1 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6 several redundant entries.
8 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
11 type from "rtx" to "rtx_insn *".
12 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
13 for local "call_insn", removing cast.
14 (ix86_expand_call): Likewise, introducing a "call_insn" local.
16 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
18 PR tree-optimization/86066
19 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
20 for BIT_INSERT_EXPR stores.
22 2018-06-06 Richard Biener <rguenther@suse.de>
24 PR tree-optimization/86062
25 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
27 of to be offsetted base.
29 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
31 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
32 to be static and remove check on interrupt attribute name.
34 gcc/testsuite/gcc.target/msp430/
35 * function-attributes-4.c: New test.
36 * static-interrupts.c: New test.
38 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
40 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
41 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
43 2018-06-05 Steve Ellcey <sellcey@cavium.com>
46 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
48 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
49 Remove second argument, change how error is called.
50 (aarch64_layout_arg): Remove second argument from
51 aarch64_err_no_fpadvsimd call.
52 (aarch64_init_cumulative_args): Ditto.
53 (aarch64_gimplify_va_arg_expr): Ditto.
54 * config/aarch64/aarch64.md (mov<mode>): Ditto.
56 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
58 * config/i386/i386.md (simple_return_indirect_internal): New expander.
59 (*simple_return_indirect_internal<mode>): Rename from
60 simple_return_indirect_internal. Use W mode iterator.
61 (rstorssp): New expander.
62 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
63 (clrssbsy): New expander.
64 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
66 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
68 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
70 (cmse_check_pointed_object): Likewise.
72 2018-06-05 Martin Liska <mliska@suse.cz>
75 * doc/invoke.texi: Document how -fprofile-dir format
78 2018-06-05 Richard Biener <rguenther@suse.de>
80 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
81 removal pretend DOM info isn't available so we do not update
82 it and only remove edges, not dominated blocks. Actually free
83 DOM info in case we removed something. Remove unreachable blocks.
84 (mfb_keep_latches): Work with either DOM info or marked backedges.
85 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
86 first. Mark backedges if DOM info isn't available.
87 (Re-)compute DOM info after cleanup_control_flow_pre.
89 2018-06-05 Richard Biener <rguenther@suse.de>
91 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
92 (locus_discrim_hasher::hash): Adjust.
93 (locus_discrim_hasher::equal): Likewise.
94 (next_discriminator_for_locus): Work on line directly.
95 (same_line_p): Pass in expanded locus1 as well.
96 (assign_discriminators): Avoid redundant location expansions.
98 2018-06-05 Richard Biener <rguenther@suse.de>
100 PR tree-optimization/86046
101 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
102 if required after clearing TREE_ADDRESSABLE.
104 2018-06-05 Richard Biener <rguenther@suse.de>
106 PR tree-optimization/86047
107 * tree-ssa-loop.c (for_each_index): Glob handling of all
108 decls and constants and really handle all of them.
110 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
113 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
114 qualifier_void_pointer and qualifier_const_void_pointer.
115 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
116 (arm_init_builtins): Handle the above.
117 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
118 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
119 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
122 2018-06-05 Martin Liska <mliska@suse.cz>
124 * auto-profile.c (read_autofdo_file): Do not use
125 gcov_ctr_summary struct.
126 (afdo_callsite_hot_enough_for_early_inline): Likewise.
127 * coverage.c (struct counts_entry): Likewise.
128 (read_counts_file): Read just single summary entry.
129 (get_coverage_counts): Use gcov_summary struct.
130 * coverage.h (get_coverage_counts): Likewise.
131 * gcov-dump.c (dump_working_sets): Likewise.
132 (tag_summary): Dump just single summary.
133 * gcov-io.c (gcov_write_summary): Write just histogram
135 (gcov_read_summary): Read just single summary.
136 (compute_working_sets): Use gcov_summary struct.
137 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
138 of GCOV_COUNTERS_SUMMABLE.
139 (GCOV_COUNTERS_SUMMABLE): Remove.
140 (GCOV_FIRST_VALUE_COUNTER): Replace with
141 GCOV_COUNTER_V_INTERVAL.
142 (struct gcov_ctr_summary): Remove.
143 (struct gcov_summary): Directly use fields of former
145 (compute_working_sets): Use gcov_summary struct.
146 * gcov.c (read_count_file): Do not use ctrs fields.
147 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
149 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
151 * profile.c: Likewise.
152 * profile.h: Likewise.
154 2018-06-05 Martin Liska <mliska@suse.cz>
156 PR gcov-profile/84846
157 * gcov.c (output_lines): Print working directory only
158 in intermediate format.
160 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
162 * config/s390/s390-builtin-types.def: Add void function type.
163 * config/s390/s390-builtins.def: Use the function type for the
166 2018-06-04 Jim Wilson <jimw@sifive.com>
168 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
170 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
171 and maybe_eh_return. Change regno to unsigned int. Use new args to
172 handle EH_RETURN_DATA_REGNO registers properly.
173 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
174 (riscv_expand_epilogue): Update comment. Change argument name and
175 type. Update code to use new name and type. Pass new args to
176 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
178 * config/riscv/riscv.md (NORMAL_RETURN): New.
179 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
180 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
181 (eh_return): Call gen_eh_return_internal and emit barrier.
182 (eh_return_internal): Call riscv_expand_epilogue.
184 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
186 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
187 bit_insertion field and declare can_be_merged_into method.
188 (merged_store_group::can_be_merged_into): New method.
189 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
190 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
191 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
193 2018-06-04 Richard Biener <rguenther@suse.de>
195 PR tree-optimization/85955
196 * builtins.c (fold_builtin_sincos): Convert pointers to
197 destination to appropriate type before dereferencing.
199 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
201 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
203 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
205 * expr.c (expand_expr_real_1): Force the operand into memory if
206 its TYPE_MODE is BLKmode and if there is no integer mode for
207 the number of bits being extracted.
209 2018-06-04 Jakub Jelinek <jakub@redhat.com>
213 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
214 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
216 2018-06-04 Richard Biener <rguenther@suse.de>
218 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
219 (cleanup_tree_cfg_noloop): ... single caller. Do
220 start_recording_case_labels later.
222 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
224 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
225 to _IMMINTRIN_H_INCLUDED.
226 * config/i386/pconfigintrin.h: Ditto.
227 * config/i386/waitpkgintrin.h: Ditto.
228 * config/i386/immintrin.h: Add includes for sgxintrin.h,
229 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
230 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
231 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
232 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
233 waitpkgintrin.h and cldemoteintrin.h.
235 2018-06-04 Richard Biener <rguenther@suse.de>
237 PR tree-optimization/86038
238 * tracer.c (find_best_successor): Check probability for
239 being initialized, bail out if not.
241 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
244 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
245 of bits to ignore when comparing architectures.
247 2018-06-04 Jakub Jelinek <jakub@redhat.com>
249 PR tree-optimization/69615
250 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
251 maximum or minimum of the type, try to merge it also as if
252 range1 is + [-, x - 1] or + [x + 1, -].
255 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
257 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
259 PR tree-optimization/86034
260 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
261 the unsigned bitfield type in a bit insertion sequence if it does not
262 have a larger precision than the bitfield size.
263 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
265 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
267 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
269 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
271 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
272 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
273 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
274 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
276 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
278 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
279 Disable -fdelete-null-pointer-checks for ELF toolchain.
281 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
282 Kito Cheng <kito.cheng@gmail.com>
284 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
285 (nds32le-*-*, nds32be-*-*): Integrate checking process.
286 (nds32*-*-*): Add glibc and uclibc conditions.
287 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
288 (TARGET_EXCEPT_UNWIND_INFO): Define.
289 * config/nds32/elf.h: New file.
290 * config/nds32/linux.h: New file.
291 * config/nds32/nds32-elf.opt: New file.
292 * config/nds32/nds32-linux.opt: New file.
293 * config/nds32/nds32-fp-as-gp.c
294 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
295 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
297 (nds32_asm_file_end): Ditto.
298 (nds32_print_operand): Ditto.
299 (nds32_insert_attributes): Ditto.
300 (nds32_init_libfuncs): New function.
301 (TARGET_HAVE_TLS): Define.
302 (TARGET_INIT_LIBFUNCS): Define.
303 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
305 (TARGET_ELF): Apply different mcmodel setting.
306 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
307 been migrated into elf.h and linux.h files.
308 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
309 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
310 (mcmodel): The content has been migrated into nds32-elf.opt and
311 nds32-linux.opt files.
312 * config/nds32/t-elf: New file.
313 * config/nds32/t-linux: New file.
315 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
316 Shiva Chen <shiva0217@gmail.com>
318 * config/nds32/constants.md (unspec_volatile_element): Add
319 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
320 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
322 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
323 (make_pass_nds32_fp_as_gp): Declare.
324 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
326 (nds32_asm_function_end_prologue): Remove unused asm output.
327 (nds32_asm_function_begin_epilogue): Remove unused asm output.
328 (nds32_asm_file_start): Output necessary fp_as_gp information.
329 (nds32_option_override): Adjust register usage.
330 (nds32_expand_prologue): Consider fp_as_gp situation.
331 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
332 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
335 (simple_return): Ditto.
336 (omit_fp_begin): Output special directive for fp_as_gp.
337 (omit_fp_end): Output special directive for fp_as_gp.
338 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
339 mforbid-fp-as-gp): New options.
341 2018-06-01 Mark Wielaard <mark@klomp.org>
343 * dwarf2out.c (dwarf2out_finish): Remove generation of
346 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
348 * gimple-ssa-store-merging.c: Include gimple-fold.h.
349 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
350 (struct merged_store_group): Add bit_insertion field.
351 (dump_char_array): Use standard hexadecimal format.
352 (merged_store_group::merged_store_group): Set bit_insertion to false.
353 (merged_store_group::apply_stores): Use optimal buffer size. Deal
354 with BIT_INSERT_EXPR stores. Move up code updating the mask and
355 also print the mask in the dump file.
356 (pass_store_merging::gate): Minor tweak.
357 (imm_store_chain_info::coalesce_immediate): Fix wrong association
358 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
359 stores with INTEGER_CST stores.
360 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
361 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
362 and use it throughout. Generate bit insertion sequences if need be.
363 (pass_store_merging::process_store): Remove redundant condition.
364 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
366 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
368 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
369 the 128-bit floating point types. Fix function comment.
371 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
373 * config/aarch64/aarch64-simd.md
374 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
376 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
379 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
381 PR tree-optimization/85989
382 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
384 (backprop::intersect_uses): Check it when deciding whether this
385 is a backedge reference.
386 (backprop::process_block): Add each phi to m_visited_phis
387 after visiting it, then clear it at the end.
389 2018-06-01 Richard Biener <rguenther@suse.de>
391 * tree-vectorizer.h (vect_dr_stmt): New function.
392 (vect_get_load_cost): Adjust.
393 (vect_get_store_cost): Likewise.
394 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
395 Use vect_dr_stmt instead of DR_SMTT.
396 (vect_record_base_alignments): Likewise.
397 (vect_calculate_target_alignment): Likewise.
398 (vect_compute_data_ref_alignment): Likewise and make static.
399 (vect_update_misalignment_for_peel): Likewise.
400 (vect_verify_datarefs_alignment): Likewise.
401 (vector_alignment_reachable_p): Likewise.
402 (vect_get_data_access_cost): Likewise. Pass down
403 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
404 (vect_get_peeling_costs_all_drs): Likewise.
405 (vect_peeling_hash_get_lowest_cost): Likewise.
406 (vect_enhance_data_refs_alignment): Likewise.
407 (vect_find_same_alignment_drs): Likewise.
408 (vect_analyze_data_refs_alignment): Likewise.
409 (vect_analyze_group_access_1): Likewise.
410 (vect_analyze_group_access): Likewise.
411 (vect_analyze_data_ref_access): Likewise.
412 (vect_analyze_data_ref_accesses): Likewise.
413 (vect_vfa_segment_size): Likewise.
414 (vect_small_gap_p): Likewise.
415 (vectorizable_with_step_bound_p): Likewise.
416 (vect_prune_runtime_alias_test_list): Likewise.
417 (vect_analyze_data_refs): Likewise.
418 (vect_supportable_dr_alignment): Likewise.
419 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
420 (vect_gen_prolog_loop_niters): Likewise.
421 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
422 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
424 (vect_recog_mask_conversion_pattern): Likewise.
425 (vect_try_gather_scatter_pattern): Likewise.
426 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
427 to vect_get_store_cost.
428 (vect_get_store_cost): Get stmt_info instead of DR.
429 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
430 (vect_get_load_cost): Get stmt_info instead of DR.
432 2018-06-01 Richard Biener <rguenther@suse.de>
435 * gimple-fold.c (var_decl_component_p): Also allow offsetted
436 vars wrapped in MEM_REFs.
438 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
440 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
441 Fix subreg tests so that we only return a choice between
442 GENERAL_REGS and FP_REGS if the original classes included both.
444 2018-06-01 Richard Biener <rguenther@suse.de>
447 * tree-ssa-structalias.c (get_function_part_constraint):
448 Handle NULL fi->decl.
449 (find_func_aliases_for_call): Properly handle indirect
451 (find_func_clobbers): Likewise.
452 (ipa_pta_execute): Likewise.
453 (create_variable_info_for): For functions that are ifunc_resolver
454 resolve to a varinfo that contains the result of the resolver call.
455 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
458 2018-05-31 Michael Collison <michael.collison@arm.com>
460 * config/aarch64/aarch64.md:
461 (*fix_to_zero_extenddfdi2): New pattern.
462 * gcc.target/aarch64/fix_extend1.c: New testcase.
464 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
468 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
469 and BUILT_IN_STRNCMP_EQ.
470 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
472 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
473 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
474 (gimple_fold_builtin): Likewise.
475 * tree-ssa-strlen.c (compute_string_length): New function.
476 (determine_min_obsize): New function.
477 (handle_builtin_string_cmp): New function to handle calls to
478 string compare functions.
479 (strlen_optimize_stmt): Add handling to builtin string compare
481 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
482 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
483 * tree.c (build_common_builtin_nodes): Add new defines of
484 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
486 2018-05-31 Jakub Jelinek <jakub@redhat.com>
489 * bb-reorder.c (pass_partition_blocks::gate): Return false for
490 functions with naked attribute.
492 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
494 * config/i386/sse.md (avx_vec_concat<mode>):
495 Substitute concat_tg_mode mode attribute with xtg_mode.
496 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
497 (concat_tg_mode): Remove mode attribute.
499 2018-05-31 Martin Sebor <msebor@redhat.com>
502 * calls.c (alloc_max_size): Correct a logic error/typo.
503 Treat excessive arguments as infinite. Warn for invalid arguments.
504 * doc/invoke.texi (-Walloc-size-larger-than): Update.
506 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
509 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
510 and movx for Haswell.
512 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
513 Cesar Philippidis <cesar@codesourcery.com>
516 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
517 when emitting error on private/firstprivate reductions.
518 * omp-low.c (lower_omp_target): Avoid reference-type processing
519 on pointers for firstprivate clause.
521 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
523 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
526 * config/aarch64/aarch64-simd.md
527 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
528 (aarch64_ld1_x3_<mode>): Likewise
529 (aarch64_st1x2<VALLDIF:mode>): Likewise
530 (aarch64_st1_x2_<mode>): Likewise
531 (aarch64_st1x3<VALLDIF:mode>): Likewise
532 (aarch64_st1_x3_<mode>): Likewise
533 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
534 (vld1_s8_x3): Likewise.
535 (vld1_u16_x3): Likewise.
536 (vld1_s16_x3): Likewise.
537 (vld1_u32_x3): Likewise.
538 (vld1_s32_x3): Likewise.
539 (vld1_u64_x3): Likewise.
540 (vld1_s64_x3): Likewise.
541 (vld1_f16_x3): Likewise.
542 (vld1_f32_x3): Likewise.
543 (vld1_f64_x3): Likewise.
544 (vld1_p8_x3): Likewise.
545 (vld1_p16_x3): Likewise.
546 (vld1_p64_x3): Likewise.
547 (vld1q_u8_x3): Likewise.
548 (vld1q_s8_x3): Likewise.
549 (vld1q_u16_x3): Likewise.
550 (vld1q_s16_x3): Likewise.
551 (vld1q_u32_x3): Likewise.
552 (vld1q_s32_x3): Likewise.
553 (vld1q_u64_x3): Likewise.
554 (vld1q_s64_x3): Likewise.
555 (vld1q_f16_x3): Likewise.
556 (vld1q_f32_x3): Likewise.
557 (vld1q_f64_x3): Likewise.
558 (vld1q_p8_x3): Likewise.
559 (vld1q_p16_x3): Likewise.
560 (vld1q_p64_x3): Likewise.
561 (vst1_s64_x2): Likewise.
562 (vst1_u64_x2): Likewise.
563 (vst1_f64_x2): Likewise.
564 (vst1_s8_x2): Likewise.
565 (vst1_p8_x2): Likewise.
566 (vst1_s16_x2): Likewise.
567 (vst1_p16_x2): Likewise.
568 (vst1_s32_x2): Likewise.
569 (vst1_u8_x2): Likewise.
570 (vst1_u16_x2): Likewise.
571 (vst1_u32_x2): Likewise.
572 (vst1_f16_x2): Likewise.
573 (vst1_f32_x2): Likewise.
574 (vst1_p64_x2): Likewise.
575 (vst1q_s8_x2): Likewise.
576 (vst1q_p8_x2): Likewise.
577 (vst1q_s16_x2): Likewise.
578 (vst1q_p16_x2): Likewise.
579 (vst1q_s32_x2): Likewise.
580 (vst1q_s64_x2): Likewise.
581 (vst1q_u8_x2): Likewise.
582 (vst1q_u16_x2): Likewise.
583 (vst1q_u32_x2): Likewise.
584 (vst1q_u64_x2): Likewise.
585 (vst1q_f16_x2): Likewise.
586 (vst1q_f32_x2): Likewise.
587 (vst1q_f64_x2): Likewise.
588 (vst1q_p64_x2): Likewise.
589 (vst1_s64_x3): Likewise.
590 (vst1_u64_x3): Likewise.
591 (vst1_f64_x3): Likewise.
592 (vst1_s8_x3): Likewise.
593 (vst1_p8_x3): Likewise.
594 (vst1_s16_x3): Likewise.
595 (vst1_p16_x3): Likewise.
596 (vst1_s32_x3): Likewise.
597 (vst1_u8_x3): Likewise.
598 (vst1_u16_x3): Likewise.
599 (vst1_u32_x3): Likewise.
600 (vst1_f16_x3): Likewise.
601 (vst1_f32_x3): Likewise.
602 (vst1_p64_x3): Likewise.
603 (vst1q_s8_x3): Likewise.
604 (vst1q_p8_x3): Likewise.
605 (vst1q_s16_x3): Likewise.
606 (vst1q_p16_x3): Likewise.
607 (vst1q_s32_x3): Likewise.
608 (vst1q_s64_x3): Likewise.
609 (vst1q_u8_x3): Likewise.
610 (vst1q_u16_x3): Likewise.
611 (vst1q_u32_x3): Likewise.
612 (vst1q_u64_x3): Likewise.
613 (vst1q_f16_x3): Likewise.
614 (vst1q_f32_x3): Likewise.
615 (vst1q_f64_x3): Likewise.
616 (vst1q_p64_x3): Likewise.
618 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
620 * config/msp430/msp430.c (msp430_output_labelref): Prepend
621 user_label_prefix to name.
623 * tree-core.h: Update comment about the format of NAME string
624 passed to handler in attribute_spec.
626 * config/msp430/msp430.md: Remove erroneous subreg expression from
627 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
628 zero_extend{q,h}isi2.
630 2018-05-30 Borislav Petkov <bp@suse.de>
632 * doc/extend.texi: Document some architecture specific
633 constraints and sort entries.
635 2018-05-30 Martin Sebor <msebor@redhat.com>
638 * builtins.c (expand_builtin_stpcpy_1): New function.
639 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
640 only if the former succeeds.
642 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
644 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
647 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
649 * doc/invoke.texi (-flinker-output): Document
651 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
653 * passes.c (ipa_write_summaries): Only modify statements if body
655 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
656 incrementally linking.
657 (ipa_passes): Likewise.
658 * lto-cgraph.c (lto_output_node): When incrementally linking do not
659 pass down resolution info.
660 * common.opt (flag_incremental_link): Update info.
661 * gcc.c (plugin specs): Turn flinker-output=* to
662 -plugin-opt=-linker-output-known
663 * toplev.c (compile_file): Also cut compilation when doing incremental
665 * flag-types. (enum lto_partition_model): Add
666 LTO_LINKER_OUTPUT_NOLTOREL.
667 (invoke.texi): Add -flinker-output docs.
668 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
669 link same way as WPA; do not stream in dead initializers.
671 * dwarf2out.c (dwarf2out_die_ref_for_decl,
672 darf2out_register_external_decl): Support incremental link.
674 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
676 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
678 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
680 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
681 it down to simple_object_copy_lto_debug_sections.
682 (run_gcc): Determine incremental LTO link time and configure
683 lto1 into non-wpa mode, disable renaming of debug sections.
685 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
687 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
688 descriptions of various incorrectly documented functions.
690 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
693 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
694 address check not strict.
696 2018-05-30 Richard Biener <rguenther@suse.de>
698 PR tree-optimization/85964
699 * tracer.c (better_p): Drop initialized count check, we only
700 call the function with initialized counts now.
701 (find_best_successor): Do find a best edge if one
702 has uninitialized count.
703 (find_best_predecessor): Likewise. Do BB frequency check only
704 if count is initialized.
706 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
708 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
709 (aarch64_ldrstr_offset_compare): New.
710 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
711 load/store orderings.
712 (aarch64_gen_adjusted_ldpstp): Likewise.
714 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
716 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
717 Check for subset of GENERAL_REGS and FP_REGS.
718 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
721 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
723 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
724 and wi::to_poly_offset. Add the current offset and then check
725 whether the sum fits, rather than using an unchecked addition of
726 a checked term. Check for a shwi rather than a uhwi.
727 * expr.c (get_bit_range): Use tree_to_poly_uint64.
728 (store_constructor): Use poly_int_tree_p.
729 (expand_expr_real_1): Likewise.
730 * function.c (assign_temp): Likewise.
731 * fold-const.c (const_binop): Use poly_int_tree_p and
733 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
735 * ipa-icf-gimple.c (func_checker::compare_operand): Use
736 to_poly_offset for MEM offsets.
737 * ipa-icf.c (sem_variable::equals): Likewise.
738 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
739 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
740 wi::to_poly_offset for BIT_FIELD_REF offsets.
741 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
743 * var-tracking.c (emit_note_insn_var_location): Use
746 2018-05-29 Jim Wilson <jimw@sifive.com>
748 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
750 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
753 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
754 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
756 (sse4_1_round<mode>2): Use nonimmediate_operand
757 for operand 1 predicate.
759 2018-05-29 Martin Sebor <msebor@redhat.com>
760 Richard Biener <rguenther@suse.de>
763 * calls.c (get_size_range): Call determine_value_range instead
765 * tree-vrp.h (determine_value_range): Declared new function.
766 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
768 2018-05-29 Richard Biener <rguenther@suse.de>
770 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
771 sure to use non-pattern stmts for get_earlier_stmt arguments.
772 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
773 called on pattern stmts.
774 (get_later_stmt): Likewise.
776 2018-05-29 Martin Liska <mliska@suse.cz>
778 PR gcov-profile/85759
779 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
782 2018-05-29 Jakub Jelinek <jakub@redhat.com>
784 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
786 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
790 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
791 VEC_PACK_FLOAT_EXPR): New tree codes.
792 * tree-pretty-print.c (op_code_prio): Handle
793 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
794 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
795 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
796 * tree-inline.c (estimate_operator_cost): Likewise.
797 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
798 * fold-const.c (const_binop): Likewise.
799 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
800 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
801 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
802 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
803 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
804 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
805 * expr.c (expand_expr_real_2): Likewise.
806 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
807 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
808 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
810 * optabs.c (expand_widen_pattern_expr): For
811 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
812 sign from result type rather than operand's type.
813 (expand_binop_directly): For vec_packu_float_optab and
814 vec_packs_float_optab allow result type to be different from operand's
816 * optabs-tree.c (optab_for_tree_code): Handle
817 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
818 VEC_PACK_FLOAT_EXPR. Formatting fixes.
819 * tree-vect-generic.c (expand_vector_operations_1): Handle
820 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
822 * tree-vect-stmts.c (supportable_widening_operation): Handle
824 (supportable_narrowing_operation): Handle FLOAT_EXPR.
825 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
826 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
827 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
828 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
830 (vec_pack<floatprefix>_float_<mode>): New expander.
831 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
833 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
834 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
835 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
836 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
837 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
839 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
840 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
841 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
842 VEC_PACK_FLOAT_EXPR): Document.
844 2018-05-29 Richard Biener <rguenther@suse.de>
846 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
848 (stmt_vec_info_vec): Make pointer.
849 (init_stmt_vec_info_vec): Remove.
850 (free_stmt_vec_info_vec): Likewise.
851 (set_stmt_vec_info_vec): New function.
852 (free_stmt_vec_infos): Likewise.
853 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
854 (set_vinfo_for_stmt): Likewise.
855 (get_earlier_stmt): Likewise.
856 (get_later_stmt): Likewise.
857 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
858 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
859 (vec_info::~vec_info): Free stmt_vec_infos.
860 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
861 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
862 (pass_slp_vectorize::execute): Likewise.
863 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
864 (free_stmt_vec_info_vec): Likewise.
865 (set_stmt_vec_info_vec): New function.
866 (free_stmt_vec_infos): Likewise.
867 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
868 the global stmt_vec_info_vec.
869 * tree-parloops.c (gather_scalar_reductions): Use
870 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
873 2018-05-29 Richard Biener <rguenther@suse.de>
875 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
877 2018-05-29 Martin Liska <mliska@suse.cz>
878 David Malcolm <dmalcolm@redhat.com>
880 * vec.c (test_reverse): New.
881 (vec_c_tests): Add new test.
882 * vec.h (vl_ptr>::reverse): New function.
884 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
886 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
888 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
891 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
893 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
895 2018-05-28 Richard Biener <rguenther@suse.de>
897 PR tree-optimization/85933
898 * tree-vect-data-refs.c (vect_record_base_alignments): Only
899 look at stmts marked as vectorizable.
901 2018-05-28 Richard Biener <rguenther@suse.de>
903 PR tree-optimization/85934
904 * tree-vect-generic.c (expand_vector_operations_1): Hoist
905 vector boolean check before scalar optimization.
907 2018-05-28 Jakub Jelinek <jakub@redhat.com>
909 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
912 2018-05-28 Mark Wielaard <mark@klomp.org>
914 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
915 if it is an expression containing a minus sign.
917 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
919 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
921 2018-05-27 Paul Koning <ni1d@arrl.net>
923 * config/pdp11/pdp11.md (truncsihi2): Remove.
925 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
926 Chung-Ju Wu <jasonwucj@gmail.com>
928 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
930 (unaligned_store_dw): Ditto.
931 * config/nds32/nds32-memory-manipulation.c
932 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
933 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
934 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
935 (emit_setmem_word_loop): Rename to ...
936 (emit_setmem_doubleword_loop): ... this.
937 (nds32_gen_dup_4_byte_to_word_value): New function.
938 (nds32_gen_dup_8_byte_to_double_word_value): New function.
939 (nds32_expand_setmem_loop): Refine implementation.
940 (nds32_expand_setmem_loop_v3m): Ditto.
941 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
944 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
946 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
948 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
950 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
951 (nds32_init_machine_status): Initialize machine->attr_naked_p and
952 machine->attr_no_prologue_p.
953 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
954 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
955 (nds32_expand_epilogue): Consider attr_naked_p.
956 (nds32_expand_epilogue_v3pop): Likewise.
957 (nds32_can_use_return_insn): Likewise.
958 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
959 attr_no_prologue_p fields.
960 * config/nds32/nds32.opt (mret-in-naked-func): New option.
962 2018-05-27 Jakub Jelinek <jakub@redhat.com>
965 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
968 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
970 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
972 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
974 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
976 (*<floatsuffix>floatv2div2sf2): Rename to ...
977 (*float<floatunssuffix>v2div2sf2): ... this.
978 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
979 (float<floatunssuffix>v2div2sf2_mask): ... this.
980 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
981 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
982 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
984 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
986 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
988 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
990 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
992 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
994 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
995 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
996 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
997 gen_ufix_truncv8dfv8si2.
998 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
999 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1000 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1001 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1002 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1003 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1004 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1005 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1007 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1011 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1013 2018-05-25 Jim Wilson <jimw@sifive.com>
1015 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1016 * config/riscv/riscv.c (struct machine_function): Add
1017 interrupt_handler_p and attribute_checked_p fields.
1018 (riscv_attribute_table): Add interrupt.
1019 (riscv_interrupt_type_p): New.
1020 (riscv_save_reg_p): Save extra regs for interrupt handler.
1021 (riscv_use_save_libcall): Return false for interrupt handler.
1022 (riscv_first_stack_step): Add forward declaration.
1023 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1024 for interrupt handler with large frame. Use it for saved reg list.
1025 (riscv_expand_prologue): Move flag_stack_usage_info support to
1026 eliminate duplication.
1027 (riscv_expand_epilogue): Generate mret for interrupt handler.
1028 (riscv_epilogue_uses): New.
1029 (riscv_can_use_return_insn): Return false for interrupt handler.
1030 (riscv_function_ok_for_sibcall): Likewise.
1031 (riscv_set_current_function): Add interrupt handler support.
1032 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1033 * config/riscv/riscv.md (UNSPECV_MRET): New.
1035 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1037 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1039 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1041 PR tree-optimization/85712
1042 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1043 this candidate has already been replaced in-situ by a copy.
1045 2018-05-25 Jason Merrill <jason@redhat.com>
1047 PR c++/80485 - inline function non-zero address.
1048 * symtab.c (nonzero_address): Check DECL_COMDAT.
1050 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1053 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1054 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1055 extension of SImode operation. Use const123_operand predicate.
1056 (*saddsi_1): Remove.
1057 (*saddl_se_1): Ditto.
1059 (*ssubl_se_1): Ditto.
1060 * config/alpha/predicates.md (const123_operand): New predicate.
1061 * config/alpha/constraints.md (P): Use IN_RANGE.
