.
4172 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4175 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4178 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4179 where the inputs overlap with the output.
4181 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4183 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4184 (=v, v) alternative and explicit "memory" attribute.
4185 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4186 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4188 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4189 "sselog1" type instead of "sselog".
4190 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4191 "sselog". Remove explicit "memory" attribute.
4192 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4193 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4195 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4196 "sselog1" type instead of "sselog". Remove explicit "memory"
4198 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4199 use "sselog1" type instead of "sselog". Remove explicit "memory"
4201 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4202 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4204 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4205 "sselog1" type instead of "sselog". Remove explicit "memory"
4207 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4208 use "sselog1" type instead of "sselog". Remove explicit "memory"
4212 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4215 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4218 2018-04-17 Martin Jambor <mjambor@suse.cz>
4221 * ipa-cp.c (create_specialized_node): Call
4222 expand_all_artificial_thunks if necessary.
4224 2018-04-17 Martin Liska <mliska@suse.cz>
4227 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4228 in message, remote space in between '_G' and '('.
4230 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4233 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4234 avx512f_vmcmp<mode>3<round_saeonly_name>,
4235 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4236 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4237 avx512f_rndscale<mode><round_saeonly_name>,
4238 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4239 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4240 Use %<iptr>2 instead of %2 for -masm=intel.
4241 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4242 avx512f_vcvttss2usi<round_saeonly_name>,
4243 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4245 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4246 avx512f_vcvttsd2usi<round_saeonly_name>,
4247 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4248 Use %q1 instead of %1 for -masm=intel.
4249 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4250 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4251 of %3 for -masm=intel.
4252 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4254 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4256 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4258 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4259 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4261 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4263 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4264 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4265 %g1 and one with %0 and %1.
4266 (avx512er_vmrcp28<mode><round_saeonly_name>,
4267 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4269 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4270 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4271 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4272 of %0 and %{%4%} for -masm=intel.
4273 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4274 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4275 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4276 order of %0 and %{%5%}%{z%} for -masm=intel.
4278 2018-04-17 Jan Hubicka <jh@suse.cz>
4281 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4283 2018-04-17 Martin Liska <mliska@suse.cz>
4286 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4287 for target_clone error message. Make default implementation
4288 clone to be a local declaration.
4289 (separate_attrs): Add new argument and check for an empty
4291 (expand_target_clones): Handle it.
4292 (ipa_target_clone): Make redirection just for target_clones
4295 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4296 Tom de Vries <tom@codesourcery.com>
4299 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4300 tiled basic blocks without omp continue statements.
4302 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4305 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4306 vec_extract expression as having side effects to make sure it gets
4309 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4312 * config/i386/i386.c (get_builtin_code_for_version): Check
4315 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4318 * gcc/config.gcc: Support "skylake".
4319 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4321 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4322 (processor_target_table): Add "skylake".
4323 (ix86_option_override_internal): Add "skylake".
4324 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4325 PROCESSOR_CANNONLAKE.
4326 (get_builtin_code_for_version): Fix priority for
4327 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4328 PROCESSOR_SKYLAKE-AVX512.
4329 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4330 (processor_type): Add PROCESSOR_SKYLAKE.
4332 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4333 Jason Merrill <jason@redhat.com>
4336 * convert.c (convert_to_integer_1): Use direct recursion for
4337 enumeral types and types with a precision less than the number
4338 of bits in their mode.
4340 2018-04-16 Julia Koval <julia.koval@intel.com>
4343 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4344 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4346 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4349 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4350 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4351 and -mno-direct-move.
4353 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4356 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4357 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4358 Ensure negative shifts result in {0}.
4360 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4362 PR rtl-optimization/79916
4363 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4364 regs (if any) to define how to gnerate SD moves when LRA is in
4367 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4369 PR rtl-optimization/85393
4370 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4371 * except.c (expand_dw2_landing_pad_for_region): Make static.
4372 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4373 a label and unconditional jump to old_bb, rather than
4374 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4377 PR rtl-optimization/85376
4378 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4379 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4380 instead of a specific value.
4382 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4383 Bin Cheng <bin.cheng@arm.com>
4385 PR tree-optimization/82965
4386 PR tree-optimization/83991
4387 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4388 by_profile_only parameter.
4389 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4390 information if the loop was predicted to iterate too many times.
4391 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4393 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4396 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4399 2018-04-13 Martin Liska <mliska@suse.cz>
4400 Jakub Jelinek <jakub@redhat.com>
4403 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4404 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4405 * builtins.c (expand_builtin_memory_copy_args): Use
4406 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4407 handle dest_addr == pc_rtx.
4409 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4412 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4413 asked to not generate direct moves.
4414 (fix_trunc<mode>si2_stfiwx): Similar.
4415 (fix_trunc<mode>si2_internal): Similar.
4417 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4420 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4421 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4422 lookup if dest in some wider mode is known to be const0_rtx and
4423 if so, record permanent equivalence for it to be ZERO_EXTEND of
4424 the narrower mode destination.
4426 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4428 * lto-streamer-out.c (output_function): Revert 259346.
4429 * omp-expand.c (expand_oacc_for): Likewise.
4431 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4433 PR rtl-optimization/85354
4434 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4435 * sel-sched.c (sel_global_init): ... here.
4437 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4440 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4441 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4442 mode for PE-COFF targets.
4443 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4444 (i386_pe_asm_lto_end): Likewise.
4445 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4446 (TARGET_ASM_LTO_END): Likewise.
4447 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4448 (i386_pe_asm_lto_start): New function.
4449 (i386_pe_asm_lto_end): Likewise.
4451 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4452 Richard Biener <rguenther@suse.de>
4455 * lto-streamer-out.c (output_function): Fix CFG loop state before
4457 * omp-expand.c (expand_oacc_for): Handle calls to internal
4458 functions like regular functions.
4460 2018-04-12 Richard Biener <rguenther@suse.de>
4463 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4464 for the early LTO debug to properly generate references to it
4465 during DIE emission. Do not re-use that for the skeleton for
4467 (dwarf2out_early_finish): Likewise.
4469 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4472 * config/i386/sse.md
4473 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4474 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4475 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4476 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4477 and output is a reg, avoid creating invalid lowpart subreg, but
4478 instead split into a 512-bit move. Don't split if not AVX512VL,
4479 input is xmm16+ reg and output is a mem.
4480 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4481 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4482 xmm16+ reg and output is a mem.
4484 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4486 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4487 also for flag_dwarf2_cfi_asm.
4489 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4491 PR rtl-optimization/85342
4492 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4493 a bool scalar var inside of the loop instead. Don't try to update
4494 recog_data.operand after failed apply_change_group.
4496 2018-04-12 Tom de Vries <tom@codesourcery.com>
4499 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4500 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4501 array with flexible array member as array without given dimension.
4502 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4503 argument for undefined param to true.
4505 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4508 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4509 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4510 from PowerPC section.
4511 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4512 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4513 help text that is too long.
4514 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4515 help text that is too long.
4516 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4517 help text that is too long.
4519 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4521 * config/alpha/alpha.md (stack_probe_internal): Rename
4522 from "probe_stack". Update all callers.
4524 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4526 PR rtl-optimization/84566
4527 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4528 sched_macro_fuse_insns.
4530 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4533 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4534 (compute_block_dependences): ... from here.
4536 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4538 PR tree-optimization/85331
4539 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4540 from int to HOST_WIDE_INT.
4542 2018-04-11 Martin Jambor <mjambor@suse.cz>
4545 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4546 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4547 not the same as the source val.
4548 (cgraph_edge_brings_value_p): New parameter.
4549 (gather_edges_for_value): Pass destination value to
4550 cgraph_edge_brings_value_p.
4551 (perhaps_add_new_callers): Likewise.
4552 (get_info_about_necessary_edges): Likewise and exclude values brought
4553 only by self-recursive edges.
4554 (create_specialized_node): Redirect only clones of self-calling edges.
4555 (+self_recursive_pass_through_p): New function.
4556 (find_more_scalar_values_for_callers_subset): Use it.
4557 (find_aggregate_values_for_callers_subset): Likewise.
4558 (known_aggs_to_agg_replacement_list): Removed.
4559 (decide_whether_version_node): Re-calculate known constants for all
4560 remaining context clones.
4562 2018-04-11 Richard Biener <rguenther@suse.de>
4565 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4566 from early DWARF output.
4567 (dwarf2out_early_finish): Output line info unconditionally into
4568 early DWARF and add reference to it.
4570 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4573 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4574 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4575 other than V2DFmode using iptr mode attribute.
4576 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4578 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4580 PR rtl-optimization/84659
4581 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4583 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4586 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4588 (output_loc_list): Pass address of a dummy size variable even in the
4589 locview handling loop.
4590 (index_location_lists): Add comment on why skip_loc_list_entry can't
4593 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4596 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4599 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4602 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4603 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4604 and -mstring-compare-inline-limit.
4606 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4609 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4610 for stack clash protection in a register whenever we need it to be in
4613 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4615 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4616 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4618 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4621 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4623 (mlong-double-): Ditto.
4624 * config/rs6000/sysv4.opt (msdata=): Ditto.
4625 (mtls-size=): Ditto.
4627 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4629 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4630 erroneous entries for
4631 "vector int vec_ldl (int, long int *)", and
4632 "vector unsigned int vec_ldl (int, unsigned long int *)".
4633 Add comments and entries for
4634 "vector bool char vec_ldl (int, bool char *)",
4635 "vector bool short vec_ldl (int, bool short *)",
4636 "vector bool int vec_ldl (int, bool int *)",
4637 "vector bool long long vec_ldl (int, bool long long *)",
4638 "vector pixel vec_ldl (int, pixel *)",
4639 "vector long long vec_ldl (int, long long *)",
4640 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4641 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4642 type tree bool_long_long_type_node and correct definition of
4643 bool_V2DI_type_node to make reference to this new type tree.
4644 (rs6000_mangle_type): Replace erroneous reference to
4645 bool_long_type_node with bool_long_long_type_node.
4646 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4647 comments to emphasize sign distinctions for char and int types and
4648 replace RS6000_BTI_bool_long constant with
4649 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4650 use of RS6000_BTI_pixel.
4651 (bool_long_type_node): Remove this macro definition.
4652 (bool_long_long_type_node): New macro definition
4654 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4656 PR rtl-optimization/85300
4657 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4658 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4659 simplify_unary_operation fails.
4661 2018-04-10 Martin Liska <mliska@suse.cz>
4663 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4664 cgraph_edge and ipa_ref.
4666 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4670 * config/i386/sse.md
4671 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4672 computation of the VEC_MERGE selector from mask.
4673 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4674 Fix decoding of the VEC_MERGE selector into mask.
4676 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4678 PR tree-optimization/85286
4679 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4681 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4683 * final.c (final_1): Set insn_last_address as well as
4684 insn_current_address.
4686 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4689 * explow.c (emit_stack_probe): Call validize_mem on memory location
4690 before passing it to gen_probe_stack. Create address operand and
4691 legitimize it for the probe_stack_address case.
4693 2018-04-09 Jan Hubicka <jh@suse.cz>
4696 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4697 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4698 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4700 (free_lang_data): Rebuild type inheritance graph.
4702 2018-04-09 Martin Sebor <msebor@redhat.com>
4704 * invoke.texi (-finline-small-functions): Mention other optimization
4706 (-findirect-inlining, -fpartial-inlining): Same.
4707 (-finline-functions-called-once): Same.
4708 (-freorder-blocks-and-partition): Same.
4710 2018-04-09 Jan Hubicka <jh@suse.cz>
4713 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4714 jumps; choose last target that matches the criteria (i.e.
4715 no partition changes for non-crossing jumps).
4716 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4717 support for redirecting crossing jumps to non-crossing.
4719 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4721 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4722 also for naked functions.
4724 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4726 * config/arc/arc.md (add_shift): New pattern.
4727 (add_shift2): Likewise.
4728 (sub_shift): Likewise.
4729 (sub_shift_cmp0_noout): Likewise.
4730 (compare_si_ashiftsi): Likewise.
4731 (xbfu_cmp0_noout): New combine pattern.
4732 (xbfu_cmp0"): Likewise.
4733 (movsi_set_cc_insn): Place the predicable variant first.
4734 (commutative_binary_cmp0_noout): Remove clobber.
4735 (commutative_binary_cmp0): New pattern.
4736 (noncommutative_binary_cmp0): Likewise.
4737 (noncommutative_binary_cmp0_noout): Likewise.
4738 (noncommutative_binary_comparison_result_used): Removed.
4739 (rsub_cmp0): New pattern.
4740 (rsub_cmp0_noout): Likewise.
4741 (extzvsi): Changed, keep only meaningful variants.
4742 (SQH, SEZ): New iterators.
4743 (SQH_postfix): New mode attribute.
4744 (SEZ_prefix): New code attribute.
4745 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4746 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4747 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4749 (noncommutative_operator): Check the availability of barrel
4752 2018-04-09 Richard Biener <rguenther@suse.de>
4754 PR tree-optimization/85284
4755 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4756 Only use the niter constraining form of simple_iv when the exit
4759 2018-04-09 Tom de Vries <tom@codesourcery.com>
4762 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4763 (define_expand "*memory_barrier"): New define_expand.
4764 (define_insn "memory_barrier"): New insn.
4766 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4768 PR rtl-optimization/80463
4769 PR rtl-optimization/83972
4770 PR rtl-optimization/83480
4772 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4773 correct producer for the insn.
4774 (tidy_control_flow): Fixup seqnos in case of debug insns.
4776 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4778 PR rtl-optimization/83913
4780 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4781 different sched-times when merging exprs.
4783 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4785 PR rtl-optimization/83962
4787 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4788 tidy_fallthru_edge and tidy_control_flow.
4790 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4792 PR rtl-optimization/83530
4794 * sel-sched.c (force_next_insn): New global variable.
4795 (remove_insn_for_debug): When force_next_insn is true, also leave only
4796 next insn in the ready list.
4797 (sel_sched_region): When the region wasn't scheduled, make another pass
4798 over it with force_next_insn set to 1.
4800 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4802 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4804 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4805 for interrupt control.
4806 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4807 functions for interrupt control.
4808 * config/nds32/nds32-intrinsic.md: Likewise.
4809 * config/nds32/nds32_intrinsic.h: Likewise.
4810 * config/nds32/nds32.h (nds32_builtins): Likewise.
4812 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4814 * config/nds32/nds32.c (nds32_init_machine_status,
4815 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4816 strict_aligned_p field.
4817 (nds32_expand_to_rtl_hook): New function.
4818 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4819 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4821 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4822 Chung-Ju Wu <jasonwucj@gmail.com>
4824 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4825 * config/nds32/nds32-n7.md: New file.
4826 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4827 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4829 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4830 * config/nds32/nds32.md (pipeline_model): Add n7.
4831 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4832 * config/nds32/pipelines.md: Include n7 settings.
4834 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4835 Chung-Ju Wu <jasonwucj@gmail.com>
4837 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4838 * config/nds32/nds32-e8.md: New file.
4839 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4840 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4842 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4843 * config/nds32/nds32.md (pipeline_model): Add e8.
4844 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4845 * config/nds32/pipelines.md: Include e8 settings.
4847 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4848 Chung-Ju Wu <jasonwucj@gmail.com>
4850 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4851 * config/nds32/nds32-n8.md: New file.
4852 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4853 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4855 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4856 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4857 * config/nds32/nds32.md (pipeline_model): Add n8.
4858 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4859 * config/nds32/pipelines.md: Include n8 settings.
4861 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4862 Chung-Ju Wu <jasonwucj@gmail.com>
4864 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4865 * config/nds32/nds32-n9-2r1w.md: New file.
4866 * config/nds32/nds32-n9-3r2w.md: New file.
4867 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4868 nds32_register_ports): New or modify for cpu n9.
4869 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4871 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4872 * config/nds32/nds32-utils.c: New file.
4873 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4874 TARGET_MUL_SLOW): Define.
4875 * config/nds32/nds32.md (pipeline_model): New attribute.
4876 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4877 New options that support cpu n9.
4878 * config/nds32/pipelines.md: Include n9 settings.
4879 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4881 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4883 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4884 information if necessary.
4885 (output_cond_branch_compare_zero): Likewise.
4886 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4887 (nds32_target_alignment): Refine for alignment.
4888 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4889 (FUNCTION_BOUNDARY): Modify.
4890 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4892 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4894 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4896 * config/nds32/constants.md (unspec_volatile_element): Add values for
4897 TLB operation and data prefetch.
4898 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4899 functions for TLB operation and data prefetch.
4900 * config/nds32/nds32-intrinsic.md: Likewise.
4901 * config/nds32/nds32_intrinsic.h: Likewise.
4902 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4903 (nds32_print_operand): Likewise.
4904 * config/nds32/nds32.h (nds32_builtins): Likewise.
4906 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4907 Andrew Pinski <pinsika@gcc.gnu.org>
4910 * match.pd: Use constant_boolean_node of correct type instead of
4911 boolean_true_node or boolean_false_node for simplifying
4912 pointer comparisons to zero.
4914 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4916 PR tree-optimization/80021
4917 * tree.c (verify_type_variant): Make error call in verify_variant_match
4918 translatable and remove final full stop.
4920 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4922 * config/nds32/constants.md (unspec_volatile_element): Add
4923 UNSPEC_VOLATILE_EH_RETURN.
4924 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4925 nds32_output_stack_pop): Support dwarf exception handling process.
4926 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4927 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4928 exception handling process.
4929 (nds32_compute_stack_frame): Likewise.
4930 (nds32_return_addr_rtx): Likewise.
4931 (nds32_initial_elimination_offset): Likewise.
4932 (nds32_expand_prologue): Likewise.
4933 (nds32_expand_epilogue): Likewise.
4934 (nds32_dynamic_chain_address): New function.
4935 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4937 (DYNAMIC_CHAIN_ADDRESS): Define.
4938 (EH_RETURN_DATA_REGNO): Define.
4939 (EH_RETURN_STACKADJ_RTX): Define.
4940 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4941 patterns for dwarf exception handling.
4943 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4945 * config/nds32/nds32.h: Clean up obsolete macros.
4947 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4949 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4950 Add enum values for particular instructions.
4951 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4952 particular intrinsic functions.
4953 * config/nds32/nds32-intrinsic.md: Likewise.
4954 * config/nds32/nds32_intrinsic.h: Likewise.
4955 * config/nds32/nds32.h (nds32_builtins): Likewise.
4956 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4957 (btst, ave): New patterns for particular instructions.
4959 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4961 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4962 Add enum values for atomic load/store and memory sync.
4963 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4965 * config/nds32/nds32-intrinsic.md: Likewise.
4966 * config/nds32/nds32_intrinsic.h: Likewise.
4967 * config/nds32/nds32.h (nds32_builtins): Likewise.
4969 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4971 PR tree-optimization/85257
4972 * fold-const.c (native_encode_vector): If not all elts could fit
4973 and off is -1, return 0 rather than offset.
4974 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4975 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4976 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4977 adjust buffer in native_interpret_expr call.
4979 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4981 * config/nds32/constants.md (unspec_volatile_element): Add cache
4982 control enum values.
4983 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4984 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4985 * config/nds32/nds32.c (nds32_cctl_names): New.
4986 (nds32_print_operand): Handle cache control register names.
4987 * config/nds32/nds32.h (nds32_builtins): New enum values.
4988 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4990 * config/nds32/nds32.md (type): Add mmu.
4991 * config/nds32/pipelines.md (simple_insn): Add mmu.
4993 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4995 * config/nds32/nds32.md (type): Remove call.
4996 * config/nds32/pipelines.md (simple_insn): Likewise.
4998 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5000 * config/nds32/constants.md (unspec_volatile_element): Add
5001 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5002 UNSPEC_VOLATILE_FMFCFG.
5003 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5004 description for fmfcfg and fmfcsr.
5005 (bdesc_1arg): Add fmtcsr.
5006 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5007 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5008 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5009 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5010 unspec_fmfcfg): New patterns.
5011 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5012 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5013 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5014 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5015 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5016 __nds32__fmfcfg): Define.
5018 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5020 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5021 intrinsic register names.
5022 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5023 intrinsic register enum values and macros.
5025 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5027 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5028 for load/store addressing form.
5029 (nds32_print_operand_address): Likewise.
5031 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5034 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5035 based on LABEL_REF. Remove useless assertion.
5036 (pic_address_needs_scratch): Fix formatting.
5037 (sparc_legitimize_pic_address): Minor tweaks.
5038 (sparc_delegitimize_address): Adjust assertion accordingly.
5039 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5040 into symbolic_operand.
5041 (movsi_high_pic_label_ref): Likewise.
5042 (movsi_lo_sum_pic_label_ref): Likewise.
5043 (movdi_pic_label_ref): Likewise.
5044 (movdi_high_pic_label_ref): Likewise.
5045 (movdi_lo_sum_pic_label_ref): Likewise.
5047 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5049 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5050 custom LIB_SPEC setup.
5052 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5053 Kito Cheng <kito.cheng@gmail.com>
5055 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5056 * config/riscv/freebsd.h: New.
5058 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5060 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5061 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5064 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5065 Kito Cheng <kito.cheng@gmail.com>
5067 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5068 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5069 * config/nds32/nds32-protos.h (nds32_output_call,
5070 nds32_output_return): Declare.
5071 * config/nds32/nds32.md: Refine all the call and return patterns.
5073 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5076 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5077 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5079 PR rtl-optimization/84872
5080 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5081 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5084 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5086 * expr.c (copy_blkmode_to_reg): Revert 254862.
5087 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5089 2018-04-06 Richard Biener <rguenther@suse.de>
5092 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5093 after seeing a component reference with an adjacent field. Treat
5094 refs to arrays at struct end of external decls similar to
5095 refs to unconstrained commons.
5097 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5100 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5101 look through SAVE_EXPRs with non-side-effects argument. Adjust
5103 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5106 2018-04-06 Richard Biener <rguenther@suse.de>
5109 * alias.c (find_base_term): New wrapper around find_base_term
5110 unwinding CSELIB_VAL_PTR changes.
5111 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5114 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5116 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5118 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5119 constant definitions.
5120 ("nop"): lr 0,0 -> nopr r0
5121 ("nop_lr0", "nop_lr1"): New insn definitions.
5123 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5125 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5126 NDS32_V3PUSH_AVAILABLE_P macro.
5128 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5129 Chung-Ju Wu <jasonwucj@gmail.com>
5131 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5132 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5133 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5134 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5135 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5136 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5137 * config/nds32/constraints.md: New constraints and checking for hard
5138 float configuration.
5139 * config/nds32/iterators.md: New mode iterator and attribute for hard
5140 float configuration.
5141 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5143 * config/nds32/nds32-fpu.md: New file.
5144 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5145 deal with hard float code generation.
5146 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5148 (abi_type, float_reg_number): New enum type.
5149 * config/nds32/nds32-predicates.c: New predicates for hard float.
5150 * config/nds32/nds32-protos.h: Declare functions for hard float.
5151 * config/nds32/nds32.c: Implementation for hard float configuration.
5152 * config/nds32/nds32.h: Definitions for hard float configuration.
5153 * config/nds32/nds32.md: Include hard float machine description and
5154 modify patterns for hard float configuration.
5155 * config/nds32/nds32.opt: New options for hard float configuration.
5156 * config/nds32/predicates.md: New predicates for hard float
5159 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5161 * common/config/nds32/nds32-common.c
5162 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5164 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5167 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5168 CONSTRUCTOR_ELT (ctor, ...)->value.
5170 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5173 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5175 2018-04-05 Tom de Vries <tom@codesourcery.com>
5178 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5181 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5182 Kito Cheng <kito.cheng@gmail.com>
5184 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5185 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5186 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5187 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5189 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5190 Kito Cheng <kito.cheng@gmail.com>
5192 * config/nds32/constraints.md (Ufe): New memory constraint.
5193 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5194 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5195 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5197 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5198 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5200 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5202 * config/nds32/nds32.md: Use optimize_size in the condition for
5203 alu-shift instructions.
5205 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5207 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5209 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5211 * config/nds32/nds32.md (negsi2): Refine pattern.
5213 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5214 Chung-Ju Wu <jasonwucj@gmail.com>
5216 * config/nds32/iterators.md (shift_rotate): New code iterator.
5217 (shift): New code attribute.
5218 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5219 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5220 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5221 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5222 bit-wise operations.
5223 (andsi3, *andsi3): Ditto.
5224 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5225 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5226 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5227 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5228 nds32_ior_operand, nds32_xor_operand): New predicates.
5230 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5232 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5233 (addsi3, subsi3): ... this.
5235 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5237 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5239 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5241 * config/nds32/nds32.md: Adjust indention.
5243 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5245 * config/nds32/nds32.md (feature): New attribute.
5247 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5249 * config/nds32/nds32.md (subtype): New attribute.
5251 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5254 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5255 expansion to perform a bitwise AND of the argument followed by a
5256 boolean negation of the result.
5258 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5260 PR rtl-optimization/84878
5261 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5262 the basic block. Assert the use reference is not artificial and that
5263 it has an associated insn.
5265 2018-04-04 Michael Matz <matz@suse.de>
5267 * builtins.c (compute_objsize): Pass correct operand
5268 to array_at_struct_end_p.
5270 2018-04-04 Richard Biener <rguenther@suse.de>
5273 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5274 from contexts for DINFO_LEVEL_TERSE and below.
5276 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5278 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5279 resiter_operand condition.
5280 * config/nds32/nds32.md (*move<mode>): Ditto.
5282 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5283 Monk Chiang <sh.chiang04@gmail.com>
5285 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5287 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5289 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5291 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5292 Kito Cheng <kito.cheng@gmail.com>
5294 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5295 nds32_cond_code_str, output_cond_branch,
5296 output_cond_branch_compare_zero, nds32_expand_cbranch,
5297 nds32_expand_cstore, nds32_expand_movcc,
5298 nds32_output_cbranchsi4_equality_zero,
5299 nds32_output_cbranchsi4_equality_reg,
5300 nds32_output_cbranchsi4_equality_reg_or_const_int,
5301 nds32_output_cbranchsi4_greater_less_zero: New functions.
5302 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5303 nds32_expand_cstore, nds32_expand_movcc,
5304 nds32_output_cbranchsi4_equality_zero,
5305 nds32_output_cbranchsi4_equality_reg,
5306 nds32_output_cbranchsi4_equality_reg_or_const_int,
5307 nds32_output_cbranchsi4_greater_less_zero): Declare.
5308 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5309 nds32_rimm11s_operand): New predicates.
5310 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5311 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5314 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5316 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5317 * config/nds32/nds32.md: Ditto.
5318 * config/nds32/pipelines.md: Ditto.
5320 2018-04-04 Richard Biener <rguenther@suse.de>
5322 PR tree-optimization/85168
5323 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5324 propagating abnormals.
5326 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5328 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5330 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5331 Kito Cheng <kito.cheng@gmail.com>
5333 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5334 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5335 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5336 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5337 * config/nds32/nds32.md (sibcall_internal): New.
5338 (sibcall_register): Remove.
5339 (sibcall_immediate): Remove.
5340 (sibcall_value_internal): New.
5341 (sibcall_value_register): Remove.
5342 (sibcall_value_immediate): Remove.
5343 * config/nds32/predicates.md (nds32_general_register_operand): New.
5344 (nds32_call_address_operand): New.
5346 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5348 PR rtl-optimization/85167
5349 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5350 bb_defs if *split_p, instead preinitialize it to NULL.
5352 PR tree-optimization/85156
5353 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5354 evaluating the argument multiple times.
5356 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5358 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5360 (_mm_cvtpd_ps): Likewise.
5361 (_mm_cvttpd_epi32): Likewise.
5362 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5363 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5364 vector, pixel, and bool following altivec.h include.
5366 2018-04-03 Martin Sebor <msebor@redhat.com>
5368 * doc/extend.texi (Common Function Attributes): Clarify.
5369 (const attribute): Likewise.
5370 (pure attribute): Likewise.
5372 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5375 * config/i386/i386.c (ix86_expand_vector_set): Use
5376 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5378 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5380 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5381 instructions when changing rounding bits to preserve precision bits
5382 in the x87 control word.
5384 2018-04-03 Martin Liska <mliska@suse.cz>
5386 PR tree-optimization/82491
5387 * rtl.h (strip_offset_and_add): Replace += suboffset with
5388 poly_uint64 () + suboffset.
5390 2018-03-29 Martin Liska <mliska@suse.cz>
5391 Martin Jambor <mjambor@suse.cz>
5394 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5395 param_type is not an integral or pointer type.
5397 2018-04-03 Richard Biener <rguenther@suse.de>
5399 * sese.h (recompute_all_dominators): Remove.
5401 2018-04-02 Martin Sebor <msebor@redhat.com>
5403 * doc/invoke.texi (-Wrestrict): Fix typos.
5405 2018-04-02 Jim Wilson <jimw@sifive.com>
5407 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5408 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5409 (<optab>di3, <optab>si3_extend): Likewise.
5410 (<optab>si3_mask, <optab>si3_mask_1): New.
5411 (<optab>di3_mask, <optab>di3_mask_1): New.
5412 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5413 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5414 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5416 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5418 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5421 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5423 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5424 (nds32_canonicalize_comparison): New function.
5426 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5427 Kito Cheng <kito.cheng@gmail.com>
5428 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5430 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5431 * config/nds32/constants.md (unspec_volatile_element): Add
5432 UNSPEC_VOLATILE_RELAX_GROUP.
5433 * config/nds32/nds32-relax-opt.c: New file.
5434 * config/nds32/nds32-predicates.c
5435 (nds32_symbol_load_store_p): New function.
5436 * config/nds32/nds32-protos.h
5437 (nds32_symbol_load_store_p): Declare function.
5438 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5439 * config/nds32/nds32.c
5440 (nds32_register_pass): New function to register pass.
5441 (nds32_register_passes): New function to register passes.
5442 * config/nds32/nds32.md (relax_group): New pattern.
5443 * config/nds32/nds32.opt (mrelax-hint): New option.
5444 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5446 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5448 * config/nds32/t-nds32: Modify files dependency.
5450 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5452 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5453 (PROFILE_HOOK): Define its implementation.
5455 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5457 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5458 type and 32-bit size.
5460 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5463 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5464 (V_128_256): New mode iterator.
5465 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5466 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5467 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5469 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5470 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5472 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5475 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5476 NaN inputs correctly.
5478 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5481 * config/rs6000/vsx.md (??r): New mode attribute.
5482 (*vsx_mov<mode>_64bit): Use it.
5483 (*vsx_mov<mode>_32bit): Likewise.
5485 2018-03-30 Martin Sebor <msebor@redhat.com>
5487 PR tree-optimization/84818
5488 * builtins.c (check_access): Use warning_n.
5490 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5493 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5495 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5498 2018-03-30 Julia Koval <julia.koval@intel.com>
5501 * x86-tune.def (movx, partial_reg_dependency): Enable for
5504 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5507 * lra-constraints.c (process_alt_operands): Move setting
5508 this_alternative_matches below.
5510 2018-03-29 Martin Liska <mliska@suse.cz>
5513 * doc/invoke.texi: Document how LTO works with debug info.
5514 Describe auto-load support of binutils. Mention 'x86-64'
5515 as valid option value of -march option.
5517 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5519 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5522 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5523 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5524 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5527 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5530 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5531 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5532 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5533 for RS6000_BTM_POWERPC64.
5534 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5535 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5536 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5541 2018-03-28 Carl Love <cel@us.ibm.com>
5544 2017-09-27 Carl Love <cel@us.ibm.com>
5546 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5547 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5548 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5551 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5553 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5554 instead of __vector bool.
5555 (_mm_max_pu8): Likewise.
5556 (_mm_min_pi16): Likewise.
5558 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5561 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5562 (DIVWEUO): Likewise.
5564 (DIVDEUO): Likewise.
5565 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5566 DIVWEUO and DIVDEUO.
5567 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5568 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5569 (div_extend): Likewise.
5570 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5572 (__builtin_divweuo): Likewise.
5573 (__builtin_divdeo): Likewise.
5574 (__builtin_divdeuo): Likewise.
5576 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5579 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5580 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5582 PR tree-optimization/82004
5583 * gimple-match-head.c (optimize_pow_to_exp): New function.
5584 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5585 Don't fold to exp if optimize_pow_to_exp is false.
5587 2018-03-28 Martin Liska <mliska@suse.cz>
5590 * calls.c (initialize_argument_information): Fix trailing space.
5591 * common.opt: Fix typo and provide better explanation for
5592 -fsanitize-coverage option.
5593 * config/i386/i386.opt: Fix typo.
5595 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5596 Martin Liska <mliska@suse.cz>
5599 * gimplify.c (asan_poison_variable): Don't do the check for
5600 gimplify_omp_ctxp here.
5601 (gimplify_decl_expr): Do it here.
5602 (gimplify_target_expr): Likewise.
5604 2018-03-28 Martin Liska <mliska@suse.cz>
5607 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5608 chkp_type_bounds_count if MPX is not enabled.
5610 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5612 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5614 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5617 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5618 function to create the function decl for complex long double
5619 multiply and divide for -mabi=ieeelongdouble.
5620 (init_float128_ieee): Call it.
5622 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5625 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5626 -fcf-protection=branch -mibt.
5627 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5629 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5632 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5634 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5637 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5638 extern array declarations.
5640 2018-03-27 Richard Biener <rguenther@suse.de>
5643 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5644 explicit single_use checks.
5646 2018-03-27 Richard Biener <rguenther@suse.de>
5648 PR tree-optimization/85082
5649 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5652 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5654 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5655 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5656 Turn on fasynchronous-unwind-tables and funwind-tables.
5658 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5661 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5662 (*bmi_blsr_<mode>_ccz): Ditto.
5664 2018-03-26 Tom de Vries <tom@codesourcery.com>
5666 PR tree-optimization/85063
5667 * omp-general.c (offloading_function_p): New function. Factor out
5669 * omp-offload.c (pass_omp_target_link::gate): ... here.
5670 * omp-general.h (offloading_function_p): Declare.
5671 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5672 with attribute omp declare target for offloading functions.
5674 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5676 PR tree-optimization/84005
5677 * tree-data-ref.h (get_base_for_alignment): Declare.
5678 * tree-data-ref.c (get_base_for_alignment_1): New function.
5679 (get_base_for_alignment): Likewise.
5680 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5681 get_base_for_alignment to find a suitable base object, instead
5682 of always using drb->base_address.
5684 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5687 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5688 known size by default.
5690 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5693 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5694 and non BLKmode operands.
5696 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5699 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5700 Clean up attributes.
5702 2018-03-23 Richard Biener <rguenther@suse.de>
5705 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5706 we are going to emit early debug for LTO.
5708 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5711 * function.c (match_asm_constraints_1): Don't optimize if input
5712 doesn't satisfy general_operand predicate for output's mode.
5715 * alias.c (write_dependence_p): Don't require for x_canonicalized
5716 non-VOIDmode if x has VOIDmode.
5719 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5720 just don't try to optimize it rather than assert it never happens.
5722 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5724 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5725 macro expansions for definition of ST_INTERNAL_<mode> and
5726 LD_INTERNAL_<mode> builtins.
5727 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5729 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5731 (altivec_expand_st_builtin): Likewise.
5732 (altivec_expand_builtin): Remove calls to deleted functions.
5733 (rs6000_address_for_altivec): Delete this function.
5734 * config/rs6000/vector.md: Remove expands for
5735 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5737 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5740 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5741 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5742 re-computing once computed.
5743 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5744 (arm_init_machine_status): Initialize
5745 machine->static_chain_stack_bytes.
5747 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5750 * doc/extend.texi: Add four new prototypes for vec_ld.
5751 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5752 definitions for more logical presentation.
5753 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5754 entries for V1TI variants of __builtin_altivec_ld builtin.
5755 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5756 handling of V1TI variant of LVX icode pattern.
5757 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5758 (rs6000_gimple_fold_builtin): Likewise.
5759 (altivec_init_builtins): Add code to define
5760 __builtin_altivec_lvx_v1ti function.
5762 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5765 * function.c (match_asm_constraints_1): Don't do the optimization
5766 if input isn't a REG, SUBREG, MEM or constant.
5768 2018-03-22 Tom de Vries <tom@codesourcery.com>
5770 PR tree-optimization/84956
5771 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5772 bb_has_abnormal_pred.
5774 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5777 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5778 DECL_INITIAL (decl) to decl at the end.
5779 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5782 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5784 * doc/extend.texi (__builtin_tgmath): Document when complex
5785 integer types are treated as _Complex _Float64.
5787 2018-03-21 Tom de Vries <tom@codesourcery.com>
5789 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5791 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5793 PR tree-optimization/84960
5794 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5795 if it is ENTRY block, move them into single succ of ENTRY in that case.
5797 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5799 PR tree-optimization/84811
5800 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5801 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5802 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5803 (known_subrange_p): Update accordingly. Cast each value involved
5804 in the size comparison, rather than casting the result of the
5807 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5809 PR tree-optimization/84982
5810 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5811 by flipping the least significant bit rather than all bits from
5812 bitpos to bitpos + bitsize - 1.
5814 2018-03-21 Nathan Sidwell <nathan@acm.org>
5816 * doc/extend.texi (Deprecated Features): Remove mention of
5817 long-deleted deprecations.
5819 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5822 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5823 * configure: Regenerate.
5825 2018-03-21 Tom de Vries <tom@codesourcery.com>
5827 PR tree-optimization/83126
5828 * tree-parloops.c (num_phis): New function.
5829 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5831 2018-03-21 Nathan Sidwell <nathan@acm.org>
5833 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5834 mention anon-struct/union members and trailing attributes.
5836 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5838 PR tree-optimization/84969
5839 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5840 builtin memset partitions if they set different rhs values.
5842 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5844 PR rtl-optimization/84989
5845 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5846 VEC_DUPLICATE with scalar result mode.
5848 2018-03-21 Martin Liska <mliska@suse.cz>
5851 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5852 not intended return statement.
5854 2018-03-21 Martin Liska <mliska@suse.cz>
5857 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5858 (chkp_find_bound_slots_1): Limit number of iterations.
5860 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5863 * Minor grammar fixes for x86 options.
5865 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5868 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5869 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5871 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5874 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5875 (altivec_lvx_<mode>_1op): Likewise.
5876 (altivec_stvx_<mode>_2op): Likewise.
5877 (altivec_stvx_<mode>_1op): Likewise.
5878 (altivec_lvx_<VM2:mode>): New define_expand.
5879 (altivec_stvx_<VM2:mode>): Likewise.
5880 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5881 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5882 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5883 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5884 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5885 (rs6000_gen_lvx): Likewise.
5886 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5887 (altivec_expand_stv_builtin): Likewise.
5888 (altivec_expand_builtin): Likewise.
5889 * config/rs6000/vector.md: Likewise.
5891 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5894 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5897 2018-03-20 Richard Biener <rguenther@suse.de>
5900 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5901 sign-conversions as zero, fall back to standard scalar_stmt
5904 2018-03-20 Martin Liska <mliska@suse.cz>
5907 * predict.c (rebuild_frequencies): Handle case when we have
5908 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5910 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5913 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5914 flag_section_anchors.
5915 * varasm.c (use_blocks_for_decl_p): Remove hack for
5916 dw2_force_const_mem.
5919 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5921 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5922 be created, use lowpart_subreg of operands[0] rather than operands[0]
5924 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5925 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5926 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5927 and n constraint instead of aarch64_shift_imm_di and Usd.
5928 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5929 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5931 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5934 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5935 to favor GPR over NEON registers.
5936 (<shift>di3_neon): Likewise.
5938 2018-03-20 Tom de Vries <tom@codesourcery.com>
5941 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5942 (nvptx_process_pars): Emit bar.sync asap and alap.
5944 2018-03-20 Tom de Vries <tom@codesourcery.com>
5947 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5948 seen_label if seen_label is already set.
5950 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5953 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5954 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5955 Use 1U instead of 1. Formatting fixes.
5958 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5959 instead of TREE_TYPE (s1) for the return value.
5961 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5963 PR tree-optimization/84946
5964 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5965 bitsize + bitsize in poly_uint64 rather than poly_int64.
5968 * dwarf2asm.c: Include fold-const.c.
5969 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5970 of decl rather than decl itself.
5972 PR rtl-optimization/84643
5973 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5975 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5978 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5979 calling assemble_variable.
5981 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5984 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5985 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5987 2018-03-19 Jim Wilson <jimw@sifive.com>
5990 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5991 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5992 (riscv_first_stack_step): Likewise.
5993 (riscv_option_override): Use STACK_BOUNDARY instead of
5995 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5997 (BIGGEST_ALIGNMENT): Set to 128.
5998 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5999 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6002 2018-03-19 Richard Biener <rguenther@suse.de>
6004 PR tree-optimization/84933
6005 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6006 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6008 2018-03-19 Richard Biener <rguenther@suse.de>
6010 PR tree-optimization/84859
6011 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6012 (cond_if_else_store_replacement): Perform sinking operation on
6013 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6014 Generalize what a BB with a single eligible store is.
6016 2018-03-19 Richard Biener <rguenther@suse.de>
6018 PR tree-optimization/84929
6019 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6020 chrec_is_positive against non-chrec arg.
6022 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6025 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6027 2018-03-18 Martin Liska <mliska@suse.cz>
6029 PR rtl-optimization/84635
6030 * regrename.c (build_def_use): Use matches_mode only when
6033 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6035 PR tree-optimization/84913
6036 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6037 vectorize chains of COND_EXPRs.
6039 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6041 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6043 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6045 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6047 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6049 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6051 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6052 Kito Cheng <kito.cheng@gmail.com>
6054 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6055 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6056 (nds32_adjust_reg_alloc_order): New function.
6057 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6059 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6061 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6062 nds32_print_operand, nds32_print_operand_address): Use
6063 HOST_WIDE_INT_PRINT_DEC instead.
6065 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6067 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6069 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6072 * config/i386/i386.c (initial_ix86_tune_features,
6073 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6075 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6076 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6077 rather than 1u << ix86_tune. Formatting fix.
6078 (ix86_option_override_internal): Change ix86_arch_mask from
6079 unsigned int to unsigned HOST_WIDE_INT, initialize to
6080 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6081 (ix86_function_specific_restore): Likewise.
6083 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6086 * postreload.c (reload_combine_recognize_pattern): Perform
6087 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6088 truncate_int_for_mode the result for the destination's mode.
6091 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6092 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6094 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6097 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6100 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6101 regno_allocno_class_array and sorted_pseudos.
6102 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6103 insns where regno is used.
6105 2018-03-16 Martin Liska <mliska@suse.cz>
6108 * multiple_target.c (create_dispatcher_calls): Redirect
6109 reference in the symbol table.
6111 2018-03-16 Martin Liska <mliska@suse.cz>
6114 * multiple_target.c (create_dispatcher_calls): Redirect also
6117 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6121 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6122 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6123 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6125 2018-03-16 Julia Koval <julia.koval@intel.com>
6127 * doc/invoke.texi (Skylake Server): Add CLWB.
6128 Cannonlake): Remove CLWB.
6130 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6132 PR tree-optimization/84841
6133 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6135 (FLOAT_ONE_CONST_TYPE): Define.
6136 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6137 (sort_by_operand_rank): Put entries with higher constant_type last
6138 rather than first to match comments.
6140 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6142 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6145 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6148 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6149 check and use address of target always.
6151 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6154 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6155 (indirect_thunk_bnd_needed): Likewise.
6156 (indirect_thunks_used): Likewise.
6157 (indirect_thunks_bnd_used): Likewise.
6158 (indirect_return_needed): New.
6159 (indirect_return_bnd_needed): Likewise.
6160 (output_indirect_thunk_function): Add a bool argument for
6162 (output_indirect_thunk_function): Don't generate alias for
6163 function return thunk.
6164 (ix86_code_end): Call output_indirect_thunk_function to generate
6165 function return thunks.
6166 (ix86_output_function_return): Set indirect_return_bnd_needed
6167 and indirect_return_needed instead of indirect_thunk_bnd_needed
6168 and indirect_thunk_needed.
6170 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6172 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6173 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6174 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6176 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6177 Paul Hua <paul.hua.gm@gmail.com>
6180 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6182 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6184 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6185 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6188 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6191 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6192 instead of GET_MODE_SIZE when comparing Units.
6194 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6197 * varasm.c (hash_section): Return an unchangeble hash value
6198 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6199 Return !aarch64_can_use_per_function_literal_pools_p ().
6201 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6204 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6205 rather than address of cmode as last argument to prepare_cmp_insn.
6207 2018-03-15 Julia Koval <julia.koval@intel.com>
6209 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6210 F_AVX512VNNI, F_AVX512BITALG): New.
6212 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6215 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6216 insn for floating-point loads and stores.
6218 2018-03-14 Carl Love <cel@us.ibm.com>
6220 * config/rs6000/rs6000-c.c: Add macro definitions for
6221 ALTIVEC_BUILTIN_VEC_PERMXOR.
6222 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6223 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6224 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6225 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6227 * config/doc/extend.texi: Add prototypes for vec_permxor.
6229 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6232 * diagnostic-show-locus.c (class layout_point): Convert m_line
6233 from int to linenum_type.
6234 (line_span::comparator): Use linenum "compare" function when
6235 comparing line numbers.
6236 (test_line_span): New function.
6237 (layout_range::contains_point): Convert param "row" from int to
6239 (layout_range::intersects_line_p): Likewise.
6240 (layout::will_show_line_p): Likewise.
6241 (layout::print_source_line): Likewise.
6242 (layout::should_print_annotation_line_p): Likewise.
6243 (layout::print_annotation_line): Likewise.
6244 (layout::print_leading_fixits): Likewise.
6245 (layout::annotation_line_showed_range_p): Likewise.
6246 (struct line_corrections): Likewise for field m_row.
6247 (line_corrections::line_corrections): Likewise for param "row".
6248 (layout::print_trailing_fixits): Likewise.
6249 (layout::get_state_at_point): Likewise.
6250 (layout::get_x_bound_for_row): Likewise.
6251 (layout::print_line): Likewise.
6252 (diagnostic_show_locus): Likewise for locals "last_line" and
6254 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6255 * input.c (selftest::test_linenum_comparisons): New function.
6256 (selftest::input_c_tests): Call it.
6257 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6258 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6259 * selftest.h (ASSERT_GT): New macro.
6260 (ASSERT_GT_AT): New macro.
6261 (ASSERT_LT): New macro.
6262 (ASSERT_LT_AT): New macro.
6264 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6266 PR rtl-optimization/84780
6267 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6269 2018-03-14 Martin Liska <mliska@suse.cz>
6271 * tree.c (record_node_allocation_statistics): Use
6272 get_stats_node_kind.
6273 (get_stats_node_kind): New function extracted from
6274 record_node_allocation_statistics.
6275 (free_node): Use get_stats_node_kind.
6277 2018-03-14 Richard Biener <rguenther@suse.de>
6279 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6280 that the value-set of ANTIC_IN doesn't grow.
6283 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6285 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6286 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6288 2018-03-14 Julia Koval <julia.koval@intel.com>
6290 * config.gcc (icelake-client, icelake-server): New.
6292 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6293 (initial_ix86_arch_features): Ditto.
6294 (PTA_SKYLAKE): Add SGX.
6295 (PTA_ICELAKE): Remove.
6296 (PTA_ICELAKE_CLIENT): New.
6297 (PTA_ICELAKE_SERVER): New.
6298 (ix86_option_override_internal): Split up icelake on icelake client and
6300 (get_builtin_code_for_version): Ditto.
6301 (fold_builtin_cpu): Ditto.
6302 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6303 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6304 * config/i386/i386.h (processor_type): Ditto.
6305 * doc/invoke.texi: Ditto.
6307 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6310 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6311 INTEGER_CST offset, add it together with bitpos / 8 and
6312 sign extend based on POINTER_SIZE.
6316 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6319 * config/i386/constraints.md (Yc): New register constraint.
6320 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6321 Use Yc constraint for alternative 2 of operand 0. Remove
6322 preferred_for_speed attribute.
6324 2018-03-14 Richard Biener <rguenther@suse.de>
6326 PR tree-optimization/84830
6327 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6328 with the old one to avoid oscillations.
6330 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6333 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6335 (assign_by_spills): Return a flag of reload assignment failure.
6336 Do not process the reload assignment failures. Do not spill other
6337 reload pseudos if they has the same reg class. Update n if
6339 (lra_assign): Add a return arg. Set up from the result of
6340 assign_by_spills call.
6341 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6342 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6343 usage_insns if it is not NULL.
6344 (spill_hard_reg_in_range): New function.
6345 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6346 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6347 function prototypes.
6348 (lra_assign): Change prototype.
6349 * lra.c (lra): Add code to deal with fails by splitting hard reg
6352 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6354 * config/riscv/riscv.opt (mrelax): New option.
6355 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6356 "norelax" when riscv_mrelax is disabled.
6357 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6359 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6362 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6363 reassociation for int modes.
6365 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6367 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6368 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6370 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6371 * config/aarch64/aarch64-sve.md
6372 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6373 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6374 (*extend<mode><Vwide>2): Rename to...
6375 (aarch64_sve_extend<mode><Vwide>2): ...this.
6376 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6377 renaming the old pattern to...
6378 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6380 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6381 define_expand, renaming the old pattern to...
6382 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6383 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6384 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6385 account when deciding which SVE instruction the optab should use.
6386 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6388 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6390 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6391 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6392 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6393 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6394 (tlsdesc_small_<mode>): Turn a define_expand and use
6395 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6396 (tlsdesc_small_advsimd_<mode>): ...this.
6397 (tlsdesc_small_sve_<mode>): New pattern.
6399 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6401 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6402 (UNSPEC_UMUL_HIGHPART): New constants.
6403 (MUL_HIGHPART): New int iteraor.
6404 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6405 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6407 (*<su>mul<mode>3_highpart): New define_insn.
6409 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6412 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6415 2018-03-13 Martin Liska <mliska@suse.cz>
6418 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6420 (sem_item_optimizer::~sem_item_optimizer): Release it.
6421 (sem_item_optimizer::merge_classes): Register variable aliases.
6422 (sem_item_optimizer::fixup_pt_set): New function.
6423 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6424 * ipa-icf.h: Declare new variables and functions.
6426 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6429 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6430 integer_pow2p@2 and test integer_pow2p in condition.
6431 (A < 0 ? C : 0): Similarly for @1.
6434 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6435 characters starting at p contain '\0' character, don't look beyond
6439 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6440 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6443 * reg-stack.c (change_stack): Change update_end var from int to
6444 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6445 also call set_block_for_insn on the newly added insns and rescan.
6448 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6449 on the last operand.
6452 * tree.c (stabilize_reference_1): Return save_expr (e) for
6453 STATEMENT_LIST even if it doesn't have side-effects.
6455 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6457 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6459 2018-03-12 Renlin Li <renlin.li@arm.com>
6461 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6462 aarch64_output_scalar_simd_mov_immediate.
6464 2018-03-12 Martin Sebor <msebor@redhat.com>
6466 PR tree-optimization/83456
6467 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6468 for perfectly overlapping calls to memcpy.
6469 (gimple_fold_builtin_memory_chk): Same.
6470 (gimple_fold_builtin_strcpy): Handle no-warning.
6471 (gimple_fold_builtin_stxcpy_chk): Same.
6472 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6474 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6476 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6477 parameter. Use it for SFmode.
6478 (rs6000_function_arg_advance_1): Adjust.
6479 (rs6000_function_arg): Adjust.
6480 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6482 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6484 PR rtl-optimization/84169
6485 PR rtl-optimization/84780
6486 * combine.c (can_combine_p): Check for a 2-insn combination whether
6487 the destination register is used between the two insns, too.
6489 2018-03-12 Richard Biener <rguenther@suse.de>
6491 PR tree-optimization/84803
6492 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6493 for refs DR analysis didn't process.
6495 2018-03-12 Richard Biener <rguenther@suse.de>
6497 PR tree-optimization/84777
6498 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6499 force-vectorize loops ignore whether we are optimizing for size.
6501 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6503 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6504 (TARGET_MD_ASM_ADJUST): Define.
6506 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6507 Kito Cheng <kito.cheng@gmail.com>
6508 Chung-Ju Wu <jasonwucj@gmail.com>
6510 * config/nds32/nds32.c (nds32_compute_stack_frame,
6511 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6512 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6513 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6514 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6515 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6516 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6517 * config/nds32/nds32.md (prologue, epilogue): Use macro
6518 NDS32_V3PUSH_AVAILABLE_P to do checking.
6520 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6523 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6524 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6525 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6526 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6527 addition of most attributes on !orig_type_die or the attribute not
6528 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6530 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6531 Chung-Ju Wu <jasonwucj@gmail.com>
6533 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6535 * config/nds32/nds32.opt (mvh): New option.
6537 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6538 Chung-Ju Wu <jasonwucj@gmail.com>
6540 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6542 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6543 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6546 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6547 Chung-Ju Wu <jasonwucj@gmail.com>
6549 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6551 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6552 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6554 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6555 Kito Cheng <kito.cheng@gmail.com>
6556 Chung-Ju Wu <jasonwucj@gmail.com>
6558 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6559 UNSPEC_FFMISM and UNSPEC_FLMISM.
6560 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6561 for ffb, ffmism and flmism.
6562 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6563 (unspec_ffmism): Ditto.
6564 (unspec_flmism): Ditto.
6565 (nds32_expand_builtin_impl): Check if string extension is available.
6566 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6567 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6569 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6572 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6575 * lra-assigns.c (assign_by_spills): Return a flag of reload
6576 assignment failure. Do not process the reload assignment
6577 failures. Do not spill other reload pseudos if they has the same
6579 (lra_assign): Add a return arg. Set up from the result of
6580 assign_by_spills call.
6581 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6582 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6583 usage_insns if it is not NULL.
6584 (spill_hard_reg_in_range): New function.
6585 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6586 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6587 function prototypes.
6588 (lra_assign): Change prototype.
6589 * lra.c (lra): Add code to deal with fails by splitting hard reg
6592 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6595 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6597 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6600 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6601 (dw_val_node): Add val_symbolic_view.
6602 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6603 (symview_upper_bound): New.
6604 (new_line_info_table): Initialize symviews_since_reset.
6605 (dwarf2out_source_line): Count symviews_since_reset and set
6606 symview_upper_bound.
6607 (dw_val_equal_p): Handle symview.
6608 (add_AT_symview): New.
6609 (print_dw_val): Handle symview.
6610 (attr_checksum, attr_checksum_ordered): Likewise.
6611 (same_dw_val_p, size_of_die): Likewise.
6612 (value_format, output_die): Likewise.
6613 (add_high_low_attributes): Use add_AT_symview for entry_view.
6614 (dwarf2out_finish): Reset symview_upper_bound, clear
6617 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6620 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6621 Add strict argument and use it.
6622 (rs6000_split_multireg_move): Update for new strict argument.
6623 (mem_operand_gpr): Disallow all non-offsettable addresses.
6624 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6626 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6629 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6630 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6631 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6634 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6635 decl, use remap_type if we want to use the type.
6637 2018-03-09 Martin Sebor <msebor@redhat.com>
6639 PR tree-optimization/84526
6640 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6642 (builtin_access::generic_overlap): Be prepared to handle non-array
6645 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6647 PR rtl-optimization/84682
6648 * lra-constraints.c (process_address_1): Check is_address flag
6649 for address constraints.
6650 (process_alt_operands): Likewise.
6651 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6652 preprocess_constraints.
6653 * recog.h (preprocess_constraints): Add oploc parameter.
6655 * recog.c (preprocess_constraints): Test address_operand for
6656 CT_ADDRESS constraints.
6658 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6661 * lra-assigns.c (assign_by_spills): Return a flag of reload
6662 assignment failure. Do not process the reload assignment
6663 failures. Do not spill other reload pseudos if they has the same
6665 (lra_assign): Add a return arg. Set up from the result of
6666 assign_by_spills call.
6667 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6668 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6669 usage_insns if it is not NULL.
6670 (spill_hard_reg_in_range): New function.
6671 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6672 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6673 function prototypes.
6674 (lra_assign): Change prototype.
6675 * lra.c (lra): Add code to deal with fails by splitting hard reg
6678 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6681 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6682 Accept complain bool parameter. Only emit errors if it is true.
6683 (arm_parse_cpu_option_name): Likewise.
6684 (arm_target_thumb_only): Adjust callers of the above.
6685 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6686 prototype to take a default true bool parameter.
6687 (arm_parse_arch_option_name): Likewise.
6689 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6690 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6694 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6695 * configure: Regenerate.
6696 * configure.ac ("linker --version-script option"): New.
6697 ("linker soname option"): New.
6699 2018-03-09 Richard Biener <rguenther@suse.de>
6701 PR tree-optimization/84775
6702 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6703 immediate uses of predicate stmts and mark them modified.
6706 PR tree-optimization/84178
6707 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6709 (version_loop_for_if_conversion): Delay update_ssa call.
6710 (tree_if_conversion): Delay update_ssa until after predicate
6713 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6716 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6717 when the function accesses prior frames.
6719 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6722 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6723 gen_llsym, otherwise call maybe_gen_llsym.
6726 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6727 has ',' character inside of it.
6729 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6732 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6733 as clobbering CC_REGNUM.
6735 2018-03-08 Richard Biener <rguenther@suse.de>
6738 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6739 (follow_copies_to_constant): Do not follow SSA names registered
6742 2018-03-08 Richard Biener <rguenther@suse.de>
6744 PR tree-optimization/84178
6745 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6747 (version_loop_for_if_conversion): Delay update_ssa call.
6748 (tree_if_conversion): Delay update_ssa until after predicate
6751 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6753 PR tree-optimization/84178
6754 * tree-if-conv.c (release_bb_predicate): Remove the
6755 the assertion that the stmts have NULL use_ops.
6756 Discard the statements, asserting that they haven't
6757 yet been added to a BB.
6759 2018-03-08 Richard Biener <rguenther@suse.de>
6761 PR tree-optimization/84746
6762 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6763 (phi_translate): Pass in destination ANTIC_OUT set.
6764 (phi_translate_1): Likewise. For a simplified result lookup
6765 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6766 (phi_translate_set): Adjust.
6767 (do_pre_regular_insertion): Likewise.
6768 (do_pre_partial_partial_insertion): Likewise.
6770 2018-03-08 Martin Liska <mliska@suse.cz>
6772 PR gcov-profile/84735
6773 * doc/gcov.texi: Document usage of profile files.
6774 * gcov-io.h: Document changes in the format.
6776 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6780 * dwarf2out.c (struct dw_line_info_table): Update comments for
6782 (FORCE_RESET_NEXT_VIEW): New.
6783 (FORCE_RESETTING_VIEW_P): New.
6784 (RESETTING_VIEW_P): Check for -1 too.
6785 (ZERO_VIEW_P): Likewise.
6786 (new_line_info_table): Force-reset next view.
6787 (dwarf2out_begin_function): Likewise.
6788 (dwarf2out_source_line): Simplify zero_view_p initialization.
6789 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6790 view directly. Omit view when omitting .loc at line 0.
6792 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6794 PR tree-optimization/84740
6795 * tree-switch-conversion.c (process_switch): Call build_constructors
6796 only if info.phi_count is non-zero.
6798 PR tree-optimization/84739
6799 * tree-tailcall.c (find_tail_calls): Check call arguments against
6800 DECL_ARGUMENTS (current_function_decl) rather than
6801 DECL_ARGUMENTS (func) when checking for tail recursion.
6803 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6805 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6806 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6807 Volker Reichelt's entry and add entries for people that perform
6808 GCC fuzzy testing and report numerous bugs.
6810 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6813 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6814 readonly data in sdata, if that is disabled.
6815 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6816 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6817 -mreadonly-in-sdata option.
6819 2018-03-07 Martin Sebor <msebor@redhat.com>
6821 PR tree-optimization/84468
6822 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6823 basic block when looking for nul assignment.
6825 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6828 * except.h (output_function_exception_table): Adjust prototype.
6829 * except.c (output_function_exception_table): Remove FNNAME parameter
6830 and add SECTION parameter. Ouput one part of the table at a time.
6831 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6832 the first part of the exception table and emit unwind directives.
6833 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6834 (i386_pe_seh_cold_init): Likewise.
6835 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6836 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6837 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6838 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6839 * config/i386/winnt.c: Include except.h.
6840 (struct seh_frame_state): Add reg_offset, after_prologue and
6841 in_cold_section fields.
6842 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6843 (i386_pe_seh_cold_init): New function.
6844 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6845 to seh->in_cold_section.
6846 (seh_emit_push): Record the offset of the push.
6847 (seh_emit_save): Record the offet of the save.
6848 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6849 Test seh->after_prologue to disregard the epilogue.
6850 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6851 (i386_pe_end_cold_function): New function.
6853 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6856 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6857 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6860 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6862 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6863 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6866 * multiple_target.c: Include tree-inline.h and intl.h.
6867 (expand_target_clones): Diagnose and fail if node->definition and
6868 !tree_versionable_function_p (node->decl).
6870 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6872 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6874 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6875 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6876 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6878 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6881 * combine.c (try_combine): Use reg_or_subregno instead of handling
6882 just paradoxical SUBREGs and REGs.
6884 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6886 * config/arc/arc.c (arc_finalize_pic): Remove function.
6887 (arc_must_save_register): We use single base PIC register, remove
6888 checks to save/restore the PIC register.
6889 (arc_expand_prologue): Likewise.
6890 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6892 (arc_verify_short): Likewise.
6893 (arc_attr_type): Likewise.
6894 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6895 (walk_stores): Likewise.
6896 (arc_address_cost): Make it static.
6897 (arc_verify_short): Likewise.
6898 (branch_dest): Likewise.
6899 (arc_attr_type): Likewise.
6900 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6901 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6902 (arc_final_prescan_insn): Remove inserting the nops due to
6903 hardware hazards. It is done in reorg step.
6904 (insn_length_variant_t): Remove.
6905 (insn_length_parameters_t): Likewise.
6906 (arc_insn_length_parameters): Likewise.
6907 (arc_get_insn_variants): Likewise.
6908 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6910 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6913 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6916 PR tree-optimization/84687
6917 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6919 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6921 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6923 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6924 Rename to ppc_speculation_barrier.
6925 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6926 __builtin_ppc_speculation_barrier.
6928 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6931 * combine.c (combine_simplify_rtx): Don't try to simplify if
6932 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6935 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6937 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6938 to 32 bytes when compiling for POWER9.
6940 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6943 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6944 regparm >= 3 with no arg reg available also for calls with
6945 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6948 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6950 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6952 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6955 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6957 2018-03-05 Richard Biener <rguenther@suse.de>
6959 PR tree-optimization/84486
6960 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6961 When inserting a __builtin_assume_aligned call set the LHS
6962 SSA name alignment info accordingly.
6964 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6966 PR tree-optimization/84114
6967 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6968 Avoid reassociation of FLOAT_MODE addition.
6970 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6972 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6973 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6974 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6975 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6976 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6977 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6978 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6980 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6981 __builtin_ia32_wbinvd): New builtins.
6982 (SPECIAL_ARGS2): New.
6983 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6984 (SPECIAL_ARGS2): New.
6985 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6986 (ix86_valid_target_attribute_inner_p): Ditto.
6987 (ix86_init_mmx_sse_builtins): Add special_args2.
6988 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6989 TARGET_WBNOINVD_P): New.
6990 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6991 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6992 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6993 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6994 * config/i386/pconfigintrin.h: New file.
6995 * config/i386/wbnoinvdintrin.h: Ditto.
6996 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6997 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6999 2018-03-05 Richard Biener <rguenther@suse.de>
7001 PR tree-optimization/84670
7002 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7004 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7005 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7006 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7007 doesn't grow if all successors (recursively) were visited at least
7010 2018-03-05 Richard Biener <rguenther@suse.de>
7012 PR tree-optimization/84650
7013 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7014 if executed in the loop pipeline.
7016 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7018 * doc/configfiles.texi (Configuration Files): Move info about
7019 conditionalizing $target-protos.h to...
7020 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7021 differs from $target-protos.h.
7023 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7024 Chung-Ju Wu <jasonwucj@gmail.com>
7026 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7027 * config/nds32/nds32-multiple.md (setmemsi): Define.
7028 * config/nds32/nds32-memory-manipulation.c
7029 (nds32_gen_dup_4_byte_to_word_value): New.
7030 (emit_setmem_word_loop): New.
7031 (emit_setmem_byte_loop): New.
7032 (nds32_expand_setmem_loop): New.
7033 (nds32_expand_setmem_loop_v3m): New.
7034 (nds32_expand_setmem_unroll): New.
7035 (nds32_expand_setmem): New.
7037 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7038 Chung-Ju Wu <jasonwucj@gmail.com>
7040 * config/nds32/nds32-memory-manipulation.c
7041 (nds32_emit_load_store): New.
7042 (nds32_emit_post_inc_load_store): New.
7043 (nds32_emit_mem_move): New.
7044 (nds32_emit_mem_move_block): New.
7045 (nds32_expand_movmemsi_loop_unknown_size): New.
7046 (nds32_expand_movmemsi_loop_known_size): New.
7047 (nds32_expand_movmemsi_loop): New.
7048 (nds32_expand_movmemsi_unroll): New.
7049 (nds32_expand_movmemqi): Rename ...
7050 (nds32_expand_movmemsi): ... to this.
7051 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7052 (movmemsi): ... to this.
7053 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7054 (nds32_expand_movmemsi): ... to this.
7056 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7057 Monk Chiang <sh.chiang04@gmail.com>
7058 Chung-Ju Wu <jasonwucj@gmail.com>
7060 * config/nds32/nds32-protos.h
7061 (nds32_expand_load_multiple): New arguments.
7062 (nds32_expand_store_multiple): Ditto.
7063 (nds32_valid_multiple_load_store): Rename ...
7064 (nds32_valid_multiple_load_store_p): ... to this.
7065 * config/nds32/nds32-memory-manipulation.c
7066 (nds32_expand_load_multiple): Refine implementation.
7067 (nds32_expand_store_multiple): Ditto.
7068 * config/nds32/nds32-multiple.md
7069 (load_multiple): Update nds32_expand_load_multiple interface.
7070 (store_multiple): Update nds32_expand_store_multiple interface.
7071 * config/nds32/nds32-predicates.c
7072 (nds32_valid_multiple_load_store): Rename ...
7073 (nds32_valid_multiple_load_store_p): ... to this and refine
7075 * config/nds32/predicates.md
7076 (nds32_load_multiple_and_update_address_operation): New predicate.
7077 (nds32_store_multiple_and_update_address_operation): New predicate.
7079 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7080 Chung-Ju Wu <jasonwucj@gmail.com>
7082 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7083 (combo): New attribute.
7084 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7086 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7088 * config/nds32/nds32.opt: Change -mcmodel= default value.
7090 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7091 Monk Chiang <sh.chiang04@gmail.com>
7092 Chung-Ju Wu <jasonwucj@gmail.com>
7094 * config/nds32/constants.md (unspec_element): New enum.
7095 * config/nds32/constraints.md (Umw): New constraint.
7096 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7097 * config/nds32/nds32-intrinsic.md: Likewise.
7098 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7099 (nds32_valid_smw_lwm_base_p): New.
7100 (nds32_output_smw_single_word): New.
7101 (nds32_output_lmw_single_word): New.
7102 (nds32_expand_unaligned_load): New.
7103 (nds32_expand_unaligned_store): New.
7104 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7105 (nds32_output_smw_single_word): Declare.
7106 (nds32_output_lmw_single_word): Declare.
7107 (nds32_expand_unaligned_load): Declare.
7108 (nds32_expand_unaligned_store): Declare.
7109 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7110 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7111 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7112 NDS32_BUILTIN_UASTORE_DW.
7113 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7116 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7117 Kito Cheng <kito.cheng@gmail.com>
7118 Chung-Ju Wu <jasonwucj@gmail.com>
7120 * config/nds32/nds32-intrinsic.c
7121 (nds32_expand_builtin_null_ftype_reg): Delete.
7122 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7123 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7124 (nds32_read_argument): New.
7125 (nds32_legitimize_target): Ditto.
7126 (nds32_legitimize_argument): Ditto.
7127 (nds32_check_constant_argument): Ditto.
7128 (nds32_expand_unop_builtin): Ditto.
7129 (nds32_expand_unopimm_builtin): Ditto.
7130 (nds32_expand_binop_builtin): Ditto.
7131 (nds32_builtin_decl_impl): Ditto.
7132 (builtin_description): Ditto.
7133 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7134 (nds32_init_builtins_impl): Ditto.
7135 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7136 (nds32_builtin_decl): New.
7137 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7138 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7140 2018-03-02 Jeff Law <law@redhat.com>
7142 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7143 (redundant_insn, fill_simple_delay_slots): Likewise.
7144 (fill_slots_from_thread): Likewise.
7145 * resource.c (mark_referenced_resources): Likewise.
7146 (mark_set_resources, find_dead_or_set_registers): Likewise.
7148 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7150 * substring-locations.h (format_warning_va): Formatting fix for
7152 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7154 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7155 * substring-locations.c: Include intl.h.
7156 (format_warning_va): Turned into small wrapper around
7157 format_warning_n_va, renamed to ...
7158 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7159 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7161 (format_warning_at_substring_n): New function.
7162 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7163 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7164 format_warning_at_substring with just a shorter name instead of
7165 const function pointer.
7166 (fmtwarn_n): New function.
7167 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7168 appropriate, get rid of all the fmtstr temporaries, move conditionals
7169 with G_() wrapped string literals directly into fmtwarn arguments,
7170 cast dir.len to (int), formatting fixes.
7172 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7174 * doc/invoke.texi: Remove "Cilk Plus" references.
7176 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7177 Richard Biener <rguenther@suse.de>
7180 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7181 for error or warning attributes if CALL_FROM_THUNK_P is set.
7184 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7187 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7188 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7191 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7192 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7194 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7195 instead of -1U in last predictors element's probability member.
7197 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7200 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7201 arguments if they are comparable.
7203 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7205 PR tree-optimization/84634
7206 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7207 masks and masked_loop_p with a single loop_masks, making sure it's
7208 null for bb vectorization.
7210 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7212 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7213 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7214 loop->force_vectorize to check whether there is no alias.
7216 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7219 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7221 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7223 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7224 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7225 instead of a loop around prev_real_insn.
7226 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7230 * config/i386/i386.c (ix86_print_operand): Use conditional
7231 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7234 2018-03-02 Richard Biener <rguenther@suse.de>
7236 PR tree-optimization/84427
7237 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7238 (bitmap_set_subtract_values): Rewrite to handle multiple
7241 (prune_clobbered_mems): Likewise.
7242 (phi_translate): Take edge instead of pred/phiblock.
7243 (phi_translate_1): Likewise.
7244 (phi_translate_set): Likewise. Insert all translated
7245 exprs for a value into the set, keeping possibly multiple
7246 expressions per value.
7247 (compute_antic_aux): Adjust for phi_translate changes.
7248 When intersecting union the expressions and prune those
7249 not in the final value set, keeping possibly multiple
7250 expressions per value. Do not use value-insertion
7251 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7252 all expressions. Add verification that the value-sets
7253 only shrink during iteration.
7254 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7255 (do_pre_regular_insertion): Likewise.
7256 (do_pre_partial_partial_insertion): Likewise.
7258 2018-03-02 Richard Biener <rguenther@suse.de>
7261 * config/darwin.c (saved_debug_info_level): New static global.
7262 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7263 (darwin_asm_lto_end): Restore debug info generation settings.
7265 2018-03-01 Martin Liska <mliska@suse.cz>
7268 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7271 2018-03-01 Richard Biener <rguenther@suse.de>
7274 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7275 type annotation with LTO when debug was disabled at compile-time.
7277 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7279 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7281 (mips_final_postscan_insn): Likewise.
7283 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7285 PR rtl-optimization/84528
7286 * alias.c (init_alias_target): Add commentary.
7287 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7288 a unique base value if the frame pointer is not eliminated
7289 to the stack pointer.
7291 2018-03-01 Tom de Vries <tom@codesourcery.com>
7293 PR rtl-optimization/83327
7294 * lra-int.h (hard_regs_spilled_into): Declare.
7295 * lra.c (hard_regs_spilled_into): Define.
7296 (init_reg_info): Init hard_regs_spilled_into.