1063 2018-05-25 Richard Biener <rguenther@suse.de>
1065 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1067 (ref_maybe_used_by_stmt_p): Likewise.
1068 (stmt_may_clobber_ref_p): Likewise.
1069 (stmt_may_clobber_ref_p_1): Likewise.
1070 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1072 (ref_maybe_used_by_stmt_p): Likewise.
1073 (stmt_may_clobber_ref_p): Likewise.
1074 (stmt_may_clobber_ref_p_1): Likewise.
1075 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1076 the alias oracle to disambiguate DRs with stmts DR analysis
1078 (vect_analyze_data_refs): Do not give up on not analyzable
1079 DRs for BB vectorization. Remove code truncating the dataref
1082 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1085 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1086 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1087 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1089 2018-05-25 Richard Biener <rguenther@suse.de>
1091 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1092 function, combining stmt data ref gathering and fatal analysis
1094 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1095 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1096 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1097 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1098 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1100 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1102 PR tree-optimization/85720
1103 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1104 SCC if all partitions are builtins.
1105 (version_loop_by_alias_check): New parameter. Generate cancelable
1106 runtime alias check if all partitions are builtins.
1107 (distribute_loop): Update call to above function.
1109 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1111 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1112 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1113 (parm_default_def_partition_arg): Ditto.
1114 (set_parm_default_def_partition): Ditto.
1115 (get_parm_default_def_partitions): Ditto and make it static.
1116 (get_undefined_value_partitions): Ditto and make it static.
1117 (remove_ssa_form): Refactor call to init_var_map here.
1118 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1119 computation for loop region.
1120 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1121 (register_default_def): Delete.
1122 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1123 (parm_default_def_partition_arg): Ditto.
1124 (set_parm_default_def_partition): Ditto.
1125 (get_parm_default_def_partitions): Ditto and make it static.
1126 (get_undefined_value_partitions): Ditto and make it static.
1127 (coalesce_with_default, coalesce_with_default): Update comment.
1128 (create_coalesce_list_for_region): New func factored out from
1129 create_outofssa_var_map.
1130 (populate_coalesce_list_for_outofssa): New func factored out from
1131 create_outofssa_var_map and coalesce_ssa_name.
1132 (create_outofssa_var_map): Delete.
1133 (coalesce_ssa_name): Refactor to support live range computation.
1134 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1135 (get_parm_default_def_partitions): Delete.
1136 (get_undefined_value_partitions): Ditto.
1137 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1138 computation for loop region.
1139 (new_tree_live_info, loe_visit_block): Ditto.
1140 (live_worklist, set_var_live_on_entry): Ditto.
1141 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1142 * tree-ssa-live.h (struct _var_map): New fields.
1143 (init_var_map): Change decl.
1144 (region_contains_p): New.
1146 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1148 * tree-ssa-live.h (live_merge_and_clear): Delete.
1150 2018-05-25 Richard Biener <rguenther@suse.de>
1153 * tree-dump.c (dequeue_and_dump): Remove access to removed
1154 operand 2 of a SWITCH_EXPR.
1156 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1158 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1159 multiplication and division.
1160 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1161 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1162 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1163 (cond_udiv_optab, cond_umod_optab): New optabs.
1164 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1165 (IFN_COND_RDIV): New internal functions.
1166 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1167 TRUNC_MOD_EXPR and RDIV_EXPR.
1168 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1169 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1171 (SVE_INT_BINARY): Include mult.
1172 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1173 (optab, sve_int_op): Handle mult.
1174 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1176 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1177 for SVE_INT_BINARY_SD.
1179 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1181 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1182 (optab, sve_int_op): Handle div and udiv.
1183 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1184 for SVE_INT_BINARY_SD.
1185 (*<optab><mode>3): New insn for the same.
1187 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1189 * tree-vect-patterns.c: Include predict.h.
1190 (vect_recog_divmod_pattern): Restrict check for division support
1191 to when optimizing for size.
1193 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1195 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1196 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1197 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1198 (gimple_match_op::set_op): Likewise.
1199 (gimple_resimplify4): Declare.
1200 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1201 (expr::gen_transform): Likewise.
1202 (decision_tree::gen): Generate a simplification routine for 4 operands.
1203 * gimple-match-head.c (gimple_simplify): Add an overload for
1204 4 operands. In the top-level function, handle up to 4 call
1205 arguments and call gimple_resimplify4.
1206 (gimple_resimplify4): New function.
1207 (build_call_internal): Pass a fourth operand.
1208 (maybe_push_to_seq): Likewise.
1209 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1210 Fold VEC_COND_EXPRs of an operation and a default value into
1211 an IFN_COND_* function if possible.
1212 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1214 (SVE_COND_FP_BINARY): Include them.
1215 (optab, sve_fp_op): Handle them.
1216 (SVE_INT_BINARY_REV): New code iterator.
1217 (SVE_COND_FP_BINARY_REV): New int iterator.
1218 (commutative): New int attribute.
1219 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1221 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1223 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1224 (*cond_<optab><mode>): New patterns for reversed operands.
1226 2018-05-25 Richard Biener <rguenther@suse.de>
1228 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1229 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1230 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1231 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1232 * tree-vect-data-refs.c (everywhere): Adjust users.
1233 * tree-vect-loop.c (everywhere): Likewise.
1234 * tree-vect-slp.c (everywhere): Likewise.
1235 * tree-vect-stmts.c (everywhere): Likewise.
1236 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1238 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1240 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1242 (gcc_cv_as_section_exclude): ... this.
1243 Try Solaris as #exclude syntax.
1244 * configure: Regenerate.
1245 * config.in: Regenerate.
1246 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1248 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1249 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1251 * varasm.c (default_elf_asm_named_section): Don't check if
1252 HAVE_GAS_SECTION_EXCLUDE is defined.
1254 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1256 * doc/md.texi: Update the documentation of the cond_* optabs
1257 to mention the new final operand. Fix GET_MODE_NUNITS call.
1258 Describe the scalar case too.
1259 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1260 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1262 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1263 (get_conditional_internal_fn): Update comment.
1264 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1265 accumulator value as a final argument to conditional functions.
1266 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1267 a define_expand and add an "else" operand. Assert for now that
1268 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1269 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1270 (*cond_<optab><mode>): New patterns.
1271 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1272 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1273 (UNSPEC_COND_EOR): Delete.
1274 (optab): Remove associated mappings.
1275 (SVE_INT_BINARY): New code iterator.
1276 (sve_int_op): Remove int attribute and add "minus" to the code
1278 (SVE_COND_INT_OP): Delete.
1279 (SVE_COND_FP_OP): Rename to...
1280 (SVE_COND_FP_BINARY): ...this.
1282 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1284 * optabs.c (can_reuse_operands_p): New function.
1285 (maybe_legitimize_operands): Try to reuse the results for
1288 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1290 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1291 Add {q} suffix to insn mnemonic.
1293 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1295 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1296 (msp430_warn_func_return): New.
1298 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1300 * fold-const.c (tree_nonzero_bits): New function.
1301 * fold-const.h (tree_nonzero_bits): Likewise.
1302 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1303 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1305 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1309 * varasm.c (assemble_alias): Check ifunc_resolver only on
1312 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1315 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1316 when memory input operand is handled.
1318 2018-05-24 Luis Machado <luis.machado@linaro.org>
1320 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1322 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1324 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1326 * match.pd: Delay FMA folds until after vectorization.
1328 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1331 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1332 address check not strict.
1334 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1336 * gimple-match.h (gimple_match_op): New class.
1337 (mprts_hook): Replace parameters with a gimple_match_op *.
1338 (maybe_build_generic_op): Likewise.
1339 (gimple_simplified_result_is_gimple_val): Replace parameters with
1340 a const gimple_match_op *.
1341 (gimple_simplify): Replace code_helper * and tree * parameters with
1342 a gimple_match_op * parameter.
1343 (gimple_resimplify1): Replace code_helper *, tree and tree *
1344 parameters with a gimple_match_op * parameter.
1345 (gimple_resimplify2): Likewise.
1346 (gimple_resimplify3): Likewise.
1347 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1348 parameters with a gimple_match_op * parameter.
1349 * gimple-match-head.c (gimple_simplify): Change prototypes of
1350 auto-generated functions to take a gimple_match_op * instead of
1351 separate code_helper * and tree * parameters. Make the same
1352 change in the top-level overload and update calls to the
1353 gimple_resimplify routines. Update calls to the auto-generated
1354 functions and to maybe_push_res_to_seq in the publicly-facing
1355 operation-specific gimple_simplify overloads.
1356 (gimple_match_op::MAX_NUM_OPS): Define.
1357 (gimple_resimplify1): Replace rcode and ops with a single res_op
1358 parameter. Update call to gimple_simplify.
1359 (gimple_resimplify2): Likewise.
1360 (gimple_resimplify3): Likewise.
1361 (mprts_hook): Replace parameters with a gimple_match_op *.
1362 (maybe_build_generic_op): Likewise.
1363 (build_call_internal): Replace type, nargs and ops with
1364 a gimple_match_op *.
1365 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1366 with a single gimple_match_op *. Update calls to mprts_hook,
1367 build_call_internal and gimple_simplified_result_is_gimple_val.
1368 Factor out code that is common to the tree_code and combined_fn cases.
1369 * genmatch.c (expr::gen_transform): Replace tem_code and
1370 tem_ops with a gimple_match_op called tem_op. Update calls
1371 to the gimple_resimplify functions and maybe_push_res_to_seq.
1372 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1373 res_ops. Update call to the gimple_resimplify functions.
1374 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1375 (decision_tree::gen): Make the functions take a gimple_match_op *
1376 called res_op instead of separate res_code and res_ops parameters.
1377 Update call accordingly.
1378 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1379 and ops with a single res_op parameter. Update calls to
1380 maybe_build_generic_op and maybe_push_res_to_seq.
1381 (fold_stmt_1): Update calls to gimple_simplify and
1382 replace_stmt_with_simplification.
1383 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1384 and gimple_simplified_result_is_gimple_val.
1385 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1387 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1388 with a gimple_match_op *.
1389 (vn_nary_build_or_lookup): Likewise. Update call to
1390 vn_nary_build_or_lookup_1.
1391 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1392 gimple_match_op *. Update calls to the gimple_resimplify routines
1393 and to gimple_simplified_result_is_gimple_val.
1394 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1395 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1396 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1397 (visit_nary_op): Likewise.
1398 (visit_reference_op_load): Likewise.
1400 2018-05-23 Luis Machado <luis.machado@linaro.org>
1402 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1403 modifier for printing the step amount.
1405 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1408 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1411 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1413 * doc/sourcebuild.texi (Endianness): New subsubsection.
1415 2018-05-23 Luis Machado <luis.machado@linaro.org>
1417 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1418 <prefetch_dynamic_strides>: New const bool field.
1419 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1420 prefetch_dynamic_strides.
1421 (exynosm1_prefetch_tune): Likewise.
1422 (thunderxt88_prefetch_tune): Likewise.
1423 (thunderx_prefetch_tune): Likewise.
1424 (thunderx2t99_prefetch_tune): Likewise.
1425 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1427 (aarch64_override_options_internal): Update to set
1428 PARAM_PREFETCH_DYNAMIC_STRIDES.
1429 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1430 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1431 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1432 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1433 prefetch-dynamic-strides setting.
1435 2018-05-23 Luis Machado <luis.machado@linaro.org>
1437 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1438 <minimum_stride>: New const int field.
1439 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1440 minimum_stride field defaulting to -1.
1441 (exynosm1_prefetch_tune): Likewise.
1442 (thunderxt88_prefetch_tune): Likewise.
1443 (thunderx_prefetch_tune): Likewise.
1444 (thunderx2t99_prefetch_tune): Likewise.
1445 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1446 <default_opt_level>: Set to 3.
1447 (aarch64_override_options_internal): Update to set
1448 PARAM_PREFETCH_MINIMUM_STRIDE.
1449 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1450 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1451 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1452 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1453 stride is constant and is below the minimum stride threshold.
1455 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1457 * config/arm/arm-cpus.in (mode26): Delete.
1458 (armv4): Delete mode26 reference.
1459 * config/arm/arm.c (arm_configure_build_target): Delete use of
1462 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1464 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1466 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1467 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1469 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1470 (floatunsdidf2): Ditto.
1472 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1474 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1475 (fixuns_trunc<mode>si2_avx512f): Ditto.
1476 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1477 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1478 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1480 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1482 PR rtl-optimization/79985
1483 * df-scan.c (df_insn_refs_collect): Remove special case for
1484 global registers and asm statements.
1486 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1488 * extend.texi (Global Register Variables): Rewrite the bullet list.
1489 Note that the register is available for allocation. Note that access
1490 via inline asm must use constraints. Add note about async-signal
1491 handlers. Remove paragraph about automagic register selection.
1493 2018-05-23 Richard Biener <rguenther@suse.de>
1495 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1496 of fixed offset from memset VN.
1498 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1500 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1502 (alloc_cand_and_find_basis): Initialize first_interp field.
1503 (slsr_process_mul): Modify first_interp field.
1504 (slsr_process_add): Likewise.
1505 (slsr_process_cast): Modify first_interp field for each new
1507 (slsr_process_copy): Likewise.
1508 (dump_candidate): Dump first_interp field.
1509 (replace_mult_candidate): Process all interpretations, not just
1511 (replace_rhs_if_not_dup): Likewise.
1512 (replace_one_candidate): Likewise.
1514 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1516 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1518 (aarch64_needs_frame_chain): New function.
1519 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1521 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1524 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1525 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1526 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1527 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1528 as true for strict-align.
1529 (aarch64_can_inline_p): Perform checks even when callee has no
1530 attributes to check for strict alignment.
1531 * doc/extend.texi (AArch64 Function Attributes): Document
1533 * doc/invoke.texi: (AArch64 Options): Likewise.
1535 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1537 PR tree-optimization/85853
1538 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1539 the handling of the root of the node to...
1540 (vect_slp_analyze_node_operations_1): ...this new function,
1541 and run the whole thing with the child nodes' def types
1542 set according to their SLP node's def type.
1544 2018-05-23 Richard Biener <rguenther@suse.de>
1547 * tree-data-ref.c (create_runtime_alias_checks): Defer
1548 and ignore overflow warnings.
1550 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1552 PR tree-optimization/85822
1553 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1556 2018-05-23 Richard Biener <rguenther@suse.de>
1558 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1559 memset constants via native_interpret_expr.
1561 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1564 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1566 (cgraph_node::create_alias): Likewise.
1567 (cgraph_node::get_availability): Check ifunc_resolver instead
1568 of looking up ifunc attribute.
1569 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1570 * varasm.c (do_assemble_alias): Likewise.
1571 (assemble_alias): Likewise.
1572 (default_binds_local_p_3): Likewise.
1573 * cgraph.h (cgraph_node): Add ifunc_resolver.
1574 (cgraph_node::only_called_directly_or_aliased_p): Return false
1576 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1578 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1579 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1580 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1581 instead of looking up ifunc attribute.
1583 2018-05-22 Luis Machado <luis.machado@linaro.org>
1585 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1587 2018-05-22 Martin Sebor <msebor@redhat.com>
1590 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1591 only when expasion succeeds.
1592 (expand_builtin_strcmp): Same.
1593 (expand_builtin_strncmp): Same.
1595 2018-05-22 Martin Sebor <msebor@redhat.com>
1597 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1599 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1600 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1602 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1603 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1605 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1606 Add check for legitimate_address.
1607 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1608 (aarch64_swap_ldrstr_operands): New.
1609 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1612 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1613 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1615 * config/aarch64/aarch64.md: New patterns to generate stp
1617 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1618 single words and double words.
1619 (load_pair_sw, load_pair_dw): Likewise.
1620 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1622 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1624 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1625 for different mode ldpstp and add peephole for merged zero stores.
1627 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1629 (aarch64_gen_store_pair): Rename calls to match new patterns.
1630 (aarch64_gen_load_pair): Rename calls to match new patterns.
1631 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1632 (load_pair<DREG:mode><DREG2:mode>): ... This.
1633 (store_pair<mode>): Rename to...
1634 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1635 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1637 (V_INT_EQUIV): Handle SImode.
1638 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1641 2018-05-22 Martin Sebor <msebor@redhat.com>
1644 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1645 or ajust the presumed bound on an operation to avoid unnecessary
1648 2018-05-22 Martin Sebor <msebor@redhat.com>
1650 PR tree-optimization/85826
1651 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1652 assuming that a DECL necesarily has a constant size.
1654 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1657 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1659 2018-05-22 Richard Biener <rguenther@suse.de>
1661 PR tree-optimization/85834
1662 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1663 non-constant and non-zero memset arguments.
1665 2018-05-22 Martin Liska <mliska@suse.cz>
1668 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1670 2018-05-22 Richard Biener <rguenther@suse.de>
1672 PR tree-optimization/85863
1673 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1674 comparisons when vectype is specified.
1675 (vectorizable_condition): Do not specify vectype for
1676 vect_is_simple_cond when SLP vectorizing.
1678 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1681 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1682 define __ibm128 as long double.
1683 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1684 as a distinct type when IEEE 128-bit support is enabled.
1685 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1686 128-bit types to use the correct functions.
1687 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1688 convert between 128-bit floating point types that have different
1689 modes but the same representation, instead of using gen_lowpart to
1691 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1693 (IFKF_reg): New attributes to give the register constraints for
1695 (extend<mode>tf2_internal): New insns to mark an explicit
1696 conversion between 128-bit floating point types that have a
1697 different mode but share the same representation.
1699 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1701 PR tree-optimization/85814
1702 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1703 a null return from get_strinfo when unsharing the next
1704 strinfo in the chain.
1706 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1709 * varasm.c (weak_finish): Clean up weak_decls.
1711 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1713 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1714 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1715 UNSPEC_UADALP values.
1716 * config/aarch64/iterators.md (ABAL): New int iterator.
1719 (sur): Add mappings for the above.
1720 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1722 (aarch64_<sur>abal<mode>_4): Likewise.
1723 (aarch64_<sur>adalp<mode>_3): Likewise.
1724 (<sur>sadv16qi): New define_expand.
1726 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1728 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1729 (*movdf_internal): Ditto.
1730 (*rcpsf2_sse): Ditto.
1731 (*rsqrtsf2_sse): Ditto.
1732 (*sqrt<mode>2_sse): Ditto.
1734 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1736 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1738 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1739 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1740 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1741 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1743 * config/aarch64/arm_neon.h: Likewise.
1744 * config/aarch64/iterators.md (VQ_I): New.
1746 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1748 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1750 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1753 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1755 * config/nds32/constraints.md (S): New constraint.
1756 * config/nds32/nds32.md (call_internal): Use constraint S.
1757 (call_value_internal): Likewise.
1758 (sibcall_internal): Likewise.
1759 (sibcall_value_internal): Likewise.
1761 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1762 Chung-Ju Wu <jasonwucj@gmail.com>
1764 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1767 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1768 Chung-Ju Wu <jasonwucj@gmail.com>
1770 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1771 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1772 (nds32_rtx_costs_impl): Simplify.
1773 (nds32_address_cost_impl): Simplify.
1774 (nds32_init_rtx_costs): New function.
1775 (nds32_rtx_costs_speed_prefer): Likewise.
1776 (nds32_rtx_costs_size_prefer): Likewise.
1777 (nds32_address_cost_speed_prefer): Likewise.
1778 (nds32_address_cost_speed_fwprop): Likewise.
1779 (nds32_address_cost_size_prefer): Likewise.
1780 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1781 * config/nds32/nds32.c (nds32_option_override): Use
1782 nds32_init_rtx_costs function.
1784 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1786 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1787 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1788 (TARGET_PIPELINE_N8): Likewise.
1789 (TARGET_PIPELINE_N10): Likewise.
1790 (TARGET_PIPELINE_N13): Likewise.
1791 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1793 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1795 * config/nds32/nds32-fpu.md: Update copyright year.
1797 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1799 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1801 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1803 * config/nds32/nds32.c
1804 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1805 * config/nds32/nds32.opt (minline-asm-r15): New option.
1807 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1809 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1811 * config/nds32/nds32.md (abssi2): New pattern.
1813 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1815 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1816 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1817 Merge insn pattern from sse_cvtsi2ss<round_name> and
1818 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1819 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1820 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1821 using SWI48 mode iterator.
1822 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1823 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1824 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1825 pattern from sse_cvttss2si<round_saeonly_name>
1826 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1827 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1828 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1829 using SWI48 mode iterator.
1830 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1831 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1832 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1833 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1834 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1835 using SWI48 mode iterator.
1836 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1837 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1838 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1839 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1840 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1841 SWI48 mode iterator.
1842 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1843 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1844 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1845 pattern from sse_cvttsd2si<round_saeonly_name>
1846 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1848 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1850 * config/nds32/nds32-md-auxiliary.c
1851 (nds32_valid_smw_lwm_base_p): Refine.
1852 (nds32_output_smw_single_word): Refine.
1853 (nds32_output_smw_double_word): New.
1854 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1856 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1858 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1859 (nds32_output_stack_pop): Refine.
1860 (nds32_expand_unaligned_load): Refine.
1861 (nds32_expand_unaligned_store): Refine.
1863 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1864 Chung-Ju Wu <jasonwucj@gmail.com>
1866 * config/nds32/constants.md: Add TP_REGNUM constant.
1867 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1868 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1870 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1871 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1872 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1873 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1874 and PIC code generation.
1875 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1877 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1879 * config/nds32/nds32.md: Support TLS and PIC.
1880 * config/nds32/nds32.c: Support TLS and PIC.
1881 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1882 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1885 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1887 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1888 mode with E_ prefix.
1890 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1891 Chung-Ju Wu <jasonwucj@gmail.com>
1893 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1894 * config/nds32/nds32-md-auxiliary.c
1895 (symbolic_reference_mentioned_p): New.
1896 (nds32_legitimize_ict_address): New.
1897 (nds32_expand_ict_move): New.
1898 (nds32_indirect_call_referenced_p): New.
1899 (nds32_symbol_binds_local_p): Delete.
1900 (nds32_long_call_p): Modify.
1901 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1902 * config/nds32/nds32-protos.h
1903 (symbolic_reference_mentioned_p): Declare.
1904 (nds32_legitimize_ict_address): Declare.
1905 (nds32_expand_ict_move): Declare.
1906 (nds32_indirect_call_referenced_p): Declare.
1907 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1908 (nds32_relax_group): Use nds32_ict_const_p as condition.
1909 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1910 (nds32_asm_file_start): Output ict_model directive in asm code.
1911 (nds32_legitimate_address_p): Consider indirect call.
1912 (nds32_print_operand): Consider indirect call.
1913 (nds32_print_operand_address): Consider indirect call.
1914 (nds32_insert_attributes): Handle "indirect_call" attribute.
1915 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1916 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1917 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1918 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1919 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1920 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1921 (TARGET_ICT_MODEL_SMALL): Define.
1922 (TARGET_ICT_MODEL_LARGE): Define.
1923 * config/nds32/nds32.md (movsi): Consider ict model.
1924 (call, call_value): Consider ict model.
1925 (sibcall, sibcall_value): Consider ict model.
1926 * config/nds32/nds32.opt (mict-model): New option.
1927 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1930 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1931 Monk Chiang <sh.chiang04@gmail.com>
1932 Jim Wilson <jimw@sifive.com>
1934 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1935 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1936 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1937 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1938 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1939 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1940 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1941 compute save_libcall_adjustment properly.
1942 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1943 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1944 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1945 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1946 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1947 (ABI_SPEC): Handle mabi=ilp32e.
1948 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1949 (RVE): Add RVE mask.
1950 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1951 <-march>: Add rv32e as an example.
1953 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1956 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1957 (intra_create_variable_infos): Handle C++ constructors.
1959 2018-05-18 Martin Liska <mliska@suse.cz>
1961 * passes.def: Remove a redundant pass.
1963 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1966 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1968 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1970 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1972 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1974 (armv2, armv2a, armv3, armv3m): Delete architectures.
1975 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1976 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1977 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1979 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1980 (*mulsidi3adddi): Likewise.
1981 (mulsidi3): Likewise.
1982 (*mulsidi3_nov6): Likewise.
1983 (umulsidi3): Likewise.
1984 (umulsidi3_nov6): Likewise.
1985 (umaddsidi4): Likewise.
1986 (*umulsidi3adddi): Likewise.
1987 (smulsi3_highpart): Likewise.
1988 (*smulsi3_highpart_nov6): Likewise.
1989 (umulsi3_highpart): Likewise.
1990 (*umulsi3_highpart_nov6): Likewise.
1991 * config/arm/arm.h (arm_arch3m): Delete.
1992 * config/arm/arm.c (arm_arch3m): Delete.
1993 (arm_option_override_internal): Update armv3-related comment.
1994 (arm_configure_build_target): Delete use of isa_bit_mode32.
1995 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1996 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1997 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1999 (mulusa3): Likewise.
2000 * config/arm/arm-protos.h (arm_arch3m): Delete.
2001 * config/arm/arm-tables.opt: Regenerate.
2002 * config/arm/arm-tune.md: Likewise.
2003 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2004 deleted architectures.
2006 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2008 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2009 (armv5t, armv5te): New features.
2010 (ARMv5, ARMv5e): Delete fgroups.
2011 (ARMv5t, ARMv5te): Adjust for above changes.
2013 (armv5, armv5e): Delete arches.
2014 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2016 (*call_reg_arm): Likewise.
2017 (*call_value_reg_armv5): Likewise.
2018 (*call_value_reg_arm): Likewise.
2019 (*call_symbol): Likewise.
2020 (*call_value_symbol): Likewise.
2021 (*sibcall_insn): Likewise.
2022 (*sibcall_value_insn): Likewise.
2024 (prefetch): Likewise.
2025 (define_split and define_peephole2 dependent on arm_arch5):
2027 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2029 (TARGET_ARM_QBIT): Likewise.
2030 (TARGET_DSP_MULTIPLY): Likewise.
2031 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2032 (arm_arch5, arm_arch5e): Delete.
2033 (arm_arch5t, arm_arch5te): Declare.
2034 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2035 (arm_arch5t): Declare.
2036 (arm_option_reconfigure_globals): Update for the above.
2037 (arm_options_perform_arch_sanity_checks): Update comment, replace
2038 use of arm_arch5 with arm_arch5t.
2039 (use_return_insn): Likewise.
2040 (arm_emit_call_insn): Likewise.
2041 (output_return_instruction): Likewise.
2042 (arm_final_prescan_insn): Likewise.
2043 (arm_coproc_builtin_available): Likewise.
2044 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2045 arm_arch5e with arm_arch5t and arm_arch5te.
2046 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2047 (arm_arch5t, arm_arch5te): Declare.
2048 * config/arm/arm-tables.opt: Regenerate.
2049 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2050 * config/arm/t-multilib: Likewise.
2051 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2052 instead of arm_arch5.
2053 (*call_reg_thumb1): Likewise.
2054 (*call_value_reg_thumb1_v5): Likewise.
2055 (*call_value_reg_thumb1): Likewise.
2056 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2058 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2060 2018-05-18 Martin Liska <mliska@suse.cz>
2062 PR gcov-profile/84846
2063 * doc/gcov.texi: Document -t option of gcov tool.
2065 2018-05-18 Martin Liska <mliska@suse.cz>
2067 PR gcov-profile/84846
2068 * gcov.c (print_usage): Add new -t option.
2069 (process_args): Handle the option.
2070 (generate_results): Use stdout as output when requested by
2073 2018-05-18 Martin Liska <mliska@suse.cz>
2075 PR gcov-profile/84846
2076 * coverage.c (coverage_init): Write PWD to .gcno file.
2077 * doc/gcov.texi: Document how working directory is printed.
2078 * gcov-dump.c (dump_gcov_file): Print PWD.
2079 * gcov.c (output_intermediate_file): Likewise.
2080 (read_graph_file): Read PWD string.
2081 (output_lines): Print PWD.
2083 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2086 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2087 for retval and return false if all args to phi are zero.
2089 2018-05-18 Richard Biener <rguenther@suse.de>
2091 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2093 (evrp_dom_walker::before_dom_children): Call it.
2095 2018-05-18 Richard Biener <rguenther@suse.de>
2097 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2098 results when processing array refs with variable index.
2100 2018-05-18 Toon Moene <toon@moene.org>
2102 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2103 directly after that of -floop-interchange. Indicate that both
2104 options are enabled by default when specifying -O3.
2106 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2108 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2109 iterator. Delete separate integer-mode vec_set<mode> expander.
2110 (aarch64_simd_vec_setv2di): Delete.
2111 (vec_setv2di): Delete.
2112 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2113 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2114 the "w, r" alternative.
2116 2018-05-18 Martin Liska <mliska@suse.cz>
2118 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2119 * tree-pass.h (make_pass_lower_switch_O0): New function.
2120 * tree-switch-conversion.c (node_has_low_bound): Remove.
2121 (node_has_high_bound): Likewise.
2122 (node_is_bounded): Likewise.
2123 (class pass_lower_switch): Make it a template type and create
2125 (pass_lower_switch::execute): Add template argument.
2126 (make_pass_lower_switch): New function.
2127 (make_pass_lower_switch_O0): New function.
2128 (do_jump_if_equal): Remove.
2129 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2130 all the hard work to tree optimization passes.
2132 2018-05-18 Martin Liska <mliska@suse.cz>
2134 * dbgcnt.c (limit_low): Renamed from limit.
2135 (limit_high): New variable.
2136 (dbg_cnt_is_enabled): Check for upper limit.
2137 (dbg_cnt): Adjust dumping.
2138 (dbg_cnt_set_limit_by_index): Add new argument for high
2140 (dbg_cnt_set_limit_by_name): Likewise.
2141 (dbg_cnt_process_single_pair): Parse new format.
2142 (dbg_cnt_process_opt): Use strtok.
2143 (dbg_cnt_list_all_counters): Remove 'value' and add
2145 * doc/invoke.texi: Document changes.
2147 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2149 * doc/sourcebuild.texi (scalar_all_fma): Document.
2150 * tree.def (FMA_EXPR): Delete.
2151 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2152 * internal-fn.c (ternary_direct): New macro.
2153 (expand_ternary_optab_fn): Likewise.
2154 (direct_ternary_optab_supported_p): Likewise.
2155 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2156 * builtins.c (fold_builtin_fma): Delete.
2157 (fold_builtin_3): Don't call it.
2158 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2159 * expr.c (expand_expr_real_2): Likewise.
2160 * fold-const.c (operand_equal_p): Likewise.
2161 (fold_ternary_loc): Likewise.
2162 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2163 * gimple.c (DEFTREECODE): Likewise.
2164 * gimplify.c (gimplify_expr): Likewise.
2165 * optabs-tree.c (optab_for_tree_code): Likewise.
2166 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2167 * tree-eh.c (operation_could_trap_p): Likewise.
2168 (stmt_could_throw_1_p): Likewise.
2169 * tree-inline.c (estimate_operator_cost): Likewise.
2170 * tree-pretty-print.c (dump_generic_node): Likewise.
2171 (op_code_prio): Likewise.
2172 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2173 * tree-ssa-operands.c (get_expr_operands): Likewise.
2174 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2175 * fold-const-call.h (fold_fma): Delete.
2176 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2177 CFN_FNMA and CFN_FNMS.
2179 * genmatch.c (combined_fn): New enum.
2180 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2181 (commutative_op): New function.
2182 (commutate): Use it. Handle more than 2 operands.
2183 (dt_operand::gen_gimple_expr): Use commutative_op.
2184 (parser::parse_expr): Allow :c to be used with non-binary
2185 operators if the commutative operand is known.
2186 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2187 CFN_FMS, CFN_FNMA and CFN_FNMS.
2188 (backprop::process_assign_use): Remove FMA_EXPR handling.
2189 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2190 (gen_hsa_fma): New function.
2191 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2192 IFN_FNMA and IFN_FNMS.
2193 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2194 * gimple-fold.h (follow_all_ssa_edges): Declare.
2195 * gimple-fold.c (follow_all_ssa_edges): New function.
2196 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2197 gimple_build interface and use follow_all_ssa_edges to fold the result.
2198 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2199 instead of checking for optabs directly.
2200 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2201 rather than FMA_EXPRs.
2202 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2203 call to IFN_FMA instead of an FMA_EXPR.
2205 2018-05-17 Jim Wilson <jimw@sifive.com>
2207 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2208 sign extended promoted subreg, and the range does not have the sign bit
2209 set, then do a sign extend.
2211 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2212 test, check for sign extended subreg and/or constant operands, and
2213 do a sign extend in that case.
2215 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2217 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2218 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2220 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2221 Change logics_shift_reg to logics_shift_imm.
2222 (thunderx2t99_fp_loadpair_basic): Delete.
2223 (thunderx2t99_fp_storepair_basic): Delete.
2224 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2225 (thunderx2t99_asimd_polynomial): Delete.
2226 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2227 and neon_fp_mul_d_scalar_q.
2228 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2229 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2230 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2231 (thunderx2t99_asimd_lut): Add missing tbl types.
2232 (thunderx2t99_asimd_ext): Delete.
2233 (thunderx2t99_asimd_load1_1_mult): Delete.
2234 (thunderx2t99_asimd_load1_2_mult): Delete.
2235 (thunderx2t99_asimd_load1_ldp): New.
2236 (thunderx2t99_asimd_load1): New.
2237 (thunderx2t99_asimd_load2): Add missing *load2* types.
2238 (thunderx2t99_asimd_load3): New.
2239 (thunderx2t99_asimd_load4): New.
2240 (thunderx2t99_asimd_store1_1_mult): Delete.
2241 (thunderx2t99_asimd_store1_2_mult): Delete.
2242 (thunderx2t99_asimd_store2_mult): Delete.
2243 (thunderx2t99_asimd_store2_onelane): Delete.
2244 (thunderx2t99_asimd_store_stp): New.
2245 (thunderx2t99_asimd_store1): New.
2246 (thunderx2t99_asimd_store2): New.
2247 (thunderx2t99_asimd_store3): New.
2248 (thunderx2t99_asimd_store4): New.
2250 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2252 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2253 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2255 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2256 Segher Boessenkool <segher@kernel.crashing.org>
2259 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2262 2018-05-17 Richard Biener <rguenther@suse.de>
2264 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2265 for pruning loop and prune defs feeding only already visited PHIs.
2267 2018-05-17 Richard Biener <rguenther@suse.de>
2269 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2271 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2272 Richard Biener <rguenther@suse.de>
2274 PR tree-optimization/85793
2275 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2276 for VMAT_ELEMENTWISE.
2278 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2280 * internal-fn.h (lookup_internal_fn): Declare
2281 * internal-fn.c (lookup_internal_fn): New function.
2282 * gimple.c (gimple_build_call_from_tree): Handle calls to
2284 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2285 internal function names.
2286 * tree-pretty-print.c (dump_generic_node): Likewise.
2287 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2289 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2291 * gimple-fold.h (gimple_build): Make the function forms take
2292 combined_fn rather than built_in_function.
2293 (gimple_simplify): Likewise.
2294 * gimple-match-head.c (gimple_simplify): Likewise.
2295 * gimple-fold.c (gimple_build): Likewise.
2296 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2297 rather than gimple_build_call_internal.
2298 (get_initial_defs_for_reduction): Likewise.
2299 (vect_create_epilog_for_reduction): Likewise.
2300 (vectorizable_live_operation): Likewise.
2302 2018-05-17 Martin Liska <mliska@suse.cz>
2304 * gimple-ssa-sprintf.c (format_directive): Do not use
2305 space in between 'G_' and '('.
2307 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2310 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2311 even if the mask is not all ones.
2314 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2316 (ix86_gimple_fold_builtin): Likewise.
2319 * config/i386/i386.c: Include tree-vector-builder.h.
2320 (ix86_vector_shift_count): New function.
2321 (ix86_fold_builtin): Fold shift builtins by scalar count.
2322 (ix86_gimple_fold_builtin): Likewise.
2324 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2325 _mm512_setzero): New intrinsics.
2327 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2328 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2330 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2331 code generation for cases where splatting a value is not useful.
2332 * simplify-rtx.c (simplify_ternary_operation): Simplify
2333 vec_merge across a vec_duplicate and a paradoxical subreg forming
2334 a vector mode to a vec_concat.
2336 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2338 * config.gcc: Support "goldmont-plus".
2339 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2341 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2342 PROCESSOR_GOLDMONT_PLUS.
2343 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2344 (processor_target_table): Add "goldmont-plus".
2345 (PTA_GOLDMONT_PLUS): Define.
2346 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2347 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2348 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2349 (fold_builtin_cpu): Add "goldmont-plus".
2350 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2351 (ix86_option_override_internal): Add "goldmont-plus".
2352 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2353 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2354 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2355 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2357 2018-05-17 Richard Biener <rguenther@suse.de>
2359 PR tree-optimization/85757
2360 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2361 remove defs that only feed that PHI from further processing.
2363 2018-05-16 Jim Wilson <jimw@sifive.com>
2365 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2367 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2369 2018-05-16 Mark Wielaard <mark@klomp.org>
2371 * dwarf2out.c (count_index_strings): New function.
2372 (output_indirect_strings): Call count_index_strings and generate
2373 header for dwarf_version >= 5.
2375 2018-05-16 Mark Wielaard <mark@klomp.org>
2377 * dwarf2out.c (dwarf_FORM): New function.
2378 (set_indirect_string): Use dwarf_FORM.
2379 (reset_indirect_string): Likewise.
2380 (size_of_die): Likewise.
2381 (value_format): Likewise.
2382 (output_die): Likewise.
2383 (add_skeleton_AT_string): Likewise.
2384 (output_macinfo_op): Likewise.
2385 (index_string): Likewise.
2386 (output_index_string_offset): Likewise.
2387 (output_index_string): Likewise.
2388 (count_index_strings): Likewise.
2390 2018-05-16 Carl Love <cel@us.ibm.com>
2392 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2393 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2395 2018-05-16 Martin Jambor <mjambor@suse.cz>
2397 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2398 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2400 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2402 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2403 (fnma<mode>4): Likewise.
2404 (fms<mode>4): Likewise.
2405 (fnms<mode>4): Likewise.
2406 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2407 (aarch64_fnma<mode>4): Likewise.
2408 (aarch64_fms<mode>4): Likewise.
2409 (aarch64_fnms<mode>4): Likewise.
2410 (aarch64_fnmadd<mode>4): Likewise.
2412 2018-05-16 Jason Merrill <jason@redhat.com>
2414 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2416 2018-05-16 Richard Biener <rguenther@suse.de>
2418 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2419 (dump_stmt_cost): Declare.
2420 (add_stmt_cost): Dump cost we add.
2421 (add_stmt_costs): New function.
2422 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2424 (vect_analyze_stmt): Adjust prototype.
2425 (vectorizable_condition): Likewise.
2426 (vectorizable_live_operation): Likewise.
2427 (vectorizable_reduction): Likewise.
2428 (vectorizable_induction): Likewise.
2429 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2430 cost vector to pass to vectorizable_ and record afterwards.
2431 (vect_model_reduction_cost): Take cost vector argument and adjust.
2432 (vect_model_induction_cost): Likewise.
2433 (vectorizable_reduction): Likewise.
2434 (vectorizable_induction): Likewise.
2435 (vectorizable_live_operation): Likewise.
2436 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2437 SLP_TREE_NUMBER_OF_VEC_STMTS.
2438 (vect_analyze_slp_cost_1): Remove.
2439 (vect_analyze_slp_cost): Likewise.
2440 (vect_slp_analyze_node_operations): Take visited args and
2441 a target cost vector. Avoid processing already visited stmt sets.
2442 (vect_slp_analyze_operations): Use a local cost vector to gather
2443 costs and register those of non-discarded instances.
2444 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2445 (vect_schedule_slp_instance): Remove copying of
2446 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2448 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2449 adding cost. Record cost entry location.
2450 (vect_prologue_cost_for_slp_op): Function to compute cost of
2451 a constant or invariant generated for SLP vect in the prologue,
2452 split out from vect_analyze_slp_cost_1.
2453 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2454 (vect_model_promotion_demotion_cost): Likewise.
2455 (vect_model_store_cost): Likewise, make static.
2456 (vect_model_load_cost): Likewise.
2457 (vectorizable_bswap): Add cost vector arg and adjust.
2458 (vectorizable_call): Likewise.
2459 (vectorizable_simd_clone_call): Likewise.
2460 (vectorizable_conversion): Likewise.
2461 (vectorizable_assignment): Likewise.
2462 (vectorizable_shift): Likewise.
2463 (vectorizable_operation): Likewise.
2464 (vectorizable_store): Likewise.
2465 (vectorizable_load): Likewise.
2466 (vectorizable_condition): Likewise.
2467 (vectorizable_comparison): Likewise.
2468 (can_vectorize_live_stmts): Likewise.
2469 (vect_analyze_stmt): Likewise.
2470 (vect_transform_stmt): Adjust calls to vectorizable_*.
2471 * tree-vectorizer.c: Include gimple-pretty-print.h.
2472 (dump_stmt_cost): New function.
2474 2018-05-16 Richard Biener <rguenther@suse.de>
2476 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2477 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2478 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2479 (check_name): New callback.
2480 (dse_classify_store): Track cycles via a visited bitmap of PHI
2481 defs and simplify handling of in-loop and across loop dead stores
2482 and properly fail for loop-variant refs. Handle byte-tracking with
2483 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2486 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2488 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2489 (vect_get_mask_type_for_stmt): Likewise.
2490 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2492 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2493 to determine the statement's vector type and the vector type that
2494 should be used for calculating nunits. Deal with cases in which
2495 the type has to be deferred.
2496 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2497 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2498 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2499 (vect_determine_vf_for_stmt): New functions, split out from...
2500 (vect_determine_vectorization_factor): ...here.
2501 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2502 (vect_get_mask_type_for_stmt): New functions, split out from
2503 vect_determine_vectorization_factor.
2505 2018-05-16 Richard Biener <rguenther@suse.de>
2507 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2508 verify the [VEC_]COND_EXPR embedded comparison.
2510 2018-05-15 Martin Sebor <msebor@redhat.com>
2512 PR tree-optimization/85753
2513 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2514 RECORD_TYPE in addition to ARRAY_TYPE.
2516 2018-05-15 Martin Sebor <msebor@redhat.com>
2519 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2521 2018-05-15 Richard Biener <rguenther@suse.de>
2523 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2524 add by_clobber_p one. Change algorithm to collect all defs
2525 representing uses we need to walk and try reducing them to
2526 a single one before failing.
2527 (dse_dom_walker::dse_optimize_stmt): Adjust.
2529 2018-05-13 Mark Wielaard <mark@klomp.org>
2531 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2532 (size_of_loc_descr): Likewise.
2533 (output_loc_operands): Likewise.
2534 (output_loc_operands_raw): Likewise.
2535 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2536 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2537 (hash_loc_operands): Likewise.
2538 (compare_loc_operands): Likewise.
2540 2018-05-14 Mark Wielaard <mark@klomp.org>
2542 * dwarf2out.c (count_index_addrs): New function.
2543 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2545 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2547 PR tree-optimization/83648
2548 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2549 return value as malloc candidate.
2551 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2554 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2555 param as true in call to suggest_attribute.
2557 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2559 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2560 -mreadonly-in-sdata.
2562 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2564 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2566 (aarch64_crypto_aesd_fused): Likewise.
2568 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2570 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2571 (movsi_aarch64): Likewise.
2572 (load_pairsi): Likewise.
2573 (load_pairdi): Likewise.
2574 (store_pairsi): Likewise.
2575 (store_pairdi): Likewise.
2576 (load_pairsf): Likewise.
2577 (load_pairdf): Likewise.
2578 (store_pairsf): Likewise.
2579 (store_pairdf): Likewise.
2580 (zero_extend): Likewise.
2581 (trunc): Swap alternatives.
2582 (fcvt_target): Add '?' to prefer w over r.
2584 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2587 * config/i386/i386.md: Disallow non-commutative arithmetics in
2588 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2589 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2590 in the peephole2 before it.
2592 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2594 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2595 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2596 (ix86_handle_option): Handle -mcldemote.
2597 * config.gcc: New header.
2598 * config/i386/cldemoteintrin.h: New file.
2599 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2600 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2602 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2603 OPTION_MASK_ISA_CLDEMOTE.
2604 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2605 (ix86_valid_target_attribute_inner_p): Ditto.
2606 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2607 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2608 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2609 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2610 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2612 * config/i386/i386.opt: Add -mcldemote.
2613 * config/i386/x86intrin.h: New header.
2614 * doc/invoke.texi: Add -mcldemote.
2616 2018-05-14 Richard Biener <rguenther@suse.de>
2618 * doc/match-and-simplify.texi: Adjust :s documentation.
2620 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2622 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2623 intended memcpy size.
2624 (REORDER_45): Likewise.
2626 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2628 * sort.cc: New file.
2629 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2630 * vec.c (qsort_chk): Use gcc_qsort.
2631 * Makefile.in (OBJS-libcommon): Add sort.o.
2632 (build/sort.o): New target. Use it...
2633 (BUILD_RTL): ... here, and...
2634 (build/gencfn-macros): ... here, and...
2635 (build/genmatch): ... here.
2637 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2638 Chung-Ju Wu <jasonwucj@gmail.com>
2640 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2641 * config/nds32/nds32-graywolf.md: New file.
2642 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2643 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2645 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2646 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2647 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2648 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2649 * config/nds32/pipelines.md: Include n15 settings.
2651 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2652 Chung-Ju Wu <jasonwucj@gmail.com>
2654 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2655 * config/nds32/nds32-n13.md: New file.
2656 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2657 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2659 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2660 * config/nds32/nds32.md (pipeline_model): Add n13.
2661 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2662 * config/nds32/pipelines.md: Include n13 settings.
2664 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2665 Chung-Ju Wu <jasonwucj@gmail.com>
2667 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2668 * config/nds32/nds32-n10.md: New file.
2669 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2670 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2672 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2673 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2674 * config/nds32/nds32.md (pipeline_model): Add n10.
2675 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2676 * config/nds32/pipelines.md: Include n10 settings.
2678 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2679 Kito Cheng <kito.cheng@gmail.com>
2680 Chung-Ju Wu <jasonwucj@gmail.com>
2682 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2683 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2684 Add enum values for DSP extension instructions.
2685 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2687 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2688 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2690 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2691 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2692 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2693 * config/nds32/nds32-intrinsic.md: Likewise.
2694 * config/nds32/nds32_intrinsic.h: Likewise.
2695 * config/nds32/nds32-md-auxiliary.c: Likewise.
2696 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2697 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2698 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2699 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2700 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2701 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2702 TYPE_DMAC in switch statement.
2703 * config/nds32/nds32.c: New checking and implementation for DSP
2704 extension instructions.
2705 * config/nds32/nds32.h: Likewise.
2706 * config/nds32/nds32.md: Likewise.
2707 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2708 * config/nds32/predicates.md: Implement new predicates for DSP
2711 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2713 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2714 Reformat alternatives and attributes so it is easier to identify
2715 which constraints/attributes go with which instruction.
2716 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2717 (mov<mode>_softfloat32, FMOVE64): Likewise.
2718 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2719 (mov<mode>_softfloat64, FMOVE64): Likewise.
2721 2018-05-11 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): Likewise.
2730 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2731 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2732 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2734 2018-05-11 Martin Jambor <mjambor@suse.cz>
2737 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2740 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2743 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2745 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2747 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2748 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2749 (ix86_handle_option): Handle -mwaitpkg.
2750 * config.gcc: New header.
2751 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2752 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2753 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2755 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2756 OPTION_MASK_ISA_WAITPKG.
2757 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2758 (ix86_option_override_internal): Add PTA_WAITPKG.
2759 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2760 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2761 IX86_BUILTIN_TPAUSE.
2762 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2763 __builtin_ia32_umwait and __builtin_ia32_tpause.
2764 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2765 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2766 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2767 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2768 UNSPECV_TPAUSE): New.
2769 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2770 * config/i386/i386.opt: Add -mwaitpkg.
2771 * config/i386/waitpkgintrin.h: New file.
2772 * config/i386/x86intrin.h: New header.
2773 * doc/invoke.texi: Add -mwaitpkg.
2775 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2778 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2780 (cortex-m0): Use armv6s-m isa.
2781 (cortex-m0plus): Likewise.
2782 (cortex-m1): Likewise.
2783 (cortex-m0.small-multiply): Likewise.
2784 (cortex-m0plus.small-multiply): Likewise.
2785 (cortex-m1.small-multiply): Likewise.
2787 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2788 Jakub Jelinek <jakub@redhat.com>
2790 PR tree-optimization/85692
2791 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2792 source permute as well.
2794 2018-05-11 Martin Liska <mliska@suse.cz>
2797 * doc/extend.texi: Document LLVM style format for no_sanitize
2800 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2802 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2803 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2804 (mode_supports_vsx_dform_quad): Likewise.
2805 (mode_supports_vmx_dform): Move these functions to be next to the
2806 other mode_supports functions.
2807 (mode_supports_dq_form): Likewise.
2808 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2809 mode_supports_dq_form.
2810 (reg_offset_addressing_ok_p): Likewise.
2811 (offsettable_ok_by_alignment): Likewise.
2812 (rs6000_legitimate_offset_address_p): Likewise.
2813 (legitimate_lo_sum_address_p): Likewise.
2814 (rs6000_legitimize_address): Likewise.
2815 (rs6000_legitimize_reload_address): Likewise.
2816 (rs6000_secondary_reload_inner): Likewise.
2817 (rs6000_preferred_reload_class): Likewise.
2818 (rs6000_output_move_128bit): Likewise.
2820 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2822 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2823 Generate SImode target register for null target.
2824 <case IX86_BUILTIN_XGETBV>: Ditto.
2825 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2826 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2828 2018-05-10 Carl Love <cel@us.ibm.com>
2830 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2831 dcbtt and dcbtstt if operands[2] is 0.
2833 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2836 * config/i386/sse.md (usadv64qi): New expander.
2838 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2840 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2841 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2842 -maltivec=be support.
2843 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2844 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2845 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2846 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2847 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2848 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2849 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2850 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2851 altivec_vsumsws): Adjust.
2852 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2853 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2854 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2856 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2857 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2858 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2859 (altivec_lve<VI_char>x): Delete expand.
2860 (*altivec_lve<VI_char>x_internal): Rename to...
2861 (altivec_lve<VI_char>x): ... this.
2862 (altivec_lvxl_<mode>): Delete expand.
2863 (*altivec_lvxl_<mode>_internal): Rename to ...
2864 (altivec_lvxl_<mode>): ... this.
2865 (altivec_stvxl_<mode>): Delete expand.
2866 (*altivec_stvxl_<mode>_internal): Rename to ...
2867 (altivec_stvxl_<mode>): ... this.
2868 (altivec_stve<VI_char>x): Delete expand.
2869 (*altivec_stve<VI_char>x_internal): Rename to ...
2870 (altivec_stve<VI_char>x): ... this.
2871 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2872 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2873 reduc_plus_scal_<mode>): Adjust.
2874 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2876 (rs6000_cpu_cpp_builtins): Adjust.
2877 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2878 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2879 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2880 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2881 -maltivec=be support.
2882 (rs6000_split_vec_extract_var): Adjust.
2883 (rs6000_split_v4si_init): Adjust.
2884 (swap_selector_for_mode): Delete.
2885 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2886 altivec_expand_stvex_be): Delete.
2887 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2888 -maltivec=be support.
2889 (rs6000_gimple_fold_builtin): Ditto.
2890 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2892 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2893 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2894 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2895 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2896 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2897 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2898 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2899 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2900 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2901 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2902 anonymous split): Adjust.
2903 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2904 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2906 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2908 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2909 when --with-gxx-include-dir is also specified.
2910 * configure: Regenerate.
2912 2018-05-09 Jim Wilson <jimw@sifive.com>
2915 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2916 * config/riscv/t-withmultilib: New.
2917 * config/riscv/withmultilib.h: New.
2918 * doc/install.texi: Document RISC-V --with-multilib-list support.
2920 2018-05-09 Richard Biener <rguenther@suse.de>
2922 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2924 (vect_bb_vectorization_profitable_p): Adjust. Compute
2925 actual scalar cost using the cost vector and the add_stmt_cost
2928 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2930 PR rtl-optimization/85645
2931 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2932 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2934 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2936 PR rtl-optimization/85645
2937 * shrink-wrap.c (spread_components): Return a boolean saying if
2938 anything was changed.
2939 (try_shrink_wrapping_separate): Iterate spread_components until
2940 nothing changes anymore.
2942 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2944 PR rtl-optimization/85645
2945 * regrename.c (build_def_use): Also kill the chains that include the
2946 destination of a REG_CFA_REGISTER note.
2948 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2950 PR rtl-optimization/85645
2951 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2952 insn that has a REG_CFA_REGISTER note.
2954 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2956 * cfgexpand.c (expand_clobber): New function.
2957 (expand_gimple_stmt_1): Use it.
2958 * tree-vect-stmts.c (vect_clobber_variable): New function,
2960 (vectorizable_simd_clone_call): ...here.
2961 (vectorizable_store): Emit a clobber either side of an
2962 IFN_STORE_LANES sequence.
2963 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2965 2018-05-09 Tom de Vries <tom@codesourcery.com>
2968 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2969 (define_insn "trap_if_false"): Add exit after trap.
2971 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2973 PR rtl-optimization/85638
2974 * bb-reorder.c: Include common/common-target.h.
2975 (create_forwarder_block): New function extracted from...
2976 (fix_up_crossing_landing_pad): ...here. Rename into...
2977 (dw2_fix_up_crossing_landing_pad): ...this.
2978 (sjlj_fix_up_crossing_landing_pad): New function.
2979 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2980 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2981 from both partitions and exit the loop after one iteration.
2983 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2986 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2988 (Basic PowerPC Built-in Functions): The new name of the
2989 subsection previously known as "PowerPC Built-in Functions".
2990 (Basic PowerPC Built-in Functions Available on all Configurations):
2992 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2994 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2996 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2998 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3001 2018-05-08 Jim Wilson <jimw@sifive.com>
3003 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3004 (LD_EMUL_SUFFIX): New.
3005 (LINK_SPEC): Use it.
3007 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3009 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3011 (Basic PowerPC Built-in Functions): The new name of the
3012 subsection previously known as "PowerPC Built-in Functions".
3013 (Basic PowerPC Built-in Functions Available on all Configurations):
3015 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3017 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3019 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3021 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3024 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3027 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3028 after cmpelim optimization.
3030 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3032 * config.gcc: Support "goldmont".
3033 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3034 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3036 * config/i386/i386.c (m_GOLDMONT): Define.
3037 (processor_target_table): Add "goldmont".
3038 (PTA_GOLDMONT): Define.
3039 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3040 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3041 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3042 (fold_builtin_cpu): Add "goldmont".
3043 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3044 (ix86_option_override_internal): Add "goldmont".
3045 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3046 (processor_type): Add PROCESSOR_GOLDMONT.
3047 * config/i386/i386.md: Add CPU "glm".
3048 * config/i386/glm.md: New file.
3049 * config/i386/x86-tune.def: Add m_GOLDMONT.
3050 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3052 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3055 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3057 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3058 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3059 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3062 * config/i386/i386.c (ix86_fold_builtin): Handle
3063 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3066 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3067 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3069 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3072 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3073 (check_arch): Likewise.
3074 (check_fpu): Return the result rather than printing it.
3075 (end arch): Fix operator precedence.
3076 (end cpu): Likewise.
3077 (END): Print the result from check_fpu.
3079 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3080 Alan Hayward <alan.hayward@arm.com>
3081 David Sherwood <david.sherwood@arm.com>
3083 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3084 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3085 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3086 (*fcmuo<mode>_and): New patterns.
3088 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3090 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3091 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3092 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3093 (cmp_op, sve_imm_con): New code attributes.
3094 (SVE_COND_INT_CMP, imm_con): Delete.
3095 (cmp_op): Remove above unspecs from int attribute.
3096 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3098 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3099 comparison-specific unspecs.
3100 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3101 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3102 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3103 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3104 (*vec_fcm<cmp_op><mode>): Rename to...
3105 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3106 (*vec_fcmuo<mode>): Rename to...
3107 (*fcmuo<mode>): ...this and adjust likewise.
3108 (*pred_fcm<cmp_op><mode>): New pattern.
3109 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3110 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3112 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3114 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3115 (aarch64_emit_sve_predicated_cond): New function.
3116 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3117 (aarch64_emit_unspec_cond_or): Replace with...
3118 (aarch64_emit_sve_or_conds): ...this new function. Use
3119 aarch64_emit_sve_ptrue_op for the individual comparisons and
3120 aarch64_emit_binop to OR them together.
3121 (aarch64_emit_inverted_unspec_cond): Replace with...
3122 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3123 aarch64_emit_sve_ptrue_op for the comparison and
3124 aarch64_emit_unop to invert the result.
3125 (aarch64_expand_sve_vec_cmp_float): Update after the above
3126 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3128 2018-05-07 Nathan Sidwell <nathan@acm.org>
3130 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3131 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3132 (Backwards Compatibility): Likewise.
3134 2018-05-07 Luis Machado <luis.machado@linaro.org>
3138 2018-05-07 Luis Machado <luis.machado@linaro.org>
3140 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3141 <prefetch_dynamic_strides>: New const bool field.
3142 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3143 prefetch_dynamic_strides.
3144 (exynosm1_prefetch_tune): Likewise.
3145 (thunderxt88_prefetch_tune): Likewise.
3146 (thunderx_prefetch_tune): Likewise.
3147 (thunderx2t99_prefetch_tune): Likewise.
3148 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3150 (aarch64_override_options_internal): Update to set
3151 PARAM_PREFETCH_DYNAMIC_STRIDES.
3152 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3153 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3154 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3155 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3156 prefetch-dynamic-strides setting.
3158 2018-05-07 Luis Machado <luis.machado@linaro.org>
3160 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3161 <minimum_stride>: New const int field.
3162 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3163 minimum_stride field.
3164 (exynosm1_prefetch_tune): Likewise.
3165 (thunderxt88_prefetch_tune): Likewise.
3166 (thunderx_prefetch_tune): Likewise.
3167 (thunderx2t99_prefetch_tune): Likewise.
3168 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3169 (aarch64_override_options_internal): Update to set
3170 PARAM_PREFETCH_MINIMUM_STRIDE.
3171 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3172 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3173 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3174 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3175 stride is constant and is below the minimum stride threshold.
3177 2018-05-07 Luis Machado <luis.machado@linaro.org>
3179 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3182 2018-05-07 Luis Machado <luis.machado@linaro.org>
3184 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3185 <prefetch_dynamic_strides>: New const bool field.
3186 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3187 prefetch_dynamic_strides.
3188 (exynosm1_prefetch_tune): Likewise.
3189 (thunderxt88_prefetch_tune): Likewise.
3190 (thunderx_prefetch_tune): Likewise.
3191 (thunderx2t99_prefetch_tune): Likewise.
3192 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3194 (aarch64_override_options_internal): Update to set
3195 PARAM_PREFETCH_DYNAMIC_STRIDES.
3196 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3197 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3198 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3199 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3200 prefetch-dynamic-strides setting.
3202 2018-05-07 Luis Machado <luis.machado@linaro.org>
3204 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3205 <minimum_stride>: New const int field.
3206 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3207 minimum_stride field.
3208 (exynosm1_prefetch_tune): Likewise.
3209 (thunderxt88_prefetch_tune): Likewise.
3210 (thunderx_prefetch_tune): Likewise.
3211 (thunderx2t99_prefetch_tune): Likewise.
3212 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3213 (aarch64_override_options_internal): Update to set
3214 PARAM_PREFETCH_MINIMUM_STRIDE.
3215 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3216 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3217 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3218 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3219 stride is constant and is below the minimum stride threshold.
3221 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3224 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3225 the type is addressable. Don't force op into register if it has
3228 2018-05-05 Roland McGrath <mcgrathr@google.com>
3231 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3232 any section for which we don't know a specific type it should have,
3233 regardless of name. Previously this was done only for the exact
3234 names ".init_array", ".fini_array", and ".preinit_array".
3235 (default_elf_asm_named_section): Add comment about
3236 relationship with default_section_type_flags and SECTION_NOTYPE.
3237 (get_section): Don't consider it a type conflict if one side has
3238 SECTION_NOTYPE and the other doesn't, as long as neither has the
3239 SECTION_BSS et al used in the default_section_type_flags logic.
3241 2018-05-05 Tom de Vries <tom@codesourcery.com>
3244 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3245 (workaround_barsyncs): New function.