7297 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7298 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7299 (process_bb_lives): Handle hard_regs_spilled_into.
7300 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7301 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7303 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7305 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7306 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7307 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7308 * config/rs6000/aix72.h: New file.
7310 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7312 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7313 instead of warning_at with conditional singular and plural messages
7317 * stor-layout.c (update_alignment_for_field): For
7318 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7319 && !DECL_PACKED (field), do the alignment update, just use
7320 only desired_align instead of MAX (type_align, desired_align)
7322 (place_field): Don't do known_align < desired_align handling
7323 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7324 is non-NULL, instead do it after rli->prev_field handling and
7325 only if not within a bitfield word. For DECL_PACKED (field)
7326 use type_align of BITS_PER_UNIT.
7328 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7330 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7331 superfluous parentheses and trailing spaces.
7333 2018-02-28 Richard Biener <rguenther@suse.de>
7335 PR tree-optimization/84584
7336 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7337 SCoPs with fake exit edge.
7339 2018-02-28 Martin Liska <mliska@suse.cz>
7342 * timevar.c (timer::print): Fix format to properly print 100%
7345 2018-02-28 Richard Biener <rguenther@suse.de>
7348 * genmatch.c (capture_info::walk_match): Do not mark
7349 captured expressions without operands as expr_p given
7350 they act more like predicates and should be subject to
7351 "lost tail" side-effect preserving.
7353 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7355 PR rtl-optimization/81611
7356 * auto-inc-dec.c (attempt_change): Move dead note from
7357 mem_insn if it's the next use of regno
7358 (find_address): Take address use of reg holding
7359 non-incremented value. Add parm to limit search to the named
7361 (merge_in_block): Attempt to use a mem insn that is the next
7362 use of the original regno.
7364 2018-02-27 Martin Sebor <msebor@redhat.com>
7367 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7368 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7370 2018-02-27 Martin Sebor <msebor@redhat.com>
7372 PR translation/84207
7373 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7374 n argument to unsigned HOST_WIDE_INT.
7375 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7376 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7377 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7378 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7380 2018-02-27 Richard Biener <rguenther@suse.de>
7382 PR tree-optimization/84512
7383 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7384 Do not use the estimate returned from record_stmt_cost for
7385 the scalar iteration cost but sum properly using add_stmt_cost.
7387 2018-02-27 Richard Biener <rguenther@suse.de>
7389 PR tree-optimization/84466
7390 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7391 Adjust last change to less strictly validate use operands.
7393 2018-02-27 Martin Liska <mliska@suse.cz>
7395 PR gcov-profile/84548
7396 * gcov.c (process_file): Allow partial overlap and consider it
7397 also as group functions.
7398 (output_lines): Properly calculate range of lines for a group.
7400 2018-02-27 Martin Liska <mliska@suse.cz>
7402 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7403 'ggc' suffixes. Change first column width.
7404 (timer::print): Fix formatting of the column.
7406 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7408 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7409 preserve inline entry blocks for the sake of debug inline
7410 entry point markers alone.
7411 (remove_unused_locals): Suggest in comments a better place to
7412 force the preservation of inline entry blocks that are
7413 otherwise unused, but do not preserve them.
7415 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7417 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7419 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7422 * config/i386/constraints.md (Bs): Replace
7423 ix86_indirect_branch_register with
7424 TARGET_INDIRECT_BRANCH_REGISTER.
7426 * config/i386/i386.md (indirect_jump): Likewise.
7427 (tablejump): Likewise.
7428 (*sibcall_memory): Likewise.
7429 (*sibcall_value_memory): Likewise.
7430 Peepholes of indirect call and jump via memory: Likewise.
7431 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7432 (*sibcall_value_GOT_32): Likewise.
7433 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7434 (GOT_memory_operand): Likewise.
7435 (call_insn_operand): Likewise.
7436 (sibcall_insn_operand): Likewise.
7437 (GOT32_symbol_operand): Likewise.
7438 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7440 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7442 PR rtl-optimization/83496
7443 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7444 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7445 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7446 redundant insn, if any.
7447 (relax_delay_slots): Likewise.
7448 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7450 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7452 PR tree-optimization/83965
7453 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7454 that grouped statements are part of a reduction chain. Return
7455 true if the statement is not marked as a reduction itself but
7457 (vect_recog_dot_prod_pattern): Don't check whether the statement
7458 is part of a group here.
7459 (vect_recog_sad_pattern): Likewise.
7460 (vect_recog_widen_sum_pattern): Likewise.
7462 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7465 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7467 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7470 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7472 (ix86_output_indirect_function_return): New prototype.
7473 (ix86_split_simple_return_pop_internal): Likewise.
7474 * config/i386/i386.c (indirect_return_via_cx): New.
7475 (indirect_return_via_cx_bnd): Likewise.
7476 (indirect_thunk_name): Handle return va CX_REG.
7477 (output_indirect_thunk_function): Create alias for
7478 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7479 (ix86_output_indirect_jmp): Remove the bool argument.
7480 (ix86_output_indirect_function_return): New function.
7481 (ix86_split_simple_return_pop_internal): Likewise.
7482 * config/i386/i386.md (*indirect_jump): Don't pass false
7483 to ix86_output_indirect_jmp.
7484 (*tablejump_1): Likewise.
7485 (simple_return_pop_internal): Change it to define_insn_and_split.
7486 Call ix86_split_simple_return_pop_internal to split it for
7488 (simple_return_indirect_internal): Call
7489 ix86_output_indirect_function_return instead of
7490 ix86_output_indirect_jmp.
7492 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7495 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7496 memset and value initialization afterwards.
7498 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7500 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7502 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7505 * common/config/aarch64/aarch64-common.c
7506 (aarch_option_optimization_table[]): Switch
7507 off fomit-frame-pointer
7509 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7510 Chung-Ju Wu <jasonwucj@gmail.com>
7512 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7514 (store_multiple): Ditto.
7516 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7518 * config.gcc: Add --with-cpu support for nds32 target.
7519 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7520 * config/nds32/nds32.opt: Add -mcpu= option.
7522 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7524 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7525 isel=yes): Warn for these deprecated options.
7527 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7529 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7530 ISA_2_5_MASKS_EMBEDDED.
7532 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7534 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7535 p->max as pointers rather than using iterative_hash_expr.
7537 2018-02-23 Carl Love <cel@us.ibm.com>
7539 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7540 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7542 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7543 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7544 P8V_BUILTIN_VEC_VUNSIGNED2.
7546 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7549 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7550 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7552 * lra-constraints.c (curr_insn_transform): Set up
7553 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7555 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7556 * lra-eliminations.c (spill_pseudos): Ditto.
7557 (process_insn_for_elimination): Ditto.
7558 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7559 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7560 LRA_NON_CLOBBERED_ALT.
7562 2018-02-22 Martin Sebor <msebor@redhat.com>
7564 PR tree-optimization/84480
7565 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7566 to maybe_diag_stxncpy_trunc. Call it.
7567 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7568 from gimple_fold_builtin_strcpy. Print inlining stack.
7569 (handle_builtin_stxncpy): Print inlining stack.
7570 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7572 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7575 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7576 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7577 and -fcheck-pointer-bounds are used together.
7578 (indirect_thunk_prefix): New enum.
7579 (indirect_thunk_need_prefix): New function.
7580 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7581 "_nt" instead of "_bnd" for NOTRACK prefix.
7582 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7583 (output_indirect_thunk_function): Likewise.
7585 (ix86_code_end): Update output_indirect_thunk_function calls.
7586 (ix86_output_indirect_branch_via_reg): Replace
7587 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7588 (ix86_output_indirect_branch_via_push): Likewise.
7589 (ix86_output_function_return): Likewise.
7590 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7591 incompatible with -fcf-protection=branch and
7592 -fcheck-pointer-bounds.
7594 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7597 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7598 Change gcc_assert call to output_operand_lossage.
7600 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7602 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7604 2018-02-22 DJ Delorie <dj@redhat.com>
7605 Sebastian Perta <sebastian.perta@renesas.com>
7606 Oleg Endo <olegendo@gcc.gnu.org>
7608 * config/rx/rx.c (rx_rtx_costs): New function.
7609 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7611 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7613 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7615 2018-02-22 Martin Liska <mliska@suse.cz>
7618 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7619 Add "native" as a possible value.
7621 2018-02-22 Martin Liska <mliska@suse.cz>
7624 * config/i386/i386.c (ix86_option_override_internal):
7625 Add "native" as a possible value for -march and -mtune.
7627 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7630 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7631 to all type variants.
7633 PR tree-optimization/84503
7634 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7635 width as info->bitpos + info->bitsize - start.
7636 (merged_store_group::merge_overlapping): Simplify width computation.
7637 (check_no_overlap): New function.
7638 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7639 start + width and last_order of the group, fail if check_no_overlap
7641 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7642 to group if check_no_overlap fails.
7644 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7646 * config/rs6000/altivec.md: Delete contraint arguments to
7647 define_expand, define_split, and define_peephole2, and in
7648 define_insn_and_split if always unused.
7649 * config/rs6000/darwin.md: Ditto.
7650 * config/rs6000/dfp.md: Ditto.
7651 * config/rs6000/rs6000.md: Ditto.
7652 * config/rs6000/sync.md: Ditto.
7653 * config/rs6000/vector.md: Ditto.
7654 * config/rs6000/vsx.md: Ditto.
7656 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7658 * config/rs6000/altivec.md: Write output control strings as braced
7659 blocks instead of double-quoted strings.
7660 * config/rs6000/darwin.md: Ditto.
7661 * config/rs6000/rs6000.md: Ditto.
7662 * config/rs6000/vector.md: Ditto.
7663 * config/rs6000/vsx.md: Ditto.
7665 2018-02-21 Jason Merrill <jason@redhat.com>
7667 PR c++/84314 - ICE with templates and fastcall attribute.
7668 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7670 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7672 * ipa-cp.c (determine_versionability): Fix comment typos.
7674 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7677 * ipa-cp.c (determine_versionability): Do not version functions caling
7680 2018-02-21 Martin Liska <mliska@suse.cz>
7683 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7684 Add "native" as a possible value.
7685 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7686 the macro when native cpu detection is available.
7688 2018-02-21 Martin Liska <mliska@suse.cz>
7691 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7692 Add "native" as a possible value.
7693 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7694 when native cpu detection is available.
7696 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7697 Martin Sebor <msebor@redhat.com>
7699 PR tree-optimization/84478
7700 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7702 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7703 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7704 support which is conservatively correct, for 2 only stay conservative
7705 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7706 argument to the 2 argument get_range_strlen, adjust 6 arg
7707 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7709 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7710 (gimple_fold_builtin_strlen): Pass true as last argument to
7713 2018-02-20 Martin Sebor <msebor@redhat.com>
7716 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7717 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7718 (builtin_memref::builtin_memref): Factor out parts into
7719 set_base_and_offset and call it.
7721 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7724 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7725 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7726 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7727 search at the associated MODE_INT.
7729 2018-02-20 Jeff Law <law@redhat.com>
7732 PR tree-optimization/81592
7734 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7735 for range data rather than using global data.
7736 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7737 range data rather than using global data.
7738 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7739 pass it to children as needed.
7740 (struct directive::fmtresult): Similarly.
7741 (struct directive::set_width): Similarly.
7742 (struct directive::set_precision): Similarly.
7743 (format_integer, format_directive, parse_directive): Similarly.
7744 (format_none): Accept unnamed vr_values parameter.
7745 (format_percent, format_floating, format_character): Similarly.
7746 (format_string, format_plain): Similarly.
7747 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7748 the EVRP range analyzer for range data rather than using global data.
7749 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7750 gimple-ssa-evrp-analyze.h
7751 (class sprintf_dom_walker): Add after_dom_children member function.
7752 Add evrp_range_analyzer member.
7753 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7754 range analyzer as needed.
7755 (sprintf_dom_walker::after_dom_children): New member function.
7756 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7758 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7759 (evrp_range_analyzer::pop_to_marker): Likewise.
7761 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7763 PR tree-optimization/84419
7764 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7765 with the required type if its current type is compatible but
7768 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7771 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7772 after vectorization.
7774 2018-02-20 Martin Liska <mliska@suse.cz>
7777 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7778 possible values if we don't have a hint.
7780 2018-02-20 Martin Liska <mliska@suse.cz>
7784 * final.c (shorten_branches): Build align_tab array with one
7786 * opts.c (finish_options): Add alignment option limit check.
7787 (MAX_CODE_ALIGN): Likewise.
7788 (MAX_CODE_ALIGN_VALUE): Likewise.
7789 * doc/invoke.texi: Document maximum allowed option value for
7790 all -falign-* options.
7792 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7795 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7796 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7797 * var-tracking.c (emit_note_insn_var_location): Remove all references
7798 to NOTE_INSN_CALL_ARG_LOCATION.
7799 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7800 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7801 Use copy_rtx_if_shared.
7802 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7803 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7804 (dwarf2out_var_location): Remove handling of
7805 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7807 * final.c (final_scan_insn): Remove all references to
7808 NOTE_INSN_CALL_ARG_LOCATION.
7809 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7810 before dumping final insns.
7811 * except.c (emit_note_eh_region_end): Remove all references to
7812 NOTE_INSN_CALL_ARG_LOCATION.
7813 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7814 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7815 * config/arc/arc.c (hwloop_optimize): Likewise.
7816 * config/arm/arm.c (create_fix_barrier): Likewise.
7817 * config/s390/s390.c (s390_chunkify_start): Likewise.
7818 * config/sh/sh.c (find_barrier): Likewise.
7819 * config/i386/i386.c (rest_of_insert_endbranch,
7820 ix86_seh_fixup_eh_fallthru): Likewise.
7821 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7822 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7823 * config/frv/frv.c (frv_function_prologue): Likewise.
7824 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7826 (note_outside_basic_block_p): Remove all references to
7827 NOTE_INSN_CALL_ARG_LOCATION.
7828 * gengtype.c (adjust_field_rtx_def): Likewise.
7829 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7831 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7832 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7835 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7838 PR tree-optimization/84452
7839 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7840 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7843 2018-02-19 Martin Liska <mliska@suse.cz>
7846 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7848 2018-02-19 Martin Liska <mliska@suse.cz>
7849 Richard Sandiford <richard.sandiford@linaro.org>
7851 PR tree-optimization/82491
7852 * gimple-fold.c (get_base_constructor): Make earlier bail out
7855 2018-02-19 Carl Love <cel@us.ibm.com>
7857 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7858 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7860 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7861 P8V_BUILTIN_VEC_NEG.
7863 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7865 * config/rl78/rl78.md (movdf): New define expand.
7867 2018-02-19 Martin Liska <mliska@suse.cz>
7870 * doc/invoke.texi: Fix typo.
7871 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7873 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7875 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7876 handle rs6000_single_float and rs6000_double_float specially for
7879 2018-02-16 Jeff Law <law@redhat.com>
7881 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7884 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7887 * ipa-inline.c (inline_small_functions): Fix a typo.
7889 2018-02-16 Nathan Sidwell <nathan@acm.org>
7891 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7893 2018-02-16 Carl Love <cel@us.ibm.com>
7895 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7896 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7897 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7898 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7899 expansion to P8V_BUILTIN_VEC_FLOAT2.
7901 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7903 PR rtl-optimization/70023
7904 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7905 src_regno into account.
7907 2018-02-16 Carl Love <cel@us.ibm.com>
7909 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7910 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7911 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7912 * config/rs6000/rs6000.c: Remove case statements for
7913 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7914 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7915 and P9V_BUILTIN_VEC_VINSERT4B.
7916 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7917 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7918 * config/rs6000/vsx.md:
7919 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7922 2018-02-16 Carl Love <cel@us.ibm.com>
7924 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7926 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7928 * config/rs6000/rs6000-c.c: Add the definitions for
7929 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7930 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7931 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7932 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7933 definition for insert4b and define insn *insert3b_internal.
7934 * doc/extend.texi: Add documentation for vec_extract4b.
7936 2018-02-16 Nathan Sidwell <nathan@acm.org>
7938 * doc/extend.texi (Backwards Compatibility): Mention friend
7939 injection. Note for-scope is deprecated.
7940 * doc/invoke.texi (-ffriend-injection): Deprecate.
7942 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7944 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7945 that moved to I2, also allow destinations that are a paradoxical
7946 subreg (instead of a normal reg).
7948 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7951 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7954 2018-02-16 Richard Biener <rguenther@suse.de>
7956 PR tree-optimization/84037
7957 PR tree-optimization/84016
7959 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7960 Adjust vec_construct for the fact we need additional higher latency
7961 128bit inserts for AVX256 and AVX512 vector builds.
7962 (ix86_add_stmt_cost): Scale vector construction cost for
7965 2018-02-16 Richard Biener <rguenther@suse.de>
7967 PR tree-optimization/84417
7968 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7969 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7970 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7972 2018-02-16 Martin Liska <mliska@suse.cz>
7975 * internal-fn.def (ASAN_CHECK): Set proper flags.
7976 (ASAN_MARK): Likewise.
7978 2018-02-16 Julia Koval <julia.koval@intel.com>
7980 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7981 from PTA_CANNONLAKE.
7983 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7986 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7987 Use ++iter rather than iter++ for std::list iterators.
7988 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7989 defer deleting them until all nodes in the forest are processed. Do
7990 free even leaf nodes. Change to_process into auto_vec.
7993 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7994 * vec.h (vec_default_construct): Use memset instead of placement new
7995 if BROKEN_VALUE_INITIALIZATION is defined.
7996 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7997 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8000 PR rtl-optimization/83723
8001 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8002 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8003 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8005 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8007 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8009 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8011 PR rtl-optimization/81443
8012 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8013 from inner REGs to paradoxical SUBREGs.
8015 2018-02-16 Richard Biener <rguenther@suse.de>
8017 PR tree-optimization/84399
8018 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8019 For operands we can analyze at their definition make sure we can
8020 analyze them at each use as well.
8022 2018-02-16 Richard Biener <rguenther@suse.de>
8024 PR tree-optimization/84190
8025 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8026 volatile accesses if the decl isn't volatile.
8028 2018-02-15 Jason Merrill <jason@redhat.com>
8030 PR c++/84314 - ICE with templates and fastcall attribute.
8031 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8032 TYPE_CANONICAL on an existing type.
8034 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8036 PR tree-optimization/84383
8037 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8038 dstoff nor call operand_equal_p if dstbase is NULL.
8040 PR tree-optimization/84334
8041 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8042 also a CONSTANT_CLASS_P, punt.
8044 2018-02-14 Jim Wilson <jimw@sifive.com>
8046 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8047 first SMALL_OPERAND check. New local min_second_step. Move assert
8048 to where locals are set. Add TARGET_RVC support.
8049 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8051 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8053 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8055 2018-02-14 Martin Sebor <msebor@redhat.com>
8057 PR tree-optimization/83698
8058 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8059 arrays constrain the offset range to their bounds.
8060 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8061 (builtin_access::overlap): Avoid setting the size of overlap if it's
8063 (maybe_diag_overlap): Also consider arrays when deciding what values
8064 of offsets to include in diagnostics.
8066 2018-02-14 Martin Sebor <msebor@redhat.com>
8069 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8070 that correspond to the kind of a declaration.
8072 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8075 * config/pa/pa.md: Load address of PIC label using the linkage table
8076 if the label is nonlocal.
8078 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8080 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8081 warning message if user requests -maltivec=be.
8082 * doc/invoke.texi: Document deprecation of -maltivec=be.
8084 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8087 * config/rs6000/rs6000-c.c: Update definitions for
8088 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8089 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8091 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8094 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8095 add _get_ssp intrinsics. Remove argument from
8096 __builtin_ia32_rdssp[d|q].
8097 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8098 * config/i386/i386-builtin.def: Remove argument from
8099 __builtin_ia32_rdssp[d|q].
8100 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8101 ix86_expand_special_args_builtin for _rdssp[d|q].
8102 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8103 Clear register before usage.
8104 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8105 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8107 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8109 PR tree-optimization/84357
8110 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8111 operand 1 of an ARRAY_REF too.
8113 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8116 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8117 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8119 (set_of_reg): New struct.
8120 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8121 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8122 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8124 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8125 Split into bitclr, bitset, bitinvert patterns if appropriate.
8126 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8127 use rx_fuse_in_memory_bitop.
8128 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8129 to named insn, correct maximum insn length.
8131 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8134 * machmode.def: Define a complex mode for PARTIAL_INT.
8135 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8137 * doc/rtl.texi: Document CSPImode.
8138 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8140 (msp430_hard_regno_nregs_with_padding): Likewise.
8142 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8145 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8147 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8149 PR rtl-optimization/84169
8150 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8151 we generated a parallel as new i3 and we split that to new i2 and i3
8152 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8153 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8154 those to i2, not i1. Partially rewrite this scan code.
8156 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8159 * stor-layout.c (place_field): For variable length fields, adjust
8160 offset_align afterwards not just based on the field's alignment,
8161 but also on the size.
8164 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8165 of exps and logs in the use_exp2 case.
8167 2018-02-13 Jeff Law <law@redhat.com>
8169 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8172 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8175 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8179 * common.opt (gas-loc-support, gas-locview-support): New.
8180 (ginline-points, ginternal-reset-location-views): New.
8181 * doc/invoke.texi: Document them. Use @itemx where intended.
8182 (gvariable-location-views): Adjust.
8183 * target.def (reset_location_view): New.
8184 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8185 (TARGET_RESET_LOCATION_VIEW): New.
8186 * doc/tm.texi: Rebuilt.
8187 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8188 (dwarf2out_default_as_locview_support): New.
8189 (output_asm_line_debug_info): Use option variables.
8190 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8191 (output_loc_list): Likewise.
8192 (add_high_low_attributes): Check option variables.
8193 Don't output entry view attribute in strict mode.
8194 (gen_inlined_subroutine_die): Check option variables.
8195 (dwarf2out_inline_entry): Likewise.
8196 (init_sections_and_labels): Likewise.
8197 (dwarf2out_early_finish): Likewise.
8198 (maybe_reset_location_view): New, from...
8199 (dwarf2out_var_location): ... here. Call it.
8200 * debug.h (dwarf2out_default_as_loc_support): Declare.
8201 (dwarf2out_default_as_locview_support): Declare.
8202 * hooks.c (hook_int_rtx_insn_0): New.
8203 * hooks.h (hook_int_rtx_insn_0): Declare.
8204 * toplev.c (process_options): Take -gas-loc-support and
8205 -gas-locview-support from dwarf2out. Enable
8206 -gvariable-location-views by default only with locview
8207 assembler support. Enable -ginternal-reset-location-views by
8208 default only if the target defines the corresponding hook.
8209 Enable -ginline-points by default if location views are
8210 enabled; force it disabled if statement frontiers are
8212 * tree-inline.c (expand_call_inline): Check option variables.
8213 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8215 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8217 PR tree-optimization/84321
8218 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8219 handling. Also check whether the anti-range contains any values
8220 that satisfy the mask; switch to a VR_RANGE if not.
8222 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8225 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8227 2018-02-13 Martin Jambor <mjambor@suse.cz>
8230 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8231 of call statements, also set location of a load to a temporary.
8233 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8235 * config/rl78/rl78.c (add_vector_labels): New function.
8236 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8237 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8238 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8239 which checks that no arguments are passed.
8240 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8241 * doc/extend.texi: Documentation for the new attribute.
8243 2018-02-13 Andreas Schwab <schwab@suse.de>
8245 * config/riscv/linux.h (CPP_SPEC): Define.
8247 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8250 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8251 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8252 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8253 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8254 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8255 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8256 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8257 temporarily for AES and PCLMUL builtins.
8259 PR tree-optimization/84339
8260 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8261 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8265 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8266 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8267 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8269 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8271 * omp-simd-clone.h: New file.
8272 * omp-simd-clone.c: Include omp-simd-clone.h.
8273 (expand_simd_clones): No longer static.
8274 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8275 cgraph.h and omp-simd-clone.h.
8276 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8277 (vect_recog_widen_shift_pattern): Formatting fix.
8278 (vect_pattern_recog_1): Don't check optab for calls.
8281 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8282 operands[2] into a REG before using gen_lowpart on it.
8284 2018-02-12 Jeff Law <law@redhat.com>
8287 * config/sh/sh.c (find_barrier): Consider a sibling call
8290 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8291 successfully back substituting a reg.
8293 2018-02-12 Richard Biener <rguenther@suse.de>
8295 PR tree-optimization/84037
8296 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8297 parameter, move visited init to caller.
8298 (vect_slp_analyze_operations): Separate cost from validity
8299 check, initialize visited once for all instances.
8300 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8302 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8304 (vect_model_promotion_demotion_cost): Likewise.
8305 (vectorizable_bswap): Guard cost modeling with !slp_node
8306 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8308 (vectorizable_call): Likewise.
8309 (vectorizable_conversion): Likewise.
8310 (vectorizable_assignment): Likewise.
8311 (vectorizable_shift): Likewise.
8312 (vectorizable_operation): Likewise.
8313 (vectorizable_store): Likewise.
8314 (vectorizable_load): Likewise.
8315 (vectorizable_condition): Likewise.
8316 (vectorizable_comparison): Likewise.
8318 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8321 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8322 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8323 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8325 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8328 * params.def (inline-min-speedup): Increase from 8 to 15.
8329 (max-inline-insns-auto): Decrease from 40 to 30.
8330 * ipa-split.c (consider_split): Add some buffer for function to
8331 be considered inlining candidate.
8332 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8335 2018-02-12 Richard Biener <rguenther@suse.de>
8337 PR tree-optimization/84037
8338 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8339 matched stmts if we cannot swap the non-matched ones.
8341 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8343 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8344 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8345 _mm_maskz_scalef_round_ss): New intrinsics.
8346 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8347 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8348 __builtin_ia32_scalefss_round): Remove.
8349 (__builtin_ia32_scalefsd_mask_round,
8350 __builtin_ia32_scalefss_mask_round): New intrinsics.
8351 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8352 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8353 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8354 "<round_constraint>")): Changed to ...
8355 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8356 "<round_scalar_constraint>")): ... this.
8357 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8358 %0, %1, %2<round_op3>}"): Changed to ...
8359 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8360 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8361 %2<round_scalar_mask_op3>}"): ... this.
8362 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8364 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8366 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8367 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8368 (_mm_maskz_sqrt_round_ss): New intrinsics.
8369 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8370 (__builtin_ia32_sqrtsd_mask_round)
8371 (__builtin_ia32_sqrtss_mask_round): New builtins.
8372 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8373 (__builtin_ia32_sqrtss_round): Remove.
8374 (__builtin_ia32_sqrtsd_mask_round)
8375 (__builtin_ia32_sqrtss_mask_round): New builtins.
8376 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8377 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8378 ((match_operand:VF_128 1 "vector_operand"
8379 "xBm,<round_constraint>")): Changed to ...
8380 ((match_operand:VF_128 1 "vector_operand"
8381 "xBm,<round_scalar_constraint>")): ... this.
8382 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8383 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8384 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8385 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8386 %<iptr>1<round_scalar_mask_op3>}): ... this.
8387 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8388 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8390 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8393 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8394 Cast vec_cmpeq result to correct type.
8395 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8396 Cast vec_cmpgt result to correct type.
8398 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8400 * final.c (final_scan_insn_1): Renamed from...
8401 (final_scan_insn): ... this. New wrapper, to recover
8402 seen from the outermost call in recursive ones.
8403 * config/sparc/sparc.c (output_return): Drop seen from call.
8404 (output_sibcall): Likewise.
8405 * config/visium/visium.c (output_branch): Likewise.
8407 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8409 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8412 2018-02-10 Alan Modra <amodra@gmail.com>
8415 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8416 Specify LR as an input.
8418 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8421 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8422 remove_member_access_dummy_vars): New functions.
8423 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8424 lower_omp_1, execute_lower_omp): Use them.
8426 PR rtl-optimization/84308
8427 * shrink-wrap.c (spread_components): Release todo vector.
8429 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8431 PR rtl-optimization/57193
8432 * ira-color.c (struct allocno_color_data): Add member
8433 conflict_allocno_hard_prefs.
8434 (update_conflict_allocno_hard_prefs): New.
8435 (bucket_allocno_compare_func): Add a preference based on
8436 conflict_allocno_hard_prefs.
8437 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8438 (color_allocnos): Remove a dead code. Initiate
8439 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8441 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8444 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8445 constraint from =wa to wa. Avoid a subreg on the output operand,
8446 instead use a pseudo and subreg it in a move.
8447 (p9_xxbrd_<mode>): Changed to ...
8448 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8449 (p9_xxbrd_v2df): New expander.
8450 (p9_xxbrw_<mode>): Changed to ...
8451 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8452 (p9_xxbrw_v4sf): New expander.
8454 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8456 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8458 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8461 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8462 multiply in 32-bit mode.
8463 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8464 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8467 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8469 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8470 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8471 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8472 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8474 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8477 * dwarf2out.c (is_trivial_indirect_ref): New function.
8478 (dwarf2out_late_global_decl): Do not generate a location
8479 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8480 and that are not defined in the current unit.
8482 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8484 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8485 instead of a libcall for UNORDERED.
8487 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8490 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8491 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8493 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8496 * config/s390/s390.c (s390_set_current_function): Invoke
8497 s390_indirect_branch_settings also if fndecl didn't change.
8499 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8501 * config/rs6000/rs6000.md (blockage): Set length to zero.
8503 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8505 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8507 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8510 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8511 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8515 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8516 PARALLEL incoming that failed vt_get_decl_and_offset check.
8519 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8520 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8522 (get_variable_section): For decls in named .bss* sections pass true as
8523 second argument to bss_initializer_p.
8525 2018-02-09 Marek Polacek <polacek@redhat.com>
8526 Jakub Jelinek <jakub@redhat.com>
8529 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8530 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8531 Sync some changes from cxx_fold_indirect_ref.
8533 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8535 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8537 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8538 (BLOCK_INLINE_ENTRY_LABEL): New.
8539 (dwarf2out_var_location): Disregard inline entry markers.
8540 (inline_entry_data): New struct.
8541 (inline_entry_data_hasher): New hashtable type.
8542 (inline_entry_data_hasher::hash): New.
8543 (inline_entry_data_hasher::equal): New.
8544 (inline_entry_data_table): New variable.
8545 (add_high_low_attributes): Add DW_AT_entry_pc and
8546 DW_AT_GNU_entry_view attributes if a pending entry is found
8547 in inline_entry_data_table. Add old entry_pc attribute only
8548 if debug nonbinding markers are disabled.
8549 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8550 markers are enabled.
8551 (block_within_block_p, dwarf2out_inline_entry): New.
8552 (dwarf2out_finish): Check that no entries remained in
8553 inline_entry_data_table.
8554 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8555 (final_scan_insn, notice_source_line): Likewise.
8556 (rest_of_clean_state): Skip inline entry markers.
8557 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8559 * gimple.c (gimple_build_debug_inline_entry): New.
8560 * gimple.h (enum gimple_debug_subcode): Add
8561 GIMPLE_DEBUG_INLINE_ENTRY.
8562 (gimple_build_debug_inline_entry): Declare.
8563 (gimple_debug_inline_entry_p): New.
8564 (gimple_debug_nonbind_marker_p): Adjust.
8565 * insn-notes.def (INLINE_ENTRY): New.
8566 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8567 inline entry marker notes.
8568 (print_insn): Likewise.
8569 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8570 (INSN_DEBUG_MARKER_KIND): Likewise.
8571 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8572 * tree-inline.c (expand_call_inline): Build and insert
8573 debug_inline_entry stmt.
8574 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8575 inline entry blocks early, if nonbind markers are enabled.
8576 (dump_scope_block): Dump fragment info.
8577 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8578 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8579 (gimple_build_debug_inline_entry): New.
8580 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8581 Enable/disable inline entry points too.
8582 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8583 (DEBUG_INSN): Describe inline entry markers.
8585 * common.opt (gvariable-location-views): New.
8586 (gvariable-location-views=incompat5): New.
8587 * config.in: Rebuilt.
8588 * configure: Rebuilt.
8589 * configure.ac: Test assembler for view support.
8590 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8591 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8592 * dwarf2out.c (var_loc_view): New typedef.
8593 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8594 (dwarf2out_locviews_in_attribute): New.
8595 (dwarf2out_locviews_in_loclist): New.
8596 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8597 (enum dw_line_info_opcode): Add LI_adv_address.
8598 (struct dw_line_info_table): Add view.
8599 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8600 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8601 (zero_view_p): New variable.
8602 (ZERO_VIEW_P): New macro.
8603 (output_asm_line_debug_info): New.
8604 (struct var_loc_node): Add view.
8605 (add_AT_view_list, AT_loc_list): New.
8606 (add_var_loc_to_decl): Add view param. Test it against last.
8607 (new_loc_list): Add view params. Record them.
8608 (AT_loc_list_ptr): Handle loc and view lists.
8609 (view_list_to_loc_list_val_node): New.
8610 (print_dw_val): Handle dw_val_class_view_list.
8611 (size_of_die): Likewise.
8612 (value_format): Likewise.
8613 (loc_list_has_views): New.
8614 (gen_llsym): Set vl_symbol too.
8615 (maybe_gen_llsym, skip_loc_list_entry): New.
8616 (dwarf2out_maybe_output_loclist_view_pair): New.
8617 (output_loc_list): Output view list or entries too.
8618 (output_view_list_offset): New.
8619 (output_die): Handle dw_val_class_view_list.
8620 (output_dwarf_version): New.
8621 (output_compilation_unit_header): Use it.
8622 (output_skeleton_debug_sections): Likewise.
8623 (output_rnglists, output_line_info): Likewise.
8624 (output_pubnames, output_aranges): Update version comments.
8625 (output_one_line_info_table): Output view numbers in asm comments.
8626 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8627 Call maybe_gen_llsym.
8628 (loc_list_from_tree_1): Adjust.
8629 (add_AT_location_description): Create view list attribute if
8630 needed, check it's absent otherwise.
8631 (convert_cfa_to_fb_loc_list): Adjust.
8632 (maybe_emit_file): Call output_asm_line_debug_info for test.
8633 (dwarf2out_var_location): Reset views as needed. Precompute
8634 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8635 attribute. Set view.
8636 (new_line_info_table): Reset next view.
8637 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8638 (dwarf2out_source_line): Likewise. Output view resets and labels to
8639 the assembler, or select appropriate line info opcodes.
8640 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8641 (optimize_string_length): Catch it. Adjust.
8642 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8643 dw_val_class_view_list, and remove it if no longer needed.
8644 (hash_loc_list): Hash view numbers.
8645 (loc_list_hasher::equal): Compare them.
8646 (optimize_location_lists): Check whether a view list symbol is
8647 needed, and whether the locview attribute is present, and
8648 whether they match. Remove the locview attribute if no longer
8650 (index_location_lists): Call skip_loc_list_entry for test.
8651 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8652 Use output_dwarf_version.
8653 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8654 (struct dw_val_node): Add val_view_list.
8655 * final.c (SEEN_NEXT_VIEW): New.
8656 (set_next_view_needed): New.
8657 (clear_next_view_needed): New.
8658 (maybe_output_next_view): New.
8659 (final_start_function): Rename to...
8660 (final_start_function_1): ... this. Take pointer to FIRST,
8661 add SEEN parameter. Emit param bindings in the initial view.
8662 (final_start_function): Reintroduce SEEN-less interface.
8663 (final): Rename to...
8664 (final_1): ... this. Take SEEN parameter. Output final pending
8665 next view at the end.
8666 (final): Reintroduce seen-less interface.
8667 (final_scan_insn): Output pending next view before switching
8668 sections or ending a block. Mark the next view as needed when
8669 outputting variable locations. Notify debug backend of section
8670 changes, and of location view changes.
8671 (rest_of_handle_final): Adjust.
8672 * toplev.c (process_options): Autodetect value for debug variable
8673 location views option. Warn on incompat5 without -gdwarf-5.
8674 * doc/invoke.texi (gvariable-location-views): New.
8675 (gvariable-location-views=incompat5): New.
8676 (gno-variable-location-views): New.
8678 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8680 PR tree-optimization/84136
8681 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8682 that the result of find_edge is non-NULL.
8684 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8687 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8688 storing integer register in SImode. Fix cost of 256 and 512
8689 byte aligned SSE register store.