3246 (nvptx_reorg): Use workaround_barsyncs.
3247 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3248 (define_expand "nvptx_membar_cta"): New define_expand.
3249 (define_insn "*nvptx_membar_cta"): New insn.
3251 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3253 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3254 To improve optimization opportunities.
3255 * builtin-types.def: The new needed builtin types for the above.
3257 2018-05-04 Richard Biener <rguenther@suse.de>
3259 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3260 * gimple-ssa-store-merging.c
3261 (imm_store_chain_info::output_merged_store): Remove redundant create,
3262 release split_store vector contents on failure.
3263 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3264 scalar stmt vector on cache hit.
3266 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3268 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3270 * config.gcc (powerpc-xilinx-eabi*): Remove.
3271 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3273 (fusion_addis_mem_combo_load): Ditto.
3274 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3276 (rs6000_cpu_cpp_builtins): Ditto.
3277 * config/rs6000/rs6000-linux.c
3278 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3279 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3280 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3282 (rs6000_setup_reg_addr_masks): Ditto.
3283 (rs6000_init_hard_regno_mode_ok): Ditto.
3284 (rs6000_option_override_internal): Ditto.
3285 (legitimate_lo_sum_address_p): Ditto.
3286 (rs6000_legitimize_address): Ditto.
3287 (rs6000_legitimize_reload_address): Ditto.
3288 (rs6000_legitimate_address_p): Ditto.
3289 (abi_v4_pass_in_fpr): Ditto.
3290 (setup_incoming_varargs): Ditto.
3291 (rs6000_gimplify_va_arg): Ditto.
3292 (rs6000_split_multireg_move): Ditto.
3293 (rs6000_savres_strategy): Ditto.
3294 (rs6000_emit_prologue_components): Ditto.
3295 (rs6000_emit_epilogue_components): Ditto.
3296 (rs6000_emit_prologue): Ditto.
3297 (rs6000_emit_epilogue): Ditto.
3298 (rs6000_elf_file_end): Ditto.
3299 (rs6000_function_value): Ditto.
3300 (rs6000_libcall_value): Ditto.
3301 * config/rs6000/rs6000.h: Ditto.
3302 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3303 (TARGET_MINMAX): ... this. New.
3304 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3305 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3306 (*movsi_internal1_single): Delete.
3307 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3308 mfpu=, mxilinx-fpu): Delete.
3309 * config/rs6000/singlefp.h: Delete.
3310 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3311 * config/rs6000/t-rs6000: Ditto.
3312 * config/rs6000/t-xilinx: Delete.
3313 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3314 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3315 (VStype_simple): Delete.
3316 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3317 * config/rs6000/xfpu.h: Delete.
3318 * config/rs6000/xfpu.md: Delete.
3319 * config/rs6000/xilinx.h: Delete.
3320 * config/rs6000/xilinx.opt: Delete.
3321 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3322 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3324 2018-05-04 Tom de Vries <tom@codesourcery.com>
3327 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3330 2018-05-04 Richard Biener <rguenther@suse.de>
3333 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3334 (expand_complex_div_wide): Likewise.
3335 (expand_complex_operations_1): Likewise.
3336 (expand_complex_libcall): Preserve EH info of the original stmt.
3337 (tree_lower_complex): Handle removed blocks.
3338 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3339 on complex multiplication and division libcall builtins.
3341 2018-05-04 Richard Biener <rguenther@suse.de>
3344 * fold-const.c (negate_expr_p): Restrict negation of operand
3345 zero of a division to when we know that can happen without
3347 (fold_negate_expr_1): Likewise.
3349 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3352 * real.h (real_nextafter): Declare.
3353 * real.c (real_nextafter): New function.
3354 * fold-const-call.c (fold_const_nextafter): New function.
3355 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3356 CASE_CFN_NEXTTOWARD.
3357 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3358 even when arg1_mode is different from arg0_mode.
3360 2018-05-03 Nathan Sidwell <nathan@acm.org>
3362 * doc/extend.texi (Deprecated Features): Remove
3364 (Backwards Compatibility): Likewise.
3365 * doc/invoke.texi (C++ Language Options): Likewise.
3366 (C++ Dialect Options): Likewise.
3368 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3371 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3372 _mm512_mask_mullox_epi64): New intrinsics.
3374 2018-05-03 Tom de Vries <tom@codesourcery.com>
3377 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3378 dump files): Add offload-tree.
3380 2018-05-03 Richard Biener <rguenther@suse.de>
3382 PR tree-optimization/85615
3383 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3384 to loops not nested in BBs loop father to avoid creating multi-entry
3387 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3389 PR tree-optimization/70291
3390 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3391 arguments. Change return type to tree. Emit libcall as a new
3392 statement rather than replacing existing one when inplace_p is true.
3393 (expand_complex_multiplication_components): New function.
3394 (expand_complex_multiplication): Expand floating-point complex
3395 multiplication using the above.
3396 (expand_complex_division): Rename inner_type parameter to type.
3397 Update expand_complex_libcall call-site.
3398 (expand_complex_operations_1): Update expand_complex_multiplication
3399 and expand_complex_division call-sites.
3401 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3404 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3405 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3406 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3407 the highest significant bit of the shift count mask is clear. In
3408 check whether and[sq]i3 is needed verify that all significant bits
3409 of the shift count other than the highest are set.
3411 2018-05-02 Tom de Vries <tom@codesourcery.com>
3414 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3415 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3416 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3417 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3418 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3419 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3420 __builtin_goacc_parlevel_size.
3422 2018-05-02 Richard Biener <rguenther@suse.de>
3424 PR tree-optimization/85597
3425 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3426 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3429 2018-05-02 Tom de Vries <tom@codesourcery.com>
3432 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3433 dump files): Add ltrans-tree.
3435 2018-05-02 Tom de Vries <tom@codesourcery.com>
3438 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3439 dump files): Add wpa-ipa.
3441 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3443 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3444 powerpc*-*-linux*paired* target.
3445 * config/rs6000/750cl.h: Delete.
3446 * config/rs6000/paired.h: Delete.
3447 * config/rs6000/paired.md: Delete.
3448 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3450 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3451 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3452 comment. Remove paired float support.
3453 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3454 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3456 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3457 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3459 * config/rs6000/rs6000.c: Remove paired float support.
3460 (paired_expand_vector_init, paired_expand_vector_move,
3461 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3462 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3463 paired_expand_builtin, paired_expand_predicate_builtin,
3464 paired_init_builtins): Delete.
3465 * config/rs6000/rs6000.h: Remove paired float support.
3466 * config/rs6000/rs6000.md: Remove paired float support.
3467 (move_from_CR_ov_bit): Delete.
3468 * config/rs6000/rs6000.opt (mpaired): Delete.
3469 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3470 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3472 2018-05-02 Richard Biener <rguenther@suse.de>
3475 * gimplify.c (gimplify_save_expr): When in SSA form allow
3476 SAVE_EXPRs to compute to SSA vars.
3478 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3481 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3482 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3483 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3484 clobber operands[2], instead use a new pseudo. Formatting fixes.
3486 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3488 PR tree-optimization/85586
3489 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3490 exit early for statements in the same group if the accesses are
3493 2018-05-02 Tom de Vries <tom@codesourcery.com>
3496 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3499 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3501 PR tree-optimization/85143
3502 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3504 2018-05-01 Tom de Vries <tom@codesourcery.com>
3507 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3508 not found" error message.
3510 2018-05-01 Tom de Vries <tom@codesourcery.com>
3513 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3514 * vec.c (test_ordered_remove_if): New function.
3515 (vec_c_tests): Call test_ordered_remove_if.
3516 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3517 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3518 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3519 VEC_ORDERED_REMOVE_IF.
3521 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3523 PR tree-optimization/82665
3524 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3525 pointer subtraction where arguments come from a memchr call.
3527 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3529 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3530 --push-state --as-needed and --pop-state instead of --as-needed and
3531 --no-as-needed if ld supports it.
3532 * configure: Regenerated.
3535 * doc/install.texi2html: Replace _002d with - and _002a with * in
3536 generated html files using sed.
3538 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3541 * gcc-rich-location.c (blank_line_before_p): New function.
3542 (use_new_line): New function.
3543 (gcc_rich_location::add_fixit_insert_formatted): New function.
3544 * gcc-rich-location.h
3545 (gcc_rich_location::add_fixit_insert_formatted): New function.
3547 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3549 * selftest.c (assert_streq): Rename "expected" and "actual" to
3550 "val1" and "val2". Extend NULL-handling to cover both inputs
3551 symmetrically, while still requiring both to be non-NULL for a pass.
3552 * selftest.h (assert_streq): Rename "expected" and "actual" to
3554 (ASSERT_EQ): Likewise.
3555 (ASSERT_EQ_AT): Likewise.
3556 (ASSERT_KNOWN_EQ): Likewise.
3557 (ASSERT_KNOWN_EQ_AT): Likewise.
3558 (ASSERT_NE): Likewise.
3559 (ASSERT_MAYBE_NE): Likewise.
3560 (ASSERT_MAYBE_NE_AT): Likewise.
3561 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3562 the assertion to pass.
3563 (ASSERT_STREQ_AT): Likewise.
3565 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3567 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3568 interaction with -pie.
3570 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3572 * selftest.h: Fix alphabetization of per-source-file selftest
3575 2018-04-30 Jason Merrill <jason@redhat.com>
3577 PR c++/61982 - dead stores to destroyed objects.
3578 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3581 2018-04-30 Jason Merrill <jason@redhat.com>
3583 * tree.c (build_clobber): New.
3584 * tree.h: Declare it.
3585 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3587 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3589 * diagnostic-show-locus.c (layout::layout): Update for
3590 location_get_source_line returning a char_span.
3591 (struct char_span): Move to input.h.
3592 (struct correction): Update for fields in char_span becoming
3594 (struct source_line): Update for location_get_source_line
3595 returning a char_span.
3596 (layout::print_line): Likewise.
3597 * edit-context.c (edited_file::print_content): Likewise.
3598 (edited_file::print_diff_hunk): Likewise.
3599 (edited_file::print_run_of_changed_lines): Likewise.
3600 (edited_file::get_num_lines): Likewise.
3601 (edited_line::edited_line): Likewise.
3602 * final.c (asm_show_source): Likewise.
3603 * input.c (location_get_source_line): Convert return type
3604 from const char * to char_span, losing the final "line_len"
3606 (dump_location_info): Update for the above.
3607 (get_substring_ranges_for_loc): Likewise. Use a char_span
3608 when handling the literal within the line.
3609 (test_reading_source_line): Update for location_get_source_line
3610 returning a char_span.
3611 * input.h (class char_span): Move here from
3612 diagnostic-show-locus.c, converting from a struct to a class.
3613 Make data members private.
3614 (char_span::operator bool): New.
3615 (char_span::length): New.
3616 (char_span::get_buffer): New.
3617 (char_span::operator[]): New.
3618 (char_span::subspan): Make const.
3619 (char_span::xstrdup): New.
3620 (location_get_source_line): Convert return type from const char *
3621 to char_span, losing the final "line_size" param.
3623 2018-04-30 Jan Hubicka <jh@suse.cz>
3625 * lto-wrapper.c (ltrans_priorities): New static var.
3626 (cmp_priority): New.
3627 (run_gcc): Read priorities and if doing parallel build order
3628 the Makefile by them.
3630 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3632 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3634 2018-04-30 Richard Biener <rguenther@suse.de>
3636 * tree-cfg.c (verify_address): Remove base argument, add
3637 flag whether to check TREE_ADDRESSABLE and do that.
3638 (verify_expr): Remove.
3639 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3640 (verify_gimple_assign_single): Likewise.
3641 (verify_gimple_switch): Likewise.
3642 (verify_expr_location_1): Dereference tp once. Add (disabled)
3643 piece from verify_expr.
3644 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3646 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3648 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3650 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3652 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3653 (small_data_pattern): Likewise.
3654 (arc_rewrite_small_data): Likewise.
3655 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3656 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3657 (get_symbol_alignment): New function.
3658 (legitimate_small_data_address_p): Likewise.
3659 (legitimate_scaled_address): Update, call
3660 legitimate_small_data_address_p.
3661 (output_sdata): New static variable.
3662 (arc_print_operand): Update how we handle small data operands.
3663 (arc_print_operand_address): Likewise.
3664 (arc_legitimate_address_p): Update, use
3665 legitimate_small_data_address_p.
3666 (arc_rewrite_small_data_p): Remove.
3667 (arc_rewrite_small_data_1): Likewise.
3668 (arc_rewrite_small_data): Likewise.
3669 (small_data_pattern): Likewise.
3670 (compact_sda_memory_operand): Update to use
3671 legitimate_small_data_address_p and get_symbol_alignment.
3672 (prepare_move_operands): Don't rewite sdata pattern.
3673 (prepare_extend_operands): Remove.
3674 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3676 (zero_extendqisi2): Likewise.
3677 (zero_extendhisi2): Likewise.
3678 (extendqihi2): Likewise.
3679 (extendqisi2): Likewise.
3680 (extendhisi2): Likewise.
3684 * config/arc/constraints.md (Usd): Change it to memory constraint.
3686 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3688 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3689 as source of std instructions.
3690 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3691 allow 6-bit constants as source for store instructions.
3692 (movdi_insn): Update instruction pattern to allow 6-bit constants
3693 as source for store instructions.
3695 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3697 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3699 2018-04-30 Nathan Sidwell <nathan@acm.org>
3700 Sandra Loosemore <sandra@codesourcery.com>
3702 * dumpfile.c (dump_open): Allow '-' for stdout.
3703 * doc/invoke.texi (Developer Options): Document dump filename
3704 determination early. Document stdin/stdout selection.
3706 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3708 Microblaze Target: PIC data text relative
3710 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3711 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3713 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3714 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3715 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3716 New addressing mode for data-text relative position indepenedent code.
3717 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3718 'ADDRESS_SYMBOLIC_TXT_REL'.
3719 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3720 (microblaze_legitimate_pic_operand): Exclude function calls from
3721 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3722 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3724 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3725 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3726 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3727 for 'address + offset'.
3728 (microblaze_expand_prologue): Add new function prologue call for
3730 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3731 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3732 table in case of TARGET_PIC_DATA_TEXT_REL.
3733 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3734 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3735 Add new macros 'UNSPEC_TEXT',
3736 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3737 + exclude function calls from 'UNSPEC_PLT' in case of data text
3739 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3740 new target hook for generating address diff vector tables in case of
3742 * doc/tm.texi : Regenerate.
3743 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3744 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3745 of addr diff vector generation.
3746 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3747 target hook definition.
3748 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3749 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3750 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3751 option for data text relative.
3753 2018-04-30 Richard Biener <rguenther@suse.de>
3755 * tree-chrec.h (evolution_function_is_constant_p): Remove
3757 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3759 2018-04-30 Richard Biener <rguenther@suse.de>
3762 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3764 2018-04-30 Richard Biener <rguenther@suse.de>
3766 PR tree-optimization/28364
3767 PR tree-optimization/85275
3768 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3769 copying first exit test.
3771 2018-04-28 Mark Wielaard <mark@klomp.org>
3773 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3775 (dwarf_AT): Handle DW_AT_addr_base.
3776 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3778 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3781 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3782 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3783 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3784 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3786 2018-04-28 Richard Biener <rguenther@suse.de>
3788 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3789 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3790 to reflect use. Only add interesting stmts.
3792 2018-04-27 Martin Jambor <mjambor@suse.cz>
3795 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3796 the jump function allows for passing through aggregate values.
3798 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3800 * input.h (in_system_header_at): Convert from macro to inline
3802 (from_macro_expansion_at): Likewise.
3803 (from_macro_definition_at): Likewise.
3805 2018-04-27 Jeff Law <law@redhat.com>
3807 * config.gcc: Mark tile* targets as deprecated/obsolete.
3809 2018-04-27 Richard Biener <rguenther@suse.de>
3811 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3814 2018-04-27 Richard Biener <rguenther@suse.de>
3816 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3818 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3820 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3821 with Yd constraint. Set "preferred_for_speed" attribute from
3822 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3824 (*movdi_internal): Ditto.
3825 (movti_interunit splitters): Remove
3826 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3827 (movdi_interunit splitters): Ditto.
3828 * config/i386/constraints.md (Ye): Remove.
3829 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3831 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3834 * config/aarch64/constraints.md (Usg): Limit to 31.
3837 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3839 PR tree-optimization/85529
3840 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3841 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3842 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3843 zero extension or masking of the MSB bit.
3844 (optimize_range_tests): Add FIRST_BB argument, pass it through
3845 to optimize_range_tests_var_bound.
3846 (maybe_optimize_range_tests, reassociate_bb): Adjust
3847 optimize_range_tests callers.
3849 2018-04-26 Richard Biener <rguenther@suse.de>
3850 Jakub Jelinek <jakub@redhat.com>
3852 * cgraph.h (symbol_table): Just declare debug method here.
3853 * symtab.c (symbol_table::debug): Define.
3855 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3857 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3859 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3861 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3862 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3863 (*movdi_internal): Substitute Yi and Yj constraint with x
3864 and Ym and Yn constraint with y constraint. Update "isa"
3865 attribute and set "preferred_for_speed" attribute from
3866 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3867 (*movsi_internal): Ditto.
3868 (*movdf_internal): Ditto.
3869 (*movsf_internal): Ditto.
3870 (*zero_extendsidi2): Ditto.
3871 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3872 (sse2_loadld): Ditto.
3873 (*vec_extract<ssevecmodelower>_0): Ditto.
3874 (*vec_extractv4si_0_zext_sse4): Ditto.
3875 (vec_concatv2di): Ditto.
3876 (*vec_dup<mode>): Ditto.
3877 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3878 * config/i386/constraints.md (Yi): Remove.
3883 2018-04-26 Nathan Sidwell <nathan@acm.org>
3885 * dumpfile.c (dump_open): New.
3886 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3887 (dump_finish): Detect stdio/stderr by value not name.
3889 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3891 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3893 2018-04-26 Tom de Vries <tom@codesourcery.com>
3896 * config/nvptx/nvptx.c (verify_neutering_jumps)
3897 (verify_neutering_labels): New function
3898 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3900 2018-04-26 Tom de Vries <tom@codesourcery.com>
3903 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3904 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3907 2018-04-26 Richard Biener <rguenther@suse.de>
3908 Tom de Vries <tom@codesourcery.com>
3911 * lto-streamer-out.c (output_function): Fixup loops if required to match
3912 discovery done in the reader.
3914 2018-04-26 Richard Biener <rguenther@suse.de>
3916 PR tree-optimization/85116
3917 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3918 have a loop exit from the single latch predecessor. Remove
3919 case of header with just condition.
3920 (ch_base::copy_headers): Exclude infinite loops from any
3922 (pass_ch::execute): Record exits.
3924 2018-04-26 Richard Biener <rguenther@suse.de>
3926 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3927 prologue cost vector and pass it to vect_get_load_cost.
3928 (vect_get_peeling_costs_all_drs): Likewise.
3929 (vect_peeling_hash_get_lowest_cost): Likewise.
3930 (vect_enhance_data_refs_alignment): Likewise.
3932 2018-04-26 Richard Biener <rguenther@suse.de>
3935 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3936 checking of integer<->pointer conversions.
3937 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3938 sign-/zero-extending pointer types.
3939 (expand_omp_for_static_chunk): Likewise.
3941 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3942 Jean Lee <xiaoyur347@gmail.com>
3944 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3945 (TARGET_ASAN_SHADOW_OFFSET): Define.
3946 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3947 true for -fsanitize=address.
3949 2018-04-25 Mark Wielaard <mark@klomp.org>
3951 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3954 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3956 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3957 than "alu", remove explicit "memory" and "imm_disp" attributes.
3958 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3961 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3962 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3963 gen_lowpart_no_emit.
3965 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3968 * config/i386/i386.c (ix86_expand_builtin): Change memory
3969 operand to XI, extend p0 to Pmode.
3970 * config/i386/i386.md: Change unspec volatile and operand
3971 1 mode to XI, change operand 0 mode to P.
3973 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3975 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3976 GET_MODE_MASK before any checking.
3977 (nds32_can_use_bset_p): Likewise.
3978 (nds32_can_use_btgl_p): Likewise.
3980 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3982 * config/nds32/nds32-doubleword.md: New define_split pattern for
3983 illegal register number.
3985 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3987 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3989 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3991 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3993 2018-04-25 Richard Biener <rguenther@suse.de>
3995 * lto-streamer.h (LTO_major_version): Bump to 8.
3997 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3999 * BASE-VER: Set to 9.0.0.
4001 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4003 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4004 in __abskf2 and __powikf2.
4006 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4009 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4010 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4011 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4012 Use the above on operand 2. Reindent.
4013 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4015 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4018 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4020 * config/i386/i386.opt (mcet): Removed.
4021 * doc/install.texi: Remove -mcet documentation.
4022 * doc/invoke.texi: Likewise.
4024 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4027 * doc/install.texi: Remove -mcet from bootstrap-cet.
4029 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4032 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4033 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4037 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4038 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4039 containing a CONST_VECTOR.
4041 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4043 * doc/install.texi: Update newlib dependency for nvptx.
4045 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4048 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4049 instead of INTVAL when shifting x left.
4051 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4053 PR tree-optimization/85478
4054 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4055 vect_grouped_store_supported for single element vectors.
4057 2018-04-24 Richard Biener <rguenther@suse.de>
4060 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4061 load cost increase to the case of non-constant step.
4063 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4066 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4067 destination if any_malformed_asm.
4069 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4072 * expr.c (store_field): In the bitfield case, if the value comes from
4073 a function call and is returned in registers by means of a PARALLEL,
4074 do not change the mode of the temporary unless BLKmode and VOIDmode.
4076 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4078 PR rtl-optimization/85423
4079 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4080 dependencies to debug insns when the previous insn is non-debug.
4082 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4084 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4085 enums into a single definition.
4086 (fls): Fix predicates and printing.
4089 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4091 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4092 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4093 and short u6 immediate.
4094 (check_if_valid_sleep_operand): Remove.
4095 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4097 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4099 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4100 flag_always_save_lp condition.
4101 * config/nds32/nds32.opt (malways-save-lp): New option.
4103 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4105 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4106 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4107 * config/nds32/nds32.h
4108 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4109 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4111 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4113 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4114 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4116 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4117 Chung-Ju Wu <jasonwucj@gmail.com>
4119 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4121 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4122 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4124 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4126 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4128 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4130 * config/nds32/nds32-protos.h (nds32_data_alignment,
4131 nds32_local_alignment): Declare.
4132 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4133 nds32_local_alignment): New functions.
4134 (TARGET_CONSTANT_ALIGNMENT): Define.
4135 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4137 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4139 * config/nds32/nds32.c
4140 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4141 (TARGET_MODES_TIEABLE_P): Likewise.
4143 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4145 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4148 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4149 Chung-Ju Wu <jasonwucj@gmail.com>
4151 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4152 for unaligned access.
4153 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4155 * config/nds32/nds32-intrinsic.md: Likewise.
4156 * config/nds32/nds32_intrinsic.h: Likewise.
4157 * config/nds32/nds32.h (nds32_builtins): Likewise.
4158 * config/nds32/nds32.opt (munaligned-access): New option.
4159 * config/nds32/nds32.c (nds32_asm_file_start): Display
4160 flag_unaligned_access status.
4162 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4164 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4165 -mno-relax is present.
4166 * config/riscv/linux.h (LINK_SPEC): Ditto.
4168 2018-04-20 Martin Sebor <msebor@redhat.com>
4171 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4173 (gimple_fold_builtin_stxcpy_chk): Same.
4174 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4176 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4179 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4180 __powikf2 when long double is IEEE 128-bit.
4182 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4184 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4185 step to make sure stack always aligned.
4187 2018-04-20 Carl Love <cel@us.ibm.com>
4190 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4191 size check for arg0.
4193 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4194 Tom de Vries <tom@codesourcery.com>
4197 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4198 Emit insns for calls too.
4199 (nvptx_find_par): Always look for worker-level predecessor insn.
4200 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4202 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4203 (nvptx_process_pars): Propagate frames for calls.
4205 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4208 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4210 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4211 (ix86_handle_option): Don't handle OPT_mibt.
4212 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4214 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4216 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4217 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4218 (ix86_target_macros): Define __CET__ with flag_cf_protection
4219 for -fcf-protection.
4220 * config/i386/i386.c (isa2_opts): Remove -mibt.
4221 * config/i386/i386.h (TARGET_IBT): Removed.
4222 (TARGET_IBT_P): Likewise.
4223 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4224 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4225 * config/i386/i386.opt (mcet): Update help message.
4228 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4229 -mcet as an alias for -mshstk.
4231 2018-04-20 Richard Biener <rguenther@suse.de>
4234 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4235 complexity by forcing a single use of the multiply operand.
4237 2018-04-20 Martin Jambor <mjambor@suse.cz>
4240 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4241 recursion dependency to only apply to non-clones.
4243 2018-04-20 Martin Jambor <mjambor@suse.cz>
4246 * ipa-cp.c (create_specialized_node): Check that clones of
4247 self-recursive edges exist during IPA-CP.
4249 2018-04-19 Toon Moene <toon@moene.org>
4251 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4254 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4256 PR tree-optimization/85467
4257 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4258 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4259 VECTOR_CST element to type.
4261 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4264 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4265 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4266 (builtin_longjmp): Likewise.
4267 (save_stack_nonlocal): New pattern.
4268 (restore_stack_nonlocal): Likewise.
4270 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4273 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4274 Replace ASM_OUTPUT_LABEL with fprintf.
4276 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4279 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4280 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4281 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4282 define __IBT__ and __SHSTK__ for -fcf-protection.
4283 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4285 (ix86_trampoline_init): Likewise.
4286 (x86_output_mi_thunk): Likewise.
4287 (ix86_notrack_prefixed_insn_p): Likewise.
4288 (ix86_option_override_internal): Don't disallow -fcf-protection.
4289 * config/i386/i386.md (rdssp<mode>): Also enable for
4291 (incssp<mode>): Likewise.
4292 (nop_endbr): Likewise.
4293 * config/i386/i386.opt (mcet): Change help message to built-in
4297 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4298 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4299 enable CET built-in functions.
4301 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4303 * common/config/i386/i386-common.c
4304 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4305 OPTION_MASK_ISA_MOVDIRI_UNSET,
4306 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4307 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4308 * config.gcc (movdirintrin.h): New header.
4309 * config/i386/cpuid.h (bit_MOVDIRI,
4310 bit_MOVDIR64B): New bits.
4311 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4313 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4314 (VOID, PVOID, PCVOID)): New function types.
4315 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4316 __builtin_ia32_directstoreu_u64,
4317 __builtin_ia32_movdir64b): New builtins.
4318 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4319 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4321 (ix86_valid_target_attribute_inner_p): Ditto.
4322 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4323 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4324 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4325 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4326 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4327 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4328 (movdiri<mode>, movdir64b_<mode>): New.
4329 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4330 * config/i386/immintrin.h: Include movdirintrin.h.
4331 * config/i386/movdirintrin.h: New file.
4332 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4334 2018-04-19 Richard Biener <rguenther@suse.de>
4337 * cfg.c (clear_bb_flags): When loop state says we have
4338 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4340 2018-04-19 Richard Biener <rguenther@suse.de>
4342 PR tree-optimization/84737
4343 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4344 copying restrict info.
4345 (vect_setup_realignment): Use it.
4346 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4347 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4348 the first DR to all generated stores.
4349 (vectorizable_load): Likewise for loads.
4351 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4353 PR tree-optimization/85446
4354 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4355 the integral and pointer types to have the same precision.
4357 * doc/install.texi: Document --disable-cet being the default and
4360 2018-04-18 Martin Liska <mliska@suse.cz>
4362 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4365 2018-04-18 Martin Liska <mliska@suse.cz>
4368 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4371 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4372 arguments if they are comparable.
4374 2018-04-18 Martin Liska <mliska@suse.cz>
4377 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4380 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4383 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4386 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4387 ENDBR after calling __morestack.
4389 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4392 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4393 by using gcc_base_ver to generate a gcc_driver_version, and use
4394 it when generating GCC_DRIVER_NAME.
4395 * configure: Regenerate.
4397 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4400 * config.gcc: Obsolete powerpc*-*-*spe*.
4402 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4405 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4406 (stabstr_D): Change type of unum from unsigned int to
4407 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4410 2018-04-17 Jim Wilson <jimw@sifive.com>
4413 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4414 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4415 Set arg_pointer_offset after using pretend_args_size.
4417 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4419 PR rtl-optimization/85431
4420 * dse.c (record_store): Ignore zero width stores.
4423 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4424 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4425 __builtin_stack_restore rather than after it.
4426 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4427 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4428 argument instead of virtual_dynamic_stack_rtx.
4430 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4432 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4434 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4435 Add note to error message to explain internal mapping of overloaded
4436 built-in function name to non-overloaded built-in function name.
4437 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4440 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4443 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4444 where the inputs overlap with the output.
4446 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4448 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4449 (=v, v) alternative and explicit "memory" attribute.
4450 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4451 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4453 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4454 "sselog1" type instead of "sselog".
4455 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4456 "sselog". Remove explicit "memory" attribute.
4457 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4458 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4460 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4461 "sselog1" type instead of "sselog". Remove explicit "memory"
4463 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4464 use "sselog1" type instead of "sselog". Remove explicit "memory"
4466 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4467 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4469 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4470 "sselog1" type instead of "sselog". Remove explicit "memory"
4472 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4473 use "sselog1" type instead of "sselog". Remove explicit "memory"
4477 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4480 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4483 2018-04-17 Martin Jambor <mjambor@suse.cz>
4486 * ipa-cp.c (create_specialized_node): Call
4487 expand_all_artificial_thunks if necessary.
4489 2018-04-17 Martin Liska <mliska@suse.cz>
4492 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4493 in message, remote space in between '_G' and '('.
4495 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4498 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4499 avx512f_vmcmp<mode>3<round_saeonly_name>,
4500 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4501 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4502 avx512f_rndscale<mode><round_saeonly_name>,
4503 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4504 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4505 Use %<iptr>2 instead of %2 for -masm=intel.
4506 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4507 avx512f_vcvttss2usi<round_saeonly_name>,
4508 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4510 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4511 avx512f_vcvttsd2usi<round_saeonly_name>,
4512 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4513 Use %q1 instead of %1 for -masm=intel.
4514 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4515 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4516 of %3 for -masm=intel.
4517 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4519 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4521 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4523 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4524 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4526 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4528 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4529 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4530 %g1 and one with %0 and %1.
4531 (avx512er_vmrcp28<mode><round_saeonly_name>,
4532 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4534 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4535 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4536 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4537 of %0 and %{%4%} for -masm=intel.
4538 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4539 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4540 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4541 order of %0 and %{%5%}%{z%} for -masm=intel.
4543 2018-04-17 Jan Hubicka <jh@suse.cz>
4546 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4548 2018-04-17 Martin Liska <mliska@suse.cz>
4551 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4552 for target_clone error message. Make default implementation
4553 clone to be a local declaration.
4554 (separate_attrs): Add new argument and check for an empty
4556 (expand_target_clones): Handle it.
4557 (ipa_target_clone): Make redirection just for target_clones
4560 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4561 Tom de Vries <tom@codesourcery.com>
4564 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4565 tiled basic blocks without omp continue statements.
4567 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4570 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4571 vec_extract expression as having side effects to make sure it gets
4574 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4577 * config/i386/i386.c (get_builtin_code_for_version): Check
4580 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4583 * gcc/config.gcc: Support "skylake".
4584 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4586 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4587 (processor_target_table): Add "skylake".
4588 (ix86_option_override_internal): Add "skylake".
4589 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4590 PROCESSOR_CANNONLAKE.
4591 (get_builtin_code_for_version): Fix priority for
4592 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4593 PROCESSOR_SKYLAKE-AVX512.
4594 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4595 (processor_type): Add PROCESSOR_SKYLAKE.
4597 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4598 Jason Merrill <jason@redhat.com>
4601 * convert.c (convert_to_integer_1): Use direct recursion for
4602 enumeral types and types with a precision less than the number
4603 of bits in their mode.
4605 2018-04-16 Julia Koval <julia.koval@intel.com>
4608 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4609 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4611 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4614 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4615 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4616 and -mno-direct-move.
4618 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4621 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4622 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4623 Ensure negative shifts result in {0}.
4625 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4627 PR rtl-optimization/79916
4628 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4629 regs (if any) to define how to gnerate SD moves when LRA is in
4632 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4634 PR rtl-optimization/85393
4635 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4636 * except.c (expand_dw2_landing_pad_for_region): Make static.
4637 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4638 a label and unconditional jump to old_bb, rather than
4639 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4642 PR rtl-optimization/85376
4643 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4644 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4645 instead of a specific value.
4647 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4648 Bin Cheng <bin.cheng@arm.com>
4650 PR tree-optimization/82965
4651 PR tree-optimization/83991
4652 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4653 by_profile_only parameter.
4654 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4655 information if the loop was predicted to iterate too many times.
4656 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4658 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4661 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4664 2018-04-13 Martin Liska <mliska@suse.cz>
4665 Jakub Jelinek <jakub@redhat.com>
4668 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4669 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4670 * builtins.c (expand_builtin_memory_copy_args): Use
4671 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4672 handle dest_addr == pc_rtx.
4674 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4677 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4678 asked to not generate direct moves.
4679 (fix_trunc<mode>si2_stfiwx): Similar.
4680 (fix_trunc<mode>si2_internal): Similar.
4682 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4685 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4686 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4687 lookup if dest in some wider mode is known to be const0_rtx and
4688 if so, record permanent equivalence for it to be ZERO_EXTEND of
4689 the narrower mode destination.
4691 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4693 * lto-streamer-out.c (output_function): Revert 259346.
4694 * omp-expand.c (expand_oacc_for): Likewise.
4696 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4698 PR rtl-optimization/85354
4699 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4700 * sel-sched.c (sel_global_init): ... here.
4702 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4705 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4706 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4707 mode for PE-COFF targets.
4708 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4709 (i386_pe_asm_lto_end): Likewise.
4710 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4711 (TARGET_ASM_LTO_END): Likewise.
4712 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4713 (i386_pe_asm_lto_start): New function.
4714 (i386_pe_asm_lto_end): Likewise.
4716 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4717 Richard Biener <rguenther@suse.de>
4720 * lto-streamer-out.c (output_function): Fix CFG loop state before
4722 * omp-expand.c (expand_oacc_for): Handle calls to internal
4723 functions like regular functions.
4725 2018-04-12 Richard Biener <rguenther@suse.de>
4728 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4729 for the early LTO debug to properly generate references to it
4730 during DIE emission. Do not re-use that for the skeleton for
4732 (dwarf2out_early_finish): Likewise.
4734 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4737 * config/i386/sse.md
4738 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4739 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4740 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4741 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4742 and output is a reg, avoid creating invalid lowpart subreg, but
4743 instead split into a 512-bit move. Don't split if not AVX512VL,
4744 input is xmm16+ reg and output is a mem.
4745 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4746 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4747 xmm16+ reg and output is a mem.
4749 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4751 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4752 also for flag_dwarf2_cfi_asm.
4754 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4756 PR rtl-optimization/85342
4757 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4758 a bool scalar var inside of the loop instead. Don't try to update
4759 recog_data.operand after failed apply_change_group.
4761 2018-04-12 Tom de Vries <tom@codesourcery.com>
4764 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4765 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4766 array with flexible array member as array without given dimension.
4767 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4768 argument for undefined param to true.
4770 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4773 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4774 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4775 from PowerPC section.
4776 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4777 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4778 help text that is too long.
4779 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4780 help text that is too long.
4781 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4782 help text that is too long.
4784 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4786 * config/alpha/alpha.md (stack_probe_internal): Rename
4787 from "probe_stack". Update all callers.
4789 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4791 PR rtl-optimization/84566
4792 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4793 sched_macro_fuse_insns.
4795 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4798 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4799 (compute_block_dependences): ... from here.
4801 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4803 PR tree-optimization/85331
4804 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4805 from int to HOST_WIDE_INT.
4807 2018-04-11 Martin Jambor <mjambor@suse.cz>
4810 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4811 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4812 not the same as the source val.
4813 (cgraph_edge_brings_value_p): New parameter.
4814 (gather_edges_for_value): Pass destination value to
4815 cgraph_edge_brings_value_p.
4816 (perhaps_add_new_callers): Likewise.
4817 (get_info_about_necessary_edges): Likewise and exclude values brought
4818 only by self-recursive edges.
4819 (create_specialized_node): Redirect only clones of self-calling edges.
4820 (+self_recursive_pass_through_p): New function.
4821 (find_more_scalar_values_for_callers_subset): Use it.
4822 (find_aggregate_values_for_callers_subset): Likewise.
4823 (known_aggs_to_agg_replacement_list): Removed.
4824 (decide_whether_version_node): Re-calculate known constants for all
4825 remaining context clones.
4827 2018-04-11 Richard Biener <rguenther@suse.de>
4830 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4831 from early DWARF output.
4832 (dwarf2out_early_finish): Output line info unconditionally into
4833 early DWARF and add reference to it.
4835 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4838 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4839 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4840 other than V2DFmode using iptr mode attribute.
4841 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4843 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4845 PR rtl-optimization/84659
4846 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4848 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4851 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4853 (output_loc_list): Pass address of a dummy size variable even in the
4854 locview handling loop.
4855 (index_location_lists): Add comment on why skip_loc_list_entry can't
4858 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4861 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4864 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4867 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4868 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4869 and -mstring-compare-inline-limit.
4871 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4874 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4875 for stack clash protection in a register whenever we need it to be in
4878 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4880 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4881 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4883 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4886 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4888 (mlong-double-): Ditto.
4889 * config/rs6000/sysv4.opt (msdata=): Ditto.
4890 (mtls-size=): Ditto.
4892 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4894 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4895 erroneous entries for
4896 "vector int vec_ldl (int, long int *)", and
4897 "vector unsigned int vec_ldl (int, unsigned long int *)".
4898 Add comments and entries for
4899 "vector bool char vec_ldl (int, bool char *)",
4900 "vector bool short vec_ldl (int, bool short *)",
4901 "vector bool int vec_ldl (int, bool int *)",
4902 "vector bool long long vec_ldl (int, bool long long *)",
4903 "vector pixel vec_ldl (int, pixel *)",
4904 "vector long long vec_ldl (int, long long *)",
4905 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4906 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4907 type tree bool_long_long_type_node and correct definition of
4908 bool_V2DI_type_node to make reference to this new type tree.
4909 (rs6000_mangle_type): Replace erroneous reference to
4910 bool_long_type_node with bool_long_long_type_node.
4911 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4912 comments to emphasize sign distinctions for char and int types and
4913 replace RS6000_BTI_bool_long constant with
4914 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4915 use of RS6000_BTI_pixel.
4916 (bool_long_type_node): Remove this macro definition.
4917 (bool_long_long_type_node): New macro definition
4919 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4921 PR rtl-optimization/85300
4922 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4923 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4924 simplify_unary_operation fails.
4926 2018-04-10 Martin Liska <mliska@suse.cz>
4928 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4929 cgraph_edge and ipa_ref.
4931 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4935 * config/i386/sse.md
4936 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4937 computation of the VEC_MERGE selector from mask.
4938 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4939 Fix decoding of the VEC_MERGE selector into mask.
4941 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4943 PR tree-optimization/85286
4944 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4946 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4948 * final.c (final_1): Set insn_last_address as well as
4949 insn_current_address.
4951 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4954 * explow.c (emit_stack_probe): Call validize_mem on memory location
4955 before passing it to gen_probe_stack. Create address operand and
4956 legitimize it for the probe_stack_address case.
4958 2018-04-09 Jan Hubicka <jh@suse.cz>
4961 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4962 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4963 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4965 (free_lang_data): Rebuild type inheritance graph.
4967 2018-04-09 Martin Sebor <msebor@redhat.com>
4969 * invoke.texi (-finline-small-functions): Mention other optimization
4971 (-findirect-inlining, -fpartial-inlining): Same.
4972 (-finline-functions-called-once): Same.
4973 (-freorder-blocks-and-partition): Same.
4975 2018-04-09 Jan Hubicka <jh@suse.cz>
4978 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4979 jumps; choose last target that matches the criteria (i.e.
4980 no partition changes for non-crossing jumps).
4981 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4982 support for redirecting crossing jumps to non-crossing.
4984 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4986 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4987 also for naked functions.
4989 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4991 * config/arc/arc.md (add_shift): New pattern.
4992 (add_shift2): Likewise.
4993 (sub_shift): Likewise.
4994 (sub_shift_cmp0_noout): Likewise.
4995 (compare_si_ashiftsi): Likewise.
4996 (xbfu_cmp0_noout): New combine pattern.
4997 (xbfu_cmp0"): Likewise.
4998 (movsi_set_cc_insn): Place the predicable variant first.
4999 (commutative_binary_cmp0_noout): Remove clobber.
5000 (commutative_binary_cmp0): New pattern.
5001 (noncommutative_binary_cmp0): Likewise.
5002 (noncommutative_binary_cmp0_noout): Likewise.
5003 (noncommutative_binary_comparison_result_used): Removed.
5004 (rsub_cmp0): New pattern.
5005 (rsub_cmp0_noout): Likewise.
5006 (extzvsi): Changed, keep only meaningful variants.
5007 (SQH, SEZ): New iterators.
5008 (SQH_postfix): New mode attribute.
5009 (SEZ_prefix): New code attribute.
5010 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5011 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5012 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5014 (noncommutative_operator): Check the availability of barrel
5017 2018-04-09 Richard Biener <rguenther@suse.de>
5019 PR tree-optimization/85284
5020 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5021 Only use the niter constraining form of simple_iv when the exit
5024 2018-04-09 Tom de Vries <tom@codesourcery.com>
5027 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5028 (define_expand "*memory_barrier"): New define_expand.
5029 (define_insn "memory_barrier"): New insn.
5031 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5033 PR rtl-optimization/80463
5034 PR rtl-optimization/83972
5035 PR rtl-optimization/83480
5037 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5038 correct producer for the insn.
5039 (tidy_control_flow): Fixup seqnos in case of debug insns.
5041 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5043 PR rtl-optimization/83913
5045 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5046 different sched-times when merging exprs.
5048 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5050 PR rtl-optimization/83962
5052 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5053 tidy_fallthru_edge and tidy_control_flow.
5055 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5057 PR rtl-optimization/83530
5059 * sel-sched.c (force_next_insn): New global variable.
5060 (remove_insn_for_debug): When force_next_insn is true, also leave only
5061 next insn in the ready list.
5062 (sel_sched_region): When the region wasn't scheduled, make another pass
5063 over it with force_next_insn set to 1.
5065 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5067 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5069 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5070 for interrupt control.
5071 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5072 functions for interrupt control.
5073 * config/nds32/nds32-intrinsic.md: Likewise.
5074 * config/nds32/nds32_intrinsic.h: Likewise.
5075 * config/nds32/nds32.h (nds32_builtins): Likewise.
5077 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5079 * config/nds32/nds32.c (nds32_init_machine_status,
5080 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5081 strict_aligned_p field.
5082 (nds32_expand_to_rtl_hook): New function.
5083 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5084 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5086 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5087 Chung-Ju Wu <jasonwucj@gmail.com>
5089 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5090 * config/nds32/nds32-n7.md: New file.
5091 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5092 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5094 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5095 * config/nds32/nds32.md (pipeline_model): Add n7.
5096 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5097 * config/nds32/pipelines.md: Include n7 settings.
5099 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5100 Chung-Ju Wu <jasonwucj@gmail.com>
5102 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5103 * config/nds32/nds32-e8.md: New file.
5104 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5105 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5107 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5108 * config/nds32/nds32.md (pipeline_model): Add e8.
5109 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5110 * config/nds32/pipelines.md: Include e8 settings.
5112 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5113 Chung-Ju Wu <jasonwucj@gmail.com>
5115 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5116 * config/nds32/nds32-n8.md: New file.
5117 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5118 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5120 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5121 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5122 * config/nds32/nds32.md (pipeline_model): Add n8.
5123 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5124 * config/nds32/pipelines.md: Include n8 settings.
5126 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5127 Chung-Ju Wu <jasonwucj@gmail.com>
5129 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5130 * config/nds32/nds32-n9-2r1w.md: New file.
5131 * config/nds32/nds32-n9-3r2w.md: New file.
5132 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5133 nds32_register_ports): New or modify for cpu n9.
5134 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5136 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5137 * config/nds32/nds32-utils.c: New file.
5138 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5139 TARGET_MUL_SLOW): Define.
5140 * config/nds32/nds32.md (pipeline_model): New attribute.
5141 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5142 New options that support cpu n9.
5143 * config/nds32/pipelines.md: Include n9 settings.
5144 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5146 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5148 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5149 information if necessary.
5150 (output_cond_branch_compare_zero): Likewise.
5151 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5152 (nds32_target_alignment): Refine for alignment.
5153 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5154 (FUNCTION_BOUNDARY): Modify.
5155 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5157 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5159 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5161 * config/nds32/constants.md (unspec_volatile_element): Add values for
5162 TLB operation and data prefetch.
5163 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5164 functions for TLB operation and data prefetch.
5165 * config/nds32/nds32-intrinsic.md: Likewise.
5166 * config/nds32/nds32_intrinsic.h: Likewise.
5167 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5168 (nds32_print_operand): Likewise.
5169 * config/nds32/nds32.h (nds32_builtins): Likewise.
5171 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5172 Andrew Pinski <pinsika@gcc.gnu.org>
5175 * match.pd: Use constant_boolean_node of correct type instead of
5176 boolean_true_node or boolean_false_node for simplifying
5177 pointer comparisons to zero.
5179 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5181 PR tree-optimization/80021
5182 * tree.c (verify_type_variant): Make error call in verify_variant_match
5183 translatable and remove final full stop.
5185 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5187 * config/nds32/constants.md (unspec_volatile_element): Add
5188 UNSPEC_VOLATILE_EH_RETURN.
5189 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5190 nds32_output_stack_pop): Support dwarf exception handling process.
5191 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5192 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5193 exception handling process.
5194 (nds32_compute_stack_frame): Likewise.
5195 (nds32_return_addr_rtx): Likewise.
5196 (nds32_initial_elimination_offset): Likewise.
5197 (nds32_expand_prologue): Likewise.
5198 (nds32_expand_epilogue): Likewise.
5199 (nds32_dynamic_chain_address): New function.
5200 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5202 (DYNAMIC_CHAIN_ADDRESS): Define.
5203 (EH_RETURN_DATA_REGNO): Define.
5204 (EH_RETURN_STACKADJ_RTX): Define.
5205 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5206 patterns for dwarf exception handling.
5208 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5210 * config/nds32/nds32.h: Clean up obsolete macros.
5212 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5214 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5215 Add enum values for particular instructions.
5216 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5217 particular intrinsic functions.
5218 * config/nds32/nds32-intrinsic.md: Likewise.
5219 * config/nds32/nds32_intrinsic.h: Likewise.
5220 * config/nds32/nds32.h (nds32_builtins): Likewise.
5221 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5222 (btst, ave): New patterns for particular instructions.
5224 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5226 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5227 Add enum values for atomic load/store and memory sync.
5228 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5230 * config/nds32/nds32-intrinsic.md: Likewise.
5231 * config/nds32/nds32_intrinsic.h: Likewise.
5232 * config/nds32/nds32.h (nds32_builtins): Likewise.
5234 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5236 PR tree-optimization/85257
5237 * fold-const.c (native_encode_vector): If not all elts could fit
5238 and off is -1, return 0 rather than offset.
5239 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5240 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5241 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5242 adjust buffer in native_interpret_expr call.
5244 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5246 * config/nds32/constants.md (unspec_volatile_element): Add cache
5247 control enum values.
5248 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5249 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5250 * config/nds32/nds32.c (nds32_cctl_names): New.
5251 (nds32_print_operand): Handle cache control register names.
5252 * config/nds32/nds32.h (nds32_builtins): New enum values.
5253 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5255 * config/nds32/nds32.md (type): Add mmu.
5256 * config/nds32/pipelines.md (simple_insn): Add mmu.
5258 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5260 * config/nds32/nds32.md (type): Remove call.
5261 * config/nds32/pipelines.md (simple_insn): Likewise.
5263 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5265 * config/nds32/constants.md (unspec_volatile_element): Add
5266 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5267 UNSPEC_VOLATILE_FMFCFG.
5268 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5269 description for fmfcfg and fmfcsr.
5270 (bdesc_1arg): Add fmtcsr.
5271 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5272 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5273 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5274 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5275 unspec_fmfcfg): New patterns.
5276 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5277 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5278 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5279 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5280 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5281 __nds32__fmfcfg): Define.
5283 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5285 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5286 intrinsic register names.
5287 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5288 intrinsic register enum values and macros.
5290 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5292 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5293 for load/store addressing form.
5294 (nds32_print_operand_address): Likewise.
5296 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5299 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5300 based on LABEL_REF. Remove useless assertion.
5301 (pic_address_needs_scratch): Fix formatting.
5302 (sparc_legitimize_pic_address): Minor tweaks.
5303 (sparc_delegitimize_address): Adjust assertion accordingly.
5304 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5305 into symbolic_operand.
5306 (movsi_high_pic_label_ref): Likewise.
5307 (movsi_lo_sum_pic_label_ref): Likewise.
5308 (movdi_pic_label_ref): Likewise.
5309 (movdi_high_pic_label_ref): Likewise.
5310 (movdi_lo_sum_pic_label_ref): Likewise.
5312 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5314 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5315 custom LIB_SPEC setup.
5317 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5318 Kito Cheng <kito.cheng@gmail.com>
5320 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5321 * config/riscv/freebsd.h: New.
5323 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5325 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5326 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5329 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5330 Kito Cheng <kito.cheng@gmail.com>
5332 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5333 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5334 * config/nds32/nds32-protos.h (nds32_output_call,
5335 nds32_output_return): Declare.
5336 * config/nds32/nds32.md: Refine all the call and return patterns.
5338 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5341 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5342 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5344 PR rtl-optimization/84872
5345 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5346 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5349 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5351 * expr.c (copy_blkmode_to_reg): Revert 254862.
5352 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5354 2018-04-06 Richard Biener <rguenther@suse.de>
5357 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5358 after seeing a component reference with an adjacent field. Treat
5359 refs to arrays at struct end of external decls similar to
5360 refs to unconstrained commons.
5362 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5365 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5366 look through SAVE_EXPRs with non-side-effects argument. Adjust
5368 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5371 2018-04-06 Richard Biener <rguenther@suse.de>
5374 * alias.c (find_base_term): New wrapper around find_base_term
5375 unwinding CSELIB_VAL_PTR changes.
5376 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5379 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5381 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5383 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5384 constant definitions.
5385 ("nop"): lr 0,0 -> nopr r0
5386 ("nop_lr0", "nop_lr1"): New insn definitions.
5388 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5390 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5391 NDS32_V3PUSH_AVAILABLE_P macro.
5393 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5394 Chung-Ju Wu <jasonwucj@gmail.com>
5396 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5397 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5398 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5399 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5400 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5401 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5402 * config/nds32/constraints.md: New constraints and checking for hard
5403 float configuration.
5404 * config/nds32/iterators.md: New mode iterator and attribute for hard
5405 float configuration.
5406 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5408 * config/nds32/nds32-fpu.md: New file.
5409 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5410 deal with hard float code generation.
5411 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5413 (abi_type, float_reg_number): New enum type.
5414 * config/nds32/nds32-predicates.c: New predicates for hard float.
5415 * config/nds32/nds32-protos.h: Declare functions for hard float.
5416 * config/nds32/nds32.c: Implementation for hard float configuration.
5417 * config/nds32/nds32.h: Definitions for hard float configuration.
5418 * config/nds32/nds32.md: Include hard float machine description and
5419 modify patterns for hard float configuration.
5420 * config/nds32/nds32.opt: New options for hard float configuration.
5421 * config/nds32/predicates.md: New predicates for hard float
5424 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5426 * common/config/nds32/nds32-common.c
5427 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5429 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5432 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5433 CONSTRUCTOR_ELT (ctor, ...)->value.
5435 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5438 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5440 2018-04-05 Tom de Vries <tom@codesourcery.com>
5443 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5446 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5447 Kito Cheng <kito.cheng@gmail.com>
5449 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5450 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5451 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5452 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5454 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5455 Kito Cheng <kito.cheng@gmail.com>
5457 * config/nds32/constraints.md (Ufe): New memory constraint.
5458 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5459 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5460 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5462 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5463 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5465 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5467 * config/nds32/nds32.md: Use optimize_size in the condition for
5468 alu-shift instructions.
5470 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5472 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5474 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5476 * config/nds32/nds32.md (negsi2): Refine pattern.
5478 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5479 Chung-Ju Wu <jasonwucj@gmail.com>
5481 * config/nds32/iterators.md (shift_rotate): New code iterator.
5482 (shift): New code attribute.
5483 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5484 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5485 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5486 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5487 bit-wise operations.
5488 (andsi3, *andsi3): Ditto.
5489 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5490 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5491 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5492 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5493 nds32_ior_operand, nds32_xor_operand): New predicates.
5495 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5497 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5498 (addsi3, subsi3): ... this.
5500 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5502 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5504 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5506 * config/nds32/nds32.md: Adjust indention.
5508 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5510 * config/nds32/nds32.md (feature): New attribute.
5512 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5514 * config/nds32/nds32.md (subtype): New attribute.
5516 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5519 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5520 expansion to perform a bitwise AND of the argument followed by a
5521 boolean negation of the result.
5523 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5525 PR rtl-optimization/84878
5526 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5527 the basic block. Assert the use reference is not artificial and that
5528 it has an associated insn.
5530 2018-04-04 Michael Matz <matz@suse.de>
5532 * builtins.c (compute_objsize): Pass correct operand
5533 to array_at_struct_end_p.
5535 2018-04-04 Richard Biener <rguenther@suse.de>
5538 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5539 from contexts for DINFO_LEVEL_TERSE and below.
5541 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5543 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5544 resiter_operand condition.
5545 * config/nds32/nds32.md (*move<mode>): Ditto.
5547 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5548 Monk Chiang <sh.chiang04@gmail.com>
5550 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5552 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5554 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5556 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5557 Kito Cheng <kito.cheng@gmail.com>
5559 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5560 nds32_cond_code_str, output_cond_branch,
5561 output_cond_branch_compare_zero, nds32_expand_cbranch,
5562 nds32_expand_cstore, nds32_expand_movcc,
5563 nds32_output_cbranchsi4_equality_zero,
5564 nds32_output_cbranchsi4_equality_reg,
5565 nds32_output_cbranchsi4_equality_reg_or_const_int,
5566 nds32_output_cbranchsi4_greater_less_zero: New functions.
5567 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5568 nds32_expand_cstore, nds32_expand_movcc,
5569 nds32_output_cbranchsi4_equality_zero,
5570 nds32_output_cbranchsi4_equality_reg,
5571 nds32_output_cbranchsi4_equality_reg_or_const_int,
5572 nds32_output_cbranchsi4_greater_less_zero): Declare.
5573 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5574 nds32_rimm11s_operand): New predicates.
5575 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5576 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5579 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5581 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5582 * config/nds32/nds32.md: Ditto.
5583 * config/nds32/pipelines.md: Ditto.
5585 2018-04-04 Richard Biener <rguenther@suse.de>
5587 PR tree-optimization/85168
5588 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5589 propagating abnormals.
5591 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5593 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5595 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5596 Kito Cheng <kito.cheng@gmail.com>
5598 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5599 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5600 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5601 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5602 * config/nds32/nds32.md (sibcall_internal): New.
5603 (sibcall_register): Remove.
5604 (sibcall_immediate): Remove.
5605 (sibcall_value_internal): New.
5606 (sibcall_value_register): Remove.
5607 (sibcall_value_immediate): Remove.
5608 * config/nds32/predicates.md (nds32_general_register_operand): New.
5609 (nds32_call_address_operand): New.
5611 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5613 PR rtl-optimization/85167
5614 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5615 bb_defs if *split_p, instead preinitialize it to NULL.
5617 PR tree-optimization/85156
5618 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5619 evaluating the argument multiple times.
5621 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5623 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5625 (_mm_cvtpd_ps): Likewise.
5626 (_mm_cvttpd_epi32): Likewise.
5627 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5628 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5629 vector, pixel, and bool following altivec.h include.
5631 2018-04-03 Martin Sebor <msebor@redhat.com>
5633 * doc/extend.texi (Common Function Attributes): Clarify.
5634 (const attribute): Likewise.
5635 (pure attribute): Likewise.
5637 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5640 * config/i386/i386.c (ix86_expand_vector_set): Use
5641 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5643 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5645 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5646 instructions when changing rounding bits to preserve precision bits
5647 in the x87 control word.
5649 2018-04-03 Martin Liska <mliska@suse.cz>
5651 PR tree-optimization/82491
5652 * rtl.h (strip_offset_and_add): Replace += suboffset with
5653 poly_uint64 () + suboffset.
5655 2018-03-29 Martin Liska <mliska@suse.cz>
5656 Martin Jambor <mjambor@suse.cz>
5659 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5660 param_type is not an integral or pointer type.
5662 2018-04-03 Richard Biener <rguenther@suse.de>
5664 * sese.h (recompute_all_dominators): Remove.
5666 2018-04-02 Martin Sebor <msebor@redhat.com>
5668 * doc/invoke.texi (-Wrestrict): Fix typos.
5670 2018-04-02 Jim Wilson <jimw@sifive.com>
5672 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5673 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5674 (<optab>di3, <optab>si3_extend): Likewise.
5675 (<optab>si3_mask, <optab>si3_mask_1): New.
5676 (<optab>di3_mask, <optab>di3_mask_1): New.
5677 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5678 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5679 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5681 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5683 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5686 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5688 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5689 (nds32_canonicalize_comparison): New function.
5691 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5692 Kito Cheng <kito.cheng@gmail.com>
5693 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5695 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5696 * config/nds32/constants.md (unspec_volatile_element): Add
5697 UNSPEC_VOLATILE_RELAX_GROUP.
5698 * config/nds32/nds32-relax-opt.c: New file.
5699 * config/nds32/nds32-predicates.c
5700 (nds32_symbol_load_store_p): New function.
5701 * config/nds32/nds32-protos.h
5702 (nds32_symbol_load_store_p): Declare function.
5703 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5704 * config/nds32/nds32.c
5705 (nds32_register_pass): New function to register pass.
5706 (nds32_register_passes): New function to register passes.
5707 * config/nds32/nds32.md (relax_group): New pattern.
5708 * config/nds32/nds32.opt (mrelax-hint): New option.
5709 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5711 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5713 * config/nds32/t-nds32: Modify files dependency.
5715 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5717 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5718 (PROFILE_HOOK): Define its implementation.
5720 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5722 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5723 type and 32-bit size.
5725 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5728 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5729 (V_128_256): New mode iterator.
5730 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5731 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5732 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5734 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5735 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5737 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5740 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5741 NaN inputs correctly.
5743 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5746 * config/rs6000/vsx.md (??r): New mode attribute.
5747 (*vsx_mov<mode>_64bit): Use it.
5748 (*vsx_mov<mode>_32bit): Likewise.
5750 2018-03-30 Martin Sebor <msebor@redhat.com>
5752 PR tree-optimization/84818
5753 * builtins.c (check_access): Use warning_n.
5755 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5758 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5760 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5763 2018-03-30 Julia Koval <julia.koval@intel.com>
5766 * x86-tune.def (movx, partial_reg_dependency): Enable for
5769 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5772 * lra-constraints.c (process_alt_operands): Move setting
5773 this_alternative_matches below.
5775 2018-03-29 Martin Liska <mliska@suse.cz>
5778 * doc/invoke.texi: Document how LTO works with debug info.
5779 Describe auto-load support of binutils. Mention 'x86-64'
5780 as valid option value of -march option.
5782 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5784 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5787 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5788 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5789 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5792 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5795 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5796 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5797 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5798 for RS6000_BTM_POWERPC64.
5799 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5800 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5801 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5806 2018-03-28 Carl Love <cel@us.ibm.com>
5809 2017-09-27 Carl Love <cel@us.ibm.com>
5811 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5812 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5813 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5816 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5818 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5819 instead of __vector bool.
5820 (_mm_max_pu8): Likewise.
5821 (_mm_min_pi16): Likewise.
5823 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5826 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5827 (DIVWEUO): Likewise.
5829 (DIVDEUO): Likewise.
5830 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5831 DIVWEUO and DIVDEUO.