8691 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8693 * config/i386/i386.c (ix86_multiplication_cost): Fix
8694 multiplication cost for TARGET_AVX512DQ.
8696 2018-02-08 Marek Polacek <polacek@redhat.com>
8698 PR tree-optimization/84238
8699 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8702 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8704 PR tree-optimization/84265
8705 * tree-vect-stmts.c (vectorizable_store): Don't treat
8706 VMAT_CONTIGUOUS accesses as grouped.
8707 (vectorizable_load): Likewise.
8709 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8711 PR tree-optimization/81635
8712 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8713 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8714 (test_round_for_mask): New functions.
8715 (wide_int_cc_tests): Call test_round_for_mask.
8716 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8717 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8718 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8719 range returned by get_range_info.
8721 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8724 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8725 * symtab.c: Include builtins.h
8726 (symtab_node::output_to_lto_symbol_table_p): Move here
8727 from lto-streamer-out.c:output_symbol_p.
8728 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8729 (output_symbol_p): Move all logic to symtab.c
8730 (produce_symtab): Update.
8732 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8734 * config/s390/s390-opts.h (enum indirect_branch): Define.
8735 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8736 (s390_indirect_branch_via_thunk)
8737 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8738 (enum s390_indirect_branch_type): Define.
8739 * config/s390/s390.c (struct s390_frame_layout, struct
8740 machine_function): Remove.
8741 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8742 (indirect_branch_table_label_no, indirect_branch_table_name):
8744 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8745 (enum s390_indirect_branch_option): Define.
8746 (s390_return_addr_from_memory): New function.
8747 (s390_handle_string_attribute): New function.
8748 (s390_attribute_table): Add new attribute handler.
8749 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8750 (s390_indirect_branch_via_thunk): New function.
8751 (s390_indirect_branch_via_inline_thunk): New function.
8752 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8753 sibling call optimization for non z10 compiles.
8754 (s390_emit_call): Force indirect branch target to be a single
8755 register. Add r1 clobber for non-z10 compiles.
8756 (s390_emit_epilogue): Emit return jump via return_use expander.
8757 (s390_reorg): Handle JUMP_INSNs as execute targets.
8758 (s390_option_override_internal): Perform validity checks for the
8759 new command line options.
8760 (s390_indirect_branch_attrvalue): New function.
8761 (s390_indirect_branch_settings): New function.
8762 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8763 (s390_output_indirect_thunk_function): New function.
8764 (s390_code_end): Implement target hook.
8765 (s390_case_values_threshold): Implement target hook.
8766 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8768 * config/s390/s390.h (struct s390_frame_layout)
8769 (struct machine_function): Move here from s390.c.
8770 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8771 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8772 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8773 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8774 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8775 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8776 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8777 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8778 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8779 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8780 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8781 (mnemonic attribute): Add values which aren't recognized
8783 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8784 pattern for branch conversion. Fix mnemonic attribute.
8785 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8786 indirect branch via thunk if requested.
8787 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8788 ("*indirect_jump"): Disable for branch conversion using out of
8790 ("indirect_jump_via_thunk<mode>_z10")
8791 ("indirect_jump_via_thunk<mode>")
8792 ("indirect_jump_via_inlinethunk<mode>_z10")
8793 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8794 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8795 ("casesi_jump_via_inlinethunk<mode>_z10")
8796 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8797 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8798 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8799 ("*indirect2_jump"): Disable for branch conversion.
8800 ("casesi_jump"): Turn into expander and expand patterns for branch
8802 ("return_use"): New expander.
8803 ("*return"): Emit return via thunk and rename it to ...
8804 ("*return<mode>"): ... this one.
8805 * config/s390/s390.opt: Add new options and and enum for the
8808 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8810 * lra-constraints.c (match_reload): Unconditionally use
8811 gen_lowpart_SUBREG, rather than selecting between that
8812 and equivalent gen_rtx_SUBREG code.
8814 2018-02-08 Richard Biener <rguenther@suse.de>
8816 PR tree-optimization/84233
8817 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8818 changed flag instead of boguously re-using phi_inserted.
8820 2018-02-08 Martin Jambor <mjambor@suse.cz>
8822 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8823 static local variables.
8825 2018-02-08 Richard Biener <rguenther@suse.de>
8827 PR tree-optimization/84278
8828 * tree-vect-stmts.c (vectorizable_store): When looking for
8829 smaller vector types to perform grouped strided loads/stores
8830 make sure the mode is supported by the target.
8831 (vectorizable_load): Likewise.
8833 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8835 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8836 Increase LDP/STP opportunities by adding adjacent callee-saves.
8838 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8840 PR rtl-optimization/84068
8841 PR rtl-optimization/83459
8842 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8844 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8846 PR tree-optimization/84224
8847 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8848 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8851 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8854 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8855 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8856 count, remove one USE.
8858 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8860 * doc/install.texi (Configuration): Document the
8861 --with-long-double-format={ibm,ieee} PowerPC configuration
8865 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8866 Convert from define_expand to be define_insn_and_split. Rework
8867 float/double/_Float128 conversions to QI/HI/SImode to work with
8868 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8869 conversions to QI/HImode types did a store and then a load to
8870 truncate the value. For conversions to VSX registers, don't split
8871 the insn, instead emit the code directly. Use the code iterator
8872 any_fix to combine signed and unsigned conversions.
8873 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8874 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8875 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8876 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8877 (fix_<mode>di2_hw): Likewise.
8878 (fixuns_<mode>di2_hw): Likewise.
8879 (fix_<mode>si2_hw): Likewise.
8880 (fixuns_<mode>si2_hw): Likewise.
8881 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8882 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8883 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8884 fix<uns>_trunc<SFDF:mode>si2_p8.
8885 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8887 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8888 (fix<uns>_<mode>_mem): Likewise.
8889 (fctiw<u>z_<mode>_mem): Likewise.
8890 (fix<uns>_<mode>_mem): Likewise.
8891 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8892 the register allocator from doing a direct move to the GPRs to do
8893 a store, and instead use the ISA 3.0 store byte/half-word from
8894 vector register instruction. For IEEE 128-bit floating point,
8895 also optimize stores of 32-bit ints.
8896 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8898 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8900 * genextract.c (push_pathstr_operand): New function to support
8902 (walk_rtx): Call push_pathstr_operand.
8903 (print_path): Support [a-zA-Z].
8905 2018-02-07 Richard Biener <rguenther@suse.de>
8907 PR tree-optimization/84037
8908 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8909 (cse_and_gimplify_to_preheader): Declare.
8910 (vect_get_place_in_interleaving_chain): Likewise.
8911 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8913 (_loop_vec_info::~_loop_vec_info): Delete it.
8914 (cse_and_gimplify_to_preheader): New function.
8915 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8916 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8917 (vectorizable_load): Likewise. For grouped stores always base
8918 the IV on the first element.
8919 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8920 condition before gimplifying.
8922 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8924 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8925 *DIV_EXPR and *MOD_EXPR.
8927 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8930 * config/i386/i386.c (ix86_option_override_internal): Mask out
8931 the CF_SET bit when checking -fcf-protection.
8933 2018-02-07 Tom de Vries <tom@codesourcery.com>
8936 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8939 2018-02-07 Richard Biener <rguenther@suse.de>
8941 PR tree-optimization/84204
8942 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8945 PR tree-optimization/84205
8946 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8947 special-case isl_ast_op_zdiv_r.
8949 PR tree-optimization/84223
8950 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8951 Only add conditions from within the region.
8952 (gather_bbs::after_dom_children): Adjust.
8954 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8957 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8958 * config/avr/avr.md: Only post-reload split REG-REG moves if
8959 either register is GENERAL_REG_P.
8961 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8963 PR tree-optimization/84235
8964 * tree-ssa-scopedtables.c
8965 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8966 if the subtraction is performed in floating point type where NaNs are
8967 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8968 build 1. Formatting fix.
8970 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8973 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8974 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8975 and skip it regardless of bb boundaries. Use CALL_P macro,
8976 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8977 unnecessarily, formatting fix.
8979 2018-02-06 Michael Collison <michael.collison@arm.com>
8981 * config/arm/thumb2.md:
8982 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8983 (*thumb_mov_notscc): Ditto.
8985 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8988 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8989 unsigned_fix, not "s".
8991 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8993 * configure.ac (gcc_fn_eh_frame_ro): New function.
8994 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8995 correct .eh_frame permissions.
8996 * configure: Regenerate.
8998 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9000 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9003 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9005 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9006 Display warning message for -mno-speculate-indirect-jumps.
9008 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9010 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9012 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9014 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9016 PR tree-optimization/84225
9017 * tree-eh.c (find_trapping_overflow): Only call
9018 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9020 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9023 * config/i386/i386.c: Reimplement the check of possible options
9024 -mibt/-mshstk conbination. Change error messages.
9025 * doc/invoke.texi: Fix a typo: remove extra '='.
9027 2018-02-06 Marek Polacek <polacek@redhat.com>
9029 PR tree-optimization/84228
9030 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9032 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9035 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9036 emitted arch directives.
9037 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9038 __ARM_FEATURE_COPROC before changing architectures.
9040 2018-02-06 Richard Biener <rguenther@suse.de>
9042 * config/i386/i386.c (print_reg): Fix typo.
9043 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9045 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9047 * configure: Regenerate.
9049 2018-02-05 Martin Sebor <msebor@redhat.com>
9051 PR tree-optimization/83369
9052 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9055 2018-02-05 Martin Liska <mliska@suse.cz>
9057 * doc/invoke.texi: Cherry-pick upstream r323995.
9059 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9061 * ira.c (ira_init_register_move_cost): Adjust comment.
9063 2018-02-05 Martin Liska <mliska@suse.cz>
9065 PR gcov-profile/84137
9066 * doc/gcov.texi: Fix typo in documentation.
9068 2018-02-05 Martin Liska <mliska@suse.cz>
9070 PR gcov-profile/83879
9071 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9072 using dlopen functionality.
9074 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9076 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9077 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9078 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9079 _mm_maskz_range_round_ss): New intrinsics.
9080 (__builtin_ia32_rangesd128_round)
9081 (__builtin_ia32_rangess128_round): Remove.
9082 (__builtin_ia32_rangesd128_mask_round,
9083 __builtin_ia32_rangess128_mask_round): New builtins.
9084 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9085 __builtin_ia32_rangess128_round): Remove.
9086 (__builtin_ia32_rangesd128_mask_round,
9087 __builtin_ia32_rangess128_mask_round): New builtins.
9088 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9089 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9090 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9091 "<round_saeonly_constraint>")): Changed to ...
9092 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9093 "<round_saeonly_scalar_constraint>")): ... this.
9094 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9095 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9096 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9097 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9098 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9100 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9102 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9104 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9105 Remove all values except native, 8540 and 8548.
9107 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9109 * config/i386/i386.c (ix86_output_function_return): Pass
9110 INVALID_REGNUM, instead of -1, as invalid register number to
9111 indirect_thunk_name and output_indirect_thunk.
9113 2018-02-02 Julia Koval <julia.koval@intel.com>
9115 * config.gcc: Add -march=icelake.
9116 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9117 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9118 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9119 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9120 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9121 (processor_target_table): Add icelake.
9122 (ix86_option_override_internal): Handle new PTAs.
9123 (get_builtin_code_for_version): Handle icelake.
9124 (M_INTEL_COREI7_ICELAKE): New.
9125 (fold_builtin_cpu): Handle icelake.
9126 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9127 * doc/invoke.texi: Add -march=icelake.
9129 2018-02-02 Julia Koval <julia.koval@intel.com>
9131 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9132 to wide_int_bitmask.
9133 * wide-int-bitmask.h: New.
9135 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9138 * config/i386/i386.md: Replace Pmode with word_mode in
9139 builtin_setjmp_setup and builtin_longjmp to support x32.
9141 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9145 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9147 (rs6000_supported_cpu_names): New static variable.
9148 (linux_cpu_translation_table): Likewise.
9149 (elf_platform) <cpu>: Define new static variable and use it.
9150 Translate kernel AT_PLATFORM name to canonical name if needed.
9151 Error if platform name is unknown.
9153 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9156 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9158 2018-02-01 Jeff Law <law@redhat.com>
9161 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9162 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9163 the scratch if RELEASE_VIA_POP is false.
9164 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9165 If we have to save a temporary register, decrement SIZE appropriately.
9166 Pass new arguments to release_scratch_register_on_entry.
9167 (ix86_adjust_stack_and_probe): Likewise.
9168 (ix86_emit_probe_stack_range): Pass new arguments to
9169 release_scratch_register_on_entry.
9171 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9173 PR rtl-optimization/84157
9174 * combine.c (change_zero_ext): Use REG_P predicate in
9175 front of HARD_REGISTER_P predicate.
9177 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9179 * config/avr/avr.c (avr_option_override): Move disabling of
9180 -fdelete-null-pointer-checks to...
9181 * common/config/avr/avr-common.c (avr_option_optimization_table):
9184 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9186 PR tree-optimization/81635
9187 * tree-data-ref.c (split_constant_offset_1): For types that
9188 wrap on overflow, try to use range info to prove that wrapping
9191 2018-02-01 Renlin Li <renlin.li@arm.com>
9194 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9196 (aarch64_register_move_cost): Likewise.
9197 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9199 (REG_CLASS_NAMES): Likewise.
9200 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9201 TAILCALL_ADDR_REGS. Remove IP registers.
9202 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9204 2018-02-01 Richard Biener <rguenther@suse.de>
9206 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9207 for specifying RPO order and allow NULL for that.
9208 * domwalk.c (dom_walker::dom_walker): Likewise.
9209 (dom_walker::walk): Handle NULL RPO order.
9210 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9212 (rewrite_update_dom_walker): Likewise.
9213 (mark_def_dom_walker): Likewise.
9215 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9217 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9218 (aarch64_maybe_expand_sve_subreg_move): Declare.
9219 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9220 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9222 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9223 that are semantically a reverse operation.
9224 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9225 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9226 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9228 (aarch64_can_change_mode_class): For big-endian, forbid changes
9229 between two SVE modes if they have different element sizes.
9231 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9233 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9234 the TImode handling for big-endian targets.
9236 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9238 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9239 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9241 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9242 Remove BSWAP handing for big-endian targets and use the form of
9243 LD1RQ appropariate for the mode.
9245 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9247 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9248 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9251 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9254 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9255 check for operands that need to go through aarch64_sve_reload_be.
9257 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9259 PR tree-optimization/81661
9260 PR tree-optimization/84117
9261 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9262 * tree-eh.c: Include gimplify.h.
9263 (find_trapping_overflow, replace_trapping_overflow,
9264 rewrite_to_non_trapping_overflow): New functions.
9265 * tree-vect-loop.c: Include tree-eh.h.
9266 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9267 * tree-data-ref.c: Include tree-eh.h.
9268 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9270 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9272 PR rtl-optimization/84123
9273 * combine.c (change_zero_ext): Check if hard register satisfies
9274 can_change_dest_mode before calling gen_lowpart_SUBREG.
9276 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9279 * ira.c (ira_init_register_move_cost): Remove assert.
9281 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9283 PR rtl-optimization/84071
9284 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9285 * doc/tm.texi: Regenerate.
9287 2018-01-31 Richard Biener <rguenther@suse.de>
9289 PR tree-optimization/84132
9290 * tree-data-ref.c (analyze_miv_subscript): Properly
9291 check whether evolution_function_is_affine_multivariate_p
9292 before calling gcd_of_steps_may_divide_p.
9294 2018-01-31 Julia Koval <julia.koval@intel.com>
9297 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9298 * config/i386/i386.md (rdpid_rex64) New.
9299 (rdpid): Make 32bit only.
9301 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9304 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9305 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9307 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9310 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9312 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9314 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9316 PR rtl-optimization/84071
9317 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9318 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9320 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9322 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9323 (arc_attribute_table): Add 'aux' attribute.
9324 (arc_in_small_data_p): Consider aux like variables.
9325 (arc_is_aux_reg_p): New function.
9326 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9327 (arc_get_aux_arg): New function.
9328 (prepare_move_operands): Handle aux-register access.
9329 (arc_handle_aux_attribute): New function.
9330 * doc/extend.texi (ARC Variable attributes): Add subsection.
9332 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9334 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9335 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9336 (arc_attribute_table): Add 'uncached' attribute.
9337 (arc_print_operand): Print '.di' flag for uncached memory
9339 (arc_in_small_data_p): Do not consider for small data the uncached
9341 (arc_is_uncached_mem_p): New function.
9342 * config/arc/predicates.md (compact_store_memory_operand): Check
9343 for uncached memory accesses.
9344 (nonvol_nonimm_operand): Likewise.
9345 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9347 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9350 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9351 falign-loops=): Add Optimization flag.
9353 2018-01-30 Jeff Law <law@redhat.com>
9356 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9357 INT_REGISTERS_SAVED. Check it prior to calling
9358 get_scratch_register_on_entry.
9359 (ix86_adjust_stack_and_probe): Similarly.
9360 (ix86_emit_probe_stack_range): Similarly.
9361 (ix86_expand_prologue): Corresponding changes.
9363 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9366 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9367 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9369 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9372 * lra-constraints.c (curr_insn_transform): Process AND in the
9375 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9377 PR rtl-optimization/83986
9378 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9379 dependence against last_pending_memory_flush in addition to
9382 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9384 PR tree-optimization/81611
9385 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9388 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9391 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9394 2018-01-30 Richard Biener <rguenther@suse.de>
9395 Jakub Jelinek <jakub@redhat.com>
9397 PR tree-optimization/84111
9398 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9399 inner loops added during recursion, as they don't have up-to-date
9402 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9405 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9406 (can_inline_edge_by_limits_p): ... here.
9407 (can_early_inline_edge_p, check_callers,
9408 update_caller_keys, update_callee_keys, recursive_inlining,
9409 add_new_edges_to_heap, speculation_useful_p,
9410 inline_small_functions,
9411 inline_small_functions, flatten_function,
9412 inline_to_all_callers_1): Update.
9414 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9416 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9419 2018-01-30 Richard Biener <rguenther@suse.de>
9421 PR tree-optimization/83008
9422 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9423 invariant and constant vector uses in stmts when they need
9426 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9429 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9430 * configure: Regenerate.
9432 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9434 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9436 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9437 Use gen_rtx_REG rather than gen_lowpart.
9439 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9441 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9442 rather than 0 when creating partial subregs.
9444 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9446 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9449 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9452 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9453 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9454 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9455 flags. This restores the settings used before the 2017-07-24.
9456 Turning off pre increment/decrement/modify allows IVOPTS to
9457 optimize DF/SF loops where the index is an int.
9459 2018-01-29 Richard Biener <rguenther@suse.de>
9460 Kelvin Nilsen <kelvin@gcc.gnu.org>
9463 * tree-vect-stmts.c (vectorizable_call): Don't call
9464 targetm.vectorize_builtin_md_vectorized_function if callee is
9467 2018-01-22 Carl Love <cel@us.ibm.com>
9469 * doc/extend.tex: Fix typo in second arg in
9470 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9472 2018-01-29 Richard Biener <rguenther@suse.de>
9474 PR tree-optimization/84086
9475 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9476 (flush_ssaname_freelist): When SSA names were released reset
9477 the SCEV hash table.
9479 2018-01-29 Richard Biener <rguenther@suse.de>
9481 PR tree-optimization/84057
9482 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9483 removed paths when removing edges.
9485 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9487 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9488 -mfunction-return=@var{choice}.
9490 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9493 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9495 (layout::print_source_line): Likewise.
9496 (test_get_line_width_without_trailing_whitespace): Add test cases.
9498 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9501 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9504 2018-01-26 Jim Wilson <jimw@sifive.com>
9506 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9508 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9511 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9513 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9514 and CMP + SUB-immediate -> SUBS.
9516 2018-01-26 Martin Sebor <msebor@redhat.com>
9518 PR tree-optimization/83896
9519 * tree-ssa-strlen.c (get_string_len): Rename...
9520 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9521 Avoid assuming length is constant.
9522 (handle_char_store): Use HOST_WIDE_INT for string length.
9524 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9527 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9528 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9530 2018-01-26 Richard Biener <rguenther@suse.de>
9532 PR rtl-optimization/84003
9533 * dse.c (record_store): Only record redundant stores when
9534 the earlier store aliases at least all accesses the later one does.
9536 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9538 PR rtl-optimization/83985
9539 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9540 REG_CFA_RESTORE insns.
9541 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9542 REG_CFA_RESTORE insns here.
9545 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9546 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9548 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9550 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9551 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9552 (arc_init): Likewise.
9553 (arc_override_options): Likewise.
9554 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9556 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9558 * config/arc/arc.h (TARGET_DBNZ): Define.
9559 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9560 properly set the tune attribute.
9561 (dbnz): Use TARGET_DBNZ guard.
9562 * config/arc/arc.opt (mtune): Add core3 option.
9564 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9566 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9567 recognize new pic like addresses.
9568 (arc_delegitimize_address): Clean up.
9570 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9572 * config/arc/arc-arches.def: Option mrf16 valid for all
9574 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9575 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9576 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9577 * config/arc/arc-tables.opt: Regenerate.
9578 * config/arc/arc.c (arc_conditional_register_usage): Handle
9579 reduced register file case.
9580 (arc_file_start): Set must have build attributes.
9581 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9583 * config/arc/arc.opt (mrf16): Add new option.
9584 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9585 * config/arc/genmultilib.awk: Handle new mrf16 option.
9586 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9587 * config/arc/t-multilib: Regenerate.
9588 * doc/invoke.texi (ARC Options): Document mrf16 option.
9590 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9592 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9593 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9594 (arc_attribute_table): Add 'secure_call' attribute.
9595 (arc_print_operand): Print secure call operand.
9596 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9598 (arc_is_secure_call_p): New function. * config/arc/arc.md
9599 (call_i): Add support for sjli instruction.
9600 (call_value_i): Likewise.
9601 * config/arc/constraints.md (Csc): New constraint.
9603 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9604 John Eric Martin <John.Martin@emmicro-us.com>
9606 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9607 * config/arc/arc.c (_arc_jli_section): New struct.
9608 (arc_jli_section): New type.
9609 (rc_jli_sections): New static variable.
9610 (arc_handle_jli_attribute): New function.
9611 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9612 (arc_file_end): New function.
9613 (TARGET_ASM_FILE_END): Define.
9614 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9615 (arc_add_jli_section): New function.
9616 (jli_call_scan): Likewise.
9617 (arc_reorg): Call jli_call_scan.
9618 (arc_output_addsi): Remove 'S' from printing asm operand.
9619 (arc_is_jli_call_p): New function.
9620 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9622 (movhi_insn): Likewise.
9623 (movsi_insn): Likewise.
9624 (movsi_set_cc_insn): Likewise.
9625 (loadqi_update): Likewise.
9626 (load_zeroextendqisi_update): Likewise.
9627 (load_signextendqisi_update): Likewise.
9628 (loadhi_update): Likewise.
9629 (load_zeroextendhisi_update): Likewise.
9630 (load_signextendhisi_update): Likewise.
9631 (loadsi_update): Likewise.
9632 (loadsf_update): Likewise.
9633 (movsicc_insn): Likewise.
9634 (bset_insn): Likewise.
9635 (bxor_insn): Likewise.
9636 (bclr_insn): Likewise.
9637 (bmsk_insn): Likewise.
9638 (bicsi3_insn): Likewise.
9639 (cmpsi_cc_c_insn): Likewise.
9640 (movsi_ne): Likewise.
9641 (movsi_cond_exec): Likewise.
9642 (clrsbsi2): Likewise.
9653 (call_i): Remove 'S' asm letter, add jli instruction.
9654 (call_value_i): Likewise.
9655 * config/arc/arc.op (mjli-always): New option.
9656 * config/arc/constraints.md (Cji): New constraint.
9657 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9659 (subsf3_fpx): Likewise.
9660 (mulsf3_fpx): Likewise.
9661 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9663 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9665 * doc/invoke.texi (ARC): Document mjli-always option.
9667 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9669 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9670 avoid addition with 0 and use incw and decw where possible.
9672 2018-01-26 Richard Biener <rguenther@suse.de>
9674 PR tree-optimization/81082
9675 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9676 association if it requires casting to unsigned.
9677 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9678 from fold_plusminus_mult_expr to catch important cases late when
9679 range info is available.
9681 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9683 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9684 * configure.ac (hidden_linkonce): New test.
9685 * configure: Regenerate.
9686 * config.in: Regenerate.
9688 2018-01-26 Julia Koval <julia.koval@intel.com>
9690 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9691 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9692 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9693 _mm_mask_bitshuffle_epi64_mask): Fix type.
9694 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9695 USI_FTYPE_V4DI_V4DI_USI): Remove.
9696 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9697 __builtin_ia32_vpshufbitqmb256_mask,
9698 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9699 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9700 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9702 2018-01-26 Alan Modra <amodra@gmail.com>
9705 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9706 UNSPEC_VBPERMQ. Sort other unspecs.
9708 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9710 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9712 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9715 * predict.c (drop_profile): Do not push/pop cfun; update also
9717 (handle_missing_profiles): Fix logic looking for zero profiles.
9719 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9722 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9723 on functions with #pragma omp declare simd or functions with simd
9725 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9726 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9727 Remove trailing \n from warning_at calls.
9729 2018-01-25 Tom de Vries <tom@codesourcery.com>
9732 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9733 for neutered workers.
9735 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9738 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9739 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9741 2018-01-24 Jeff Law <law@redhat.com>
9744 * i386.c (get_probe_interval): Move to earlier point.
9745 (ix86_compute_frame_layout): If -fstack-clash-protection and
9746 the frame is larger than the probe interval, then use pushes
9747 to save registers rather than reg->mem moves.
9748 (ix86_expand_prologue): Remove conditional for int_registers_saved
9751 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9754 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9755 min/max for never referenced object.
9757 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9760 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9762 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9763 attributes from DECL_ATTRIBUTES (decl) without affecting
9764 DECL_ATTRIBUTES (current_function_decl).
9765 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9766 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9768 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9770 PR tree-optimization/83979
9771 * fold-const.c (fold_comparison): Use constant_boolean_node
9772 instead of boolean_{true,false}_node.
9774 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9776 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9779 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9781 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9782 Simplify the clause that sets the length attribute.
9783 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9784 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9785 clause that sets the length attribute.
9786 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9788 2018-01-24 Tom de Vries <tom@codesourcery.com>
9791 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9792 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9793 Add strict parameter.
9794 (prevent_branch_around_nothing): Insert dummy insn between branch to
9795 label and label with no ptx insn inbetween.
9796 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9798 2018-01-24 Tom de Vries <tom@codesourcery.com>
9801 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9802 for neutered threads in warp.
9803 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9805 2018-01-24 Richard Biener <rguenther@suse.de>
9807 PR tree-optimization/83176
9808 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9811 2018-01-24 Richard Biener <rguenther@suse.de>
9813 PR tree-optimization/82819
9814 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9815 code generating pluses that are no-ops in the target precision.
9817 2018-01-24 Richard Biener <rguenther@suse.de>
9820 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9822 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9824 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9825 to merge probabilities.
9826 * predict.c (probably_never_executed): Also mark as cold functions
9827 with global 0 profile and guessed local profile.
9828 * profile-count.c (profile_probability::combine_with_count): New
9830 * profile-count.h (profile_probability::operator*,
9831 profile_probability::operator*=, profile_probability::operator/,
9832 profile_probability::operator/=): Reduce precision to adjusted
9833 and set value to guessed on contradictory divisions.
9834 (profile_probability::combine_with_freq): Remove.
9835 (profile_probability::combine_wiht_count): Declare.
9836 (profile_count::force_nonzero):: Set to adjusted.
9837 (profile_count::probability_in):: Set quality to adjusted.
9838 * tree-ssa-tail-merge.c (replace_block_by): Use
9841 2018-01-23 Andrew Waterman <andrew@sifive.com>
9842 Jim Wilson <jimw@sifive.com>
9844 * config/riscv/riscv.c (riscv_stack_boundary): New.
9845 (riscv_option_override): Set riscv_stack_boundary. Handle
9846 riscv_preferred_stack_boundary_arg.
9847 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9848 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9849 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9850 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9851 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9852 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9854 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9857 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9858 of struct ix86_frame.
9859 (ix86_expand_epilogue): Likewise. Add a local variable for
9860 the reg_save_offset field in struct ix86_frame.
9862 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9864 PR tree-optimization/82604
9865 * tree-loop-distribution.c (enum partition_kind): New enum item
9866 PKIND_PARTIAL_MEMSET.
9867 (partition_builtin_p): Support above new enum item.
9868 (generate_code_for_partition): Ditto.
9869 (compute_access_range): Differentiate cases that equality can be
9870 proven at all loops, the innermost loops or no loops.
9871 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9872 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9873 (finalize_partitions, distribute_loop): Don't fuse partition of
9874 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9875 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9878 2018-01-23 Martin Liska <mliska@suse.cz>
9880 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9881 order to ignore the predictor.
9882 (PRED_POLYMORPHIC_CALL): Likewise.
9883 (PRED_RECURSIVE_CALL): Likewise.
9885 2018-01-23 Martin Liska <mliska@suse.cz>
9887 * tree-profile.c (tree_profiling): Print function header to
9888 aware reader which function we are working on.
9889 * value-prof.c (gimple_find_values_to_profile): Do not print
9890 not interesting value histograms.
9892 2018-01-23 Martin Liska <mliska@suse.cz>
9894 * profile-count.h (enum profile_quality): Add
9895 profile_uninitialized as the first value. Do not number values
9896 as they are zero based.
9897 (profile_count::verify): Update sanity check.
9898 (profile_probability::verify): Likewise.
9900 2018-01-23 Nathan Sidwell <nathan@acm.org>
9902 * doc/invoke.texi (ffor-scope): Deprecate.
9904 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9906 PR tree-optimization/83510
9907 * domwalk.c (set_all_edges_as_executable): New function.
9908 (dom_walker::dom_walker): Convert bool param
9909 "skip_unreachable_blocks" to enum reachability. Move setup of
9910 edge flags to set_all_edges_as_executable and only do it when
9911 reachability is REACHABLE_BLOCKS.
9912 * domwalk.h (enum dom_walker::reachability): New enum.
9913 (dom_walker::dom_walker): Convert bool param
9914 "skip_unreachable_blocks" to enum reachability.
9915 (set_all_edges_as_executable): New decl.
9916 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9917 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9919 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9920 but converting true to REACHABLE_BLOCKS.
9921 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9923 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9924 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9925 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9927 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9928 if check_all_array_refs will be called.
9930 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9932 * tree.c (selftest::test_location_wrappers): Add more test
9935 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9937 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9938 (selftest::test_bit_in_range): Likewise.
9940 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9943 * doc/sourcebuild.texi (vect_float): Say that the selector
9944 only describes the situation when -funsafe-math-optimizations is on.
9945 (vect_float_strict): Document.
9947 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9949 PR tree-optimization/83965
9950 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9951 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9952 instead of checking only for a reduction.
9953 (vect_recog_widen_sum_pattern): Likewise.
9955 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9957 * predict.c (probably_never_executed): Only use precise profile info.
9958 (compute_function_frequency): Skip after inlining hack since we now
9959 have quality checking.
9961 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9963 * profile-count.h (profile_probability::very_unlikely,
9964 profile_probability::unlikely, profile_probability::even): Set
9965 precision to guessed.
9967 2018-01-23 Richard Biener <rguenther@suse.de>
9969 PR tree-optimization/83963
9970 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9971 Properly terminate dominator walk when crossing the exit edge not
9972 when visiting its source block.
9974 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9977 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9978 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9980 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9982 PR tree-optimization/83957
9983 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9984 semicolon after for body surrounded by braces.
9986 PR tree-optimization/83081
9987 * profile-count.h (profile_probability::split): New method.
9988 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9989 Use profile_probability::split.
9990 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9991 when splitting a single conditional jump into 2.
9993 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9995 PR tree-optimization/69452
9996 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9999 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10001 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10002 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10003 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10005 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10007 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10008 * config/rl78/rl78.md (movdi): New define_expand.
10009 * config/rl78/rl78.c (rl78_split_movdi): New function.
10011 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10014 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10016 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10017 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10018 128-bit to produce an UNSPEC move to get the double word with the
10019 signbit and then a shift directly to do signbit.
10020 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10021 implementation with a new version that just does either a direct
10022 move or a regular move. Move memory interface to separate insns.
10023 Move insns so they are next to the expander.
10024 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10025 with signbit move. Split big and little endian case.
10026 (signbit<mode>2_dm_mem_le): Likewise.
10027 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10028 (signbit<mode>2_dm2): Likewise.
10030 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10032 * config/rl78/rl78.md (anddi3): New define_expand.
10034 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10036 * config/rl78/rl78.md (umindi3): New define_expand.
10038 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10040 * config/rl78/rl78.md (smindi3): New define_expand.
10042 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10044 * config/rl78/rl78.md (smaxdi3): New define_expand.
10046 2018-01-22 Carl Love <cel@us.ibm.com>
10048 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10049 LVX_V1TI): Add macro expansion.
10050 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10051 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10052 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10053 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10054 Change check to determine if the instruction is a byte reversing
10055 entry. Fix typo in comment.
10056 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10057 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10058 Add def_builtin calls for new builtins.
10059 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10060 Add define_insn expansion.
10062 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10064 * config/rl78/rl78.md (umaxdi3): New define_expand.
10066 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10068 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10069 for non-QImode registers.
10071 2018-01-22 Richard Biener <rguenther@suse.de>
10073 PR tree-optimization/83963
10074 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10075 including the loop exit block.
10076 (scop_detection::merge_sese): Likewise.
10077 (scop_detection::add_scop): Do it here instead.
10079 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10081 * doc/sourcebuild.texi (arm_softfloat): Document.
10083 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10086 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10087 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10088 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10090 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10091 David Edelsohn <dje.gcc@gmail.com>
10094 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10095 Change "crset eq" to "crset 2".
10096 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10097 (*call_indirect_aix<mode>_nospec): Likewise.
10098 (*call_value_indirect_aix<mode>_nospec): Likewise.
10099 (*call_indirect_elfv2<mode>_nospec): Likewise.
10100 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10101 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10102 change assembly output from . to $.
10103 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10104 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10105 (*tablejump<mode>_internal1_nospec): Likewise.
10107 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10110 * config/sh/sh_optimize_sett_clrt.cc:
10111 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10113 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10115 PR tree-optimization/83940
10116 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10117 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10118 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10119 use it to pass back the definition type.
10120 (vect_check_store_rhs): Likewise rhs_dt_out.
10121 (vect_build_gather_load_calls): Add a mask_dt argument and use
10122 it instead of a call to vect_is_simple_use.
10123 (vectorizable_store): Update calls to vect_check_load_store_mask
10124 and vect_check_store_rhs. Use the dt returned by the latter instead
10125 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10126 instead of calls to vect_is_simple_use. Pass the scalar rather
10127 than the vector operand to vect_is_simple_use when handling
10128 second and subsequent copies of an rhs value.
10129 (vectorizable_load): Update calls to vect_check_load_store_mask
10130 and vect_build_gather_load_calls. Use the cached mask_dt and
10131 gs_info.offset_dt instead of calls to vect_is_simple_use.
10133 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10135 PR middle-end/83945
10136 * tree-emutls.c: Include gimplify.h.
10137 (lower_emutls_2): New function.
10138 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10139 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10140 it before further processing.
10143 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10144 UINTVAL (trueop1) instead of INTVAL (op1).
10146 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10150 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10151 INCOMING_FRAME_SP_OFFSET if not defined.
10152 (scan_trace): Add ENTRY argument. If true and
10153 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10154 emit a note to adjust the CFA offset.
10155 (create_cfi_notes): Adjust scan_trace callers.
10156 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10157 INCOMING_FRAME_SP_OFFSET in the CIE.
10158 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10159 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10161 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10162 * doc/tm.texi: Regenerated.
10164 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10166 PR rtl-optimization/83147
10167 * lra-constraints.c (remove_inheritance_pseudos): Use
10168 lra_substitute_pseudo_within_insn.
10170 2018-01-19 Tom de Vries <tom@codesourcery.com>
10171 Cesar Philippidis <cesar@codesourcery.com>
10174 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10176 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10179 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10180 spaces for function labels.
10182 2018-01-19 Martin Liska <mliska@suse.cz>
10184 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10185 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10186 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10187 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10188 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10189 (PRED_CONST_RETURN): Change from 69 to 65.
10190 (PRED_NULL_RETURN): Change from 91 to 71.
10191 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10192 (PRED_LOOP_GUARD): Change from 66 to 73.
10194 2018-01-19 Martin Liska <mliska@suse.cz>
10196 * predict.c (predict_insn_def): Add new assert.
10197 (struct branch_predictor): Change type to signed integer.
10198 (test_prediction_value_range): Amend test to cover
10199 PROB_UNINITIALIZED.
10200 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10201 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10202 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10203 (PRED_LOOP_IV_COMPARE): Likewise.
10204 * predict.h (PROB_UNINITIALIZED): Define new constant.
10206 2018-01-19 Martin Liska <mliska@suse.cz>
10208 * predict.c (dump_prediction): Add new format for
10209 analyze_brprob.py script which is enabled with -details
10211 * profile-count.h (precise_p): New function.
10213 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10215 PR tree-optimization/83922
10216 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10217 there are no statements that need masking.
10218 (vect_active_double_reduction_p): New function.
10219 (vect_analyze_loop_operations): Use it when handling phis that
10220 are not in the loop header.
10222 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10224 PR tree-optimization/83914
10225 * tree-vect-loop.c (vectorizable_induction): Don't convert
10226 init_expr or apply the peeling adjustment for inductions
10227 that are nested within the vectorized loop.
10229 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10231 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10234 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10238 * function.h (gimplify_parameters): Add gimple_seq * argument.
10239 * function.c: Include gimple.h and options.h.
10240 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10241 for the added local temporaries if needed.
10242 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10243 if there are any parameter cleanups, wrap whole body into a
10244 try/finally with the cleanups.
10246 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10249 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10250 Use GET_MODE_CLASS for scalar floating point.
10252 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10256 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10257 Fix call of call_cgraph_insertion_hooks.
10259 2018-01-18 Martin Sebor <msebor@redhat.com>
10261 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10263 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10266 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10269 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10272 * common.opt: (-ffile-prefix-map): New option.
10273 * opts.c (common_handle_option): Defer it.
10274 * opts-global.c (handle_common_deferred_options): Handle it.
10275 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10276 * file-prefix-map.h: New file.
10277 (remap_debug_filename, add_debug_prefix_map): ...here.
10278 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10279 * final.c (debug_prefix_map, add_debug_prefix_map
10280 remap_debug_filename): Move to...
10281 * file-prefix-map.c: New file.
10282 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10283 generalize, get rid of alloca(), use strrchr() instead of strchr().
10284 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10285 Implement in terms of add_prefix_map().
10286 (remap_macro_filename, remap_debug_filename): Implement in term of
10288 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10289 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10290 * dbxout.c: Include file-prefix-map.h.
10291 * varasm.c: Likewise.
10292 * vmsdbgout.c: Likewise.
10293 * xcoffout.c: Likewise.
10294 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10295 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10296 * doc/invoke.texi (-ffile-prefix-map): Document.
10297 (-fdebug-prefix-map): Update description.
10299 2018-01-18 Martin Liska <mliska@suse.cz>
10301 * config/i386/i386.c (indirect_thunk_name): Document that also
10303 (output_indirect_thunk): Document why both instructions
10304 (pause and lfence) are generated.
10306 2018-01-18 Richard Biener <rguenther@suse.de>
10308 PR tree-optimization/83887
10309 * graphite-scop-detection.c
10310 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10311 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10312 (scop_detection::merge_sese): Re-implement with a flood-fill
10313 algorithm that properly finds a SESE region if it exists.
10315 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10318 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10319 pointer_diff optimizations use view_convert instead of convert.
10321 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10323 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10324 Generate different code for -mno-speculate-indirect-jumps.
10325 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10326 (*call_indirect_aix<mode>): Disable for
10327 -mno-speculate-indirect-jumps.
10328 (*call_indirect_aix<mode>_nospec): New define_insn.
10329 (*call_value_indirect_aix<mode>): Disable for
10330 -mno-speculate-indirect-jumps.
10331 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10332 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10333 -mno-speculate-indirect-jumps.
10334 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10336 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10338 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10339 long double type, set the flags for noting the default long double
10340 type, even if we don't pass or return a long double type.
10342 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10345 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10348 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10350 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10351 support for merge[hl].
10352 (fold_mergehl_helper): New helper function.
10353 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10354 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10355 (altivec_vmrglw_direct): Add xxmrglw insn.
10357 2018-01-17 Andrew Waterman <andrew@sifive.com>
10359 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10360 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10362 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10365 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10366 call the lto_location_cache before reading the
10367 DECL_SOURCE_LOCATION of the types.
10369 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10370 Richard Sandiford <richard.sandiford@linaro.org>
10372 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10373 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10374 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10375 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10376 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10378 * config/aarch64/constraints.md (aarch64_movti_operand):
10380 * config/aarch64/predicates.md (Uti): Add new constraint.
10382 2018-01-17 Carl Love <cel@us.ibm.com>
10384 * config/rs6000/vsx.md (define_expand xl_len_r,
10385 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10386 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10388 (define_expand, define_insn): Move the shift left from the
10389 define_insn to the define_expand for lxvl and stxvl instructions.
10390 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10391 and XL_LEN_R definitions to PURE.
10393 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10395 * config/i386/i386.c (indirect_thunk_name): Declare regno
10396 as unsigned int. Compare regno with INVALID_REGNUM.
10397 (output_indirect_thunk): Ditto.
10398 (output_indirect_thunk_function): Ditto.
10399 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10400 in the call to output_indirect_thunk_function.
10402 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10404 PR middle-end/83884
10405 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10406 rather than the size of inner_type to determine the stack slot size
10407 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10409 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10412 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10415 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10417 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10418 endian Linux systems to optionally enable multilibs for selecting
10419 the long double type if the user configured an explicit type.
10420 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10421 have no long double multilibs if not defined.
10422 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10423 warn if the user used -mabi={ieee,ibm}longdouble and we built
10424 multilibs for long double.
10425 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10426 appropriate multilib option.
10427 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10429 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10430 for building long double multilibs.
10431 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10433 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10435 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10438 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10440 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10443 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10446 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10449 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10451 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10452 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10454 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10456 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10457 different rtl trees depending on TARGET_64BIT.
10458 (rs6000_gen_lvx): Likewise.
10460 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10462 * config/visium/visium.md (nop): Tweak comment.
10463 (hazard_nop): Likewise.
10465 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10467 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10468 -mspeculate-indirect-jumps.
10469 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10470 for -mno-speculate-indirect-jumps.
10471 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10472 (*call_value_indirect_elfv2<mode>): Disable for
10473 -mno-speculate-indirect-jumps.
10474 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10475 (indirect_jump): Emit different RTL for
10476 -mno-speculate-indirect-jumps.
10477 (*indirect_jump<mode>): Disable for
10478 -mno-speculate-indirect-jumps.
10479 (*indirect_jump<mode>_nospec): New define_insn.
10480 (tablejump): Emit different RTL for
10481 -mno-speculate-indirect-jumps.
10482 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10483 (tablejumpsi_nospec): New define_expand.
10484 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10485 (tablejumpdi_nospec): New define_expand.
10486 (*tablejump<mode>_internal1): Disable for
10487 -mno-speculate-indirect-jumps.
10488 (*tablejump<mode>_internal1_nospec): New define_insn.
10489 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10492 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10494 * caller-save.c (insert_save): Drop unnecessary parameter. All
10497 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10498 Richard Biener <rguenth@suse.de>
10501 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10502 return early, inline manually is_gimple_sizepos. Make sure if we
10503 call gimplify_expr we don't end up with a gimple constant.
10504 * tree.c (variably_modified_type_p): Don't return true for
10505 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10506 * gimplify.h (is_gimple_sizepos): Remove.
10508 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10510 PR tree-optimization/83857
10511 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10512 vectorizable_live_operation for pure SLP statements.
10513 (vectorizable_live_operation): Handle PHIs.
10515 2018-01-16 Richard Biener <rguenther@suse.de>
10517 PR tree-optimization/83867
10518 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10519 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10521 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10524 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10525 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10526 If off is not INTEGER_CST, issue a may not be aligned warning
10527 rather than isn't aligned. Use isn%'t rather than isn't.
10528 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10530 <case MULT_EXPR>: Improve the case when bottom and one of the
10531 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10532 operand, in that case check if the other operand is multiple of
10533 bottom divided by the INTEGER_CST operand.
10535 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10538 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10539 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10540 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10541 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10542 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10543 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10544 (pa_function_arg_size): New function.
10546 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10548 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10549 in a separate statement.
10551 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10553 PR tree-optimization/83847
10554 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10555 group gathers and scatters.
10557 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10559 PR rtl-optimization/86620
10560 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10562 PR rtl-optimization/83213
10563 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10564 to last if both are JUMP_INSNs.
10566 PR tree-optimization/83843
10567 * gimple-ssa-store-merging.c
10568 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10569 store_immediate_info for bswap/nop orig_stores.
10571 2018-01-15 Andrew Waterman <andrew@sifive.com>
10573 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10575 <UDIV>: Increase cost if !TARGET_DIV.
10577 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10579 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10580 (define_attr "cr_logical_3op"): New.
10581 (cceq_ior_compare): Adjust.
10582 (cceq_ior_compare_complement): Adjust.
10583 (*cceq_rev_compare): Adjust.
10584 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10585 (is_cracked_insn): Adjust.
10586 (insn_must_be_first_in_group): Adjust.
10587 * config/rs6000/40x.md: Adjust.
10588 * config/rs6000/440.md: Adjust.
10589 * config/rs6000/476.md: Adjust.
10590 * config/rs6000/601.md: Adjust.
10591 * config/rs6000/603.md: Adjust.
10592 * config/rs6000/6xx.md: Adjust.
10593 * config/rs6000/7450.md: Adjust.
10594 * config/rs6000/7xx.md: Adjust.
10595 * config/rs6000/8540.md: Adjust.
10596 * config/rs6000/cell.md: Adjust.
10597 * config/rs6000/e300c2c3.md: Adjust.
10598 * config/rs6000/e500mc.md: Adjust.
10599 * config/rs6000/e500mc64.md: Adjust.
10600 * config/rs6000/e5500.md: Adjust.
10601 * config/rs6000/e6500.md: Adjust.
10602 * config/rs6000/mpc.md: Adjust.
10603 * config/rs6000/power4.md: Adjust.
10604 * config/rs6000/power5.md: Adjust.
10605 * config/rs6000/power6.md: Adjust.
10606 * config/rs6000/power7.md: Adjust.
10607 * config/rs6000/power8.md: Adjust.
10608 * config/rs6000/power9.md: Adjust.
10609 * config/rs6000/rs64.md: Adjust.
10610 * config/rs6000/titan.md: Adjust.
10612 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10614 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10615 ix86_indirect_branch_register logic.
10617 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10619 * config/i386/constraints.md (Bs): Update
10620 ix86_indirect_branch_register check. Don't check
10621 ix86_indirect_branch_register with GOT_memory_operand.
10623 * config/i386/predicates.md (GOT_memory_operand): Don't check
10624 ix86_indirect_branch_register here.
10625 (GOT32_symbol_operand): Likewise.
10627 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10629 * config/i386/predicates.md (constant_call_address_operand):
10630 Rewrite ix86_indirect_branch_register logic.
10631 (sibcall_insn_operand): Likewise.
10633 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10635 * config/i386/constraints.md (Bs): Replace
10636 ix86_indirect_branch_thunk_register with
10637 ix86_indirect_branch_register.
10639 * config/i386/i386.md (indirect_jump): Likewise.
10640 (tablejump): Likewise.
10641 (*sibcall_memory): Likewise.
10642 (*sibcall_value_memory): Likewise.
10643 Peepholes of indirect call and jump via memory: Likewise.
10644 * config/i386/i386.opt: Likewise.
10645 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10646 (GOT_memory_operand): Likewise.
10647 (call_insn_operand): Likewise.
10648 (sibcall_insn_operand): Likewise.
10649 (GOT32_symbol_operand): Likewise.
10651 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10653 PR middle-end/83837
10654 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10655 type rather than type addr's type points to.
10656 (expand_omp_atomic_mutex): Likewise.
10657 (expand_omp_atomic): Likewise.
10659 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10662 * config/i386/i386.c (output_indirect_thunk_function): Use
10663 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10664 for __x86_return_thunk.
10666 2018-01-15 Richard Biener <rguenther@suse.de>
10668 PR middle-end/83850
10669 * expmed.c (extract_bit_field_1): Fix typo.
10671 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10674 * config/arm/iterators.md (VF): New mode iterator.
10675 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10676 Remove integer-related logic from pattern.
10677 (neon_vabd<mode>_3): Likewise.
10679 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10681 PR middle-end/82694
10682 * common.opt (fstrict-overflow): No longer an alias.
10683 (fwrapv-pointer): New option.
10684 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10685 also for pointer types based on flag_wrapv_pointer.
10686 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10687 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10688 opts->x_flag_wrapv got set.
10689 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10690 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10691 POINTER_TYPE_OVERFLOW_UNDEFINED.
10692 * match.pd: Likewise in address comparison pattern.
10693 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10695 2018-01-15 Richard Biener <rguenther@suse.de>
10698 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10699 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10700 Reset type names to their identifier if their TYPE_DECL doesn't
10701 have linkage (and thus is used for ODR and devirt).
10702 (save_debug_info_for_decl): Remove.
10703 (save_debug_info_for_type): Likewise.
10704 (add_tree_to_fld_list): Adjust.
10705 * tree-pretty-print.c (dump_generic_node): Make dumping of
10706 type names more robust.
10708 2018-01-15 Richard Biener <rguenther@suse.de>
10710 * BASE-VER: Bump to 8.0.1.
10712 2018-01-14 Martin Sebor <msebor@redhat.com>
10715 * builtins.c (check_access): Avoid warning when the no-warning bit
10718 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10720 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10721 * ira-color (allocno_hard_regs_compare): Likewise.
10723 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10726 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10727 Use .pushsection/.popsection.
10729 2018-01-14 Martin Sebor <msebor@redhat.com>
10732 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10734 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10736 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10737 entry from extra_headers.
10738 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10739 extra_headers, make the list bitwise identical to the i?86-*-* one.
10741 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10743 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10744 -mcmodel=large with -mindirect-branch=thunk,
10745 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10746 -mfunction-return=thunk-extern.
10747 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10748 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10749 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10751 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10753 * config/i386/i386.c (print_reg): Print the name of the full
10754 integer register without '%'.
10755 (ix86_print_operand): Handle 'V'.
10756 * doc/extend.texi: Document 'V' modifier.
10758 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10760 * config/i386/constraints.md (Bs): Disallow memory operand for
10761 -mindirect-branch-register.
10763 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10764 (GOT_memory_operand): Likewise.
10765 (call_insn_operand): Likewise.
10766 (sibcall_insn_operand): Likewise.
10767 (GOT32_symbol_operand): Likewise.
10768 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10769 for -mindirect-branch-register.
10770 (tablejump): Likewise.
10771 (*sibcall_memory): Likewise.
10772 (*sibcall_value_memory): Likewise.
10773 Disallow peepholes of indirect call and jump via memory for
10774 -mindirect-branch-register.
10775 (*call_pop): Replace m with Bw.
10776 (*call_value_pop): Likewise.
10777 (*sibcall_pop_memory): Replace m with Bs.
10778 * config/i386/i386.opt (mindirect-branch-register): New option.
10779 * doc/invoke.texi: Document -mindirect-branch-register option.
10781 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10783 * config/i386/i386-protos.h (ix86_output_function_return): New.
10784 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10785 set function_return_type.
10786 (indirect_thunk_name): Add ret_p to indicate thunk for function
10788 (output_indirect_thunk_function): Pass false to
10789 indirect_thunk_name.
10790 (ix86_output_indirect_branch_via_reg): Likewise.
10791 (ix86_output_indirect_branch_via_push): Likewise.
10792 (output_indirect_thunk_function): Create alias for function
10793 return thunk if regno < 0.
10794 (ix86_output_function_return): New function.
10795 (ix86_handle_fndecl_attribute): Handle function_return.
10796 (ix86_attribute_table): Add function_return.
10797 * config/i386/i386.h (machine_function): Add
10798 function_return_type.
10799 * config/i386/i386.md (simple_return_internal): Use
10800 ix86_output_function_return.
10801 (simple_return_internal_long): Likewise.
10802 * config/i386/i386.opt (mfunction-return=): New option.
10803 (indirect_branch): Mention -mfunction-return=.
10804 * doc/extend.texi: Document function_return function attribute.
10805 * doc/invoke.texi: Document -mfunction-return= option.
10807 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10809 * config/i386/i386-opts.h (indirect_branch): New.
10810 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10811 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10812 with local indirect jump when converting indirect call and jump.
10813 (ix86_set_indirect_branch_type): New.
10814 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10815 (indirectlabelno): New.
10816 (indirect_thunk_needed): Likewise.
10817 (indirect_thunk_bnd_needed): Likewise.
10818 (indirect_thunks_used): Likewise.
10819 (indirect_thunks_bnd_used): Likewise.
10820 (INDIRECT_LABEL): Likewise.
10821 (indirect_thunk_name): Likewise.
10822 (output_indirect_thunk): Likewise.
10823 (output_indirect_thunk_function): Likewise.
10824 (ix86_output_indirect_branch_via_reg): Likewise.
10825 (ix86_output_indirect_branch_via_push): Likewise.
10826 (ix86_output_indirect_branch): Likewise.
10827 (ix86_output_indirect_jmp): Likewise.
10828 (ix86_code_end): Call output_indirect_thunk_function if needed.
10829 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10831 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10832 (ix86_attribute_table): Add indirect_branch.
10833 * config/i386/i386.h (machine_function): Add indirect_branch_type
10834 and has_local_indirect_jump.
10835 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10837 (tablejump): Likewise.
10838 (*indirect_jump): Use ix86_output_indirect_jmp.
10839 (*tablejump_1): Likewise.
10840 (simple_return_indirect_internal): Likewise.
10841 * config/i386/i386.opt (mindirect-branch=): New option.
10842 (indirect_branch): New.
10845 (thunk-inline): Likewise.
10846 (thunk-extern): Likewise.
10847 * doc/extend.texi: Document indirect_branch function attribute.
10848 * doc/invoke.texi: Document -mindirect-branch= option.
10850 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10853 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10855 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10857 * ipa-inline.c (want_inline_small_function_p): Return false if
10858 inlining has already failed with CIF_FINAL_ERROR.
10859 (update_caller_keys): Call want_inline_small_function_p before
10861 (update_callee_keys): Likewise.
10863 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10865 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10867 (rs6000_quadword_masked_address_p): Likewise.
10868 (quad_aligned_load_p): Likewise.
10869 (quad_aligned_store_p): Likewise.
10870 (const_load_sequence_p): Add comment to describe the outer-most loop.
10871 (mimic_memory_attributes_and_flags): New function.
10872 (rs6000_gen_stvx): Likewise.
10873 (replace_swapped_aligned_store): Likewise.
10874 (rs6000_gen_lvx): Likewise.
10875 (replace_swapped_aligned_load): Likewise.
10876 (replace_swapped_load_constant): Capitalize argument name in
10877 comment describing this function.
10878 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10879 and stores that access quad-word aligned addresses and replace
10880 with stvx or lvx instructions when appropriate.
10881 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10882 New function prototype.
10883 (rs6000_quadword_masked_address_p): Likewise.
10884 (rs6000_gen_lvx): Likewise.
10885 (rs6000_gen_stvx): Likewise.
10886 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10887 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10888 when memory address is aligned.
10889 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10890 this split to select lvx instruction when memory address is aligned.
10891 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10892 instruction when memory address is aligned.
10893 (*vsx_le_perm_load_v16qi): Likewise.
10894 (four unnamed splitters): Modify to select the stvx instruction
10895 when memory is aligned.
10897 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10899 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10900 which appears in the queue multiple times.
10902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10903 Alan Hayward <alan.hayward@arm.com>
10904 David Sherwood <david.sherwood@arm.com>
10906 * tree-vectorizer.h (vec_lower_bound): New structure.
10907 (_loop_vec_info): Add check_nonzero and lower_bounds.
10908 (LOOP_VINFO_CHECK_NONZERO): New macro.
10909 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10910 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10911 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10912 fields. Make seg_len the distance travelled, not including the
10914 (dr_direction_indicator): Declare.
10915 (dr_zero_step_indicator): Likewise.
10916 (dr_known_forward_stride_p): Likewise.
10917 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10919 (runtime_alias_check_p): Allow runtime alias checks with
10921 (operator ==): Compare access_size and align.
10922 (prune_runtime_alias_test_list): Rework for new distinction between
10923 the access_size and seg_len.
10924 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10926 (get_segment_min_max): New function.
10927 (create_intersect_range_checks): Use it.
10928 (dr_step_indicator): New function.
10929 (dr_direction_indicator): Likewise.
10930 (dr_zero_step_indicator): Likewise.
10931 (dr_known_forward_stride_p): Likewise.
10932 * tree-loop-distribution.c (data_ref_segment_size): Return
10933 DR_STEP * (niters - 1).
10934 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10936 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10937 (vect_preserves_scalar_order_p): New function, split out from...
10938 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10939 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10940 (vect_vfa_access_size): New function.
10941 (vect_vfa_align): Likewise.
10942 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10943 (dump_lower_bound): New function.
10944 (vect_check_lower_bound): Likewise.
10945 (vect_small_gap_p): Likewise.
10946 (vectorizable_with_step_bound_p): Likewise.
10947 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10948 depencies if the vectorization factor is 1. Convert the checks
10949 for nonzero steps into checks on the bounds of DR_STEP. Try using
10950 a bunds check for variable steps if the minimum required step is
10951 relatively small. Update calls to the dr_with_seg_len
10952 constructor and to vect_compile_time_alias.
10953 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10955 (vect_loop_versioning): Call it.
10956 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10958 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10961 Alan Hayward <alan.hayward@arm.com>
10962 David Sherwood <david.sherwood@arm.com>
10964 * doc/sourcebuild.texi (vect_scatter_store): Document.
10965 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10967 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10969 * genopinit.c (main): Add supports_vec_scatter_store and
10970 supports_vec_scatter_store_cached to target_optabs.
10971 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10972 IFN_MASK_SCATTER_STORE.
10973 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10975 * internal-fn.h (internal_store_fn_p): Declare.
10976 (internal_fn_stored_value_index): Likewise.
10977 * internal-fn.c (scatter_store_direct): New macro.
10978 (expand_scatter_store_optab_fn): New function.
10979 (direct_scatter_store_optab_supported_p): New macro.
10980 (internal_store_fn_p): New function.
10981 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10982 IFN_MASK_SCATTER_STORE.
10983 (internal_fn_mask_index): Likewise.
10984 (internal_fn_stored_value_index): New function.
10985 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10986 for scatter stores.
10987 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10988 * optabs-query.c (supports_vec_scatter_store_p): New function.
10989 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10990 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10991 true for scatter stores.
10992 (vect_gather_scatter_fn_p): Handle scatter stores too.
10993 (vect_check_gather_scatter): Consider using scatter stores if
10994 supports_vec_scatter_store_p.
10995 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10996 scatter stores too.
10997 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10998 internal_fn_stored_value_index.
10999 (check_load_store_masking): Handle scatter stores too.
11000 (vect_get_store_rhs): Make public.
11001 (vectorizable_call): Use internal_store_fn_p.
11002 (vectorizable_store): Handle scatter store internal functions.
11003 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11004 when deciding whether the end of the group has been reached.
11005 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11006 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11007 (mask_scatter_store<mode>): New insns.
11009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11010 Alan Hayward <alan.hayward@arm.com>
11011 David Sherwood <david.sherwood@arm.com>
11013 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11014 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11015 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11017 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11018 Use vect_truncate_gather_scatter_offset if we can't treat the
11019 operation as a normal gather load or scatter store.
11020 (get_group_load_store_type): Take the gather_scatter_info
11021 as argument. Try using a gather load or scatter store for
11022 single-element groups.
11023 (get_load_store_type): Update calls to get_group_load_store_type
11024 and vect_use_strided_gather_scatters_p.
11026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11027 Alan Hayward <alan.hayward@arm.com>
11028 David Sherwood <david.sherwood@arm.com>
11030 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11031 optional tree argument.
11032 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11034 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11035 but continue to use the current value as a fallback.
11036 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11037 to compare the updates.
11038 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11039 (get_load_store_type): Use it when handling a strided access.
11040 (vect_get_strided_load_store_ops): New function.
11041 (vect_get_data_ptr_increment): Likewise.
11042 (vectorizable_load): Handle strided gather loads. Always pass
11043 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11046 Alan Hayward <alan.hayward@arm.com>
11047 David Sherwood <david.sherwood@arm.com>
11049 * doc/md.texi (gather_load@var{m}): Document.
11050 (mask_gather_load@var{m}): Likewise.
11051 * genopinit.c (main): Add supports_vec_gather_load and
11052 supports_vec_gather_load_cached to target_optabs.
11053 * optabs-tree.c (init_tree_optimization_optabs): Use
11054 ggc_cleared_alloc to allocate target_optabs.
11055 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11056 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11058 * internal-fn.h (internal_load_fn_p): Declare.
11059 (internal_gather_scatter_fn_p): Likewise.
11060 (internal_fn_mask_index): Likewise.
11061 (internal_gather_scatter_fn_supported_p): Likewise.
11062 * internal-fn.c (gather_load_direct): New macro.
11063 (expand_gather_load_optab_fn): New function.
11064 (direct_gather_load_optab_supported_p): New macro.
11065 (direct_internal_fn_optab): New function.
11066 (internal_load_fn_p): Likewise.
11067 (internal_gather_scatter_fn_p): Likewise.
11068 (internal_fn_mask_index): Likewise.
11069 (internal_gather_scatter_fn_supported_p): Likewise.
11070 * optabs-query.c (supports_at_least_one_mode_p): New function.
11071 (supports_vec_gather_load_p): Likewise.
11072 * optabs-query.h (supports_vec_gather_load_p): Declare.
11073 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11074 and memory_type field.
11075 (NUM_PATTERNS): Bump to 15.
11076 * tree-vect-data-refs.c: Include internal-fn.h.
11077 (vect_gather_scatter_fn_p): New function.
11078 (vect_describe_gather_scatter_call): Likewise.
11079 (vect_check_gather_scatter): Try using internal functions for
11080 gather loads. Recognize existing calls to a gather load function.
11081 (vect_analyze_data_refs): Consider using gather loads if
11082 supports_vec_gather_load_p.
11083 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11084 (vect_get_gather_scatter_offset_type): Likewise.
11085 (vect_convert_mask_for_vectype): Likewise.
11086 (vect_add_conversion_to_patterm): Likewise.
11087 (vect_try_gather_scatter_pattern): Likewise.
11088 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11089 (vect_vect_recog_func_ptrs): Add it.
11090 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11091 internal_fn_mask_index and internal_gather_scatter_fn_p.
11092 (check_load_store_masking): Take the gather_scatter_info as an
11093 argument and handle gather loads.
11094 (vect_get_gather_scatter_ops): New function.
11095 (vectorizable_call): Check internal_load_fn_p.
11096 (vectorizable_load): Likewise. Handle gather load internal
11098 (vectorizable_store): Update call to check_load_store_masking.
11099 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11100 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11101 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11102 (aarch64_gather_scale_operand_d): New predicates.
11103 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11104 (mask_gather_load<mode>): New insns.
11106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11107 Alan Hayward <alan.hayward@arm.com>
11108 David Sherwood <david.sherwood@arm.com>
11110 * optabs.def (fold_left_plus_optab): New optab.
11111 * doc/md.texi (fold_left_plus_@var{m}): Document.
11112 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11113 * internal-fn.c (fold_left_direct): Define.
11114 (expand_fold_left_optab_fn): Likewise.
11115 (direct_fold_left_optab_supported_p): Likewise.
11116 * fold-const-call.c (fold_const_fold_left): New function.
11117 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11118 * tree-parloops.c (valid_reduction_p): New function.
11119 (gather_scalar_reductions): Use it.
11120 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11121 (vect_finish_replace_stmt): Declare.
11122 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11123 (needs_fold_left_reduction_p): New function, split out from...
11124 (vect_is_simple_reduction): ...here. Accept reductions that
11125 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11126 (vect_force_simple_reduction): Also store the reduction type in
11127 the assignment's STMT_VINFO_REDUC_TYPE.
11128 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11129 (merge_with_identity): New function.
11130 (vect_expand_fold_left): Likewise.
11131 (vectorize_fold_left_reduction): Likewise.
11132 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11133 scalar phi in place for it. Check for target support and reject
11134 cases that would reassociate the operation. Defer the transform
11135 phase to vectorize_fold_left_reduction.
11136 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11137 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11138 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11142 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11143 call to ifc_temp_var.
11145 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11146 Alan Hayward <alan.hayward@arm.com>
11147 David Sherwood <david.sherwood@arm.com>
11149 * target.def (legitimize_address_displacement): Take the original
11150 offset as a poly_int.
11151 * targhooks.h (default_legitimize_address_displacement): Update
11153 * targhooks.c (default_legitimize_address_displacement): Likewise.
11154 * doc/tm.texi: Regenerate.
11155 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11156 as an argument, moving assert of ad->disp == ad->disp_term to...
11157 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11158 Try calling targetm.legitimize_address_displacement before expanding
11159 the address rather than afterwards, and adjust for the new interface.
11160 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11161 Match the new hook interface. Handle SVE addresses.
11162 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11163 new hook interface.
11165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11167 * Makefile.in (OBJS): Add early-remat.o.
11168 * target.def (select_early_remat_modes): New hook.
11169 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11170 * doc/tm.texi: Regenerate.
11171 * targhooks.h (default_select_early_remat_modes): Declare.
11172 * targhooks.c (default_select_early_remat_modes): New function.
11173 * timevar.def (TV_EARLY_REMAT): New timevar.
11174 * passes.def (pass_early_remat): New pass.
11175 * tree-pass.h (make_pass_early_remat): Declare.
11176 * early-remat.c: New file.
11177 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11179 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11182 Alan Hayward <alan.hayward@arm.com>
11183 David Sherwood <david.sherwood@arm.com>
11185 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11186 vfm1 with a bound_epilog parameter.
11187 (vect_do_peeling): Update calls accordingly, and move the prologue
11188 call earlier in the function. Treat the base bound_epilog as 0 for
11189 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11190 this base when peeling for gaps.
11191 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11192 with fully-masked loops.
11193 (vect_estimate_min_profitable_iters): Handle the single peeled
11194 iteration in that case.
11196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11197 Alan Hayward <alan.hayward@arm.com>
11198 David Sherwood <david.sherwood@arm.com>
11200 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11201 single-element interleaving even if the size is not a power of 2.
11202 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11203 accesses for single-element interleaving if the group size is
11206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11207 Alan Hayward <alan.hayward@arm.com>
11208 David Sherwood <david.sherwood@arm.com>
11210 * doc/md.texi (fold_extract_last_@var{m}): Document.
11211 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11212 * optabs.def (fold_extract_last_optab): New optab.
11213 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11214 * internal-fn.c (fold_extract_direct): New macro.
11215 (expand_fold_extract_optab_fn): Likewise.
11216 (direct_fold_extract_optab_supported_p): Likewise.
11217 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11218 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11219 EXTRACT_LAST_REDUCTION.
11220 (get_initial_def_for_reduction): Do not create an initial vector
11221 for EXTRACT_LAST_REDUCTION reductions.
11222 (vectorizable_reduction): Leave the scalar phi in place for
11223 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11224 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11225 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11226 transform phase to vectorizable_condition.
11227 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11229 (vect_finish_stmt_generation): ...here.
11230 (vect_finish_replace_stmt): New function.
11231 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11232 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11234 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11237 Alan Hayward <alan.hayward@arm.com>
11238 David Sherwood <david.sherwood@arm.com>
11240 * doc/md.texi (extract_last_@var{m}): Document.
11241 * optabs.def (extract_last_optab): New optab.
11242 * internal-fn.def (EXTRACT_LAST): New internal function.
11243 * internal-fn.c (cond_unary_direct): New macro.
11244 (expand_cond_unary_optab_fn): Likewise.
11245 (direct_cond_unary_optab_supported_p): Likewise.
11246 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11247 loops using EXTRACT_LAST.
11248 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11249 (extract_last_<mode>): ...this optab.
11250 (vec_extract<mode><Vel>): Update accordingly.
11252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11253 Alan Hayward <alan.hayward@arm.com>
11254 David Sherwood <david.sherwood@arm.com>
11256 * target.def (empty_mask_is_expensive): New hook.
11257 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11258 * doc/tm.texi: Regenerate.
11259 * targhooks.h (default_empty_mask_is_expensive): Declare.
11260 * targhooks.c (default_empty_mask_is_expensive): New function.
11261 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11262 if the target says that empty masks are expensive.
11263 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11265 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11268 Alan Hayward <alan.hayward@arm.com>
11269 David Sherwood <david.sherwood@arm.com>
11271 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11272 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11273 (vect_use_loop_mask_for_alignment_p): New function.
11274 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11275 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11276 niters_skip argument. Make sure that the first niters_skip elements
11277 of the first iteration are inactive.
11278 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11279 Update call to vect_set_loop_masks_directly.
11280 (get_misalign_in_elems): New function, split out from...
11281 (vect_gen_prolog_loop_niters): ...here.
11282 (vect_update_init_of_dr): Take a code argument that specifies whether
11283 the adjustment should be added or subtracted.
11284 (vect_update_init_of_drs): Likewise.
11285 (vect_prepare_for_masked_peels): New function.
11286 (vect_do_peeling): Skip prologue peeling if we're using a mask
11287 instead. Update call to vect_update_inits_of_drs.
11288 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11290 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11291 alignment. Do not include the number of peeled iterations in
11292 the minimum threshold in that case.
11293 (vectorizable_induction): Adjust the start value down by
11294 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11295 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11296 Take the number of skipped iterations into account when calculating
11298 * tree-vect-stmts.c (vect_gen_while_not): New function.
11300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11301 Alan Hayward <alan.hayward@arm.com>
11302 David Sherwood <david.sherwood@arm.com>
11304 * doc/sourcebuild.texi (vect_fully_masked): Document.
11305 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11306 default value to 0.
11307 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11309 (vect_analyze_loop_2): ...here. Don't check the vectorization
11310 factor against the number of loop iterations if the loop is
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11318 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11319 (dump_groups): Update accordingly.
11320 (iv_use::mem_type): New member variable.
11321 (address_p): New function.
11322 (record_use): Add a mem_type argument and initialize the new
11324 (record_group_use): Add a mem_type argument. Use address_p.
11325 Remove obsolete null checks of base_object. Update call to record_use.
11326 (find_interesting_uses_op): Update call to record_group_use.
11327 (find_interesting_uses_cond): Likewise.
11328 (find_interesting_uses_address): Likewise.
11329 (get_mem_type_for_internal_fn): New function.
11330 (find_address_like_use): Likewise.
11331 (find_interesting_uses_stmt): Try find_address_like_use before
11332 calling find_interesting_uses_op.
11333 (addr_offset_valid_p): Use the iv mem_type field as the type
11334 of the addressed memory.
11335 (add_autoinc_candidates): Likewise.
11336 (get_address_cost): Likewise.
11337 (split_small_address_groups_p): Use address_p.
11338 (split_address_groups): Likewise.
11339 (add_iv_candidate_for_use): Likewise.
11340 (autoinc_possible_for_pair): Likewise.
11341 (rewrite_groups): Likewise.
11342 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11343 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11344 (get_alias_ptr_type_for_ptr_address): New function.
11345 (rewrite_use_address): Rewrite address uses in calls that were
11346 identified by find_address_like_use.
11348 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11349 Alan Hayward <alan.hayward@arm.com>
11350 David Sherwood <david.sherwood@arm.com>
11352 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11354 * gimple-expr.h (is_gimple_addressable: Likewise.
11355 * gimple-expr.c (is_gimple_address): Likewise.
11356 * internal-fn.c (expand_call_mem_ref): New function.