5832 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5833 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5834 (div_extend): Likewise.
5835 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5837 (__builtin_divweuo): Likewise.
5838 (__builtin_divdeo): Likewise.
5839 (__builtin_divdeuo): Likewise.
5841 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5844 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5845 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5847 PR tree-optimization/82004
5848 * gimple-match-head.c (optimize_pow_to_exp): New function.
5849 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5850 Don't fold to exp if optimize_pow_to_exp is false.
5852 2018-03-28 Martin Liska <mliska@suse.cz>
5855 * calls.c (initialize_argument_information): Fix trailing space.
5856 * common.opt: Fix typo and provide better explanation for
5857 -fsanitize-coverage option.
5858 * config/i386/i386.opt: Fix typo.
5860 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5861 Martin Liska <mliska@suse.cz>
5864 * gimplify.c (asan_poison_variable): Don't do the check for
5865 gimplify_omp_ctxp here.
5866 (gimplify_decl_expr): Do it here.
5867 (gimplify_target_expr): Likewise.
5869 2018-03-28 Martin Liska <mliska@suse.cz>
5872 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5873 chkp_type_bounds_count if MPX is not enabled.
5875 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5877 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5879 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5882 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5883 function to create the function decl for complex long double
5884 multiply and divide for -mabi=ieeelongdouble.
5885 (init_float128_ieee): Call it.
5887 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5890 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5891 -fcf-protection=branch -mibt.
5892 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5894 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5897 * config/arm/arm.c (arm_valid_symbolic_address): Handle
5898 arm_word_relocations.
5900 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5903 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5904 extern array declarations.
5906 2018-03-27 Richard Biener <rguenther@suse.de>
5909 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5910 explicit single_use checks.
5912 2018-03-27 Richard Biener <rguenther@suse.de>
5914 PR tree-optimization/85082
5915 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5918 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5920 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5921 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5922 Turn on fasynchronous-unwind-tables and funwind-tables.
5924 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5927 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5928 (*bmi_blsr_<mode>_ccz): Ditto.
5930 2018-03-26 Tom de Vries <tom@codesourcery.com>
5932 PR tree-optimization/85063
5933 * omp-general.c (offloading_function_p): New function. Factor out
5935 * omp-offload.c (pass_omp_target_link::gate): ... here.
5936 * omp-general.h (offloading_function_p): Declare.
5937 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5938 with attribute omp declare target for offloading functions.
5940 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5942 PR tree-optimization/84005
5943 * tree-data-ref.h (get_base_for_alignment): Declare.
5944 * tree-data-ref.c (get_base_for_alignment_1): New function.
5945 (get_base_for_alignment): Likewise.
5946 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5947 get_base_for_alignment to find a suitable base object, instead
5948 of always using drb->base_address.
5950 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5953 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5954 known size by default.
5956 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5959 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5960 and non BLKmode operands.
5962 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5965 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5966 Clean up attributes.
5968 2018-03-23 Richard Biener <rguenther@suse.de>
5971 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5972 we are going to emit early debug for LTO.
5974 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5977 * function.c (match_asm_constraints_1): Don't optimize if input
5978 doesn't satisfy general_operand predicate for output's mode.
5981 * alias.c (write_dependence_p): Don't require for x_canonicalized
5982 non-VOIDmode if x has VOIDmode.
5985 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5986 just don't try to optimize it rather than assert it never happens.
5988 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5990 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5991 macro expansions for definition of ST_INTERNAL_<mode> and
5992 LD_INTERNAL_<mode> builtins.
5993 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5995 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5997 (altivec_expand_st_builtin): Likewise.
5998 (altivec_expand_builtin): Remove calls to deleted functions.
5999 (rs6000_address_for_altivec): Delete this function.
6000 * config/rs6000/vector.md: Remove expands for
6001 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6003 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6006 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6007 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6008 re-computing once computed.
6009 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6010 (arm_init_machine_status): Initialize
6011 machine->static_chain_stack_bytes.
6013 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6016 * doc/extend.texi: Add four new prototypes for vec_ld.
6017 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6018 definitions for more logical presentation.
6019 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6020 entries for V1TI variants of __builtin_altivec_ld builtin.
6021 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6022 handling of V1TI variant of LVX icode pattern.
6023 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6024 (rs6000_gimple_fold_builtin): Likewise.
6025 (altivec_init_builtins): Add code to define
6026 __builtin_altivec_lvx_v1ti function.
6028 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6031 * function.c (match_asm_constraints_1): Don't do the optimization
6032 if input isn't a REG, SUBREG, MEM or constant.
6034 2018-03-22 Tom de Vries <tom@codesourcery.com>
6036 PR tree-optimization/84956
6037 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6038 bb_has_abnormal_pred.
6040 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6043 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6044 DECL_INITIAL (decl) to decl at the end.
6045 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6048 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6050 * doc/extend.texi (__builtin_tgmath): Document when complex
6051 integer types are treated as _Complex _Float64.
6053 2018-03-21 Tom de Vries <tom@codesourcery.com>
6055 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6057 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6059 PR tree-optimization/84960
6060 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6061 if it is ENTRY block, move them into single succ of ENTRY in that case.
6063 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6065 PR tree-optimization/84811
6066 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6067 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6068 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6069 (known_subrange_p): Update accordingly. Cast each value involved
6070 in the size comparison, rather than casting the result of the
6073 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6075 PR tree-optimization/84982
6076 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6077 by flipping the least significant bit rather than all bits from
6078 bitpos to bitpos + bitsize - 1.
6080 2018-03-21 Nathan Sidwell <nathan@acm.org>
6082 * doc/extend.texi (Deprecated Features): Remove mention of
6083 long-deleted deprecations.
6085 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6088 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6089 * configure: Regenerate.
6091 2018-03-21 Tom de Vries <tom@codesourcery.com>
6093 PR tree-optimization/83126
6094 * tree-parloops.c (num_phis): New function.
6095 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6097 2018-03-21 Nathan Sidwell <nathan@acm.org>
6099 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6100 mention anon-struct/union members and trailing attributes.
6102 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6104 PR tree-optimization/84969
6105 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6106 builtin memset partitions if they set different rhs values.
6108 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6110 PR rtl-optimization/84989
6111 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6112 VEC_DUPLICATE with scalar result mode.
6114 2018-03-21 Martin Liska <mliska@suse.cz>
6117 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6118 not intended return statement.
6120 2018-03-21 Martin Liska <mliska@suse.cz>
6123 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6124 (chkp_find_bound_slots_1): Limit number of iterations.
6126 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6129 * Minor grammar fixes for x86 options.
6131 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6134 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6135 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6137 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6140 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6141 (altivec_lvx_<mode>_1op): Likewise.
6142 (altivec_stvx_<mode>_2op): Likewise.
6143 (altivec_stvx_<mode>_1op): Likewise.
6144 (altivec_lvx_<VM2:mode>): New define_expand.
6145 (altivec_stvx_<VM2:mode>): Likewise.
6146 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6147 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6148 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6149 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6150 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6151 (rs6000_gen_lvx): Likewise.
6152 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6153 (altivec_expand_stv_builtin): Likewise.
6154 (altivec_expand_builtin): Likewise.
6155 * config/rs6000/vector.md: Likewise.
6157 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6160 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6163 2018-03-20 Richard Biener <rguenther@suse.de>
6166 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6167 sign-conversions as zero, fall back to standard scalar_stmt
6170 2018-03-20 Martin Liska <mliska@suse.cz>
6173 * predict.c (rebuild_frequencies): Handle case when we have
6174 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6176 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6179 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6180 flag_section_anchors.
6181 * varasm.c (use_blocks_for_decl_p): Remove hack for
6182 dw2_force_const_mem.
6185 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6187 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6188 be created, use lowpart_subreg of operands[0] rather than operands[0]
6190 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6191 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6192 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6193 and n constraint instead of aarch64_shift_imm_di and Usd.
6194 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6195 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6197 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6200 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6201 to favor GPR over NEON registers.
6202 (<shift>di3_neon): Likewise.
6204 2018-03-20 Tom de Vries <tom@codesourcery.com>
6207 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6208 (nvptx_process_pars): Emit bar.sync asap and alap.
6210 2018-03-20 Tom de Vries <tom@codesourcery.com>
6213 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6214 seen_label if seen_label is already set.
6216 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6219 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6220 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6221 Use 1U instead of 1. Formatting fixes.
6224 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6225 instead of TREE_TYPE (s1) for the return value.
6227 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6229 PR tree-optimization/84946
6230 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6231 bitsize + bitsize in poly_uint64 rather than poly_int64.
6234 * dwarf2asm.c: Include fold-const.c.
6235 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6236 of decl rather than decl itself.
6238 PR rtl-optimization/84643
6239 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6241 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6244 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6245 calling assemble_variable.
6247 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6250 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6251 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6253 2018-03-19 Jim Wilson <jimw@sifive.com>
6256 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6257 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6258 (riscv_first_stack_step): Likewise.
6259 (riscv_option_override): Use STACK_BOUNDARY instead of
6261 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6263 (BIGGEST_ALIGNMENT): Set to 128.
6264 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6265 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6268 2018-03-19 Richard Biener <rguenther@suse.de>
6270 PR tree-optimization/84933
6271 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6272 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6274 2018-03-19 Richard Biener <rguenther@suse.de>
6276 PR tree-optimization/84859
6277 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6278 (cond_if_else_store_replacement): Perform sinking operation on
6279 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6280 Generalize what a BB with a single eligible store is.
6282 2018-03-19 Richard Biener <rguenther@suse.de>
6284 PR tree-optimization/84929
6285 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6286 chrec_is_positive against non-chrec arg.
6288 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6291 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6293 2018-03-18 Martin Liska <mliska@suse.cz>
6295 PR rtl-optimization/84635
6296 * regrename.c (build_def_use): Use matches_mode only when
6299 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6301 PR tree-optimization/84913
6302 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6303 vectorize chains of COND_EXPRs.
6305 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6307 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6309 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6311 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6313 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6315 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6317 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6318 Kito Cheng <kito.cheng@gmail.com>
6320 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6321 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6322 (nds32_adjust_reg_alloc_order): New function.
6323 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6325 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6327 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6328 nds32_print_operand, nds32_print_operand_address): Use
6329 HOST_WIDE_INT_PRINT_DEC instead.
6331 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6333 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6335 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6338 * config/i386/i386.c (initial_ix86_tune_features,
6339 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6341 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6342 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6343 rather than 1u << ix86_tune. Formatting fix.
6344 (ix86_option_override_internal): Change ix86_arch_mask from
6345 unsigned int to unsigned HOST_WIDE_INT, initialize to
6346 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6347 (ix86_function_specific_restore): Likewise.
6349 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6352 * postreload.c (reload_combine_recognize_pattern): Perform
6353 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6354 truncate_int_for_mode the result for the destination's mode.
6357 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6358 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6360 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6363 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6366 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6367 regno_allocno_class_array and sorted_pseudos.
6368 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6369 insns where regno is used.
6371 2018-03-16 Martin Liska <mliska@suse.cz>
6374 * multiple_target.c (create_dispatcher_calls): Redirect
6375 reference in the symbol table.
6377 2018-03-16 Martin Liska <mliska@suse.cz>
6380 * multiple_target.c (create_dispatcher_calls): Redirect also
6383 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6387 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6388 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6389 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6391 2018-03-16 Julia Koval <julia.koval@intel.com>
6393 * doc/invoke.texi (Skylake Server): Add CLWB.
6394 Cannonlake): Remove CLWB.
6396 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6398 PR tree-optimization/84841
6399 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6401 (FLOAT_ONE_CONST_TYPE): Define.
6402 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6403 (sort_by_operand_rank): Put entries with higher constant_type last
6404 rather than first to match comments.
6406 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6408 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6411 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6414 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6415 check and use address of target always.
6417 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6420 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6421 (indirect_thunk_bnd_needed): Likewise.
6422 (indirect_thunks_used): Likewise.
6423 (indirect_thunks_bnd_used): Likewise.
6424 (indirect_return_needed): New.
6425 (indirect_return_bnd_needed): Likewise.
6426 (output_indirect_thunk_function): Add a bool argument for
6428 (output_indirect_thunk_function): Don't generate alias for
6429 function return thunk.
6430 (ix86_code_end): Call output_indirect_thunk_function to generate
6431 function return thunks.
6432 (ix86_output_function_return): Set indirect_return_bnd_needed
6433 and indirect_return_needed instead of indirect_thunk_bnd_needed
6434 and indirect_thunk_needed.
6436 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6438 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6439 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6440 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6442 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6443 Paul Hua <paul.hua.gm@gmail.com>
6446 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6448 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6450 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6451 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6454 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6457 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6458 instead of GET_MODE_SIZE when comparing Units.
6460 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6463 * varasm.c (hash_section): Return an unchangeble hash value
6464 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6465 Return !aarch64_can_use_per_function_literal_pools_p ().
6467 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6470 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6471 rather than address of cmode as last argument to prepare_cmp_insn.
6473 2018-03-15 Julia Koval <julia.koval@intel.com>
6475 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6476 F_AVX512VNNI, F_AVX512BITALG): New.
6478 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6481 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6482 insn for floating-point loads and stores.
6484 2018-03-14 Carl Love <cel@us.ibm.com>
6486 * config/rs6000/rs6000-c.c: Add macro definitions for
6487 ALTIVEC_BUILTIN_VEC_PERMXOR.
6488 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6489 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6490 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6491 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6493 * config/doc/extend.texi: Add prototypes for vec_permxor.
6495 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6498 * diagnostic-show-locus.c (class layout_point): Convert m_line
6499 from int to linenum_type.
6500 (line_span::comparator): Use linenum "compare" function when
6501 comparing line numbers.
6502 (test_line_span): New function.
6503 (layout_range::contains_point): Convert param "row" from int to
6505 (layout_range::intersects_line_p): Likewise.
6506 (layout::will_show_line_p): Likewise.
6507 (layout::print_source_line): Likewise.
6508 (layout::should_print_annotation_line_p): Likewise.
6509 (layout::print_annotation_line): Likewise.
6510 (layout::print_leading_fixits): Likewise.
6511 (layout::annotation_line_showed_range_p): Likewise.
6512 (struct line_corrections): Likewise for field m_row.
6513 (line_corrections::line_corrections): Likewise for param "row".
6514 (layout::print_trailing_fixits): Likewise.
6515 (layout::get_state_at_point): Likewise.
6516 (layout::get_x_bound_for_row): Likewise.
6517 (layout::print_line): Likewise.
6518 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6519 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6520 * input.c (selftest::test_linenum_comparisons): New function.
6521 (selftest::input_c_tests): Call it.
6522 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6523 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6524 * selftest.h (ASSERT_GT): New macro.
6525 (ASSERT_GT_AT): New macro.
6526 (ASSERT_LT): New macro.
6527 (ASSERT_LT_AT): New macro.
6529 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6531 PR rtl-optimization/84780
6532 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6534 2018-03-14 Martin Liska <mliska@suse.cz>
6536 * tree.c (record_node_allocation_statistics): Use
6537 get_stats_node_kind.
6538 (get_stats_node_kind): New function extracted from
6539 record_node_allocation_statistics.
6540 (free_node): Use get_stats_node_kind.
6542 2018-03-14 Richard Biener <rguenther@suse.de>
6544 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6545 that the value-set of ANTIC_IN doesn't grow.
6548 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6550 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6551 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6553 2018-03-14 Julia Koval <julia.koval@intel.com>
6555 * config.gcc (icelake-client, icelake-server): New.
6557 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6558 (initial_ix86_arch_features): Ditto.
6559 (PTA_SKYLAKE): Add SGX.
6560 (PTA_ICELAKE): Remove.
6561 (PTA_ICELAKE_CLIENT): New.
6562 (PTA_ICELAKE_SERVER): New.
6563 (ix86_option_override_internal): Split up icelake on icelake client and
6565 (get_builtin_code_for_version): Ditto.
6566 (fold_builtin_cpu): Ditto.
6567 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6568 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6569 * config/i386/i386.h (processor_type): Ditto.
6570 * doc/invoke.texi: Ditto.
6572 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6575 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6576 INTEGER_CST offset, add it together with bitpos / 8 and
6577 sign extend based on POINTER_SIZE.
6581 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6584 * config/i386/constraints.md (Yc): New register constraint.
6585 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6586 Use Yc constraint for alternative 2 of operand 0. Remove
6587 preferred_for_speed attribute.
6589 2018-03-14 Richard Biener <rguenther@suse.de>
6591 PR tree-optimization/84830
6592 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6593 with the old one to avoid oscillations.
6595 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6598 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6600 (assign_by_spills): Return a flag of reload assignment failure.
6601 Do not process the reload assignment failures. Do not spill other
6602 reload pseudos if they has the same reg class. Update n if
6604 (lra_assign): Add a return arg. Set up from the result of
6605 assign_by_spills call.
6606 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6607 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6608 usage_insns if it is not NULL.
6609 (spill_hard_reg_in_range): New function.
6610 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6611 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6612 function prototypes.
6613 (lra_assign): Change prototype.
6614 * lra.c (lra): Add code to deal with fails by splitting hard reg
6617 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6619 * config/riscv/riscv.opt (mrelax): New option.
6620 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6621 "norelax" when riscv_mrelax is disabled.
6622 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6624 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6627 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6628 reassociation for int modes.
6630 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6632 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6633 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6635 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6636 * config/aarch64/aarch64-sve.md
6637 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6638 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6639 (*extend<mode><Vwide>2): Rename to...
6640 (aarch64_sve_extend<mode><Vwide>2): ...this.
6641 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6642 renaming the old pattern to...
6643 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6645 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6646 define_expand, renaming the old pattern to...
6647 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6648 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6649 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6650 account when deciding which SVE instruction the optab should use.
6651 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6653 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6655 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6656 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6657 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6658 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6659 (tlsdesc_small_<mode>): Turn a define_expand and use
6660 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6661 (tlsdesc_small_advsimd_<mode>): ...this.
6662 (tlsdesc_small_sve_<mode>): New pattern.
6664 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6666 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6667 (UNSPEC_UMUL_HIGHPART): New constants.
6668 (MUL_HIGHPART): New int iteraor.
6669 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6670 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6672 (*<su>mul<mode>3_highpart): New define_insn.
6674 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6677 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6680 2018-03-13 Martin Liska <mliska@suse.cz>
6683 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6685 (sem_item_optimizer::~sem_item_optimizer): Release it.
6686 (sem_item_optimizer::merge_classes): Register variable aliases.
6687 (sem_item_optimizer::fixup_pt_set): New function.
6688 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6689 * ipa-icf.h: Declare new variables and functions.
6691 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6694 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6695 integer_pow2p@2 and test integer_pow2p in condition.
6696 (A < 0 ? C : 0): Similarly for @1.
6699 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6700 characters starting at p contain '\0' character, don't look beyond
6704 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6705 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6708 * reg-stack.c (change_stack): Change update_end var from int to
6709 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6710 also call set_block_for_insn on the newly added insns and rescan.
6713 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6714 on the last operand.
6717 * tree.c (stabilize_reference_1): Return save_expr (e) for
6718 STATEMENT_LIST even if it doesn't have side-effects.
6720 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6722 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6724 2018-03-12 Renlin Li <renlin.li@arm.com>
6726 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6727 aarch64_output_scalar_simd_mov_immediate.
6729 2018-03-12 Martin Sebor <msebor@redhat.com>
6731 PR tree-optimization/83456
6732 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6733 for perfectly overlapping calls to memcpy.
6734 (gimple_fold_builtin_memory_chk): Same.
6735 (gimple_fold_builtin_strcpy): Handle no-warning.
6736 (gimple_fold_builtin_stxcpy_chk): Same.
6737 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6739 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6741 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6742 parameter. Use it for SFmode.
6743 (rs6000_function_arg_advance_1): Adjust.
6744 (rs6000_function_arg): Adjust.
6745 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6747 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6749 PR rtl-optimization/84169
6750 PR rtl-optimization/84780
6751 * combine.c (can_combine_p): Check for a 2-insn combination whether
6752 the destination register is used between the two insns, too.
6754 2018-03-12 Richard Biener <rguenther@suse.de>
6756 PR tree-optimization/84803
6757 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6758 for refs DR analysis didn't process.
6760 2018-03-12 Richard Biener <rguenther@suse.de>
6762 PR tree-optimization/84777
6763 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6764 force-vectorize loops ignore whether we are optimizing for size.
6766 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6768 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6769 (TARGET_MD_ASM_ADJUST): Define.
6771 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6772 Kito Cheng <kito.cheng@gmail.com>
6773 Chung-Ju Wu <jasonwucj@gmail.com>
6775 * config/nds32/nds32.c (nds32_compute_stack_frame,
6776 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6777 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6778 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6779 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6780 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6781 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6782 * config/nds32/nds32.md (prologue, epilogue): Use macro
6783 NDS32_V3PUSH_AVAILABLE_P to do checking.
6785 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6788 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6789 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6790 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6791 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6792 addition of most attributes on !orig_type_die or the attribute not
6793 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6795 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6796 Chung-Ju Wu <jasonwucj@gmail.com>
6798 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6800 * config/nds32/nds32.opt (mvh): New option.
6802 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6803 Chung-Ju Wu <jasonwucj@gmail.com>
6805 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6807 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6808 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6811 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6812 Chung-Ju Wu <jasonwucj@gmail.com>
6814 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6816 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6817 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6819 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6820 Kito Cheng <kito.cheng@gmail.com>
6821 Chung-Ju Wu <jasonwucj@gmail.com>
6823 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6824 UNSPEC_FFMISM and UNSPEC_FLMISM.
6825 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6826 for ffb, ffmism and flmism.
6827 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6828 (unspec_ffmism): Ditto.
6829 (unspec_flmism): Ditto.
6830 (nds32_expand_builtin_impl): Check if string extension is available.
6831 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6832 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6834 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6837 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6840 * lra-assigns.c (assign_by_spills): Return a flag of reload
6841 assignment failure. Do not process the reload assignment
6842 failures. Do not spill other reload pseudos if they has the same
6844 (lra_assign): Add a return arg. Set up from the result of
6845 assign_by_spills call.
6846 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6847 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6848 usage_insns if it is not NULL.
6849 (spill_hard_reg_in_range): New function.
6850 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6851 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6852 function prototypes.
6853 (lra_assign): Change prototype.
6854 * lra.c (lra): Add code to deal with fails by splitting hard reg
6857 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6860 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6862 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6865 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6866 (dw_val_node): Add val_symbolic_view.
6867 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6868 (symview_upper_bound): New.
6869 (new_line_info_table): Initialize symviews_since_reset.
6870 (dwarf2out_source_line): Count symviews_since_reset and set
6871 symview_upper_bound.
6872 (dw_val_equal_p): Handle symview.
6873 (add_AT_symview): New.
6874 (print_dw_val): Handle symview.
6875 (attr_checksum, attr_checksum_ordered): Likewise.
6876 (same_dw_val_p, size_of_die): Likewise.
6877 (value_format, output_die): Likewise.
6878 (add_high_low_attributes): Use add_AT_symview for entry_view.
6879 (dwarf2out_finish): Reset symview_upper_bound, clear
6882 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6885 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6886 Add strict argument and use it.
6887 (rs6000_split_multireg_move): Update for new strict argument.
6888 (mem_operand_gpr): Disallow all non-offsettable addresses.
6889 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6891 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6894 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6895 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6896 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6899 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6900 decl, use remap_type if we want to use the type.
6902 2018-03-09 Martin Sebor <msebor@redhat.com>
6904 PR tree-optimization/84526
6905 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6907 (builtin_access::generic_overlap): Be prepared to handle non-array
6910 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6912 PR rtl-optimization/84682
6913 * lra-constraints.c (process_address_1): Check is_address flag
6914 for address constraints.
6915 (process_alt_operands): Likewise.
6916 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6917 preprocess_constraints.
6918 * recog.h (preprocess_constraints): Add oploc parameter.
6920 * recog.c (preprocess_constraints): Test address_operand for
6921 CT_ADDRESS constraints.
6923 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6926 * lra-assigns.c (assign_by_spills): Return a flag of reload
6927 assignment failure. Do not process the reload assignment
6928 failures. Do not spill other reload pseudos if they has the same
6930 (lra_assign): Add a return arg. Set up from the result of
6931 assign_by_spills call.
6932 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6933 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6934 usage_insns if it is not NULL.
6935 (spill_hard_reg_in_range): New function.
6936 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6937 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6938 function prototypes.
6939 (lra_assign): Change prototype.
6940 * lra.c (lra): Add code to deal with fails by splitting hard reg
6943 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6946 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6947 Accept complain bool parameter. Only emit errors if it is true.
6948 (arm_parse_cpu_option_name): Likewise.
6949 (arm_target_thumb_only): Adjust callers of the above.
6950 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6951 prototype to take a default true bool parameter.
6952 (arm_parse_arch_option_name): Likewise.
6954 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6955 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6959 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6960 * configure: Regenerate.
6961 * configure.ac ("linker --version-script option"): New.
6962 ("linker soname option"): New.
6964 2018-03-09 Richard Biener <rguenther@suse.de>
6966 PR tree-optimization/84775
6967 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6968 immediate uses of predicate stmts and mark them modified.
6971 PR tree-optimization/84178
6972 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6974 (version_loop_for_if_conversion): Delay update_ssa call.
6975 (tree_if_conversion): Delay update_ssa until after predicate
6978 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6981 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6982 when the function accesses prior frames.
6984 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6987 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6988 gen_llsym, otherwise call maybe_gen_llsym.
6991 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6992 has ',' character inside of it.
6994 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6997 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6998 as clobbering CC_REGNUM.
7000 2018-03-08 Richard Biener <rguenther@suse.de>
7003 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7004 (follow_copies_to_constant): Do not follow SSA names registered
7007 2018-03-08 Richard Biener <rguenther@suse.de>
7009 PR tree-optimization/84178
7010 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7012 (version_loop_for_if_conversion): Delay update_ssa call.
7013 (tree_if_conversion): Delay update_ssa until after predicate
7016 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7018 PR tree-optimization/84178
7019 * tree-if-conv.c (release_bb_predicate): Remove the
7020 the assertion that the stmts have NULL use_ops.
7021 Discard the statements, asserting that they haven't
7022 yet been added to a BB.
7024 2018-03-08 Richard Biener <rguenther@suse.de>
7026 PR tree-optimization/84746
7027 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7028 (phi_translate): Pass in destination ANTIC_OUT set.
7029 (phi_translate_1): Likewise. For a simplified result lookup
7030 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7031 (phi_translate_set): Adjust.
7032 (do_pre_regular_insertion): Likewise.
7033 (do_pre_partial_partial_insertion): Likewise.
7035 2018-03-08 Martin Liska <mliska@suse.cz>
7037 PR gcov-profile/84735
7038 * doc/gcov.texi: Document usage of profile files.
7039 * gcov-io.h: Document changes in the format.
7041 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7045 * dwarf2out.c (struct dw_line_info_table): Update comments for
7047 (FORCE_RESET_NEXT_VIEW): New.
7048 (FORCE_RESETTING_VIEW_P): New.
7049 (RESETTING_VIEW_P): Check for -1 too.
7050 (ZERO_VIEW_P): Likewise.
7051 (new_line_info_table): Force-reset next view.
7052 (dwarf2out_begin_function): Likewise.
7053 (dwarf2out_source_line): Simplify zero_view_p initialization.
7054 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7055 view directly. Omit view when omitting .loc at line 0.
7057 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7059 PR tree-optimization/84740
7060 * tree-switch-conversion.c (process_switch): Call build_constructors
7061 only if info.phi_count is non-zero.
7063 PR tree-optimization/84739
7064 * tree-tailcall.c (find_tail_calls): Check call arguments against
7065 DECL_ARGUMENTS (current_function_decl) rather than
7066 DECL_ARGUMENTS (func) when checking for tail recursion.
7068 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7070 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7071 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7072 Volker Reichelt's entry and add entries for people that perform
7073 GCC fuzzy testing and report numerous bugs.
7075 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7078 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7079 readonly data in sdata, if that is disabled.
7080 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7081 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7082 -mreadonly-in-sdata option.
7084 2018-03-07 Martin Sebor <msebor@redhat.com>
7086 PR tree-optimization/84468
7087 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7088 basic block when looking for nul assignment.
7090 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7093 * except.h (output_function_exception_table): Adjust prototype.
7094 * except.c (output_function_exception_table): Remove FNNAME parameter
7095 and add SECTION parameter. Ouput one part of the table at a time.
7096 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7097 the first part of the exception table and emit unwind directives.
7098 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7099 (i386_pe_seh_cold_init): Likewise.
7100 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7101 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7102 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7103 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7104 * config/i386/winnt.c: Include except.h.
7105 (struct seh_frame_state): Add reg_offset, after_prologue and
7106 in_cold_section fields.
7107 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7108 (i386_pe_seh_cold_init): New function.
7109 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7110 to seh->in_cold_section.
7111 (seh_emit_push): Record the offset of the push.
7112 (seh_emit_save): Record the offet of the save.
7113 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7114 Test seh->after_prologue to disregard the epilogue.
7115 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7116 (i386_pe_end_cold_function): New function.
7118 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7121 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7122 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7125 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7127 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7128 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7131 * multiple_target.c: Include tree-inline.h and intl.h.
7132 (expand_target_clones): Diagnose and fail if node->definition and
7133 !tree_versionable_function_p (node->decl).
7135 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7137 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7139 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7140 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7141 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7143 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7146 * combine.c (try_combine): Use reg_or_subregno instead of handling
7147 just paradoxical SUBREGs and REGs.
7149 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7151 * config/arc/arc.c (arc_finalize_pic): Remove function.
7152 (arc_must_save_register): We use single base PIC register, remove
7153 checks to save/restore the PIC register.
7154 (arc_expand_prologue): Likewise.
7155 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7157 (arc_verify_short): Likewise.
7158 (arc_attr_type): Likewise.
7159 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7160 (walk_stores): Likewise.
7161 (arc_address_cost): Make it static.
7162 (arc_verify_short): Likewise.
7163 (branch_dest): Likewise.
7164 (arc_attr_type): Likewise.
7165 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7166 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7167 (arc_final_prescan_insn): Remove inserting the nops due to
7168 hardware hazards. It is done in reorg step.
7169 (insn_length_variant_t): Remove.
7170 (insn_length_parameters_t): Likewise.
7171 (arc_insn_length_parameters): Likewise.