11357 (expand_mask_load_optab_fn): Use it.
11358 (expand_mask_store_optab_fn): Likewise.
11360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11361 Alan Hayward <alan.hayward@arm.com>
11362 David Sherwood <david.sherwood@arm.com>
11364 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11365 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11366 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11367 (cond_umax@var{mode}): Document.
11368 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11369 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11370 (cond_umin_optab, cond_umax_optab): New optabs.
11371 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11372 (COND_IOR, COND_XOR): New internal functions.
11373 * internal-fn.h (get_conditional_internal_fn): Declare.
11374 * internal-fn.c (cond_binary_direct): New macro.
11375 (expand_cond_binary_optab_fn): Likewise.
11376 (direct_cond_binary_optab_supported_p): Likewise.
11377 (get_conditional_internal_fn): New function.
11378 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11379 Cope with reduction statements that are vectorized as calls rather
11381 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11382 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11383 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11384 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11385 (UNSPEC_COND_EOR): New unspecs.
11386 (optab): Add mappings for them.
11387 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11388 (sve_int_op, sve_fp_op): New int attributes.
11390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11391 Alan Hayward <alan.hayward@arm.com>
11392 David Sherwood <david.sherwood@arm.com>
11394 * optabs.def (while_ult_optab): New optab.
11395 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11396 * internal-fn.def (WHILE_ULT): New internal function.
11397 * internal-fn.h (direct_internal_fn_supported_p): New override
11398 that takes two types as argument.
11399 * internal-fn.c (while_direct): New macro.
11400 (expand_while_optab_fn): New function.
11401 (convert_optab_supported_p): Likewise.
11402 (direct_while_optab_supported_p): New macro.
11403 * wide-int.h (wi::udiv_ceil): New function.
11404 * tree-vectorizer.h (rgroup_masks): New structure.
11405 (vec_loop_masks): New typedef.
11406 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11407 and fully_masked_p.
11408 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11409 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11410 (vect_max_vf): New function.
11411 (slpeel_make_loop_iterate_ntimes): Delete.
11412 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11413 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11414 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11415 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11416 internal-fn.h, stor-layout.h and optabs-query.h.
11417 (vect_set_loop_mask): New function.
11418 (add_preheader_seq): Likewise.
11419 (add_header_seq): Likewise.
11420 (interleave_supported_p): Likewise.
11421 (vect_maybe_permute_loop_masks): Likewise.
11422 (vect_set_loop_masks_directly): Likewise.
11423 (vect_set_loop_condition_masked): Likewise.
11424 (vect_set_loop_condition_unmasked): New function, split out from
11425 slpeel_make_loop_iterate_ntimes.
11426 (slpeel_make_loop_iterate_ntimes): Rename to..
11427 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11428 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11429 (vect_do_peeling): Update call accordingly.
11430 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11432 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11433 mask_compare_type, can_fully_mask_p and fully_masked_p.
11434 (release_vec_loop_masks): New function.
11435 (_loop_vec_info): Use it to free the loop masks.
11436 (can_produce_all_loop_masks_p): New function.
11437 (vect_get_max_nscalars_per_iter): Likewise.
11438 (vect_verify_full_masking): Likewise.
11439 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11440 retries, and free the mask rgroups before retrying. Check loop-wide
11441 reasons for disallowing fully-masked loops. Make the final decision
11442 about whether use a fully-masked loop or not.
11443 (vect_estimate_min_profitable_iters): Do not assume that peeling
11444 for the number of iterations will be needed for fully-masked loops.
11445 (vectorizable_reduction): Disable fully-masked loops.
11446 (vectorizable_live_operation): Likewise.
11447 (vect_halve_mask_nunits): New function.
11448 (vect_double_mask_nunits): Likewise.
11449 (vect_record_loop_mask): Likewise.
11450 (vect_get_loop_mask): Likewise.
11451 (vect_transform_loop): Handle the case in which the final loop
11452 iteration might handle a partial vector. Call vect_set_loop_condition
11453 instead of slpeel_make_loop_iterate_ntimes.
11454 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11455 (check_load_store_masking): New function.
11456 (prepare_load_store_mask): Likewise.
11457 (vectorizable_store): Handle fully-masked loops.
11458 (vectorizable_load): Likewise.
11459 (supportable_widening_operation): Use vect_halve_mask_nunits for
11461 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11462 (vect_gen_while): New function.
11463 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11464 (aarch64_uqdec<mode>): New insn.
11466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11467 Alan Hayward <alan.hayward@arm.com>
11468 David Sherwood <david.sherwood@arm.com>
11470 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11471 (reduc_xor_scal_optab): New optabs.
11472 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11473 (reduc_xor_scal_@var{m}): Document.
11474 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11475 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11476 internal functions.
11477 * fold-const-call.c (fold_const_call): Handle them.
11478 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11479 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11480 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11481 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11482 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11483 (UNSPEC_XORV): New unspecs.
11484 (optab): Add entries for them.
11485 (BITWISEV): New int iterator.
11486 (bit_reduc_op): New int attributes.
11488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11489 Alan Hayward <alan.hayward@arm.com>
11490 David Sherwood <david.sherwood@arm.com>
11492 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11493 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11494 * optabs.def (vec_shl_insert_optab): New optab.
11495 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11496 (duplicate_and_interleave): Likewise.
11497 * tree-vect-loop.c: Include internal-fn.h.
11498 (neutral_op_for_slp_reduction): New function, split out from
11499 get_initial_defs_for_reduction.
11500 (get_initial_def_for_reduction): Handle option 2 for variable-length
11501 vectors by loading the neutral value into a vector and then shifting
11502 the initial value into element 0.
11503 (get_initial_defs_for_reduction): Replace the code argument with
11504 the neutral value calculated by neutral_op_for_slp_reduction.
11505 Use gimple_build_vector for constant-length vectors.
11506 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11507 but the first group_size elements have a neutral value.
11508 Use duplicate_and_interleave otherwise.
11509 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11510 Update call to get_initial_defs_for_reduction. Handle SLP
11511 reductions for variable-length vectors by creating one vector
11512 result for each scalar result, with the elements associated
11513 with other scalar results stubbed out with the neutral value.
11514 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11515 Require IFN_VEC_SHL_INSERT for double reductions on
11516 variable-length vectors, or SLP reductions that have
11517 a neutral value. Require can_duplicate_and_interleave_p
11518 support for variable-length unchained SLP reductions if there
11519 is no neutral value, such as for MIN/MAX reductions. Also require
11520 the number of vector elements to be a multiple of the number of
11521 SLP statements when doing variable-length unchained SLP reductions.
11522 Update call to vect_create_epilog_for_reduction.
11523 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11524 and remove initial values.
11525 (duplicate_and_interleave): Make public.
11526 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11527 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11530 Alan Hayward <alan.hayward@arm.com>
11531 David Sherwood <david.sherwood@arm.com>
11533 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11534 (can_duplicate_and_interleave_p): New function.
11535 (vect_get_and_check_slp_defs): Take the vector of statements
11536 rather than just the current one. Remove excess parentheses.
11537 Restriction rejectinon of vect_constant_def and vect_external_def
11538 for variable-length vectors to boolean types, or types for which
11539 can_duplicate_and_interleave_p is false.
11540 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11541 (duplicate_and_interleave): New function.
11542 (vect_get_constant_vectors): Use gimple_build_vector for
11543 constant-length vectors and suitable variable-length constant
11544 vectors. Use duplicate_and_interleave for other variable-length
11545 vectors. Don't defer the update when inserting new statements.
11547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11548 Alan Hayward <alan.hayward@arm.com>
11549 David Sherwood <david.sherwood@arm.com>
11551 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11552 min_profitable_iters doesn't go negative.
11554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11555 Alan Hayward <alan.hayward@arm.com>
11556 David Sherwood <david.sherwood@arm.com>
11558 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11559 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11560 * optabs.def (vec_mask_load_lanes_optab): New optab.
11561 (vec_mask_store_lanes_optab): Likewise.
11562 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11563 (MASK_STORE_LANES): Likewise.
11564 * internal-fn.c (mask_load_lanes_direct): New macro.
11565 (mask_store_lanes_direct): Likewise.
11566 (expand_mask_load_optab_fn): Handle masked operations.
11567 (expand_mask_load_lanes_optab_fn): New macro.
11568 (expand_mask_store_optab_fn): Handle masked operations.
11569 (expand_mask_store_lanes_optab_fn): New macro.
11570 (direct_mask_load_lanes_optab_supported_p): Likewise.
11571 (direct_mask_store_lanes_optab_supported_p): Likewise.
11572 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11574 (vect_load_lanes_supported): Likewise.
11575 * tree-vect-data-refs.c (strip_conversion): New function.
11576 (can_group_stmts_p): Likewise.
11577 (vect_analyze_data_ref_accesses): Use it instead of checking
11578 for a pair of assignments.
11579 (vect_store_lanes_supported): Take a masked_p parameter.
11580 (vect_load_lanes_supported): Likewise.
11581 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11582 vect_store_lanes_supported and vect_load_lanes_supported.
11583 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11584 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11585 parameter. Don't allow gaps for masked accesses.
11586 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11587 and vect_load_lanes_supported.
11588 (get_load_store_type): Take a masked_p parameter and update
11589 call to get_group_load_store_type.
11590 (vectorizable_store): Update call to get_load_store_type.
11591 Handle IFN_MASK_STORE_LANES.
11592 (vectorizable_load): Update call to get_load_store_type.
11593 Handle IFN_MASK_LOAD_LANES.
11595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11596 Alan Hayward <alan.hayward@arm.com>
11597 David Sherwood <david.sherwood@arm.com>
11599 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11601 * config/aarch64/aarch64-protos.h
11602 (aarch64_sve_struct_memory_operand_p): Declare.
11603 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11604 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11605 (VPRED, vpred): Handle SVE structure modes.
11606 * config/aarch64/constraints.md (Utx): New constraint.
11607 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11608 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11609 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11610 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11611 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11612 structure modes. Split into pieces after RA.
11613 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11614 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11616 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11617 SVE structure modes.
11618 (aarch64_classify_address): Likewise.
11619 (sizetochar): Move earlier in file.
11620 (aarch64_print_operand): Handle SVE register lists.
11621 (aarch64_array_mode): New function.
11622 (aarch64_sve_struct_memory_operand_p): Likewise.
11623 (TARGET_ARRAY_MODE): Redefine.
11625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11626 Alan Hayward <alan.hayward@arm.com>
11627 David Sherwood <david.sherwood@arm.com>
11629 * target.def (array_mode): New target hook.
11630 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11631 * doc/tm.texi: Regenerate.
11632 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11633 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11634 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11635 targetm.array_mode.
11636 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11639 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11640 Alan Hayward <alan.hayward@arm.com>
11641 David Sherwood <david.sherwood@arm.com>
11643 * fold-const.c (fold_binary_loc): Check the argument types
11644 rather than the result type when testing for a vector operation.
11646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11648 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11649 * doc/tm.texi: Regenerate.
11651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11652 Alan Hayward <alan.hayward@arm.com>
11653 David Sherwood <david.sherwood@arm.com>
11655 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11656 (sve): Document new AArch64 extension.
11657 * doc/md.texi (w): Extend the description of the AArch64
11658 constraint to include SVE vectors.
11659 (Upl, Upa): Document new AArch64 predicate constraints.
11660 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11662 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11663 (msve-vector-bits=): New option.
11664 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11665 SVE when these are disabled.
11666 (sve): New extension.
11667 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11668 modes. Adjust their number of units based on aarch64_sve_vg.
11669 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11670 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11671 aarch64_addr_query_type.
11672 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11673 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11674 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11675 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11676 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11677 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11678 (aarch64_simd_imm_zero_p): Delete.
11679 (aarch64_check_zero_based_sve_index_immediate): Declare.
11680 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11681 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11682 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11683 (aarch64_sve_float_mul_immediate_p): Likewise.
11684 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11685 rather than an rtx.
11686 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11687 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11688 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11689 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11690 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11691 (aarch64_regmode_natural_size): Likewise.
11692 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11693 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11695 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11696 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11697 for VG and the SVE predicate registers.
11698 (V_ALIASES): Add a "z"-prefixed alias.
11699 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11700 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11701 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11702 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11703 (REG_CLASS_NAMES): Add entries for them.
11704 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11705 and the predicate registers.
11706 (aarch64_sve_vg): Declare.
11707 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11708 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11709 (REGMODE_NATURAL_SIZE): Define.
11710 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11712 * config/aarch64/aarch64.c: Include cfgrtl.h.
11713 (simd_immediate_info): Add a constructor for series vectors,
11714 and an associated step field.
11715 (aarch64_sve_vg): New variable.
11716 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11717 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11718 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11719 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11720 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11721 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11722 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11723 (aarch64_get_mask_mode): New functions.
11724 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11725 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11726 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11727 predicate modes and predicate registers. Explicitly restrict
11728 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11729 to store a vector mode if it is recognized by
11730 aarch64_classify_vector_mode.
11731 (aarch64_regmode_natural_size): New function.
11732 (aarch64_hard_regno_caller_save_mode): Return the original mode
11734 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11735 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11736 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11737 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11739 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11740 does not overlap dest if the function is frame-related. Handle
11742 (aarch64_split_add_offset): New function.
11743 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11744 them aarch64_add_offset.
11745 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11746 and update call to aarch64_sub_sp.
11747 (aarch64_add_cfa_expression): New function.
11748 (aarch64_expand_prologue): Pass extra temporary registers to the
11749 functions above. Handle the case in which we need to emit new
11750 DW_CFA_expressions for registers that were originally saved
11751 relative to the stack pointer, but now have to be expressed
11752 relative to the frame pointer.
11753 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11755 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11756 IP0 and IP1 values for SVE frames.
11757 (aarch64_expand_vec_series): New function.
11758 (aarch64_expand_sve_widened_duplicate): Likewise.
11759 (aarch64_expand_sve_const_vector): Likewise.
11760 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11761 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11762 into the register, rather than emitting a SET directly.
11763 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11764 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11765 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11766 (offset_9bit_signed_scaled_p): New functions.
11767 (aarch64_replicate_bitmask_imm): New function.
11768 (aarch64_bitmask_imm): Use it.
11769 (aarch64_cannot_force_const_mem): Reject expressions involving
11770 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11771 (aarch64_classify_index): Handle SVE indices, by requiring
11772 a plain register index with a scale that matches the element size.
11773 (aarch64_classify_address): Handle SVE addresses. Assert that
11774 the mode of the address is VOIDmode or an integer mode.
11775 Update call to aarch64_classify_symbol.
11776 (aarch64_classify_symbolic_expression): Update call to
11777 aarch64_classify_symbol.
11778 (aarch64_const_vec_all_in_range_p): New function.
11779 (aarch64_print_vector_float_operand): Likewise.
11780 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11781 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11782 and the FP immediates 1.0 and 0.5.
11783 (aarch64_print_address_internal): Handle SVE addresses.
11784 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11785 (aarch64_regno_regclass): Handle predicate registers.
11786 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11788 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11789 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11790 (aarch64_convert_sve_vector_bits): New function.
11791 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11792 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11793 rather than an rtx.
11794 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11795 Handle SVE vector and predicate modes. Accept VL-based constants
11796 that need only one temporary register, and VL offsets that require
11797 no temporary registers.
11798 (aarch64_conditional_register_usage): Mark the predicate registers
11799 as fixed if SVE isn't available.
11800 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11801 Return true for SVE vector and predicate modes.
11802 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11803 rather than an unsigned int. Handle SVE modes.
11804 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11806 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11808 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11809 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11810 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11811 (aarch64_sve_float_mul_immediate_p): New functions.
11812 (aarch64_sve_valid_immediate): New function.
11813 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11814 Explicitly reject structure modes. Check for INDEX constants.
11815 Handle PTRUE and PFALSE constants.
11816 (aarch64_check_zero_based_sve_index_immediate): New function.
11817 (aarch64_simd_imm_zero_p): Delete.
11818 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11819 vector modes. Accept constants in the range of CNT[BHWD].
11820 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11821 ask for an Advanced SIMD mode.
11822 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11823 (aarch64_simd_vector_alignment): Handle SVE predicates.
11824 (aarch64_vectorize_preferred_vector_alignment): New function.
11825 (aarch64_simd_vector_alignment_reachable): Use it instead of
11827 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11828 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11830 (MAX_VECT_LEN): Delete.
11831 (expand_vec_perm_d): Add a vec_flags field.
11832 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11833 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11834 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11836 (aarch64_evpc_rev): Rename to...
11837 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11838 (aarch64_evpc_rev_global): New function.
11839 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11840 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11842 (aarch64_evpc_sve_tbl): New function.
11843 (aarch64_expand_vec_perm_const_1): Update after rename of
11844 aarch64_evpc_rev. Handle SVE permutes too, trying
11845 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11846 than aarch64_evpc_tbl.
11847 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11848 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11849 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11850 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11851 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11852 (aarch64_expand_sve_vcond): New functions.
11853 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11854 of aarch64_vector_mode_p.
11855 (aarch64_dwarf_poly_indeterminate_value): New function.
11856 (aarch64_compute_pressure_classes): Likewise.
11857 (aarch64_can_change_mode_class): Likewise.
11858 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11859 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11860 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11861 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11862 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11863 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11864 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11865 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11867 (Dn, Dl, Dr): Accept const as well as const_vector.
11868 (Dz): Likewise. Compare against CONST0_RTX.
11869 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11870 of "vector" where appropriate.
11871 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11872 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11873 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11874 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11875 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11876 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11877 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11878 (v_int_equiv): Extend to SVE modes.
11879 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11881 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11882 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11883 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11884 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11885 (SVE_COND_FP_CMP): New int iterators.
11886 (perm_hilo): Handle the new unpack unspecs.
11887 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11889 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11890 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11891 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11892 (aarch64_equality_operator, aarch64_constant_vector_operand)
11893 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11894 (aarch64_sve_nonimmediate_operand): Likewise.
11895 (aarch64_sve_general_operand): Likewise.
11896 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11897 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11898 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11899 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11900 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11901 (aarch64_sve_float_arith_immediate): Likewise.
11902 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11903 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11904 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11905 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11906 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11907 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11908 (aarch64_sve_float_arith_operand): Likewise.
11909 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11910 (aarch64_sve_float_mul_operand): Likewise.
11911 (aarch64_sve_vec_perm_operand): Likewise.
11912 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11913 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11914 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11915 as well as const_vector.
11916 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11917 in file. Use CONST0_RTX and CONSTM1_RTX.
11918 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11919 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11920 Use aarch64_simd_imm_zero.
11921 * config/aarch64/aarch64-sve.md: New file.
11922 * config/aarch64/aarch64.md: Include it.
11923 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11924 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11925 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11926 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11927 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11928 (sve): New attribute.
11929 (enabled): Disable instructions with the sve attribute unless
11931 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11932 aarch64_expand_mov_immediate.
11933 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11934 CNT[BHSD] immediates.
11935 (movti): Split CONST_POLY_INT moves into two halves.
11936 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11937 Split additions that need a temporary here if the destination
11938 is the stack pointer.
11939 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11940 (*add<mode>3_poly_1): New instruction.
11941 (set_clobber_cc): New expander.
11943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11945 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11946 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11947 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11948 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11949 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11950 Change innermode from fixed_mode_size to machine_mode.
11951 (simplify_subreg): Update call accordingly. Handle a constant-sized
11952 subreg of a variable-length CONST_VECTOR.
11954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11955 Alan Hayward <alan.hayward@arm.com>
11956 David Sherwood <david.sherwood@arm.com>
11958 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11959 (add_offset_to_base): New function, split out from...
11960 (create_mem_ref): ...here. When handling a scale other than 1,
11961 check first whether the address is valid without the offset.
11962 Add it into the base if so, leaving the index and scale as-is.
11964 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11967 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11968 fold_for_warn before checking if arg2 is INTEGER_CST.
11970 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11972 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11973 (store_multiple_operation): Delete.
11974 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11975 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11976 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11977 guarded by TARGET_STRING.
11978 (rs6000_output_load_multiple): Delete.
11979 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11980 OPTION_MASK_STRING / TARGET_STRING handling.
11981 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11982 (const rs6000_opt_masks) <"string">: Change mask to 0.
11983 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11984 (MASK_STRING): Delete.
11985 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11987 (load_multiple): Delete.
11994 (store_multiple): Delete.
12001 (movmemsi_8reg): Delete.
12002 (corresponding unnamed define_insn): Delete.
12003 (movmemsi_6reg): Delete.
12004 (corresponding unnamed define_insn): Delete.
12005 (movmemsi_4reg): Delete.
12006 (corresponding unnamed define_insn): Delete.
12007 (movmemsi_2reg): Delete.
12008 (corresponding unnamed define_insn): Delete.
12009 (movmemsi_1reg): Delete.
12010 (corresponding unnamed define_insn): Delete.
12011 * config/rs6000/rs6000.opt (mno-string): New.
12012 (mstring): Replace by deprecation warning stub.
12013 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12015 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12017 * regrename.c (regrename_do_replace): If replacing the same
12018 reg multiple times, try to reuse last created gen_raw_REG.
12021 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12022 main to workaround a bug in GDB.
12024 2018-01-12 Tom de Vries <tom@codesourcery.com>
12027 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12029 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12031 PR rtl-optimization/80481
12032 * ira-color.c (get_cap_member): New function.
12033 (allocnos_conflict_by_live_ranges_p): Use it.
12034 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12035 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12037 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12040 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12041 (*saddl_se_1): Ditto.
12042 (*ssubsi_1): Ditto.
12043 (*ssubl_se_1): Ditto.
12045 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12047 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12048 rather than wi::to_widest for DR_INITs.
12049 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12050 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12051 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12053 (vect_analyze_group_access_1): Note that here.
12055 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12057 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12058 polynomial type sizes.
12060 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12062 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12063 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12064 (gimple_add_tmp_var): Likewise.
12066 2018-01-12 Martin Liska <mliska@suse.cz>
12068 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12069 (gimple_alloc_sizes): Likewise.
12070 (dump_gimple_statistics): Use PRIu64 in printf format.
12071 * gimple.h: Change uint64_t to int.
12073 2018-01-12 Martin Liska <mliska@suse.cz>
12075 * tree-core.h: Use uint64_t instead of int.
12076 * tree.c (tree_node_counts): Likewise.
12077 (tree_node_sizes): Likewise.
12078 (dump_tree_statistics): Use PRIu64 in printf format.
12080 2018-01-12 Martin Liska <mliska@suse.cz>
12082 * Makefile.in: As qsort_chk is implemented in vec.c, add
12083 vec.o to linkage of gencfn-macros.
12084 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12085 passing the info to record_node_allocation_statistics.
12086 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12088 * ggc-common.c (struct ggc_usage): Add operator== and use
12089 it in operator< and compare function.
12090 * mem-stats.h (struct mem_usage): Likewise.
12091 * vec.c (struct vec_usage): Remove operator< and compare
12092 function. Can be simply inherited.
12094 2018-01-12 Martin Jambor <mjambor@suse.cz>
12097 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12098 * tree-ssa-math-opts.c: Include domwalk.h.
12099 (convert_mult_to_fma_1): New function.
12100 (fma_transformation_info): New type.
12101 (fma_deferring_state): Likewise.
12102 (cancel_fma_deferring): New function.
12103 (result_of_phi): Likewise.
12104 (last_fma_candidate_feeds_initial_phi): Likewise.
12105 (convert_mult_to_fma): Added deferring logic, split actual
12106 transformation to convert_mult_to_fma_1.
12107 (math_opts_dom_walker): New type.
12108 (math_opts_dom_walker::after_dom_children): New method, body moved
12109 here from pass_optimize_widening_mul::execute, added deferring logic
12111 (pass_optimize_widening_mul::execute): Moved most of code to
12112 math_opts_dom_walker::after_dom_children.
12113 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12114 * config/i386/i386.c (ix86_option_override_internal): Added
12115 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12117 2018-01-12 Richard Biener <rguenther@suse.de>
12120 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12121 inline instance vars.
12123 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12126 * config/rx/rx.c (rx_is_restricted_memory_address):
12127 Handle SUBREG case.
12129 2018-01-12 Richard Biener <rguenther@suse.de>
12131 PR tree-optimization/80846
12132 * target.def (split_reduction): New target hook.
12133 * targhooks.c (default_split_reduction): New function.
12134 * targhooks.h (default_split_reduction): Declare.
12135 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12136 target requests first reduce vectors by combining low and high
12138 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12139 (get_vectype_for_scalar_type_and_size): Export.
12140 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12141 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12142 * doc/tm.texi: Regenerate.
12143 * config/i386/i386.c (ix86_split_reduction): Implement
12144 TARGET_VECTORIZE_SPLIT_REDUCTION.
12146 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12149 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12150 in PIC mode except for TARGET_VXWORKS_RTP.
12151 * config/sparc/sparc.c: Include cfgrtl.h.
12152 (TARGET_INIT_PIC_REG): Define.
12153 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12154 (sparc_pic_register_p): New predicate.
12155 (sparc_legitimate_address_p): Use it.
12156 (sparc_legitimize_pic_address): Likewise.
12157 (sparc_delegitimize_address): Likewise.
12158 (sparc_mode_dependent_address_p): Likewise.
12159 (gen_load_pcrel_sym): Remove 4th parameter.
12160 (load_got_register): Adjust call to above. Remove obsolete stuff.
12161 (sparc_expand_prologue): Do not call load_got_register here.
12162 (sparc_flat_expand_prologue): Likewise.
12163 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12164 (sparc_use_pseudo_pic_reg): New function.
12165 (sparc_init_pic_reg): Likewise.
12166 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12167 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12169 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12171 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12172 Add item for branch_cost.
12174 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12176 PR rtl-optimization/83565
12177 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12178 not extend the result to a larger mode for rotate operations.
12179 (num_sign_bit_copies1): Likewise.
12181 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12184 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12186 Use values-Xc.o for -pedantic.
12187 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12189 2018-01-12 Martin Liska <mliska@suse.cz>
12192 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12194 (possible_polymorphic_call_targets): Use it.
12195 (ipa_devirt): Likewise.
12197 2018-01-12 Martin Liska <mliska@suse.cz>
12199 * profile-count.h (enum profile_quality): Use 0 as invalid
12200 enum value of profile_quality.
12202 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12204 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12205 -mext-string options.
12207 2018-01-12 Richard Biener <rguenther@suse.de>
12209 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12210 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12211 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12213 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12215 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12217 * configure.ac (--with-long-double-format): Add support for the
12218 configuration option to change the default long double format on
12220 * config.gcc (powerpc*-linux*-*): Likewise.
12221 * configure: Regenerate.
12222 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12223 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12224 used without modification.
12226 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12228 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12229 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12230 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12231 MISC_BUILTIN_SPEC_BARRIER.
12232 (rs6000_init_builtins): Likewise.
12233 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12235 (speculation_barrier): New define_insn.
12236 * doc/extend.texi: Document __builtin_speculation_barrier.
12238 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12241 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12242 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12243 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12245 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12246 integral modes instead of "ss" and "sd".
12247 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12248 vectors with 32-bit and 64-bit elements.
12249 (vecdupssescalarmodesuffix): New mode attribute.
12250 (vec_dup<mode>): Use it.
12252 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12255 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12256 frame if argument is passed on stack.
12258 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12261 * ree.c (combine_reaching_defs): Optimize also
12262 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12263 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12265 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12267 PR middle-end/83189
12268 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12270 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12272 PR middle-end/83718
12273 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12274 after they are computed.
12276 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12278 PR tree-optimization/83695
12279 * gimple-loop-linterchange.cc
12280 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12281 reset cached scev information after interchange.
12282 (pass_linterchange::execute): Remove call to scev_reset_htab.
12284 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12286 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12287 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12288 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12289 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12290 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12291 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12292 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12293 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12294 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12295 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12296 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12297 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12298 (V_lane_reg): Likewise.
12299 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12301 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12302 (vfmal_lane_low<mode>_intrinsic,
12303 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12304 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12305 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12306 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12307 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12308 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12310 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12312 * config/arm/arm-cpus.in (fp16fml): New feature.
12313 (ALL_SIMD): Add fp16fml.
12314 (armv8.2-a): Add fp16fml as an option.
12315 (armv8.3-a): Likewise.
12316 (armv8.4-a): Add fp16fml as part of fp16.
12317 * config/arm/arm.h (TARGET_FP16FML): Define.
12318 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12320 * config/arm/arm-modes.def (V2HF): Define.
12321 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12322 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12323 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12324 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12325 vfmsl_low, vfmsl_high): New set of builtins.
12326 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12327 (vfml_op): New code attribute.
12328 (VFMLHALVES): New int iterator.
12329 (VFML, VFMLSEL): New mode attributes.
12330 (V_reg): Define mapping for V2HF.
12331 (V_hi, V_lo): New mode attributes.
12332 (VF_constraint): Likewise.
12333 (vfml_half, vfml_half_selector): New int attributes.
12334 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12336 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12337 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12339 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12340 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12341 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12342 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12344 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12345 Document new effective target and option set.
12347 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12349 * config/arm/arm-cpus.in (armv8_4): New feature.
12350 (ARMv8_4a): New fgroup.
12351 (armv8.4-a): New arch.
12352 * config/arm/arm-tables.opt: Regenerate.
12353 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12354 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12355 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12356 Add matching rules for -march=armv8.4-a and extensions.
12357 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12359 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12362 * config/rx/rx.md (BW): New mode attribute.
12363 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12365 2018-01-11 Richard Biener <rguenther@suse.de>
12367 PR tree-optimization/83435
12368 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12369 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12370 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12372 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12373 Alan Hayward <alan.hayward@arm.com>
12374 David Sherwood <david.sherwood@arm.com>
12376 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12378 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12379 (aarch64_print_address_internal): Use it to check for a zero offset.
12381 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12382 Alan Hayward <alan.hayward@arm.com>
12383 David Sherwood <david.sherwood@arm.com>
12385 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12386 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12387 Return a poly_int64 rather than a HOST_WIDE_INT.
12388 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12389 rather than a HOST_WIDE_INT.
12390 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12391 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12392 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12393 final_offset from HOST_WIDE_INT to poly_int64.
12394 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12395 to_constant when getting the number of units in an Advanced SIMD
12397 (aarch64_builtin_vectorized_function): Check for a constant number
12399 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12401 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12402 attribute instead of GET_MODE_NUNITS.
12403 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12404 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12405 GET_MODE_SIZE for fixed-size registers.
12406 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12407 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12408 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12409 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12410 (aarch64_print_operand, aarch64_print_address_internal)
12411 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12412 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12413 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12414 Handle polynomial GET_MODE_SIZE.
12415 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12416 wider than SImode without modification.
12417 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12418 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12419 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12420 passing and returning SVE modes.
12421 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12422 rather than GEN_INT.
12423 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12424 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12425 (aarch64_allocate_and_probe_stack_space): Likewise.
12426 (aarch64_layout_frame): Cope with polynomial offsets.
12427 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12428 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12429 polynomial offsets.
12430 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12431 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12432 poly_int64 rather than a HOST_WIDE_INT.
12433 (aarch64_get_separate_components, aarch64_process_components)
12434 (aarch64_expand_prologue, aarch64_expand_epilogue)
12435 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12436 (aarch64_anchor_offset): New function, split out from...
12437 (aarch64_legitimize_address): ...here.
12438 (aarch64_builtin_vectorization_cost): Handle polynomial
12439 TYPE_VECTOR_SUBPARTS.
12440 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12442 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12443 number of elements from the PARALLEL rather than the mode.
12444 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12445 rather than GET_MODE_BITSIZE.
12446 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12447 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12448 (aarch64_expand_vec_perm_const_1): Handle polynomial
12449 d->perm.length () and d->perm elements.
12450 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12451 Apply to_constant to d->perm elements.
12452 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12453 polynomial CONST_VECTOR_NUNITS.
12454 (aarch64_move_pointer): Take amount as a poly_int64 rather
12456 (aarch64_progress_pointer): Avoid temporary variable.
12457 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12458 the mode attribute instead of GET_MODE.
12460 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12461 Alan Hayward <alan.hayward@arm.com>
12462 David Sherwood <david.sherwood@arm.com>
12464 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12465 x exists before using it.
12466 (aarch64_add_constant_internal): Rename to...
12467 (aarch64_add_offset_1): ...this. Replace regnum with separate
12468 src and dest rtxes. Handle the case in which they're different,
12469 including when the offset is zero. Replace scratchreg with an rtx.
12470 Use 2 additions if there is no spare register into which we can
12471 move a 16-bit constant.
12472 (aarch64_add_constant): Delete.
12473 (aarch64_add_offset): Replace reg with separate src and dest
12474 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12475 Use aarch64_add_offset_1.
12476 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12477 an rtx rather than an int. Take the delta as a poly_int64
12478 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12479 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12480 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12481 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12482 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12483 and aarch64_add_sp.
12484 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12485 aarch64_add_constant.
12487 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12489 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12490 Use scalar_float_mode.
12492 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12494 * config/aarch64/aarch64-simd.md
12495 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12496 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12497 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12498 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12499 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12500 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12501 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12502 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12503 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12504 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12506 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12509 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12510 targ_options->x_arm_arch_string is non NULL.
12512 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12514 * config/aarch64/aarch64.h
12515 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12517 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12520 * expmed.c (emit_store_flag_force): Swap if const op0
12521 and change VOIDmode to mode of op0.
12523 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12525 PR rtl-optimization/83761
12526 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12527 than bytes to mode_for_size.
12529 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12531 PR middle-end/83189
12532 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12533 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12536 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12538 PR middle-end/83575
12539 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12540 when in layout mode.
12541 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12542 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12545 2018-01-10 Michael Collison <michael.collison@arm.com>
12547 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12548 * config/aarch64/aarch64-option-extension.def: Add
12549 AARCH64_OPT_EXTENSION of 'fp16fml'.
12550 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12551 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12552 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12553 * config/aarch64/constraints.md (Ui7): New constraint.
12554 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12555 (VFMLA_SEL_W): Ditto.
12558 (VFMLA16_LOW): New int iterator.
12559 (VFMLA16_HIGH): Ditto.
12560 (UNSPEC_FMLAL): New unspec.
12561 (UNSPEC_FMLSL): Ditto.
12562 (UNSPEC_FMLAL2): Ditto.
12563 (UNSPEC_FMLSL2): Ditto.
12564 (f16mac): New code attribute.
12565 * config/aarch64/aarch64-simd-builtins.def
12566 (aarch64_fmlal_lowv2sf): Ditto.
12567 (aarch64_fmlsl_lowv2sf): Ditto.
12568 (aarch64_fmlalq_lowv4sf): Ditto.
12569 (aarch64_fmlslq_lowv4sf): Ditto.
12570 (aarch64_fmlal_highv2sf): Ditto.
12571 (aarch64_fmlsl_highv2sf): Ditto.
12572 (aarch64_fmlalq_highv4sf): Ditto.
12573 (aarch64_fmlslq_highv4sf): Ditto.
12574 (aarch64_fmlal_lane_lowv2sf): Ditto.
12575 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12576 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12577 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12578 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12579 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12580 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12581 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12582 (aarch64_fmlal_lane_highv2sf): Ditto.
12583 (aarch64_fmlsl_lane_highv2sf): Ditto.
12584 (aarch64_fmlal_laneq_highv2sf): Ditto.
12585 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12586 (aarch64_fmlalq_lane_highv4sf): Ditto.
12587 (aarch64_fmlsl_lane_highv4sf): Ditto.
12588 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12589 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12590 * config/aarch64/aarch64-simd.md:
12591 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12592 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12593 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12594 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12595 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12596 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12597 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12598 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12599 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12600 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12601 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12602 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12603 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12604 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12605 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12606 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12607 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12608 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12609 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12610 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12611 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12612 (vfmlsl_low_u32): Ditto.
12613 (vfmlalq_low_u32): Ditto.
12614 (vfmlslq_low_u32): Ditto.
12615 (vfmlal_high_u32): Ditto.
12616 (vfmlsl_high_u32): Ditto.
12617 (vfmlalq_high_u32): Ditto.
12618 (vfmlslq_high_u32): Ditto.
12619 (vfmlal_lane_low_u32): Ditto.
12620 (vfmlsl_lane_low_u32): Ditto.