7172 (arc_get_insn_variants): Likewise.
7173 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7175 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7178 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7181 PR tree-optimization/84687
7182 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7184 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7186 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7188 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7189 Rename to ppc_speculation_barrier.
7190 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7191 __builtin_ppc_speculation_barrier.
7193 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7196 * combine.c (combine_simplify_rtx): Don't try to simplify if
7197 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7200 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7202 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7203 to 32 bytes when compiling for POWER9.
7205 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7208 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7209 regparm >= 3 with no arg reg available also for calls with
7210 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7213 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7215 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7217 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7220 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7222 2018-03-05 Richard Biener <rguenther@suse.de>
7224 PR tree-optimization/84486
7225 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7226 When inserting a __builtin_assume_aligned call set the LHS
7227 SSA name alignment info accordingly.
7229 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7231 PR tree-optimization/84114
7232 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7233 Avoid reassociation of FLOAT_MODE addition.
7235 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7237 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7238 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7239 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7240 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7241 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7242 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7243 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7245 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7246 __builtin_ia32_wbinvd): New builtins.
7247 (SPECIAL_ARGS2): New.
7248 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7249 (SPECIAL_ARGS2): New.
7250 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7251 (ix86_valid_target_attribute_inner_p): Ditto.
7252 (ix86_init_mmx_sse_builtins): Add special_args2.
7253 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7254 TARGET_WBNOINVD_P): New.
7255 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7256 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7257 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7258 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7259 * config/i386/pconfigintrin.h: New file.
7260 * config/i386/wbnoinvdintrin.h: Ditto.
7261 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7263 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7265 2018-03-05 Richard Biener <rguenther@suse.de>
7267 PR tree-optimization/84670
7268 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7270 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7271 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7272 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7273 doesn't grow if all successors (recursively) were visited at least
7276 2018-03-05 Richard Biener <rguenther@suse.de>
7278 PR tree-optimization/84650
7279 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7280 if executed in the loop pipeline.
7282 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7284 * doc/configfiles.texi (Configuration Files): Move info about
7285 conditionalizing $target-protos.h to...
7286 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7287 differs from $target-protos.h.
7289 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7290 Chung-Ju Wu <jasonwucj@gmail.com>
7292 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7293 * config/nds32/nds32-multiple.md (setmemsi): Define.
7294 * config/nds32/nds32-memory-manipulation.c
7295 (nds32_gen_dup_4_byte_to_word_value): New.
7296 (emit_setmem_word_loop): New.
7297 (emit_setmem_byte_loop): New.
7298 (nds32_expand_setmem_loop): New.
7299 (nds32_expand_setmem_loop_v3m): New.
7300 (nds32_expand_setmem_unroll): New.
7301 (nds32_expand_setmem): New.
7303 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7304 Chung-Ju Wu <jasonwucj@gmail.com>
7306 * config/nds32/nds32-memory-manipulation.c
7307 (nds32_emit_load_store): New.
7308 (nds32_emit_post_inc_load_store): New.
7309 (nds32_emit_mem_move): New.
7310 (nds32_emit_mem_move_block): New.
7311 (nds32_expand_movmemsi_loop_unknown_size): New.
7312 (nds32_expand_movmemsi_loop_known_size): New.
7313 (nds32_expand_movmemsi_loop): New.
7314 (nds32_expand_movmemsi_unroll): New.
7315 (nds32_expand_movmemqi): Rename ...
7316 (nds32_expand_movmemsi): ... to this.
7317 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7318 (movmemsi): ... to this.
7319 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7320 (nds32_expand_movmemsi): ... to this.
7322 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7323 Monk Chiang <sh.chiang04@gmail.com>
7324 Chung-Ju Wu <jasonwucj@gmail.com>
7326 * config/nds32/nds32-protos.h
7327 (nds32_expand_load_multiple): New arguments.
7328 (nds32_expand_store_multiple): Ditto.
7329 (nds32_valid_multiple_load_store): Rename ...
7330 (nds32_valid_multiple_load_store_p): ... to this.
7331 * config/nds32/nds32-memory-manipulation.c
7332 (nds32_expand_load_multiple): Refine implementation.
7333 (nds32_expand_store_multiple): Ditto.
7334 * config/nds32/nds32-multiple.md
7335 (load_multiple): Update nds32_expand_load_multiple interface.
7336 (store_multiple): Update nds32_expand_store_multiple interface.
7337 * config/nds32/nds32-predicates.c
7338 (nds32_valid_multiple_load_store): Rename ...
7339 (nds32_valid_multiple_load_store_p): ... to this and refine
7341 * config/nds32/predicates.md
7342 (nds32_load_multiple_and_update_address_operation): New predicate.
7343 (nds32_store_multiple_and_update_address_operation): New predicate.
7345 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7346 Chung-Ju Wu <jasonwucj@gmail.com>
7348 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7349 (combo): New attribute.
7350 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7352 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7354 * config/nds32/nds32.opt: Change -mcmodel= default value.
7356 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7357 Monk Chiang <sh.chiang04@gmail.com>
7358 Chung-Ju Wu <jasonwucj@gmail.com>
7360 * config/nds32/constants.md (unspec_element): New enum.
7361 * config/nds32/constraints.md (Umw): New constraint.
7362 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7363 * config/nds32/nds32-intrinsic.md: Likewise.
7364 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7365 (nds32_valid_smw_lwm_base_p): New.
7366 (nds32_output_smw_single_word): New.
7367 (nds32_output_lmw_single_word): New.
7368 (nds32_expand_unaligned_load): New.
7369 (nds32_expand_unaligned_store): New.
7370 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7371 (nds32_output_smw_single_word): Declare.
7372 (nds32_output_lmw_single_word): Declare.
7373 (nds32_expand_unaligned_load): Declare.
7374 (nds32_expand_unaligned_store): Declare.
7375 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7376 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7377 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7378 NDS32_BUILTIN_UASTORE_DW.
7379 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7382 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7383 Kito Cheng <kito.cheng@gmail.com>
7384 Chung-Ju Wu <jasonwucj@gmail.com>
7386 * config/nds32/nds32-intrinsic.c
7387 (nds32_expand_builtin_null_ftype_reg): Delete.
7388 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7389 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7390 (nds32_read_argument): New.
7391 (nds32_legitimize_target): Ditto.
7392 (nds32_legitimize_argument): Ditto.
7393 (nds32_check_constant_argument): Ditto.
7394 (nds32_expand_unop_builtin): Ditto.
7395 (nds32_expand_unopimm_builtin): Ditto.
7396 (nds32_expand_binop_builtin): Ditto.
7397 (nds32_builtin_decl_impl): Ditto.
7398 (builtin_description): Ditto.
7399 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7400 (nds32_init_builtins_impl): Ditto.
7401 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7402 (nds32_builtin_decl): New.
7403 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7404 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7406 2018-03-02 Jeff Law <law@redhat.com>
7408 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7409 (redundant_insn, fill_simple_delay_slots): Likewise.
7410 (fill_slots_from_thread): Likewise.
7411 * resource.c (mark_referenced_resources): Likewise.
7412 (mark_set_resources, find_dead_or_set_registers): Likewise.
7414 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7416 * substring-locations.h (format_warning_va): Formatting fix for
7418 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7420 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7421 * substring-locations.c: Include intl.h.
7422 (format_warning_va): Turned into small wrapper around
7423 format_warning_n_va, renamed to ...
7424 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7425 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7427 (format_warning_at_substring_n): New function.
7428 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7429 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7430 format_warning_at_substring with just a shorter name instead of
7431 const function pointer.
7432 (fmtwarn_n): New function.
7433 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7434 appropriate, get rid of all the fmtstr temporaries, move conditionals
7435 with G_() wrapped string literals directly into fmtwarn arguments,
7436 cast dir.len to (int), formatting fixes.
7438 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7440 * doc/invoke.texi: Remove "Cilk Plus" references.
7442 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7443 Richard Biener <rguenther@suse.de>
7446 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7447 for error or warning attributes if CALL_FROM_THUNK_P is set.
7450 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7453 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7454 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7457 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7458 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7460 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7461 instead of -1U in last predictors element's probability member.
7463 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7466 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7467 arguments if they are comparable.
7469 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7471 PR tree-optimization/84634
7472 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7473 masks and masked_loop_p with a single loop_masks, making sure it's
7474 null for bb vectorization.
7476 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7478 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7479 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7480 loop->force_vectorize to check whether there is no alias.
7482 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7485 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7487 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7489 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7490 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7491 instead of a loop around prev_real_insn.
7492 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7496 * config/i386/i386.c (ix86_print_operand): Use conditional
7497 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7500 2018-03-02 Richard Biener <rguenther@suse.de>
7502 PR tree-optimization/84427
7503 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7504 (bitmap_set_subtract_values): Rewrite to handle multiple
7507 (prune_clobbered_mems): Likewise.
7508 (phi_translate): Take edge instead of pred/phiblock.
7509 (phi_translate_1): Likewise.
7510 (phi_translate_set): Likewise. Insert all translated
7511 exprs for a value into the set, keeping possibly multiple
7512 expressions per value.
7513 (compute_antic_aux): Adjust for phi_translate changes.
7514 When intersecting union the expressions and prune those
7515 not in the final value set, keeping possibly multiple
7516 expressions per value. Do not use value-insertion
7517 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7518 all expressions. Add verification that the value-sets
7519 only shrink during iteration.
7520 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7521 (do_pre_regular_insertion): Likewise.
7522 (do_pre_partial_partial_insertion): Likewise.
7524 2018-03-02 Richard Biener <rguenther@suse.de>
7527 * config/darwin.c (saved_debug_info_level): New static global.
7528 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7529 (darwin_asm_lto_end): Restore debug info generation settings.
7531 2018-03-01 Martin Liska <mliska@suse.cz>
7534 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7537 2018-03-01 Richard Biener <rguenther@suse.de>
7540 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7541 type annotation with LTO when debug was disabled at compile-time.
7543 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7545 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7547 (mips_final_postscan_insn): Likewise.
7549 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7551 PR rtl-optimization/84528
7552 * alias.c (init_alias_target): Add commentary.
7553 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7554 a unique base value if the frame pointer is not eliminated
7555 to the stack pointer.
7557 2018-03-01 Tom de Vries <tom@codesourcery.com>
7559 PR rtl-optimization/83327
7560 * lra-int.h (hard_regs_spilled_into): Declare.
7561 * lra.c (hard_regs_spilled_into): Define.
7562 (init_reg_info): Init hard_regs_spilled_into.
7563 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7564 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7565 (process_bb_lives): Handle hard_regs_spilled_into.
7566 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7567 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7569 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7571 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7572 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7573 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7574 * config/rs6000/aix72.h: New file.
7576 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7578 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7579 instead of warning_at with conditional singular and plural messages
7583 * stor-layout.c (update_alignment_for_field): For
7584 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7585 && !DECL_PACKED (field), do the alignment update, just use
7586 only desired_align instead of MAX (type_align, desired_align)
7588 (place_field): Don't do known_align < desired_align handling
7589 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7590 is non-NULL, instead do it after rli->prev_field handling and
7591 only if not within a bitfield word. For DECL_PACKED (field)
7592 use type_align of BITS_PER_UNIT.
7594 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7596 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7597 superfluous parentheses and trailing spaces.
7599 2018-02-28 Richard Biener <rguenther@suse.de>
7601 PR tree-optimization/84584
7602 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7603 SCoPs with fake exit edge.
7605 2018-02-28 Martin Liska <mliska@suse.cz>
7608 * timevar.c (timer::print): Fix format to properly print 100%
7611 2018-02-28 Richard Biener <rguenther@suse.de>
7614 * genmatch.c (capture_info::walk_match): Do not mark
7615 captured expressions without operands as expr_p given
7616 they act more like predicates and should be subject to
7617 "lost tail" side-effect preserving.
7619 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7621 PR rtl-optimization/81611
7622 * auto-inc-dec.c (attempt_change): Move dead note from
7623 mem_insn if it's the next use of regno
7624 (find_address): Take address use of reg holding
7625 non-incremented value. Add parm to limit search to the named
7627 (merge_in_block): Attempt to use a mem insn that is the next
7628 use of the original regno.
7630 2018-02-27 Martin Sebor <msebor@redhat.com>
7633 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7634 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7636 2018-02-27 Martin Sebor <msebor@redhat.com>
7638 PR translation/84207
7639 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7640 n argument to unsigned HOST_WIDE_INT.
7641 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7642 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7643 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7644 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7646 2018-02-27 Richard Biener <rguenther@suse.de>
7648 PR tree-optimization/84512
7649 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7650 Do not use the estimate returned from record_stmt_cost for
7651 the scalar iteration cost but sum properly using add_stmt_cost.
7653 2018-02-27 Richard Biener <rguenther@suse.de>
7655 PR tree-optimization/84466
7656 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7657 Adjust last change to less strictly validate use operands.
7659 2018-02-27 Martin Liska <mliska@suse.cz>
7661 PR gcov-profile/84548
7662 * gcov.c (process_file): Allow partial overlap and consider it
7663 also as group functions.
7664 (output_lines): Properly calculate range of lines for a group.
7666 2018-02-27 Martin Liska <mliska@suse.cz>
7668 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7669 'ggc' suffixes. Change first column width.
7670 (timer::print): Fix formatting of the column.
7672 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7674 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7675 preserve inline entry blocks for the sake of debug inline
7676 entry point markers alone.
7677 (remove_unused_locals): Suggest in comments a better place to
7678 force the preservation of inline entry blocks that are
7679 otherwise unused, but do not preserve them.
7681 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7683 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7685 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7688 * config/i386/constraints.md (Bs): Replace
7689 ix86_indirect_branch_register with
7690 TARGET_INDIRECT_BRANCH_REGISTER.
7692 * config/i386/i386.md (indirect_jump): Likewise.
7693 (tablejump): Likewise.
7694 (*sibcall_memory): Likewise.
7695 (*sibcall_value_memory): Likewise.
7696 Peepholes of indirect call and jump via memory: Likewise.
7697 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7698 (*sibcall_value_GOT_32): Likewise.
7699 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7700 (GOT_memory_operand): Likewise.
7701 (call_insn_operand): Likewise.
7702 (sibcall_insn_operand): Likewise.
7703 (GOT32_symbol_operand): Likewise.
7704 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7706 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7708 PR rtl-optimization/83496
7709 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7710 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7711 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7712 redundant insn, if any.
7713 (relax_delay_slots): Likewise.
7714 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7716 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7718 PR tree-optimization/83965
7719 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7720 that grouped statements are part of a reduction chain. Return
7721 true if the statement is not marked as a reduction itself but
7723 (vect_recog_dot_prod_pattern): Don't check whether the statement
7724 is part of a group here.
7725 (vect_recog_sad_pattern): Likewise.
7726 (vect_recog_widen_sum_pattern): Likewise.
7728 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7731 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7733 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7736 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7738 (ix86_output_indirect_function_return): New prototype.
7739 (ix86_split_simple_return_pop_internal): Likewise.
7740 * config/i386/i386.c (indirect_return_via_cx): New.
7741 (indirect_return_via_cx_bnd): Likewise.
7742 (indirect_thunk_name): Handle return va CX_REG.
7743 (output_indirect_thunk_function): Create alias for
7744 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7745 (ix86_output_indirect_jmp): Remove the bool argument.
7746 (ix86_output_indirect_function_return): New function.
7747 (ix86_split_simple_return_pop_internal): Likewise.
7748 * config/i386/i386.md (*indirect_jump): Don't pass false
7749 to ix86_output_indirect_jmp.
7750 (*tablejump_1): Likewise.
7751 (simple_return_pop_internal): Change it to define_insn_and_split.
7752 Call ix86_split_simple_return_pop_internal to split it for
7754 (simple_return_indirect_internal): Call
7755 ix86_output_indirect_function_return instead of
7756 ix86_output_indirect_jmp.
7758 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7761 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7762 memset and value initialization afterwards.
7764 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7766 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7768 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7771 * common/config/aarch64/aarch64-common.c
7772 (aarch_option_optimization_table[]): Switch
7773 off fomit-frame-pointer
7775 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7776 Chung-Ju Wu <jasonwucj@gmail.com>
7778 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7780 (store_multiple): Ditto.
7782 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7784 * config.gcc: Add --with-cpu support for nds32 target.
7785 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7786 * config/nds32/nds32.opt: Add -mcpu= option.
7788 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7790 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7791 isel=yes): Warn for these deprecated options.
7793 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7795 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7796 ISA_2_5_MASKS_EMBEDDED.
7798 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7800 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7801 p->max as pointers rather than using iterative_hash_expr.
7803 2018-02-23 Carl Love <cel@us.ibm.com>
7805 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7806 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7808 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7809 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7810 P8V_BUILTIN_VEC_VUNSIGNED2.
7812 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7815 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7816 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7818 * lra-constraints.c (curr_insn_transform): Set up
7819 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7821 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7822 * lra-eliminations.c (spill_pseudos): Ditto.
7823 (process_insn_for_elimination): Ditto.
7824 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7825 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7826 LRA_NON_CLOBBERED_ALT.
7828 2018-02-22 Martin Sebor <msebor@redhat.com>
7830 PR tree-optimization/84480
7831 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7832 to maybe_diag_stxncpy_trunc. Call it.
7833 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7834 from gimple_fold_builtin_strcpy. Print inlining stack.
7835 (handle_builtin_stxncpy): Print inlining stack.
7836 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7838 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7841 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7842 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7843 and -fcheck-pointer-bounds are used together.
7844 (indirect_thunk_prefix): New enum.
7845 (indirect_thunk_need_prefix): New function.
7846 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7847 "_nt" instead of "_bnd" for NOTRACK prefix.
7848 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7849 (output_indirect_thunk_function): Likewise.
7851 (ix86_code_end): Update output_indirect_thunk_function calls.
7852 (ix86_output_indirect_branch_via_reg): Replace
7853 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7854 (ix86_output_indirect_branch_via_push): Likewise.
7855 (ix86_output_function_return): Likewise.
7856 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7857 incompatible with -fcf-protection=branch and
7858 -fcheck-pointer-bounds.
7860 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7863 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7864 Change gcc_assert call to output_operand_lossage.
7866 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7868 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7870 2018-02-22 DJ Delorie <dj@redhat.com>
7871 Sebastian Perta <sebastian.perta@renesas.com>
7872 Oleg Endo <olegendo@gcc.gnu.org>
7874 * config/rx/rx.c (rx_rtx_costs): New function.
7875 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7877 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7879 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7881 2018-02-22 Martin Liska <mliska@suse.cz>
7884 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7885 Add "native" as a possible value.
7887 2018-02-22 Martin Liska <mliska@suse.cz>
7890 * config/i386/i386.c (ix86_option_override_internal):
7891 Add "native" as a possible value for -march and -mtune.
7893 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7896 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7897 to all type variants.
7899 PR tree-optimization/84503
7900 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7901 width as info->bitpos + info->bitsize - start.
7902 (merged_store_group::merge_overlapping): Simplify width computation.
7903 (check_no_overlap): New function.
7904 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7905 start + width and last_order of the group, fail if check_no_overlap
7907 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7908 to group if check_no_overlap fails.
7910 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7912 * config/rs6000/altivec.md: Delete contraint arguments to
7913 define_expand, define_split, and define_peephole2, and in
7914 define_insn_and_split if always unused.
7915 * config/rs6000/darwin.md: Ditto.
7916 * config/rs6000/dfp.md: Ditto.
7917 * config/rs6000/rs6000.md: Ditto.
7918 * config/rs6000/sync.md: Ditto.
7919 * config/rs6000/vector.md: Ditto.
7920 * config/rs6000/vsx.md: Ditto.
7922 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7924 * config/rs6000/altivec.md: Write output control strings as braced
7925 blocks instead of double-quoted strings.
7926 * config/rs6000/darwin.md: Ditto.
7927 * config/rs6000/rs6000.md: Ditto.
7928 * config/rs6000/vector.md: Ditto.
7929 * config/rs6000/vsx.md: Ditto.
7931 2018-02-21 Jason Merrill <jason@redhat.com>
7933 PR c++/84314 - ICE with templates and fastcall attribute.
7934 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7936 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7938 * ipa-cp.c (determine_versionability): Fix comment typos.
7940 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7943 * ipa-cp.c (determine_versionability): Do not version functions caling
7946 2018-02-21 Martin Liska <mliska@suse.cz>
7949 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7950 Add "native" as a possible value.
7951 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7952 the macro when native cpu detection is available.
7954 2018-02-21 Martin Liska <mliska@suse.cz>
7957 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7958 Add "native" as a possible value.
7959 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7960 when native cpu detection is available.
7962 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7963 Martin Sebor <msebor@redhat.com>
7965 PR tree-optimization/84478
7966 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7968 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7969 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7970 support which is conservatively correct, for 2 only stay conservative
7971 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7972 argument to the 2 argument get_range_strlen, adjust 6 arg
7973 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7975 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7976 (gimple_fold_builtin_strlen): Pass true as last argument to
7979 2018-02-20 Martin Sebor <msebor@redhat.com>
7982 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7983 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7984 (builtin_memref::builtin_memref): Factor out parts into
7985 set_base_and_offset and call it.
7987 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7990 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7991 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7992 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7993 search at the associated MODE_INT.
7995 2018-02-20 Jeff Law <law@redhat.com>
7998 PR tree-optimization/81592
8000 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8001 for range data rather than using global data.
8002 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8003 range data rather than using global data.
8004 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8005 pass it to children as needed.
8006 (struct directive::fmtresult): Similarly.
8007 (struct directive::set_width): Similarly.
8008 (struct directive::set_precision): Similarly.
8009 (format_integer, format_directive, parse_directive): Similarly.
8010 (format_none): Accept unnamed vr_values parameter.
8011 (format_percent, format_floating, format_character): Similarly.
8012 (format_string, format_plain): Similarly.
8013 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8014 the EVRP range analyzer for range data rather than using global data.
8015 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8016 gimple-ssa-evrp-analyze.h
8017 (class sprintf_dom_walker): Add after_dom_children member function.
8018 Add evrp_range_analyzer member.
8019 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8020 range analyzer as needed.
8021 (sprintf_dom_walker::after_dom_children): New member function.
8022 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8024 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8025 (evrp_range_analyzer::pop_to_marker): Likewise.
8027 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8029 PR tree-optimization/84419
8030 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8031 with the required type if its current type is compatible but
8034 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8037 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8038 after vectorization.
8040 2018-02-20 Martin Liska <mliska@suse.cz>
8043 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8044 possible values if we don't have a hint.
8046 2018-02-20 Martin Liska <mliska@suse.cz>
8050 * final.c (shorten_branches): Build align_tab array with one
8052 * opts.c (finish_options): Add alignment option limit check.
8053 (MAX_CODE_ALIGN): Likewise.
8054 (MAX_CODE_ALIGN_VALUE): Likewise.
8055 * doc/invoke.texi: Document maximum allowed option value for
8056 all -falign-* options.
8058 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8061 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8062 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8063 * var-tracking.c (emit_note_insn_var_location): Remove all references
8064 to NOTE_INSN_CALL_ARG_LOCATION.
8065 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8066 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8067 Use copy_rtx_if_shared.
8068 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8069 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8070 (dwarf2out_var_location): Remove handling of
8071 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8073 * final.c (final_scan_insn): Remove all references to
8074 NOTE_INSN_CALL_ARG_LOCATION.
8075 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8076 before dumping final insns.
8077 * except.c (emit_note_eh_region_end): Remove all references to
8078 NOTE_INSN_CALL_ARG_LOCATION.
8079 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8080 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8081 * config/arc/arc.c (hwloop_optimize): Likewise.
8082 * config/arm/arm.c (create_fix_barrier): Likewise.
8083 * config/s390/s390.c (s390_chunkify_start): Likewise.
8084 * config/sh/sh.c (find_barrier): Likewise.
8085 * config/i386/i386.c (rest_of_insert_endbranch,
8086 ix86_seh_fixup_eh_fallthru): Likewise.
8087 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8088 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8089 * config/frv/frv.c (frv_function_prologue): Likewise.
8090 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8092 (note_outside_basic_block_p): Remove all references to
8093 NOTE_INSN_CALL_ARG_LOCATION.
8094 * gengtype.c (adjust_field_rtx_def): Likewise.
8095 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8097 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8098 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8101 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8104 PR tree-optimization/84452
8105 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8106 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8109 2018-02-19 Martin Liska <mliska@suse.cz>
8112 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8114 2018-02-19 Martin Liska <mliska@suse.cz>
8115 Richard Sandiford <richard.sandiford@linaro.org>
8117 PR tree-optimization/82491
8118 * gimple-fold.c (get_base_constructor): Make earlier bail out
8121 2018-02-19 Carl Love <cel@us.ibm.com>
8123 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8124 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8126 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8127 P8V_BUILTIN_VEC_NEG.
8129 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8131 * config/rl78/rl78.md (movdf): New define expand.
8133 2018-02-19 Martin Liska <mliska@suse.cz>
8136 * doc/invoke.texi: Fix typo.
8137 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8139 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8141 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8142 handle rs6000_single_float and rs6000_double_float specially for
8145 2018-02-16 Jeff Law <law@redhat.com>
8147 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8150 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8153 * ipa-inline.c (inline_small_functions): Fix a typo.
8155 2018-02-16 Nathan Sidwell <nathan@acm.org>
8157 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8159 2018-02-16 Carl Love <cel@us.ibm.com>
8161 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8162 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8163 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8164 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8165 expansion to P8V_BUILTIN_VEC_FLOAT2.
8167 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8169 PR rtl-optimization/70023
8170 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8171 src_regno into account.
8173 2018-02-16 Carl Love <cel@us.ibm.com>
8175 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8176 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8177 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8178 * config/rs6000/rs6000.c: Remove case statements for
8179 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8180 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8181 and P9V_BUILTIN_VEC_VINSERT4B.
8182 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8183 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8184 * config/rs6000/vsx.md:
8185 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8188 2018-02-16 Carl Love <cel@us.ibm.com>
8190 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8192 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8194 * config/rs6000/rs6000-c.c: Add the definitions for
8195 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8196 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8197 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8198 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8199 definition for insert4b and define insn *insert3b_internal.
8200 * doc/extend.texi: Add documentation for vec_extract4b.
8202 2018-02-16 Nathan Sidwell <nathan@acm.org>
8204 * doc/extend.texi (Backwards Compatibility): Mention friend
8205 injection. Note for-scope is deprecated.
8206 * doc/invoke.texi (-ffriend-injection): Deprecate.
8208 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8210 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8211 that moved to I2, also allow destinations that are a paradoxical
8212 subreg (instead of a normal reg).
8214 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8217 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8220 2018-02-16 Richard Biener <rguenther@suse.de>
8222 PR tree-optimization/84037
8223 PR tree-optimization/84016
8225 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8226 Adjust vec_construct for the fact we need additional higher latency
8227 128bit inserts for AVX256 and AVX512 vector builds.
8228 (ix86_add_stmt_cost): Scale vector construction cost for
8231 2018-02-16 Richard Biener <rguenther@suse.de>
8233 PR tree-optimization/84417
8234 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8235 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8236 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8238 2018-02-16 Martin Liska <mliska@suse.cz>
8241 * internal-fn.def (ASAN_CHECK): Set proper flags.
8242 (ASAN_MARK): Likewise.
8244 2018-02-16 Julia Koval <julia.koval@intel.com>
8246 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8247 from PTA_CANNONLAKE.
8249 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8252 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8253 Use ++iter rather than iter++ for std::list iterators.
8254 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8255 defer deleting them until all nodes in the forest are processed. Do
8256 free even leaf nodes. Change to_process into auto_vec.
8259 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8260 * vec.h (vec_default_construct): Use memset instead of placement new
8261 if BROKEN_VALUE_INITIALIZATION is defined.
8262 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8263 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8266 PR rtl-optimization/83723
8267 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8268 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8269 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8271 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8273 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8275 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8277 PR rtl-optimization/81443
8278 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8279 from inner REGs to paradoxical SUBREGs.
8281 2018-02-16 Richard Biener <rguenther@suse.de>
8283 PR tree-optimization/84399
8284 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8285 For operands we can analyze at their definition make sure we can
8286 analyze them at each use as well.
8288 2018-02-16 Richard Biener <rguenther@suse.de>
8290 PR tree-optimization/84190
8291 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8292 volatile accesses if the decl isn't volatile.
8294 2018-02-15 Jason Merrill <jason@redhat.com>
8296 PR c++/84314 - ICE with templates and fastcall attribute.
8297 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8298 TYPE_CANONICAL on an existing type.
8300 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8302 PR tree-optimization/84383
8303 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8304 dstoff nor call operand_equal_p if dstbase is NULL.
8306 PR tree-optimization/84334
8307 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8308 also a CONSTANT_CLASS_P, punt.
8310 2018-02-14 Jim Wilson <jimw@sifive.com>
8312 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8313 first SMALL_OPERAND check. New local min_second_step. Move assert
8314 to where locals are set. Add TARGET_RVC support.
8315 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8317 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8319 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8321 2018-02-14 Martin Sebor <msebor@redhat.com>
8323 PR tree-optimization/83698
8324 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8325 arrays constrain the offset range to their bounds.
8326 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8327 (builtin_access::overlap): Avoid setting the size of overlap if it's
8329 (maybe_diag_overlap): Also consider arrays when deciding what values
8330 of offsets to include in diagnostics.
8332 2018-02-14 Martin Sebor <msebor@redhat.com>
8335 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8336 that correspond to the kind of a declaration.
8338 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8341 * config/pa/pa.md: Load address of PIC label using the linkage table
8342 if the label is nonlocal.
8344 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8346 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8347 warning message if user requests -maltivec=be.
8348 * doc/invoke.texi: Document deprecation of -maltivec=be.
8350 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8353 * config/rs6000/rs6000-c.c: Update definitions for
8354 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8355 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8357 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8360 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8361 add _get_ssp intrinsics. Remove argument from
8362 __builtin_ia32_rdssp[d|q].
8363 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8364 * config/i386/i386-builtin.def: Remove argument from
8365 __builtin_ia32_rdssp[d|q].
8366 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8367 ix86_expand_special_args_builtin for _rdssp[d|q].
8368 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8369 Clear register before usage.
8370 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8371 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8373 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8375 PR tree-optimization/84357
8376 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8377 operand 1 of an ARRAY_REF too.
8379 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8382 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8383 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8385 (set_of_reg): New struct.