12621 (vfmlal_laneq_low_u32): Ditto.
12622 (vfmlsl_laneq_low_u32): Ditto.
12623 (vfmlalq_lane_low_u32): Ditto.
12624 (vfmlslq_lane_low_u32): Ditto.
12625 (vfmlalq_laneq_low_u32): Ditto.
12626 (vfmlslq_laneq_low_u32): Ditto.
12627 (vfmlal_lane_high_u32): Ditto.
12628 (vfmlsl_lane_high_u32): Ditto.
12629 (vfmlal_laneq_high_u32): Ditto.
12630 (vfmlsl_laneq_high_u32): Ditto.
12631 (vfmlalq_lane_high_u32): Ditto.
12632 (vfmlslq_lane_high_u32): Ditto.
12633 (vfmlalq_laneq_high_u32): Ditto.
12634 (vfmlslq_laneq_high_u32): Ditto.
12635 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12636 (AARCH64_FL_FOR_ARCH8_4): New.
12637 (AARCH64_ISA_F16FML): New ISA flag.
12638 (TARGET_F16FML): New feature flag for fp16fml.
12639 (doc/invoke.texi): Document new fp16fml option.
12641 2018-01-10 Michael Collison <michael.collison@arm.com>
12643 * config/aarch64/aarch64-builtins.c:
12644 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12645 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12646 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12647 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12648 (AARCH64_ISA_SHA3): New ISA flag.
12649 (TARGET_SHA3): New feature flag for sha3.
12650 * config/aarch64/iterators.md (sha512_op): New int attribute.
12651 (CRYPTO_SHA512): New int iterator.
12652 (UNSPEC_SHA512H): New unspec.
12653 (UNSPEC_SHA512H2): Ditto.
12654 (UNSPEC_SHA512SU0): Ditto.
12655 (UNSPEC_SHA512SU1): Ditto.
12656 * config/aarch64/aarch64-simd-builtins.def
12657 (aarch64_crypto_sha512hqv2di): New builtin.
12658 (aarch64_crypto_sha512h2qv2di): Ditto.
12659 (aarch64_crypto_sha512su0qv2di): Ditto.
12660 (aarch64_crypto_sha512su1qv2di): Ditto.
12661 (aarch64_eor3qv8hi): Ditto.
12662 (aarch64_rax1qv2di): Ditto.
12663 (aarch64_xarqv2di): Ditto.
12664 (aarch64_bcaxqv8hi): Ditto.
12665 * config/aarch64/aarch64-simd.md:
12666 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12667 (aarch64_crypto_sha512su0qv2di): Ditto.
12668 (aarch64_crypto_sha512su1qv2di): Ditto.
12669 (aarch64_eor3qv8hi): Ditto.
12670 (aarch64_rax1qv2di): Ditto.
12671 (aarch64_xarqv2di): Ditto.
12672 (aarch64_bcaxqv8hi): Ditto.
12673 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12674 (vsha512h2q_u64): Ditto.
12675 (vsha512su0q_u64): Ditto.
12676 (vsha512su1q_u64): Ditto.
12677 (veor3q_u16): Ditto.
12678 (vrax1q_u64): Ditto.
12679 (vxarq_u64): Ditto.
12680 (vbcaxq_u16): Ditto.
12681 * config/arm/types.md (crypto_sha512): New type attribute.
12682 (crypto_sha3): Ditto.
12683 (doc/invoke.texi): Document new sha3 option.
12685 2018-01-10 Michael Collison <michael.collison@arm.com>
12687 * config/aarch64/aarch64-builtins.c:
12688 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12689 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12690 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12691 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12692 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12693 (AARCH64_ISA_SM4): New ISA flag.
12694 (TARGET_SM4): New feature flag for sm4.
12695 * config/aarch64/aarch64-simd-builtins.def
12696 (aarch64_sm3ss1qv4si): Ditto.
12697 (aarch64_sm3tt1aq4si): Ditto.
12698 (aarch64_sm3tt1bq4si): Ditto.
12699 (aarch64_sm3tt2aq4si): Ditto.
12700 (aarch64_sm3tt2bq4si): Ditto.
12701 (aarch64_sm3partw1qv4si): Ditto.
12702 (aarch64_sm3partw2qv4si): Ditto.
12703 (aarch64_sm4eqv4si): Ditto.
12704 (aarch64_sm4ekeyqv4si): Ditto.
12705 * config/aarch64/aarch64-simd.md:
12706 (aarch64_sm3ss1qv4si): Ditto.
12707 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12708 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12709 (aarch64_sm4eqv4si): Ditto.
12710 (aarch64_sm4ekeyqv4si): Ditto.
12711 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12712 (sm3part_op): Ditto.
12713 (CRYPTO_SM3TT): Ditto.
12714 (CRYPTO_SM3PART): Ditto.
12715 (UNSPEC_SM3SS1): New unspec.
12716 (UNSPEC_SM3TT1A): Ditto.
12717 (UNSPEC_SM3TT1B): Ditto.
12718 (UNSPEC_SM3TT2A): Ditto.
12719 (UNSPEC_SM3TT2B): Ditto.
12720 (UNSPEC_SM3PARTW1): Ditto.
12721 (UNSPEC_SM3PARTW2): Ditto.
12722 (UNSPEC_SM4E): Ditto.
12723 (UNSPEC_SM4EKEY): Ditto.
12724 * config/aarch64/constraints.md (Ui2): New constraint.
12725 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12726 * config/arm/types.md (crypto_sm3): New type attribute.
12727 (crypto_sm4): Ditto.
12728 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12729 (vsm3tt1aq_u32): Ditto.
12730 (vsm3tt1bq_u32): Ditto.
12731 (vsm3tt2aq_u32): Ditto.
12732 (vsm3tt2bq_u32): Ditto.
12733 (vsm3partw1q_u32): Ditto.
12734 (vsm3partw2q_u32): Ditto.
12735 (vsm4eq_u32): Ditto.
12736 (vsm4ekeyq_u32): Ditto.
12737 (doc/invoke.texi): Document new sm4 option.
12739 2018-01-10 Michael Collison <michael.collison@arm.com>
12741 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12742 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12743 (AARCH64_FL_FOR_ARCH8_4): New.
12744 (AARCH64_FL_V8_4): New flag.
12745 (doc/invoke.texi): Document new armv8.4-a option.
12747 2018-01-10 Michael Collison <michael.collison@arm.com>
12749 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12750 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12751 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12752 * config/aarch64/aarch64-option-extension.def: Add
12753 AARCH64_OPT_EXTENSION of 'sha2'.
12754 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12755 (crypto): Disable sha2 and aes if crypto disabled.
12756 (crypto): Enable aes and sha2 if enabled.
12757 (simd): Disable sha2 and aes if simd disabled.
12758 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12760 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12761 (TARGET_SHA2): New feature flag for sha2.
12762 (TARGET_AES): New feature flag for aes.
12763 * config/aarch64/aarch64-simd.md:
12764 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12765 conditional on TARGET_AES.
12766 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12767 (aarch64_crypto_sha1hsi): Make pattern conditional
12769 (aarch64_crypto_sha1hv4si): Ditto.
12770 (aarch64_be_crypto_sha1hv4si): Ditto.
12771 (aarch64_crypto_sha1su1v4si): Ditto.
12772 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12773 (aarch64_crypto_sha1su0v4si): Ditto.
12774 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12775 (aarch64_crypto_sha256su0v4si): Ditto.
12776 (aarch64_crypto_sha256su1v4si): Ditto.
12777 (doc/invoke.texi): Document new aes and sha2 options.
12779 2018-01-10 Martin Sebor <msebor@redhat.com>
12781 PR tree-optimization/83781
12782 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12785 2018-01-11 Martin Sebor <msebor@gmail.com>
12786 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12788 PR tree-optimization/83501
12789 PR tree-optimization/81703
12791 * tree-ssa-strlen.c (get_string_cst): Rename...
12792 (get_string_len): ...to this. Handle global constants.
12793 (handle_char_store): Adjust.
12795 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12796 Jim Wilson <jimw@sifive.com>
12798 * config/riscv/riscv-protos.h (riscv_output_return): New.
12799 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12800 (riscv_attribute_table, riscv_output_return),
12801 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12802 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12803 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12805 (riscv_expand_prologue): Add early return for naked function.
12806 (riscv_expand_epilogue): Likewise.
12807 (riscv_function_ok_for_sibcall): Return false for naked function.
12808 (riscv_set_current_function): New.
12809 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12810 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12811 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12812 * doc/extend.texi (RISC-V Function Attributes): New.
12814 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12816 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12817 check for 128-bit long double before checking TCmode.
12818 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12819 128-bit long doubles before checking TFmode or TCmode.
12820 (FLOAT128_IBM_P): Likewise.
12822 2018-01-10 Martin Sebor <msebor@redhat.com>
12824 PR tree-optimization/83671
12825 * builtins.c (c_strlen): Unconditionally return zero for the empty
12827 Use -Warray-bounds for warnings.
12828 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12829 for non-constant array indices with COMPONENT_REF, arrays of
12830 arrays, and pointers to arrays.
12831 (gimple_fold_builtin_strlen): Determine and set length range for
12832 non-constant character arrays.
12834 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12836 PR middle-end/81897
12837 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12840 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12842 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12844 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12847 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12848 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12849 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12850 indexed_or_indirect_operand predicate.
12851 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12852 (*vsx_le_perm_load_v8hi): Likewise.
12853 (*vsx_le_perm_load_v16qi): Likewise.
12854 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12855 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12856 (*vsx_le_perm_store_v8hi): Likewise.
12857 (*vsx_le_perm_store_v16qi): Likewise.
12858 (eight unnamed splitters): Likewise.
12860 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12862 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12863 * config/rs6000/emmintrin.h: Likewise.
12864 * config/rs6000/mmintrin.h: Likewise.
12865 * config/rs6000/xmmintrin.h: Likewise.
12867 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12870 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12872 * tree.c (tree_nop_conversion): Return true for location wrapper
12874 (maybe_wrap_with_location): New function.
12875 (selftest::check_strip_nops): New function.
12876 (selftest::test_location_wrappers): New function.
12877 (selftest::tree_c_tests): Call it.
12878 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12879 (maybe_wrap_with_location): New decl.
12880 (EXPR_LOCATION_WRAPPER_P): New macro.
12881 (location_wrapper_p): New inline function.
12882 (tree_strip_any_location_wrapper): New inline function.
12884 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12887 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12888 stack_realign_offset for the largest alignment of stack slot
12890 (ix86_find_max_used_stack_alignment): New function.
12891 (ix86_finalize_stack_frame_flags): Use it. Set
12892 max_used_stack_alignment if we don't realign stack.
12893 * config/i386/i386.h (machine_function): Add
12894 max_used_stack_alignment.
12896 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12898 * config/arm/arm.opt (-mbranch-cost): New option.
12899 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12902 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12905 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12906 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12908 2018-01-10 Richard Biener <rguenther@suse.de>
12911 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12912 early out so it also covers the case where we have a non-NULL
12915 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12917 PR tree-optimization/83753
12918 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12919 for non-strided grouped accesses if the number of elements is 1.
12921 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12924 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12925 * i386.h (TARGET_USE_GATHER): Define.
12926 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12928 2018-01-10 Martin Liska <mliska@suse.cz>
12931 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12932 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12934 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12935 CLEANUP_NO_PARTITIONING is not set.
12937 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12939 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12940 for vectors, as a partial revert of r254296.
12941 * rtl.h (const_vec_p): Delete.
12942 (const_vec_duplicate_p): Don't test for vector CONSTs.
12943 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12944 * expmed.c (make_tree): Likewise.
12947 * common.md (E, F): Use CONSTANT_P instead of checking for
12949 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12950 checking for CONST_VECTOR.
12952 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12954 PR middle-end/83575
12955 * predict.c (force_edge_cold): Handle in more sane way edges
12956 with no prediction.
12958 2018-01-09 Carl Love <cel@us.ibm.com>
12960 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12962 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12963 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12964 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12965 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12966 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12967 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12968 * config/rs6000/rs6000-protos.h: Add extern defition for
12969 rs6000_generate_float2_double_code.
12970 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12972 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12973 (float2_v2df): Add define_expand.
12975 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12978 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12979 op_mode in the force_to_mode call.
12981 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12983 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12984 instead of checking each element individually.
12985 (aarch64_evpc_uzp): Likewise.
12986 (aarch64_evpc_zip): Likewise.
12987 (aarch64_evpc_ext): Likewise.
12988 (aarch64_evpc_rev): Likewise.
12989 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12990 instead of checking each element individually. Return true without
12992 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12993 whether all selected elements come from the same input, instead of
12994 checking each element individually. Remove calls to gen_rtx_REG,
12995 start_sequence and end_sequence and instead assert that no rtl is
12998 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13000 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13001 order of HIGH and CONST checks.
13003 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13005 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13006 if the destination isn't an SSA_NAME.
13008 2018-01-09 Richard Biener <rguenther@suse.de>
13010 PR tree-optimization/83668
13011 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13013 (canonicalize_loop_form): ... here, renamed from ...
13014 (canonicalize_loop_closed_ssa_form): ... this and amended to
13015 swap successor edges for loop exit blocks to make us use
13016 the RPO order we need for initial schedule generation.
13018 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13020 PR tree-optimization/64811
13021 * match.pd: When optimizing comparisons with Inf, avoid
13022 introducing or losing exceptions from comparisons with NaN.
13024 2018-01-09 Martin Liska <mliska@suse.cz>
13027 * asan.c (shadow_mem_size): Add gcc_assert.
13029 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13031 Don't save registers in main().
13034 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13035 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13036 * config/avr/avr.c (avr_set_current_function): Don't error if
13037 naked, OS_task or OS_main are specified at the same time.
13038 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13040 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13042 * common/config/avr/avr-common.c (avr_option_optimization_table):
13043 Switch on -mmain-is-OS_task for optimizing compilations.
13045 2018-01-09 Richard Biener <rguenther@suse.de>
13047 PR tree-optimization/83572
13048 * graphite.c: Include cfganal.h.
13049 (graphite_transform_loops): Connect infinite loops to exit
13050 and remove fake edges at the end.
13052 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13054 * ipa-inline.c (edge_badness): Revert accidental checkin.
13056 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13059 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13060 symbols; not inline clones.
13062 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13065 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13066 hard registers. Formatting fixes.
13068 PR preprocessor/83722
13069 * gcc.c (try_generate_repro): Pass
13070 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13071 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13074 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13075 Kito Cheng <kito.cheng@gmail.com>
13077 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13078 (riscv_leaf_function_p): Delete.
13079 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13081 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13083 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13085 (do_ifelse): New function.
13086 (do_isel): New function.
13087 (do_sub3): New function.
13088 (do_add3): New function.
13089 (do_load_mask_compare): New function.
13090 (do_overlap_load_compare): New function.
13091 (expand_compare_loop): New function.
13092 (expand_block_compare): Call expand_compare_loop() when appropriate.
13093 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13094 option description.
13095 (-mblock-compare-inline-loop-limit): New option.
13097 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13100 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13101 Reverse order of second and third operands in first alternative.
13102 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13103 of first and second elements in UNSPEC_VPERMR vector.
13104 (altivec_expand_vec_perm_le): Likewise.
13106 2018-01-08 Jeff Law <law@redhat.com>
13108 PR rtl-optimizatin/81308
13109 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13110 (process_switch): If group_case_labels makes a change, then set
13112 (pass_convert_switch::execute): If a switch is converted, then
13113 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13115 PR rtl-optimization/81308
13116 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13119 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13121 PR target/83663 - Revert r255946
13122 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13123 generation for cases where splatting a value is not useful.
13124 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13125 across a vec_duplicate and a paradoxical subreg forming a vector
13126 mode to a vec_concat.
13128 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13130 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13131 -march=armv8.3-a variants.
13132 * config/arm/t-multilib: Likewise.
13133 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13135 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13137 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13139 (cceq_ior_compare_complement): Give it a name so I can use it, and
13140 change boolean_or_operator predicate to boolean_operator so it can
13141 be used to generate a crand.
13142 (eqne): New code iterator.
13143 (bd/bd_neg): New code_attrs.
13144 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13145 a single define_insn.
13146 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13147 decrement (bdnzt/bdnzf/bdzt/bdzf).
13148 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13149 with the new names of the branch decrement patterns, and added the
13150 names of the branch decrement conditional patterns.
13152 2018-01-08 Richard Biener <rguenther@suse.de>
13154 PR tree-optimization/83563
13155 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13158 2018-01-08 Richard Biener <rguenther@suse.de>
13160 PR middle-end/83713
13161 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13163 2018-01-08 Richard Biener <rguenther@suse.de>
13165 PR tree-optimization/83685
13166 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13167 references to abnormals.
13169 2018-01-08 Richard Biener <rguenther@suse.de>
13172 * dwarf2out.c (output_indirect_strings): Handle empty
13173 skeleton_debug_str_hash.
13174 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13176 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13178 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13179 (emit_store_direct): Likewise.
13180 (arc_trampoline_adjust_address): Likewise.
13181 (arc_asm_trampoline_template): New function.
13182 (arc_initialize_trampoline): Use asm_trampoline_template.
13183 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13184 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13185 * config/arc/arc.md (flush_icache): Delete pattern.
13187 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13189 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13190 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13193 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13196 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13197 by not USED_FOR_TARGET.
13198 (make_pass_resolve_sw_modes): Likewise.
13200 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13202 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13205 2018-01-08 Richard Biener <rguenther@suse.de>
13207 PR middle-end/83580
13208 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13210 2018-01-08 Richard Biener <rguenther@suse.de>
13212 PR middle-end/83517
13213 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13215 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13217 PR middle-end/81897
13218 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13219 basic blocks with a small number of successors.
13220 (convert_control_dep_chain_into_preds): Improve handling of
13222 (dump_predicates): Split apart into...
13223 (dump_pred_chain): ...here...
13224 (dump_pred_info): ...and here.
13225 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13226 (can_chain_union_be_invalidated_p): Improve check for invalidation
13228 (uninit_uses_cannot_happen): Avoid unnecessary if
13229 convert_control_dep_chain_into_preds yielded nothing.
13231 2018-01-06 Martin Sebor <msebor@redhat.com>
13233 PR tree-optimization/83640
13234 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13235 subtracting negative offset from size.
13236 (builtin_access::overlap): Adjust offset bounds of the access to fall
13237 within the size of the object if possible.
13239 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13241 PR rtl-optimization/83699
13242 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13243 extract_bit_field_as_subreg to cases in which the extracted
13244 value is also a vector.
13246 * lra-constraints.c (process_alt_operands): Test for the equivalence
13247 substitutions when detecting a possible reload cycle.
13249 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13252 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13253 by default if flag_selective_schedling{,2}. Formatting fixes.
13255 PR rtl-optimization/83682
13256 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13257 if it has non-VECTOR_MODE element mode.
13258 (vec_duplicate_p): Likewise.
13260 PR middle-end/83694
13261 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13262 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13264 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13267 * config/i386/i386-builtin.def
13268 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13269 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13270 Require also OPTION_MASK_ISA_AVX512F in addition to
13271 OPTION_MASK_ISA_GFNI.
13272 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13273 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13274 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13275 to OPTION_MASK_ISA_GFNI.
13276 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13277 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13278 OPTION_MASK_ISA_AVX512BW.
13279 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13280 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13281 addition to OPTION_MASK_ISA_GFNI.
13282 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13283 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13284 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13285 to OPTION_MASK_ISA_GFNI.
13286 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13287 a requirement for all ISAs rather than any of them with a few
13289 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13291 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13292 bitmasks to be enabled with 3 exceptions, instead of requiring any
13293 enabled ISA with lots of exceptions.
13294 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13295 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13296 Change avx512bw in isa attribute to avx512f.
13297 * config/i386/sgxintrin.h: Add license boilerplate.
13298 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13299 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13300 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13301 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13303 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13304 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13305 temporarily sse2 rather than sse if not enabled already.
13308 * config/i386/sse.md (VI248_VLBW): Rename to ...
13309 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13310 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13311 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13312 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13313 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13314 mode iterator instead of VI248_VLBW.
13316 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13318 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13319 (record_modified): Skip clobbers; add debug output.
13320 (param_change_prob): Use sreal frequencies.
13322 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13324 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13325 punt for user-aligned variables.
13327 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13329 * tree-chrec.c (chrec_contains_symbols): Return true for
13332 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13335 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13336 of (x|y) == x for BICS pattern.
13338 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13340 PR tree-optimization/83605
13341 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13342 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13345 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13347 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13348 * config/epiphany/rtems.h: New file.
13350 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13351 Uros Bizjak <ubizjak@gmail.com>
13354 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13355 QIreg_operand instead of register_operand predicate.
13356 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13357 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13358 comments instead of -fmitigate[-_]rop.
13360 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13363 * cgraphunit.c (symbol_table::compile): Switch to text_section
13364 before calling assembly_start debug hook.
13365 * run-rtl-passes.c (run_rtl_passes): Likewise.
13368 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13370 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13371 range_int_cst_p rather than !symbolic_range_p before calling
13372 extract_range_from_multiplicative_op_1.
13374 2018-01-04 Jeff Law <law@redhat.com>
13376 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13377 redundant test in assertion.
13379 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13381 * doc/rtl.texi: Document machine_mode wrapper classes.
13383 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13385 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13386 using tree_to_uhwi.
13388 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13390 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13391 the VEC_PERM_EXPR fold to fail.
13393 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13396 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13397 to switched_sections.
13399 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13402 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13403 test for d.testing.
13405 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13408 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13409 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13411 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13414 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13415 is BLKmode and bitpos not zero or mode change is needed.
13417 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13420 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13423 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13426 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13427 instead of MULT rtx. Update all corresponding splitters.
13428 (*saddl_se): Ditto.
13429 (*ssub<modesuffix>): Ditto.
13430 (*ssubl_se): Ditto.
13431 (*cmp_sadd_di): Update split patterns.
13432 (*cmp_sadd_si): Ditto.
13433 (*cmp_sadd_sidi): Ditto.
13434 (*cmp_ssub_di): Ditto.
13435 (*cmp_ssub_si): Ditto.
13436 (*cmp_ssub_sidi): Ditto.
13437 * config/alpha/predicates.md (const23_operand): New predicate.
13438 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13439 Look for ASHIFT, not MULT inner operand.
13440 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13442 2018-01-04 Martin Liska <mliska@suse.cz>
13444 PR gcov-profile/83669
13445 * gcov.c (output_intermediate_file): Add version to intermediate
13447 * doc/gcov.texi: Document new field 'version' in intermediate
13448 file format. Fix location of '-k' option of gcov command.
13450 2018-01-04 Martin Liska <mliska@suse.cz>
13453 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13455 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13457 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13459 2018-01-03 Martin Sebor <msebor@redhat.com>
13461 PR tree-optimization/83655
13462 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13463 checking calls with invalid arguments.
13465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13467 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13468 (vectorizable_mask_load_store): Delete.
13469 (vectorizable_call): Return false for masked loads and stores.
13470 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13471 instead of gimple_assign_rhs1.
13472 (vectorizable_load): Handle IFN_MASK_LOAD.
13473 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13477 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13479 (vectorizable_mask_load_store): ...here.
13480 (vectorizable_load): ...and here.
13482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13484 * tree-vect-stmts.c (vect_build_all_ones_mask)
13485 (vect_build_zero_merge_argument): New functions, split out from...
13486 (vectorizable_load): ...here.
13488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13490 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13492 (vectorizable_mask_load_store): ...here.
13493 (vectorizable_store): ...and here.
13495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13497 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13499 (vectorizable_mask_load_store): ...here.
13501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13503 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13504 (vect_model_store_cost): Take a vec_load_store_type instead of a
13506 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13507 (vect_model_store_cost): Take a vec_load_store_type instead of a
13509 (vectorizable_mask_load_store): Update accordingly.
13510 (vectorizable_store): Likewise.
13511 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13515 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13516 IFN_MASK_LOAD calls here rather than...
13517 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13520 Alan Hayward <alan.hayward@arm.com>
13521 David Sherwood <david.sherwood@arm.com>
13523 * expmed.c (extract_bit_field_1): For vector extracts,
13524 fall back to extract_bit_field_as_subreg if vec_extract
13527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13528 Alan Hayward <alan.hayward@arm.com>
13529 David Sherwood <david.sherwood@arm.com>
13531 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13532 they are variable or constant sized.
13533 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13534 slots for constant-sized data.
13536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13537 Alan Hayward <alan.hayward@arm.com>
13538 David Sherwood <david.sherwood@arm.com>
13540 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13541 handling COND_EXPRs with boolean comparisons, try to find a better
13542 basis for the mask type than the boolean itself.
13544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13546 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13547 is calculated and how it can be overridden.
13548 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13549 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13551 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13555 Alan Hayward <alan.hayward@arm.com>
13556 David Sherwood <david.sherwood@arm.com>
13558 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13559 Remove the mode argument.
13560 (aarch64_simd_valid_immediate): Remove the mode and inverse
13562 * config/aarch64/iterators.md (bitsize): New iterator.
13563 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13564 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13565 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13566 aarch64_simd_valid_immediate.
13567 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13568 (aarch64_reg_or_bic_imm): Likewise.
13569 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13570 with an insn_type enum and msl with a modifier_type enum.
13571 Replace element_width with a scalar_mode. Change the shift
13572 to unsigned int. Add constructors for scalar_float_mode and
13573 scalar_int_mode elements.
13574 (aarch64_vect_float_const_representable_p): Delete.
13575 (aarch64_can_const_movi_rtx_p)
13576 (aarch64_simd_scalar_immediate_valid_for_move)
13577 (aarch64_simd_make_constant): Update call to
13578 aarch64_simd_valid_immediate.
13579 (aarch64_advsimd_valid_immediate_hs): New function.
13580 (aarch64_advsimd_valid_immediate): Likewise.
13581 (aarch64_simd_valid_immediate): Remove mode and inverse
13582 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13583 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13584 and aarch64_float_const_representable_p on the result.
13585 (aarch64_output_simd_mov_immediate): Remove mode argument.
13586 Update call to aarch64_simd_valid_immediate and use of
13587 simd_immediate_info.
13588 (aarch64_output_scalar_simd_mov_immediate): Update call
13591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13592 Alan Hayward <alan.hayward@arm.com>
13593 David Sherwood <david.sherwood@arm.com>
13595 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13596 (mode_nunits): Likewise CONST_MODE_NUNITS.
13597 * machmode.def (ADJUST_NUNITS): Document.
13598 * genmodes.c (mode_data::need_nunits_adj): New field.
13599 (blank_mode): Update accordingly.
13600 (adj_nunits): New variable.
13601 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13603 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13604 listed in adj_nunits.
13605 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13606 listed in adj_nunits. Don't emit case statements for such modes.
13607 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13608 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13609 nothing if adj_nunits is nonnull.
13610 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13611 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13612 (emit_mode_fbit): Update use of print_maybe_const_decl.
13613 (emit_move_size): Likewise. Treat the array as non-const
13615 (emit_mode_adjustments): Handle adj_nunits.
13617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13619 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13620 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13621 (VECTOR_MODES): Use it.
13622 (make_vector_modes): Take the prefix as an argument.
13624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13625 Alan Hayward <alan.hayward@arm.com>
13626 David Sherwood <david.sherwood@arm.com>
13628 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13629 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13630 for MODE_VECTOR_BOOL.
13631 * machmode.def (VECTOR_BOOL_MODE): Document.
13632 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13633 (make_vector_bool_mode): New function.
13634 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13636 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13637 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13639 * stor-layout.c (int_mode_for_mode): Likewise.
13640 * tree.c (build_vector_type_for_mode): Likewise.
13641 * varasm.c (output_constant_pool_2): Likewise.
13642 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13643 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13644 for MODE_VECTOR_BOOL.
13645 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13646 of mode class checks.
13647 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13648 instead of a list of mode class checks.
13649 (expand_vector_scalar_condition): Likewise.
13650 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13653 Alan Hayward <alan.hayward@arm.com>
13654 David Sherwood <david.sherwood@arm.com>
13656 * machmode.h (mode_size): Change from unsigned short to
13658 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13659 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13660 or if measurement_type is not polynomial.
13661 (fixed_size_mode::includes_p): Check for constant-sized modes.
13662 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13663 return a poly_uint16 rather than an unsigned short.
13664 (emit_mode_size): Change the type of mode_size from unsigned short
13665 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13666 (emit_mode_adjustments): Cope with polynomial vector sizes.
13667 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13669 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13671 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13672 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13673 * caller-save.c (setup_save_areas): Likewise.
13674 (replace_reg_with_saved_mem): Likewise.
13675 * calls.c (emit_library_call_value_1): Likewise.
13676 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13677 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13678 (gen_lowpart_for_combine): Likewise.
13679 * convert.c (convert_to_integer_1): Likewise.
13680 * cse.c (equiv_constant, cse_insn): Likewise.
13681 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13682 (cselib_subst_to_values): Likewise.
13683 * dce.c (word_dce_process_block): Likewise.
13684 * df-problems.c (df_word_lr_mark_ref): Likewise.
13685 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13686 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13687 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13688 (rtl_for_decl_location): Likewise.
13689 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13690 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13691 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13692 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13693 (expand_expr_real_1): Likewise.
13694 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13695 (pad_below): Likewise.
13696 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13697 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13698 * ira.c (get_subreg_tracking_sizes): Likewise.
13699 * ira-build.c (ira_create_allocno_objects): Likewise.
13700 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13701 (ira_sort_regnos_for_alter_reg): Likewise.
13702 * ira-costs.c (record_operand_costs): Likewise.
13703 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13704 (resolve_simple_move): Likewise.
13705 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13706 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13707 (lra_constraints): Likewise.
13708 (CONST_POOL_OK_P): Reject variable-sized modes.
13709 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13710 (add_pseudo_to_slot, lra_spill): Likewise.
13711 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13712 * optabs-query.c (get_best_extraction_insn): Likewise.
13713 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13714 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13715 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13716 * recog.c (offsettable_address_addr_space_p): Likewise.
13717 * regcprop.c (maybe_mode_change): Likewise.
13718 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13719 * regrename.c (build_def_use): Likewise.
13720 * regstat.c (dump_reg_info): Likewise.
13721 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13722 (find_reloads, find_reloads_subreg_address): Likewise.
13723 * reload1.c (eliminate_regs_1): Likewise.
13724 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13725 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13726 (simplify_binary_operation_1, simplify_subreg): Likewise.
13727 * targhooks.c (default_function_arg_padding): Likewise.
13728 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13729 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13730 (verify_gimple_assign_ternary): Likewise.
13731 * tree-inline.c (estimate_move_cost): Likewise.
13732 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13733 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13734 (get_address_cost_ainc): Likewise.
13735 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13736 (vect_supportable_dr_alignment): Likewise.
13737 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13738 (vectorizable_reduction): Likewise.
13739 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13740 (vectorizable_operation, vectorizable_load): Likewise.
13741 * tree.c (build_same_sized_truth_vector_type): Likewise.
13742 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13743 * var-tracking.c (emit_note_insn_var_location): Likewise.
13744 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13745 (ADDR_VEC_ALIGN): Likewise.
13747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13748 Alan Hayward <alan.hayward@arm.com>
13749 David Sherwood <david.sherwood@arm.com>
13751 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13753 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13754 or if measurement_type is polynomial.
13755 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13756 * combine.c (make_extraction): Likewise.
13757 * dse.c (find_shift_sequence): Likewise.
13758 * dwarf2out.c (mem_loc_descriptor): Likewise.
13759 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13760 (extract_bit_field, extract_low_bits): Likewise.
13761 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13762 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13763 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13764 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13765 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13766 * reload.c (find_reloads): Likewise.
13767 * reload1.c (alter_reg): Likewise.
13768 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13769 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13770 * tree-if-conv.c (predicate_mem_writes): Likewise.
13771 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13772 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13773 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13774 * valtrack.c (dead_debug_insert_temp): Likewise.
13775 * varasm.c (mergeable_constant_section): Likewise.
13776 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13779 Alan Hayward <alan.hayward@arm.com>
13780 David Sherwood <david.sherwood@arm.com>
13782 * expr.c (expand_assignment): Cope with polynomial mode sizes
13783 when assigning to a CONCAT.
13785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13786 Alan Hayward <alan.hayward@arm.com>
13787 David Sherwood <david.sherwood@arm.com>
13789 * machmode.h (mode_precision): Change from unsigned short to
13791 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13793 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13794 or if measurement_type is not polynomial.
13795 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13796 in which the mode is already known to be a scalar_int_mode.
13797 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13798 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13800 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13801 for GET_MODE_PRECISION.
13802 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13803 for GET_MODE_PRECISION.
13804 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13806 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13807 (expand_field_assignment, make_extraction): Likewise.
13808 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13809 (get_last_value): Likewise.
13810 * convert.c (convert_to_integer_1): Likewise.
13811 * cse.c (cse_insn): Likewise.
13812 * expr.c (expand_expr_real_1): Likewise.
13813 * lra-constraints.c (simplify_operand_subreg): Likewise.
13814 * optabs-query.c (can_atomic_load_p): Likewise.
13815 * optabs.c (expand_atomic_load): Likewise.
13816 (expand_atomic_store): Likewise.
13817 * ree.c (combine_reaching_defs): Likewise.
13818 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13819 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13820 * tree.h (type_has_mode_precision_p): Likewise.
13821 * ubsan.c (instrument_si_overflow): Likewise.
13823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13824 Alan Hayward <alan.hayward@arm.com>
13825 David Sherwood <david.sherwood@arm.com>
13827 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13828 polynomial numbers of units.
13829 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13830 (valid_vector_subparts_p): New function.
13831 (build_vector_type): Remove temporary shim and take the number
13832 of units as a poly_uint64 rather than an int.
13833 (build_opaque_vector_type): Take the number of units as a
13834 poly_uint64 rather than an int.
13835 * tree.c (build_vector_from_ctor): Handle polynomial
13836 TYPE_VECTOR_SUBPARTS.
13837 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13838 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13839 (build_vector_from_val): If the number of units is variable,
13840 use build_vec_duplicate_cst for constant operands and
13841 VEC_DUPLICATE_EXPR otherwise.
13842 (make_vector_type): Remove temporary is_constant ().
13843 (build_vector_type, build_opaque_vector_type): Take the number of
13844 units as a poly_uint64 rather than an int.
13845 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13847 * cfgexpand.c (expand_debug_expr): Likewise.
13848 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13849 (store_constructor, expand_expr_real_1): Likewise.
13850 (const_scalar_mask_from_tree): Likewise.
13851 * fold-const-call.c (fold_const_reduction): Likewise.
13852 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13853 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13854 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13855 (fold_relational_const): Likewise.
13856 (native_interpret_vector): Likewise. Change the size from an
13857 int to an unsigned int.
13858 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13859 TYPE_VECTOR_SUBPARTS.
13860 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13861 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13862 duplicating a non-constant operand into a variable-length vector.
13863 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13864 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13865 * ipa-icf.c (sem_variable::equals): Likewise.
13866 * match.pd: Likewise.
13867 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13868 * print-tree.c (print_node): Likewise.
13869 * stor-layout.c (layout_type): Likewise.
13870 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13871 * tree-cfg.c (verify_gimple_comparison): Likewise.
13872 (verify_gimple_assign_binary): Likewise.
13873 (verify_gimple_assign_ternary): Likewise.
13874 (verify_gimple_assign_single): Likewise.
13875 * tree-pretty-print.c (dump_generic_node): Likewise.
13876 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13877 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13878 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13879 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13880 (vect_shift_permute_load_chain): Likewise.
13881 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13882 (expand_vector_condition, optimize_vector_constructor): Likewise.
13883 (lower_vec_perm, get_compute_type): Likewise.
13884 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13885 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13886 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13887 (vect_recog_mask_conversion_pattern): Likewise.
13888 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13889 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13890 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13891 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13892 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13893 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13894 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13895 (supportable_widening_operation): Likewise.
13896 (supportable_narrowing_operation): Likewise.
13897 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13899 * varasm.c (output_constant): Likewise.
13901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13902 Alan Hayward <alan.hayward@arm.com>
13903 David Sherwood <david.sherwood@arm.com>
13905 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13906 so that both the length == 3 and length != 3 cases set up their
13907 own permute vectors. Add comments explaining why we know the
13908 number of elements is constant.
13909 (vect_permute_load_chain): Likewise.
13911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13912 Alan Hayward <alan.hayward@arm.com>
13913 David Sherwood <david.sherwood@arm.com>
13915 * machmode.h (mode_nunits): Change from unsigned char to
13917 (ONLY_FIXED_SIZE_MODES): New macro.
13918 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13919 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13920 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13922 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13923 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13924 or if measurement_type is not polynomial.
13925 * genmodes.c (ZERO_COEFFS): New macro.
13926 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13928 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13929 Use ZERO_COEFFS when emitting initializers.
13930 * data-streamer.h (bp_pack_poly_value): New function.
13931 (bp_unpack_poly_value): Likewise.
13932 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13933 for GET_MODE_NUNITS.
13934 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13935 for GET_MODE_NUNITS.
13936 * tree.c (make_vector_type): Remove temporary shim and make
13937 the real function take the number of units as a poly_uint64
13938 rather than an int.
13939 (build_vector_type_for_mode): Handle polynomial nunits.
13940 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13941 * emit-rtl.c (const_vec_series_p_1): Likewise.
13942 (gen_rtx_CONST_VECTOR): Likewise.
13943 * fold-const.c (test_vec_duplicate_folding): Likewise.
13944 * genrecog.c (validate_pattern): Likewise.
13945 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13946 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13947 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13948 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13949 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13950 * rtlanal.c (subreg_get_info): Likewise.
13951 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13952 (vect_grouped_load_supported): Likewise.
13953 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13954 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13955 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13956 (simplify_const_unary_operation, simplify_binary_operation_1)
13957 (simplify_const_binary_operation, simplify_ternary_operation)
13958 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13959 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13960 instead of CONST_VECTOR_NUNITS.
13961 * varasm.c (output_constant_pool_2): Likewise.
13962 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13963 explicit-encoded elements in the XVEC for variable-length vectors.
13965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13967 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13970 Alan Hayward <alan.hayward@arm.com>
13971 David Sherwood <david.sherwood@arm.com>
13973 * coretypes.h (fixed_size_mode): Declare.
13974 (fixed_size_mode_pod): New typedef.
13975 * builtins.h (target_builtins::x_apply_args_mode)
13976 (target_builtins::x_apply_result_mode): Change type to
13977 fixed_size_mode_pod.
13978 * builtins.c (apply_args_size, apply_result_size, result_vector)
13979 (expand_builtin_apply_args_1, expand_builtin_apply)
13980 (expand_builtin_return): Update accordingly.
13982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13984 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13985 * cselib.c (cselib_hash_rtx): Likewise.
13986 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13987 CONST_VECTOR encoding.
13989 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13990 Jeff Law <law@redhat.com>
13993 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13994 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13995 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13996 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13999 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14000 explicitly probe *sp in a noreturn function if there were any callee
14001 register saves or frame pointer is needed.
14003 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14006 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14007 BLKmode for ternary, binary or unary expressions.
14010 * var-tracking.c (delete_vta_debug_insn): New inline function.
14011 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14012 insns from get_insns () to NULL instead of each bb separately.
14013 Use delete_vta_debug_insn. No longer static.
14014 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14015 delete_vta_debug_insns callers.
14016 * rtl.h (delete_vta_debug_insns): Declare.
14017 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14018 instead of variable_tracking_main.
14020 2018-01-03 Martin Sebor <msebor@redhat.com>
14022 PR tree-optimization/83603
14023 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14024 arguments past the endof the argument list in functions declared
14025 without a prototype.
14026 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14027 Avoid checking when arguments are null.
14029 2018-01-03 Martin Sebor <msebor@redhat.com>
14032 * doc/extend.texi (attribute const): Fix a typo.
14033 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14034 issuing -Wsuggest-attribute for void functions.
14036 2018-01-03 Martin Sebor <msebor@redhat.com>
14038 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14039 offset_int::from instead of wide_int::to_shwi.
14040 (maybe_diag_overlap): Remove assertion.
14041 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14042 * gimple-ssa-sprintf.c (format_directive): Same.
14043 (parse_directive): Same.
14044 (sprintf_dom_walker::compute_format_length): Same.
14045 (try_substitute_return_value): Same.
14047 2018-01-03 Jeff Law <law@redhat.com>
14049 PR middle-end/83654
14050 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14051 non-constant residual for zero at runtime and avoid probing in
14052 that case. Reorganize code for trailing problem to mirror handling
14055 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14057 PR tree-optimization/83501
14058 * tree-ssa-strlen.c (get_string_cst): New.
14059 (handle_char_store): Call get_string_cst.
14061 2018-01-03 Martin Liska <mliska@suse.cz>
14063 PR tree-optimization/83593
14064 * tree-ssa-strlen.c: Include tree-cfg.h.
14065 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14066 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14067 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14069 (strlen_dom_walker::before_dom_children): Call
14070 gimple_purge_dead_eh_edges. Dump tranformation with details
14072 (strlen_dom_walker::before_dom_children): Update call by adding
14073 new argument cleanup_eh.
14074 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14076 2018-01-03 Martin Liska <mliska@suse.cz>
14079 * cif-code.def (VARIADIC_THUNK): New enum value.
14080 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14083 2018-01-03 Jan Beulich <jbeulich@suse.com>
14085 * sse.md (mov<mode>_internal): Tighten condition for when to use
14086 vmovdqu<ssescalarsize> for TI and OI modes.
14088 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14090 Update copyright years.
14092 2018-01-03 Martin Liska <mliska@suse.cz>
14095 * ipa-visibility.c (function_and_variable_visibility): Skip
14096 functions with noipa attribure.
14098 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14100 * gcc.c (process_command): Update copyright notice dates.
14101 * gcov-dump.c (print_version): Ditto.
14102 * gcov.c (print_version): Ditto.
14103 * gcov-tool.c (print_version): Ditto.
14104 * gengtype.c (create_file): Ditto.
14105 * doc/cpp.texi: Bump @copying's copyright year.
14106 * doc/cppinternals.texi: Ditto.
14107 * doc/gcc.texi: Ditto.
14108 * doc/gccint.texi: Ditto.
14109 * doc/gcov.texi: Ditto.
14110 * doc/install.texi: Ditto.
14111 * doc/invoke.texi: Ditto.
14113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14115 * vector-builder.h (vector_builder::m_full_nelts): Change from
14116 unsigned int to poly_uint64.
14117 (vector_builder::full_nelts): Update prototype accordingly.
14118 (vector_builder::new_vector): Likewise.
14119 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14120 (vector_builder::operator ==): Likewise.
14121 (vector_builder::finalize): Likewise.
14122 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14123 Take the number of elements as a poly_uint64 rather than an
14125 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14126 from unsigned int to poly_uint64.
14127 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14128 (vec_perm_indices::new_vector): Likewise.
14129 (vec_perm_indices::length): Likewise.
14130 (vec_perm_indices::nelts_per_input): Likewise.
14131 (vec_perm_indices::input_nelts): Likewise.
14132 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14133 number of elements per input as a poly_uint64 rather than an
14134 unsigned int. Use the original encoding for variable-length
14135 vectors, rather than clamping each individual element.
14136 For the second and subsequent elements in each pattern,
14137 clamp the step and base before clamping their sum.
14138 (vec_perm_indices::series_p): Handle polynomial element counts.
14139 (vec_perm_indices::all_in_range_p): Likewise.
14140 (vec_perm_indices_to_tree): Likewise.
14141 (vec_perm_indices_to_rtx): Likewise.
14142 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14143 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14144 (tree_vector_builder::new_binary_operation): Handle polynomial
14145 element counts. Return false if we need to know the number
14146 of elements at compile time.
14147 * fold-const.c (fold_vec_perm): Punt if the number of elements
14148 isn't known at compile time.
14150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14152 * vec-perm-indices.h (vec_perm_builder): Change element type
14153 from HOST_WIDE_INT to poly_int64.
14154 (vec_perm_indices::element_type): Update accordingly.
14155 (vec_perm_indices::clamp): Handle polynomial element_types.
14156 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14157 (vec_perm_indices::all_in_range_p): Likewise.
14158 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14160 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14161 polynomial vec_perm_indices element types.
14162 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14163 * fold-const.c (fold_vec_perm): Likewise.
14164 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14165 * tree-vect-generic.c (lower_vec_perm): Likewise.
14166 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14167 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14168 element type to HOST_WIDE_INT.
14170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14171 Alan Hayward <alan.hayward@arm.com>
14172 David Sherwood <david.sherwood@arm.com>
14174 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14175 rather than an int. Use plus_constant.
14176 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14177 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14180 Alan Hayward <alan.hayward@arm.com>
14181 David Sherwood <david.sherwood@arm.com>
14183 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14184 a HOST_WIDE_INT to a poly_int64.
14186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14187 Alan Hayward <alan.hayward@arm.com>
14188 David Sherwood <david.sherwood@arm.com>
14190 * calls.c (load_register_parameters): Cope with polynomial
14191 mode sizes. Require a constant size for BLKmode parameters
14192 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14193 forces a parameter to be padded at the lsb end in order to
14194 fill a complete number of words, require the parameter size
14195 to be ordered wrt UNITS_PER_WORD.
14197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14198 Alan Hayward <alan.hayward@arm.com>
14199 David Sherwood <david.sherwood@arm.com>
14201 * reload1.c (spill_stack_slot_width): Change element type
14202 from unsigned int to poly_uint64_pod.
14203 (alter_reg): Treat mode sizes as polynomial.
14205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14206 Alan Hayward <alan.hayward@arm.com>
14207 David Sherwood <david.sherwood@arm.com>
14209 * reload.c (complex_word_subreg_p): New function.
14210 (reload_inner_reg_of_subreg, push_reload): Use it.
14212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14213 Alan Hayward <alan.hayward@arm.com>
14214 David Sherwood <david.sherwood@arm.com>
14216 * lra-constraints.c (process_alt_operands): Reject matched
14217 operands whose sizes aren't ordered.
14218 (match_reload): Refer to this check here.
14220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14221 Alan Hayward <alan.hayward@arm.com>
14222 David Sherwood <david.sherwood@arm.com>
14224 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14225 that the mode size is in the set {1, 2, 4, 8, 16}.
14227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14228 Alan Hayward <alan.hayward@arm.com>
14229 David Sherwood <david.sherwood@arm.com>
14231 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14232 Use plus_constant instead of gen_rtx_PLUS.
14234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14235 Alan Hayward <alan.hayward@arm.com>
14236 David Sherwood <david.sherwood@arm.com>
14238 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14239 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14240 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14241 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14242 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14243 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14244 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14245 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14246 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14247 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14249 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14250 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14251 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14252 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14253 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14254 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14255 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14256 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14257 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14258 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14260 * expr.c (emit_move_resolve_push): Treat the input and result
14261 of PUSH_ROUNDING as a poly_int64.
14262 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14263 (emit_push_insn): Likewise.
14264 * lra-eliminations.c (mark_not_eliminable): Likewise.
14265 * recog.c (push_operand): Likewise.
14266 * reload1.c (elimination_effects): Likewise.
14267 * rtlanal.c (nonzero_bits1): Likewise.
14268 * calls.c (store_one_arg): Likewise. Require the padding to be
14269 known at compile time.
14271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14272 Alan Hayward <alan.hayward@arm.com>
14273 David Sherwood <david.sherwood@arm.com>
14275 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14276 Use plus_constant instead of gen_rtx_PLUS.
14278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14279 Alan Hayward <alan.hayward@arm.com>
14280 David Sherwood <david.sherwood@arm.com>
14282 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14283 rather than an int.
14285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14286 Alan Hayward <alan.hayward@arm.com>
14287 David Sherwood <david.sherwood@arm.com>
14289 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14290 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14291 via stack temporaries. Treat the mode size as polynomial too.
14293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14294 Alan Hayward <alan.hayward@arm.com>
14295 David Sherwood <david.sherwood@arm.com>
14297 * expr.c (expand_expr_real_2): When handling conversions involving
14298 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14299 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14300 as a poly_uint64 too.
14302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14303 Alan Hayward <alan.hayward@arm.com>
14304 David Sherwood <david.sherwood@arm.com>
14306 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14309 Alan Hayward <alan.hayward@arm.com>
14310 David Sherwood <david.sherwood@arm.com>
14312 * combine.c (can_change_dest_mode): Handle polynomial
14313 REGMODE_NATURAL_SIZE.
14314 * expmed.c (store_bit_field_1): Likewise.
14315 * expr.c (store_constructor): Likewise.
14316 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14317 and polynomial REGMODE_NATURAL_SIZE.
14318 (gen_lowpart_common): Likewise.
14319 * reginfo.c (record_subregs_of_mode): Likewise.
14320 * rtlanal.c (read_modify_subreg_p): Likewise.
14322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14323 Alan Hayward <alan.hayward@arm.com>
14324 David Sherwood <david.sherwood@arm.com>
14326 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14327 numbers of elements.
14329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14330 Alan Hayward <alan.hayward@arm.com>
14331 David Sherwood <david.sherwood@arm.com>
14333 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14340 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
14347 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14348 (ipa_simd_modify_function_body): Likewise.
14350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14351 Alan Hayward <alan.hayward@arm.com>
14352 David Sherwood <david.sherwood@arm.com>
14354 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14355 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14356 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14357 (expand_vector_condition, vector_element): Likewise.
14358 (subparts_gt): New function.
14359 (get_compute_type): Use subparts_gt.
14360 (count_type_subparts): Delete.
14361 (expand_vector_operations_1): Use subparts_gt instead of
14362 count_type_subparts.
14364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14365 Alan Hayward <alan.hayward@arm.com>
14366 David Sherwood <david.sherwood@arm.com>
14368 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14369 (vect_compile_time_alias): ...this new function. Do the calculation
14370 on poly_ints rather than trees.
14371 (vect_prune_runtime_alias_test_list): Update call accordingly.
14373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14374 Alan Hayward <alan.hayward@arm.com>
14375 David Sherwood <david.sherwood@arm.com>
14377 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14379 (vect_schedule_slp_instance): Likewise.
14381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14382 Alan Hayward <alan.hayward@arm.com>
14383 David Sherwood <david.sherwood@arm.com>
14385 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14386 constant and extern definitions for variable-length vectors.
14387 (vect_get_constant_vectors): Note that the number of units
14388 is known to be constant.
14390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14391 Alan Hayward <alan.hayward@arm.com>
14392 David Sherwood <david.sherwood@arm.com>
14394 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14395 of units as polynomial. Choose between WIDE and NARROW based
14398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14399 Alan Hayward <alan.hayward@arm.com>
14400 David Sherwood <david.sherwood@arm.com>
14402 * tree-vect-stmts.c (simd_clone_subparts): New function.
14403 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14406 Alan Hayward <alan.hayward@arm.com>
14407 David Sherwood <david.sherwood@arm.com>
14409 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14410 vectors as polynomial. Use build_index_vector for
14411 IFN_GOMP_SIMD_LANE.
14413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14414 Alan Hayward <alan.hayward@arm.com>
14415 David Sherwood <david.sherwood@arm.com>
14417 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14418 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14419 for variable-length vectors.
14420 (vectorizable_mask_load_store): Treat the number of units as
14421 polynomial, asserting that it is constant if the condition has
14422 already been enforced.
14423 (vectorizable_store, vectorizable_load): Likewise.
14425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14426 Alan Hayward <alan.hayward@arm.com>
14427 David Sherwood <david.sherwood@arm.com>
14429 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14430 of units as polynomial. Punt if we can't tell at compile time
14431 which vector contains the final result.
14433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14434 Alan Hayward <alan.hayward@arm.com>
14435 David Sherwood <david.sherwood@arm.com>
14437 * tree-vect-loop.c (vectorizable_induction): Treat the number
14438 of units as polynomial. Punt on SLP inductions. Use an integer
14439 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14440 cast of such a series for variable-length floating-point
14443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14444 Alan Hayward <alan.hayward@arm.com>
14445 David Sherwood <david.sherwood@arm.com>
14447 * tree.h (build_index_vector): Declare.
14448 * tree.c (build_index_vector): New function.
14449 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14450 of units as polynomial, forcibly converting it to a constant if
14451 vectorizable_reduction has already enforced the condition.
14452 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14453 to create a {1,2,3,...} vector.
14454 (vectorizable_reduction): Treat the number of units as polynomial.
14455 Choose vectype_in based on the largest scalar element size rather
14456 than the smallest number of units. Enforce the restrictions
14459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14460 Alan Hayward <alan.hayward@arm.com>
14461 David Sherwood <david.sherwood@arm.com>
14463 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14464 number of units as polynomial.
14466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14467 Alan Hayward <alan.hayward@arm.com>
14468 David Sherwood <david.sherwood@arm.com>
14470 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14471 * target.def (autovectorize_vector_sizes): Return the vector sizes
14472 by pointer, using vector_sizes rather than a bitmask.
14473 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14474 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14475 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14477 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14478 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14479 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14480 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14481 * omp-general.c (omp_max_vf): Likewise.
14482 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14483 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14484 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14485 * tree-vect-slp.c (vect_slp_bb): Likewise.
14486 * doc/tm.texi: Regenerate.
14487 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14489 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14490 the vector size as a poly_uint64 rather than an unsigned int.
14491 (current_vector_size): Change from an unsigned int to a poly_uint64.
14492 (get_vectype_for_scalar_type): Update accordingly.
14493 * tree.h (build_truth_vector_type): Take the size and number of
14494 units as a poly_uint64 rather than an unsigned int.
14495 (build_vector_type): Add a temporary overload that takes
14496 the number of units as a poly_uint64 rather than an unsigned int.
14497 * tree.c (make_vector_type): Likewise.
14498 (build_truth_vector_type): Take the number of units as a poly_uint64
14499 rather than an unsigned int.
14501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14502 Alan Hayward <alan.hayward@arm.com>
14503 David Sherwood <david.sherwood@arm.com>
14505 * target.def (get_mask_mode): Take the number of units and length
14506 as poly_uint64s rather than unsigned ints.
14507 * targhooks.h (default_get_mask_mode): Update accordingly.
14508 * targhooks.c (default_get_mask_mode): Likewise.
14509 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14510 * doc/tm.texi: Regenerate.
14512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14513 Alan Hayward <alan.hayward@arm.com>
14514 David Sherwood <david.sherwood@arm.com>
14516 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14517 * omp-general.c (omp_max_vf): Likewise.
14518 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14519 (expand_omp_simd): Handle polynomial safelen.
14520 * omp-low.c (omplow_simd_context): Add a default constructor.
14521 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14522 (lower_rec_simd_input_clauses): Update accordingly.
14523 (lower_rec_input_clauses): Likewise.
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 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14530 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14531 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14532 (vect_analyze_slp_cost): Likewise.
14533 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14534 (vect_model_load_cost): Likewise.
14536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14537 Alan Hayward <alan.hayward@arm.com>
14538 David Sherwood <david.sherwood@arm.com>
14540 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14541 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14542 from an unsigned int * to a poly_uint64_pod *.
14543 (calculate_unrolling_factor): New function.
14544 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14547 Alan Hayward <alan.hayward@arm.com>
14548 David Sherwood <david.sherwood@arm.com>
14550 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14551 from an unsigned int to a poly_uint64.
14552 (_loop_vec_info::slp_unrolling_factor): Likewise.
14553 (_loop_vec_info::vectorization_factor): Change from an int
14555 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14556 (vect_get_num_vectors): New function.
14557 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14558 (vect_get_num_copies): Use vect_get_num_vectors.
14559 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14560 to an unsigned int *.
14561 (vect_analyze_data_refs): Change min_vf from an int * to a
14563 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14564 than an unsigned HOST_WIDE_INT.
14565 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14566 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14567 to an unsigned int *.
14568 (vect_analyze_data_ref_dependences): Likewise.
14569 (vect_compute_data_ref_alignment): Handle polynomial vf.
14570 (vect_enhance_data_refs_alignment): Likewise.
14571 (vect_prune_runtime_alias_test_list): Likewise.
14572 (vect_shift_permute_load_chain): Likewise.
14573 (vect_supportable_dr_alignment): Likewise.
14574 (dependence_distance_ge_vf): Take the vectorization factor as a
14575 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14576 (vect_analyze_data_refs): Change min_vf from an int * to a
14578 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14579 vfm1 as a poly_uint64 rather than an int. Make the same change
14580 for the returned bound_scalar.
14581 (vect_gen_vector_loop_niters): Handle polynomial vf.
14582 (vect_do_peeling): Likewise. Update call to
14583 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14584 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14586 * tree-vect-loop.c (vect_determine_vectorization_factor)
14587 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14588 (vect_get_known_peeling_cost): Likewise.
14589 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14590 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14591 (vect_transform_loop): Likewise. Use the lowest possible VF when
14592 updating the upper bounds of the loop.
14593 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14594 rather than an int.
14595 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14596 polynomial unroll factors.
14597 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14598 (vect_make_slp_decision): Likewise.
14599 (vect_supported_load_permutation_p): Likewise, and polynomial
14601 (vect_analyze_slp_cost): Handle polynomial vf.
14602 (vect_slp_analyze_node_operations): Likewise.
14603 (vect_slp_analyze_bb_1): Likewise.
14604 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14605 than an unsigned HOST_WIDE_INT.
14606 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14607 (vectorizable_load): Handle polynomial vf.
14608 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14610 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14613 Alan Hayward <alan.hayward@arm.com>
14614 David Sherwood <david.sherwood@arm.com>
14616 * match.pd: Handle bit operations involving three constants
14617 and try to fold one pair.
14619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14621 * tree-vect-loop-manip.c: Include gimple-fold.h.
14622 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14623 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14624 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14625 Add a path that uses a step of VF instead of 1, but disable it
14627 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14628 and niters_no_overflow parameters. Update calls to
14629 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14630 Create a new SSA name if the latter choses to use a ste other
14631 than zero, and return it via niters_vector_mult_vf_var.
14632 * tree-vect-loop.c (vect_transform_loop): Update calls to
14633 vect_do_peeling, vect_gen_vector_loop_niters and
14634 slpeel_make_loop_iterate_ntimes.
14635 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14636 (vect_gen_vector_loop_niters): Update declarations after above changes.
14638 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14640 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14641 128-bit round to integer instructions.
14642 (ceil<mode>2): Likewise.
14643 (btrunc<mode>2): Likewise.
14644 (round<mode>2): Likewise.
14646 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14648 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14649 unaligned VSX load/store on P8/P9.
14650 (expand_block_clear): Allow the use of unaligned VSX
14651 load/store on P8/P9.
14653 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14655 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14657 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14658 swap associated with both a load and a store.
14660 2018-01-02 Andrew Waterman <andrew@sifive.com>
14662 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14663 * config/riscv/riscv.md (clear_cache): Use it.
14665 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14667 * web.c: Remove out-of-date comment.
14669 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14671 * expr.c (fixup_args_size_notes): Check that any existing
14672 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14673 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14674 (emit_single_push_insn): ...here.
14676 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14678 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14679 (const_vector_encoded_nelts): New function.
14680 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14681 (const_vector_int_elt, const_vector_elt): Declare.
14682 * emit-rtl.c (const_vector_int_elt_1): New function.
14683 (const_vector_elt): Likewise.
14684 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14685 of CONST_VECTOR_ELT.
14687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14689 * expr.c: Include rtx-vector-builder.h.
14690 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14691 directly on the tree encoding.
14692 (const_vector_from_tree): Likewise.
14693 * optabs.c: Include rtx-vector-builder.h.
14694 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14695 sequence of "u" values.
14696 * vec-perm-indices.c: Include rtx-vector-builder.h.
14697 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14698 directly on the vec_perm_indices encoding.
14700 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14702 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14703 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14704 * rtx-vector-builder.h: New file.
14705 * rtx-vector-builder.c: Likewise.
14706 * rtl.h (rtx_def::u2): Add a const_vector field.
14707 (CONST_VECTOR_NPATTERNS): New macro.
14708 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14709 (CONST_VECTOR_DUPLICATE_P): Likewise.
14710 (CONST_VECTOR_STEPPED_P): Likewise.
14711 (CONST_VECTOR_ENCODED_ELT): Likewise.
14712 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14713 (unwrap_const_vec_duplicate): Likewise.
14714 (const_vec_series_p): Check for a non-duplicated vector encoding.
14715 Say that the function only returns true for integer vectors.
14716 * emit-rtl.c: Include rtx-vector-builder.h.
14717 (gen_const_vec_duplicate_1): Delete.
14718 (gen_const_vector): Call gen_const_vec_duplicate instead of
14719 gen_const_vec_duplicate_1.
14720 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14721 (gen_const_vec_duplicate): Use rtx_vector_builder.
14722 (gen_const_vec_series): Likewise.
14723 (gen_rtx_CONST_VECTOR): Likewise.
14724 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14725 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14726 Build a new vector rather than modifying a CONST_VECTOR in-place.
14727 (handle_special_swappables): Update call accordingly.
14728 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14729 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14730 Build a new vector rather than modifying a CONST_VECTOR in-place.
14731 (handle_special_swappables): Update call accordingly.
14733 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14735 * simplify-rtx.c (simplify_const_binary_operation): Use
14736 CONST_VECTOR_ELT instead of XVECEXP.
14738 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14740 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14741 the selector elements to be different from the data elements
14742 if the selector is a VECTOR_CST.
14743 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14744 ssizetype for the selector.
14746 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14748 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14749 before testing each element individually.
14750 * tree-vect-generic.c (lower_vec_perm): Likewise.
14752 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14754 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14755 * selftest-run-tests.c (selftest::run_tests): Call it.
14756 * vector-builder.h (vector_builder::operator ==): New function.
14757 (vector_builder::operator !=): Likewise.
14758 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14759 (vec_perm_indices::all_from_input_p): New function.
14760 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14761 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14762 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14763 instead of reading the VECTOR_CST directly. Detect whether both
14764 vector inputs are the same before constructing the vec_perm_indices,
14765 and update the number of inputs argument accordingly. Use the
14766 utility functions added above. Only construct sel2 if we need to.
14768 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14770 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14771 the broadcast of the low byte.
14772 (expand_mult_highpart): Use an explicit encoding for the permutes.
14773 * optabs-query.c (can_mult_highpart_p): Likewise.
14774 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14775 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14776 (vectorizable_bswap): Likewise.
14777 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14778 explicit encoding for the power-of-2 permutes.
14779 (vect_permute_store_chain): Likewise.
14780 (vect_grouped_load_supported): Likewise.
14781 (vect_permute_load_chain): Likewise.
14783 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14785 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14786 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14787 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14788 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14789 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14790 (vect_gen_perm_mask_any): Likewise.
14792 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14794 * int-vector-builder.h: New file.
14795 * vec-perm-indices.h: Include int-vector-builder.h.
14796 (vec_perm_indices): Redefine as an int_vector_builder.
14797 (auto_vec_perm_indices): Delete.
14798 (vec_perm_builder): Redefine as a stand-alone class.
14799 (vec_perm_indices::vec_perm_indices): New function.
14800 (vec_perm_indices::clamp): Likewise.
14801 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14802 (vec_perm_indices::new_vector): New function.
14803 (vec_perm_indices::new_expanded_vector): Update for new
14804 vec_perm_indices class.
14805 (vec_perm_indices::rotate_inputs): New function.
14806 (vec_perm_indices::all_in_range_p): Operate directly on the
14807 encoded form, without computing elided elements.
14808 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14809 encoding. Update for new vec_perm_indices class.
14810 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14811 the given vec_perm_builder.
14812 (expand_vec_perm_var): Update vec_perm_builder constructor.
14813 (expand_mult_highpart): Use vec_perm_builder instead of
14814 auto_vec_perm_indices.
14815 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14816 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14817 or double series encoding as appropriate.
14818 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14819 vec_perm_indices instead of auto_vec_perm_indices.
14820 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14821 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14822 (vect_permute_store_chain): Likewise.
14823 (vect_grouped_load_supported): Likewise.
14824 (vect_permute_load_chain): Likewise.
14825 (vect_shift_permute_load_chain): Likewise.
14826 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14827 (vect_transform_slp_perm_load): Likewise.
14828 (vect_schedule_slp_instance): Likewise.
14829 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14830 (vectorizable_mask_load_store): Likewise.
14831 (vectorizable_bswap): Likewise.
14832 (vectorizable_store): Likewise.
14833 (vectorizable_load): Likewise.
14834 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14835 vec_perm_indices instead of auto_vec_perm_indices. Use
14836 tree_to_vec_perm_builder to read the vector from a tree.
14837 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14838 vec_perm_builder instead of a vec_perm_indices.
14839 (have_whole_vector_shift): Use vec_perm_builder and
14840 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14841 truncation to calc_vec_perm_mask_for_shift.
14842 (vect_create_epilog_for_reduction): Likewise.
14843 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14844 from auto_vec_perm_indices to vec_perm_indices.
14845 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14846 instead of changing individual elements.
14847 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14848 the vector in d.perm.
14849 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14850 from auto_vec_perm_indices to vec_perm_indices.
14851 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14852 instead of changing individual elements.
14853 (arm_vectorize_vec_perm_const): Use new_vector to install
14854 the vector in d.perm.
14855 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14856 Update vec_perm_builder constructor.
14857 (rs6000_expand_interleave): Likewise.
14858 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14859 (rs6000_expand_interleave): Likewise.
14861 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14863 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14864 to qimode could truncate the indices.
14865 * optabs.c (expand_vec_perm_var): Likewise.
14867 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14869 * Makefile.in (OBJS): Add vec-perm-indices.o.
14870 * vec-perm-indices.h: New file.
14871 * vec-perm-indices.c: Likewise.
14872 * target.h (vec_perm_indices): Replace with a forward class
14874 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14875 * optabs.h: Include vec-perm-indices.h.
14876 (expand_vec_perm): Delete.
14877 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14878 (expand_vec_perm_const): Declare.
14879 * target.def (vec_perm_const_ok): Replace with...
14880 (vec_perm_const): ...this new hook.
14881 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14882 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14883 * doc/tm.texi: Regenerate.
14884 * optabs.def (vec_perm_const): Delete.
14885 * doc/md.texi (vec_perm_const): Likewise.
14886 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14887 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14888 expand_vec_perm for constant permutation vectors. Assert that
14889 the mode of variable permutation vectors is the integer equivalent
14890 of the mode that is being permuted.
14891 * optabs-query.h (selector_fits_mode_p): Declare.
14892 * optabs-query.c: Include vec-perm-indices.h.
14893 (selector_fits_mode_p): New function.
14894 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14895 is defined, instead of checking whether the vec_perm_const_optab
14896 exists. Use targetm.vectorize.vec_perm_const instead of
14897 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14898 fit in the vector mode before using a variable permute.
14899 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14900 vec_perm_indices instead of an rtx.
14901 (expand_vec_perm): Replace with...
14902 (expand_vec_perm_const): ...this new function. Take the selector
14903 as a vec_perm_indices rather than an rtx. Also take the mode of
14904 the selector. Update call to shift_amt_for_vec_perm_mask.
14905 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14906 Use vec_perm_indices::new_expanded_vector to expand the original
14907 selector into bytes. Check whether the indices fit in the vector
14908 mode before using a variable permute.
14909 (expand_vec_perm_var): Make global.
14910 (expand_mult_highpart): Use expand_vec_perm_const.
14911 * fold-const.c: Includes vec-perm-indices.h.
14912 * tree-ssa-forwprop.c: Likewise.
14913 * tree-vect-data-refs.c: Likewise.
14914 * tree-vect-generic.c: Likewise.
14915 * tree-vect-loop.c: Likewise.
14916 * tree-vect-slp.c: Likewise.
14917 * tree-vect-stmts.c: Likewise.
14918 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14920 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14921 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14922 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14923 (aarch64_vectorize_vec_perm_const): ...this new function.
14924 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14925 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14926 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14927 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14928 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14929 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14930 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14932 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14933 check for NEON modes.
14934 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14935 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14936 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14937 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14939 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14940 the old VEC_PERM_CONST conditions.
14941 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14942 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14943 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14944 (ia64_vectorize_vec_perm_const_ok): Merge into...
14945 (ia64_vectorize_vec_perm_const): ...this new function.
14946 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14947 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14948 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14949 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14950 * config/mips/mips.c (mips_expand_vec_perm_const)
14951 (mips_vectorize_vec_perm_const_ok): Merge into...
14952 (mips_vectorize_vec_perm_const): ...this new function.
14953 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14954 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14955 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14956 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14957 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14958 (rs6000_expand_vec_perm_const): Delete.
14959 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14961 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14962 (altivec_expand_vec_perm_const_le): Take each operand individually.
14963 Operate on constant selectors rather than rtxes.
14964 (altivec_expand_vec_perm_const): Likewise. Update call to
14965 altivec_expand_vec_perm_const_le.
14966 (rs6000_expand_vec_perm_const): Delete.
14967 (rs6000_vectorize_vec_perm_const_ok): Delete.
14968 (rs6000_vectorize_vec_perm_const): New function.
14969 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14970 an element count and rtx array.
14971 (rs6000_expand_extract_even): Update call accordingly.
14972 (rs6000_expand_interleave): Likewise.
14973 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14974 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14975 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14976 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14977 (rs6000_expand_vec_perm_const): Delete.
14978 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14979 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14980 (altivec_expand_vec_perm_const_le): Take each operand individually.
14981 Operate on constant selectors rather than rtxes.
14982 (altivec_expand_vec_perm_const): Likewise. Update call to
14983 altivec_expand_vec_perm_const_le.
14984 (rs6000_expand_vec_perm_const): Delete.
14985 (rs6000_vectorize_vec_perm_const_ok): Delete.
14986 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14987 reference to the SPE evmerge intructions.
14988 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14989 an element count and rtx array.
14990 (rs6000_expand_extract_even): Update call accordingly.
14991 (rs6000_expand_interleave): Likewise.
14992 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14993 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14995 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14997 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14999 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15000 vector mode and that that mode matches the mode of the data
15002 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15003 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15004 directly using expand_vec_perm_1 when forcing selectors into
15006 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15008 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15010 * optabs-query.h (can_vec_perm_p): Delete.
15011 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15012 * optabs-query.c (can_vec_perm_p): Split into...
15013 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15014 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15015 particular selector is valid.
15016 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15017 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15018 (vect_grouped_load_supported): Likewise.
15019 (vect_shift_permute_load_chain): Likewise.
15020 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15021 (vect_transform_slp_perm_load): Likewise.
15022 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15023 (vectorizable_bswap): Likewise.
15024 (vect_gen_perm_mask_checked): Likewise.
15025 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15026 implementations of variable permutation vectors into account
15027 when deciding which selector to use.
15028 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15029 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15030 with a false third argument.
15031 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15032 to test whether the constant selector is valid and can_vec_perm_var_p
15033 to test whether a variable selector is valid.
15035 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15037 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15038 * optabs-query.c (can_vec_perm_p): Likewise.
15039 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15040 instead of vec_perm_indices.
15041 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15042 (vect_gen_perm_mask_checked): Likewise,
15043 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15044 (vect_gen_perm_mask_checked): Likewise,
15046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15048 * optabs-query.h (qimode_for_vec_perm): Declare.
15049 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15050 (qimode_for_vec_perm): ...this new function.
15051 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15053 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15055 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15056 does not have a conditional at the top.
15058 2018-01-02 Richard Biener <rguenther@suse.de>
15060 * ipa-inline.c (big_speedup_p): Fix expression.
15062 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15065 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15068 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15072 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15073 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15074 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15075 cond_taken_branch_cost 3->4.
15077 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15079 PR tree-optimization/83581
15080 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15081 TODO_cleanup_cfg if any changes have been made.
15083 PR middle-end/83608
15084 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15085 convert_modes if target mode has the right side, but different mode
15088 PR middle-end/83609
15089 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15090 last argument when extracting from CONCAT. If either from_real or
15091 from_imag is NULL, use expansion through memory. If result is not
15092 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15093 the parts directly to inner mode, if even that fails, use expansion
15096 PR middle-end/83623
15097 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15098 check for bswap in mode rather than HImode and use that in expand_unop
15101 Copyright (C) 2018 Free Software Foundation, Inc.
15103 Copying and distribution of this file, with or without modification,
15104 are permitted in any medium without royalty provided the copyright
15105 notice and this notice are preserved.