8386 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8387 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8388 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8390 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8391 Split into bitclr, bitset, bitinvert patterns if appropriate.
8392 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8393 use rx_fuse_in_memory_bitop.
8394 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8395 to named insn, correct maximum insn length.
8397 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8400 * machmode.def: Define a complex mode for PARTIAL_INT.
8401 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8403 * doc/rtl.texi: Document CSPImode.
8404 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8406 (msp430_hard_regno_nregs_with_padding): Likewise.
8408 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8411 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8413 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8415 PR rtl-optimization/84169
8416 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8417 we generated a parallel as new i3 and we split that to new i2 and i3
8418 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8419 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8420 those to i2, not i1. Partially rewrite this scan code.
8422 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8425 * stor-layout.c (place_field): For variable length fields, adjust
8426 offset_align afterwards not just based on the field's alignment,
8427 but also on the size.
8430 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8431 of exps and logs in the use_exp2 case.
8433 2018-02-13 Jeff Law <law@redhat.com>
8435 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8438 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8441 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8445 * common.opt (gas-loc-support, gas-locview-support): New.
8446 (ginline-points, ginternal-reset-location-views): New.
8447 * doc/invoke.texi: Document them. Use @itemx where intended.
8448 (gvariable-location-views): Adjust.
8449 * target.def (reset_location_view): New.
8450 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8451 (TARGET_RESET_LOCATION_VIEW): New.
8452 * doc/tm.texi: Rebuilt.
8453 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8454 (dwarf2out_default_as_locview_support): New.
8455 (output_asm_line_debug_info): Use option variables.
8456 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8457 (output_loc_list): Likewise.
8458 (add_high_low_attributes): Check option variables.
8459 Don't output entry view attribute in strict mode.
8460 (gen_inlined_subroutine_die): Check option variables.
8461 (dwarf2out_inline_entry): Likewise.
8462 (init_sections_and_labels): Likewise.
8463 (dwarf2out_early_finish): Likewise.
8464 (maybe_reset_location_view): New, from...
8465 (dwarf2out_var_location): ... here. Call it.
8466 * debug.h (dwarf2out_default_as_loc_support): Declare.
8467 (dwarf2out_default_as_locview_support): Declare.
8468 * hooks.c (hook_int_rtx_insn_0): New.
8469 * hooks.h (hook_int_rtx_insn_0): Declare.
8470 * toplev.c (process_options): Take -gas-loc-support and
8471 -gas-locview-support from dwarf2out. Enable
8472 -gvariable-location-views by default only with locview
8473 assembler support. Enable -ginternal-reset-location-views by
8474 default only if the target defines the corresponding hook.
8475 Enable -ginline-points by default if location views are
8476 enabled; force it disabled if statement frontiers are
8478 * tree-inline.c (expand_call_inline): Check option variables.
8479 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8481 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8483 PR tree-optimization/84321
8484 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8485 handling. Also check whether the anti-range contains any values
8486 that satisfy the mask; switch to a VR_RANGE if not.
8488 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8491 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8493 2018-02-13 Martin Jambor <mjambor@suse.cz>
8496 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8497 of call statements, also set location of a load to a temporary.
8499 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8501 * config/rl78/rl78.c (add_vector_labels): New function.
8502 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8503 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8504 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8505 which checks that no arguments are passed.
8506 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8507 * doc/extend.texi: Documentation for the new attribute.
8509 2018-02-13 Andreas Schwab <schwab@suse.de>
8511 * config/riscv/linux.h (CPP_SPEC): Define.
8513 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8516 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8517 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8518 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8519 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8520 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8521 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8522 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8523 temporarily for AES and PCLMUL builtins.
8525 PR tree-optimization/84339
8526 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8527 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8531 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8532 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8533 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8535 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8537 * omp-simd-clone.h: New file.
8538 * omp-simd-clone.c: Include omp-simd-clone.h.
8539 (expand_simd_clones): No longer static.
8540 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8541 cgraph.h and omp-simd-clone.h.
8542 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8543 (vect_recog_widen_shift_pattern): Formatting fix.
8544 (vect_pattern_recog_1): Don't check optab for calls.
8547 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8548 operands[2] into a REG before using gen_lowpart on it.
8550 2018-02-12 Jeff Law <law@redhat.com>
8553 * config/sh/sh.c (find_barrier): Consider a sibling call
8556 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8557 successfully back substituting a reg.
8559 2018-02-12 Richard Biener <rguenther@suse.de>
8561 PR tree-optimization/84037
8562 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8563 parameter, move visited init to caller.
8564 (vect_slp_analyze_operations): Separate cost from validity
8565 check, initialize visited once for all instances.
8566 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8568 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8570 (vect_model_promotion_demotion_cost): Likewise.
8571 (vectorizable_bswap): Guard cost modeling with !slp_node
8572 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8574 (vectorizable_call): Likewise.
8575 (vectorizable_conversion): Likewise.
8576 (vectorizable_assignment): Likewise.
8577 (vectorizable_shift): Likewise.
8578 (vectorizable_operation): Likewise.
8579 (vectorizable_store): Likewise.
8580 (vectorizable_load): Likewise.
8581 (vectorizable_condition): Likewise.
8582 (vectorizable_comparison): Likewise.
8584 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8587 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8588 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8589 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8591 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8594 * params.def (inline-min-speedup): Increase from 8 to 15.
8595 (max-inline-insns-auto): Decrease from 40 to 30.
8596 * ipa-split.c (consider_split): Add some buffer for function to
8597 be considered inlining candidate.
8598 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8601 2018-02-12 Richard Biener <rguenther@suse.de>
8603 PR tree-optimization/84037
8604 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8605 matched stmts if we cannot swap the non-matched ones.
8607 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8609 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8610 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8611 _mm_maskz_scalef_round_ss): New intrinsics.
8612 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8613 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8614 __builtin_ia32_scalefss_round): Remove.
8615 (__builtin_ia32_scalefsd_mask_round,
8616 __builtin_ia32_scalefss_mask_round): New intrinsics.
8617 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8618 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8619 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8620 "<round_constraint>")): Changed to ...
8621 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8622 "<round_scalar_constraint>")): ... this.
8623 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8624 %0, %1, %2<round_op3>}"): Changed to ...
8625 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8626 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8627 %2<round_scalar_mask_op3>}"): ... this.
8628 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8630 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8632 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8633 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8634 (_mm_maskz_sqrt_round_ss): New intrinsics.
8635 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8636 (__builtin_ia32_sqrtsd_mask_round)
8637 (__builtin_ia32_sqrtss_mask_round): New builtins.
8638 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8639 (__builtin_ia32_sqrtss_round): Remove.
8640 (__builtin_ia32_sqrtsd_mask_round)
8641 (__builtin_ia32_sqrtss_mask_round): New builtins.
8642 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8643 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8644 ((match_operand:VF_128 1 "vector_operand"
8645 "xBm,<round_constraint>")): Changed to ...
8646 ((match_operand:VF_128 1 "vector_operand"
8647 "xBm,<round_scalar_constraint>")): ... this.
8648 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8649 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8650 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8651 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8652 %<iptr>1<round_scalar_mask_op3>}): ... this.
8653 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8654 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8656 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8659 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8660 Cast vec_cmpeq result to correct type.
8661 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8662 Cast vec_cmpgt result to correct type.
8664 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8666 * final.c (final_scan_insn_1): Renamed from...
8667 (final_scan_insn): ... this. New wrapper, to recover
8668 seen from the outermost call in recursive ones.
8669 * config/sparc/sparc.c (output_return): Drop seen from call.
8670 (output_sibcall): Likewise.
8671 * config/visium/visium.c (output_branch): Likewise.
8673 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8675 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8678 2018-02-10 Alan Modra <amodra@gmail.com>
8681 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8682 Specify LR as an input.
8684 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8687 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8688 remove_member_access_dummy_vars): New functions.
8689 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8690 lower_omp_1, execute_lower_omp): Use them.
8692 PR rtl-optimization/84308
8693 * shrink-wrap.c (spread_components): Release todo vector.
8695 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8697 PR rtl-optimization/57193
8698 * ira-color.c (struct allocno_color_data): Add member
8699 conflict_allocno_hard_prefs.
8700 (update_conflict_allocno_hard_prefs): New.
8701 (bucket_allocno_compare_func): Add a preference based on
8702 conflict_allocno_hard_prefs.
8703 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8704 (color_allocnos): Remove a dead code. Initiate
8705 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8707 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8710 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8711 constraint from =wa to wa. Avoid a subreg on the output operand,
8712 instead use a pseudo and subreg it in a move.
8713 (p9_xxbrd_<mode>): Changed to ...
8714 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8715 (p9_xxbrd_v2df): New expander.
8716 (p9_xxbrw_<mode>): Changed to ...
8717 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8718 (p9_xxbrw_v4sf): New expander.
8720 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8722 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8724 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8727 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8728 multiply in 32-bit mode.
8729 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8730 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8733 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8735 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8736 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8737 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8738 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8740 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8743 * dwarf2out.c (is_trivial_indirect_ref): New function.
8744 (dwarf2out_late_global_decl): Do not generate a location
8745 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8746 and that are not defined in the current unit.
8748 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8750 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8751 instead of a libcall for UNORDERED.
8753 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8756 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8757 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8759 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8762 * config/s390/s390.c (s390_set_current_function): Invoke
8763 s390_indirect_branch_settings also if fndecl didn't change.
8765 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8767 * config/rs6000/rs6000.md (blockage): Set length to zero.
8769 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8771 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8773 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8776 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8777 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8781 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8782 PARALLEL incoming that failed vt_get_decl_and_offset check.
8785 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8786 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8788 (get_variable_section): For decls in named .bss* sections pass true as
8789 second argument to bss_initializer_p.
8791 2018-02-09 Marek Polacek <polacek@redhat.com>
8792 Jakub Jelinek <jakub@redhat.com>
8795 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8796 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8797 Sync some changes from cxx_fold_indirect_ref.
8799 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8801 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8803 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8804 (BLOCK_INLINE_ENTRY_LABEL): New.
8805 (dwarf2out_var_location): Disregard inline entry markers.
8806 (inline_entry_data): New struct.
8807 (inline_entry_data_hasher): New hashtable type.
8808 (inline_entry_data_hasher::hash): New.
8809 (inline_entry_data_hasher::equal): New.
8810 (inline_entry_data_table): New variable.
8811 (add_high_low_attributes): Add DW_AT_entry_pc and
8812 DW_AT_GNU_entry_view attributes if a pending entry is found
8813 in inline_entry_data_table. Add old entry_pc attribute only
8814 if debug nonbinding markers are disabled.
8815 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8816 markers are enabled.
8817 (block_within_block_p, dwarf2out_inline_entry): New.
8818 (dwarf2out_finish): Check that no entries remained in
8819 inline_entry_data_table.
8820 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8821 (final_scan_insn, notice_source_line): Likewise.
8822 (rest_of_clean_state): Skip inline entry markers.
8823 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8825 * gimple.c (gimple_build_debug_inline_entry): New.
8826 * gimple.h (enum gimple_debug_subcode): Add
8827 GIMPLE_DEBUG_INLINE_ENTRY.
8828 (gimple_build_debug_inline_entry): Declare.
8829 (gimple_debug_inline_entry_p): New.
8830 (gimple_debug_nonbind_marker_p): Adjust.
8831 * insn-notes.def (INLINE_ENTRY): New.
8832 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8833 inline entry marker notes.
8834 (print_insn): Likewise.
8835 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8836 (INSN_DEBUG_MARKER_KIND): Likewise.
8837 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8838 * tree-inline.c (expand_call_inline): Build and insert
8839 debug_inline_entry stmt.
8840 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8841 inline entry blocks early, if nonbind markers are enabled.
8842 (dump_scope_block): Dump fragment info.
8843 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8844 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8845 (gimple_build_debug_inline_entry): New.
8846 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8847 Enable/disable inline entry points too.
8848 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8849 (DEBUG_INSN): Describe inline entry markers.
8851 * common.opt (gvariable-location-views): New.
8852 (gvariable-location-views=incompat5): New.
8853 * config.in: Rebuilt.
8854 * configure: Rebuilt.
8855 * configure.ac: Test assembler for view support.
8856 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8857 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8858 * dwarf2out.c (var_loc_view): New typedef.
8859 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8860 (dwarf2out_locviews_in_attribute): New.
8861 (dwarf2out_locviews_in_loclist): New.
8862 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8863 (enum dw_line_info_opcode): Add LI_adv_address.
8864 (struct dw_line_info_table): Add view.
8865 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8866 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8867 (zero_view_p): New variable.
8868 (ZERO_VIEW_P): New macro.
8869 (output_asm_line_debug_info): New.
8870 (struct var_loc_node): Add view.
8871 (add_AT_view_list, AT_loc_list): New.
8872 (add_var_loc_to_decl): Add view param. Test it against last.
8873 (new_loc_list): Add view params. Record them.
8874 (AT_loc_list_ptr): Handle loc and view lists.
8875 (view_list_to_loc_list_val_node): New.
8876 (print_dw_val): Handle dw_val_class_view_list.
8877 (size_of_die): Likewise.
8878 (value_format): Likewise.
8879 (loc_list_has_views): New.
8880 (gen_llsym): Set vl_symbol too.
8881 (maybe_gen_llsym, skip_loc_list_entry): New.
8882 (dwarf2out_maybe_output_loclist_view_pair): New.
8883 (output_loc_list): Output view list or entries too.
8884 (output_view_list_offset): New.
8885 (output_die): Handle dw_val_class_view_list.
8886 (output_dwarf_version): New.
8887 (output_compilation_unit_header): Use it.
8888 (output_skeleton_debug_sections): Likewise.
8889 (output_rnglists, output_line_info): Likewise.
8890 (output_pubnames, output_aranges): Update version comments.
8891 (output_one_line_info_table): Output view numbers in asm comments.
8892 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8893 Call maybe_gen_llsym.
8894 (loc_list_from_tree_1): Adjust.
8895 (add_AT_location_description): Create view list attribute if
8896 needed, check it's absent otherwise.
8897 (convert_cfa_to_fb_loc_list): Adjust.
8898 (maybe_emit_file): Call output_asm_line_debug_info for test.
8899 (dwarf2out_var_location): Reset views as needed. Precompute
8900 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8901 attribute. Set view.
8902 (new_line_info_table): Reset next view.
8903 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8904 (dwarf2out_source_line): Likewise. Output view resets and labels to
8905 the assembler, or select appropriate line info opcodes.
8906 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8907 (optimize_string_length): Catch it. Adjust.
8908 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8909 dw_val_class_view_list, and remove it if no longer needed.
8910 (hash_loc_list): Hash view numbers.
8911 (loc_list_hasher::equal): Compare them.
8912 (optimize_location_lists): Check whether a view list symbol is
8913 needed, and whether the locview attribute is present, and
8914 whether they match. Remove the locview attribute if no longer
8916 (index_location_lists): Call skip_loc_list_entry for test.
8917 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8918 Use output_dwarf_version.
8919 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8920 (struct dw_val_node): Add val_view_list.
8921 * final.c (SEEN_NEXT_VIEW): New.
8922 (set_next_view_needed): New.
8923 (clear_next_view_needed): New.
8924 (maybe_output_next_view): New.
8925 (final_start_function): Rename to...
8926 (final_start_function_1): ... this. Take pointer to FIRST,
8927 add SEEN parameter. Emit param bindings in the initial view.
8928 (final_start_function): Reintroduce SEEN-less interface.
8929 (final): Rename to...
8930 (final_1): ... this. Take SEEN parameter. Output final pending
8931 next view at the end.
8932 (final): Reintroduce seen-less interface.
8933 (final_scan_insn): Output pending next view before switching
8934 sections or ending a block. Mark the next view as needed when
8935 outputting variable locations. Notify debug backend of section
8936 changes, and of location view changes.
8937 (rest_of_handle_final): Adjust.
8938 * toplev.c (process_options): Autodetect value for debug variable
8939 location views option. Warn on incompat5 without -gdwarf-5.
8940 * doc/invoke.texi (gvariable-location-views): New.
8941 (gvariable-location-views=incompat5): New.
8942 (gno-variable-location-views): New.
8944 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8946 PR tree-optimization/84136
8947 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8948 that the result of find_edge is non-NULL.
8950 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8953 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8954 storing integer register in SImode. Fix cost of 256 and 512
8955 byte aligned SSE register store.
8957 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8959 * config/i386/i386.c (ix86_multiplication_cost): Fix
8960 multiplication cost for TARGET_AVX512DQ.
8962 2018-02-08 Marek Polacek <polacek@redhat.com>
8964 PR tree-optimization/84238
8965 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8968 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8970 PR tree-optimization/84265
8971 * tree-vect-stmts.c (vectorizable_store): Don't treat
8972 VMAT_CONTIGUOUS accesses as grouped.
8973 (vectorizable_load): Likewise.
8975 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8977 PR tree-optimization/81635
8978 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8979 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8980 (test_round_for_mask): New functions.
8981 (wide_int_cc_tests): Call test_round_for_mask.
8982 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8983 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8984 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8985 range returned by get_range_info.
8987 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8990 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8991 * symtab.c: Include builtins.h
8992 (symtab_node::output_to_lto_symbol_table_p): Move here
8993 from lto-streamer-out.c:output_symbol_p.
8994 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8995 (output_symbol_p): Move all logic to symtab.c
8996 (produce_symtab): Update.
8998 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9000 * config/s390/s390-opts.h (enum indirect_branch): Define.
9001 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9002 (s390_indirect_branch_via_thunk)
9003 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9004 (enum s390_indirect_branch_type): Define.
9005 * config/s390/s390.c (struct s390_frame_layout, struct
9006 machine_function): Remove.
9007 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9008 (indirect_branch_table_label_no, indirect_branch_table_name):
9010 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9011 (enum s390_indirect_branch_option): Define.
9012 (s390_return_addr_from_memory): New function.
9013 (s390_handle_string_attribute): New function.
9014 (s390_attribute_table): Add new attribute handler.
9015 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9016 (s390_indirect_branch_via_thunk): New function.
9017 (s390_indirect_branch_via_inline_thunk): New function.
9018 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9019 sibling call optimization for non z10 compiles.
9020 (s390_emit_call): Force indirect branch target to be a single
9021 register. Add r1 clobber for non-z10 compiles.
9022 (s390_emit_epilogue): Emit return jump via return_use expander.
9023 (s390_reorg): Handle JUMP_INSNs as execute targets.
9024 (s390_option_override_internal): Perform validity checks for the
9025 new command line options.
9026 (s390_indirect_branch_attrvalue): New function.
9027 (s390_indirect_branch_settings): New function.
9028 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9029 (s390_output_indirect_thunk_function): New function.
9030 (s390_code_end): Implement target hook.
9031 (s390_case_values_threshold): Implement target hook.
9032 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9034 * config/s390/s390.h (struct s390_frame_layout)
9035 (struct machine_function): Move here from s390.c.
9036 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9037 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9038 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9039 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9040 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9041 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9042 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9043 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9044 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9045 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9046 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9047 (mnemonic attribute): Add values which aren't recognized
9049 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9050 pattern for branch conversion. Fix mnemonic attribute.
9051 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9052 indirect branch via thunk if requested.
9053 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9054 ("*indirect_jump"): Disable for branch conversion using out of
9056 ("indirect_jump_via_thunk<mode>_z10")
9057 ("indirect_jump_via_thunk<mode>")
9058 ("indirect_jump_via_inlinethunk<mode>_z10")
9059 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9060 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9061 ("casesi_jump_via_inlinethunk<mode>_z10")
9062 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9063 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9064 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9065 ("*indirect2_jump"): Disable for branch conversion.
9066 ("casesi_jump"): Turn into expander and expand patterns for branch
9068 ("return_use"): New expander.
9069 ("*return"): Emit return via thunk and rename it to ...
9070 ("*return<mode>"): ... this one.
9071 * config/s390/s390.opt: Add new options and and enum for the
9074 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9076 * lra-constraints.c (match_reload): Unconditionally use
9077 gen_lowpart_SUBREG, rather than selecting between that
9078 and equivalent gen_rtx_SUBREG code.
9080 2018-02-08 Richard Biener <rguenther@suse.de>
9082 PR tree-optimization/84233
9083 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9084 changed flag instead of boguously re-using phi_inserted.
9086 2018-02-08 Martin Jambor <mjambor@suse.cz>
9088 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9089 static local variables.
9091 2018-02-08 Richard Biener <rguenther@suse.de>
9093 PR tree-optimization/84278
9094 * tree-vect-stmts.c (vectorizable_store): When looking for
9095 smaller vector types to perform grouped strided loads/stores
9096 make sure the mode is supported by the target.
9097 (vectorizable_load): Likewise.
9099 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9101 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9102 Increase LDP/STP opportunities by adding adjacent callee-saves.
9104 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9106 PR rtl-optimization/84068
9107 PR rtl-optimization/83459
9108 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9110 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9112 PR tree-optimization/84224
9113 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9114 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9117 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9120 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9121 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9122 count, remove one USE.
9124 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9126 * doc/install.texi (Configuration): Document the
9127 --with-long-double-format={ibm,ieee} PowerPC configuration
9131 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9132 Convert from define_expand to be define_insn_and_split. Rework
9133 float/double/_Float128 conversions to QI/HI/SImode to work with
9134 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9135 conversions to QI/HImode types did a store and then a load to
9136 truncate the value. For conversions to VSX registers, don't split
9137 the insn, instead emit the code directly. Use the code iterator
9138 any_fix to combine signed and unsigned conversions.
9139 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9140 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9141 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9142 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9143 (fix_<mode>di2_hw): Likewise.
9144 (fixuns_<mode>di2_hw): Likewise.
9145 (fix_<mode>si2_hw): Likewise.
9146 (fixuns_<mode>si2_hw): Likewise.
9147 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9148 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9149 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9150 fix<uns>_trunc<SFDF:mode>si2_p8.
9151 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9152 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9153 (fix<uns>_<mode>_mem): Likewise.
9154 (fctiw<u>z_<mode>_mem): Likewise.
9155 (fix<uns>_<mode>_mem): Likewise.
9156 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9157 the register allocator from doing a direct move to the GPRs to do
9158 a store, and instead use the ISA 3.0 store byte/half-word from
9159 vector register instruction. For IEEE 128-bit floating point,
9160 also optimize stores of 32-bit ints.
9161 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9163 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9165 * genextract.c (push_pathstr_operand): New function to support
9167 (walk_rtx): Call push_pathstr_operand.
9168 (print_path): Support [a-zA-Z].
9170 2018-02-07 Richard Biener <rguenther@suse.de>
9172 PR tree-optimization/84037
9173 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9174 (cse_and_gimplify_to_preheader): Declare.
9175 (vect_get_place_in_interleaving_chain): Likewise.
9176 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9178 (_loop_vec_info::~_loop_vec_info): Delete it.
9179 (cse_and_gimplify_to_preheader): New function.
9180 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9181 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9182 (vectorizable_load): Likewise. For grouped stores always base
9183 the IV on the first element.
9184 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9185 condition before gimplifying.
9187 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9189 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9190 *DIV_EXPR and *MOD_EXPR.
9192 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9195 * config/i386/i386.c (ix86_option_override_internal): Mask out
9196 the CF_SET bit when checking -fcf-protection.
9198 2018-02-07 Tom de Vries <tom@codesourcery.com>
9201 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9204 2018-02-07 Richard Biener <rguenther@suse.de>
9206 PR tree-optimization/84204
9207 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9210 PR tree-optimization/84205
9211 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9212 special-case isl_ast_op_zdiv_r.
9214 PR tree-optimization/84223
9215 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9216 Only add conditions from within the region.
9217 (gather_bbs::after_dom_children): Adjust.
9219 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9222 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9223 * config/avr/avr.md: Only post-reload split REG-REG moves if
9224 either register is GENERAL_REG_P.
9226 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9228 PR tree-optimization/84235
9229 * tree-ssa-scopedtables.c
9230 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9231 if the subtraction is performed in floating point type where NaNs are
9232 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9233 build 1. Formatting fix.
9235 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9238 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9239 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9240 and skip it regardless of bb boundaries. Use CALL_P macro,
9241 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9242 unnecessarily, formatting fix.
9244 2018-02-06 Michael Collison <michael.collison@arm.com>
9246 * config/arm/thumb2.md:
9247 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9248 (*thumb_mov_notscc): Ditto.
9250 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9253 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9254 unsigned_fix, not "s".
9256 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9258 * configure.ac (gcc_fn_eh_frame_ro): New function.
9259 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9260 correct .eh_frame permissions.
9261 * configure: Regenerate.
9263 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9265 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9268 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9270 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9271 Display warning message for -mno-speculate-indirect-jumps.
9273 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9275 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9277 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9279 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9281 PR tree-optimization/84225
9282 * tree-eh.c (find_trapping_overflow): Only call
9283 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9285 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9288 * config/i386/i386.c: Reimplement the check of possible options
9289 -mibt/-mshstk conbination. Change error messages.
9290 * doc/invoke.texi: Fix a typo: remove extra '='.
9292 2018-02-06 Marek Polacek <polacek@redhat.com>
9294 PR tree-optimization/84228
9295 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9297 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9300 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9301 emitted arch directives.
9302 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9303 __ARM_FEATURE_COPROC before changing architectures.
9305 2018-02-06 Richard Biener <rguenther@suse.de>
9307 * config/i386/i386.c (print_reg): Fix typo.
9308 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9310 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9312 * configure: Regenerate.
9314 2018-02-05 Martin Sebor <msebor@redhat.com>
9316 PR tree-optimization/83369
9317 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9320 2018-02-05 Martin Liska <mliska@suse.cz>
9322 * doc/invoke.texi: Cherry-pick upstream r323995.
9324 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9326 * ira.c (ira_init_register_move_cost): Adjust comment.
9328 2018-02-05 Martin Liska <mliska@suse.cz>
9330 PR gcov-profile/84137
9331 * doc/gcov.texi: Fix typo in documentation.
9333 2018-02-05 Martin Liska <mliska@suse.cz>
9335 PR gcov-profile/83879
9336 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9337 using dlopen functionality.
9339 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9341 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9342 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9343 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9344 _mm_maskz_range_round_ss): New intrinsics.
9345 (__builtin_ia32_rangesd128_round)
9346 (__builtin_ia32_rangess128_round): Remove.
9347 (__builtin_ia32_rangesd128_mask_round,
9348 __builtin_ia32_rangess128_mask_round): New builtins.
9349 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9350 __builtin_ia32_rangess128_round): Remove.
9351 (__builtin_ia32_rangesd128_mask_round,
9352 __builtin_ia32_rangess128_mask_round): New builtins.
9353 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9354 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9355 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9356 "<round_saeonly_constraint>")): Changed to ...
9357 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9358 "<round_saeonly_scalar_constraint>")): ... this.
9359 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9360 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9361 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9362 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9363 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9365 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9367 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9369 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9370 Remove all values except native, 8540 and 8548.
9372 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9374 * config/i386/i386.c (ix86_output_function_return): Pass
9375 INVALID_REGNUM, instead of -1, as invalid register number to
9376 indirect_thunk_name and output_indirect_thunk.
9378 2018-02-02 Julia Koval <julia.koval@intel.com>
9380 * config.gcc: Add -march=icelake.
9381 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9382 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9383 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9384 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9385 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9386 (processor_target_table): Add icelake.
9387 (ix86_option_override_internal): Handle new PTAs.
9388 (get_builtin_code_for_version): Handle icelake.
9389 (M_INTEL_COREI7_ICELAKE): New.
9390 (fold_builtin_cpu): Handle icelake.
9391 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9392 * doc/invoke.texi: Add -march=icelake.
9394 2018-02-02 Julia Koval <julia.koval@intel.com>
9396 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9397 to wide_int_bitmask.
9398 * wide-int-bitmask.h: New.
9400 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9403 * config/i386/i386.md: Replace Pmode with word_mode in
9404 builtin_setjmp_setup and builtin_longjmp to support x32.
9406 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9410 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9412 (rs6000_supported_cpu_names): New static variable.
9413 (linux_cpu_translation_table): Likewise.
9414 (elf_platform) <cpu>: Define new static variable and use it.
9415 Translate kernel AT_PLATFORM name to canonical name if needed.
9416 Error if platform name is unknown.
9418 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9421 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9423 2018-02-01 Jeff Law <law@redhat.com>
9426 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9427 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9428 the scratch if RELEASE_VIA_POP is false.
9429 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9430 If we have to save a temporary register, decrement SIZE appropriately.
9431 Pass new arguments to release_scratch_register_on_entry.
9432 (ix86_adjust_stack_and_probe): Likewise.
9433 (ix86_emit_probe_stack_range): Pass new arguments to
9434 release_scratch_register_on_entry.
9436 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9438 PR rtl-optimization/84157
9439 * combine.c (change_zero_ext): Use REG_P predicate in
9440 front of HARD_REGISTER_P predicate.
9442 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9444 * config/avr/avr.c (avr_option_override): Move disabling of
9445 -fdelete-null-pointer-checks to...
9446 * common/config/avr/avr-common.c (avr_option_optimization_table):
9449 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9451 PR tree-optimization/81635
9452 * tree-data-ref.c (split_constant_offset_1): For types that
9453 wrap on overflow, try to use range info to prove that wrapping
9456 2018-02-01 Renlin Li <renlin.li@arm.com>
9459 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9461 (aarch64_register_move_cost): Likewise.
9462 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9464 (REG_CLASS_NAMES): Likewise.
9465 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9466 TAILCALL_ADDR_REGS. Remove IP registers.
9467 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9469 2018-02-01 Richard Biener <rguenther@suse.de>
9471 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9472 for specifying RPO order and allow NULL for that.
9473 * domwalk.c (dom_walker::dom_walker): Likewise.
9474 (dom_walker::walk): Handle NULL RPO order.
9475 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9477 (rewrite_update_dom_walker): Likewise.
9478 (mark_def_dom_walker): Likewise.
9480 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9482 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9483 (aarch64_maybe_expand_sve_subreg_move): Declare.
9484 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9485 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9487 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9488 that are semantically a reverse operation.
9489 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9490 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9491 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9493 (aarch64_can_change_mode_class): For big-endian, forbid changes
9494 between two SVE modes if they have different element sizes.
9496 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9498 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9499 the TImode handling for big-endian targets.
9501 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9503 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9504 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9506 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9507 Remove BSWAP handing for big-endian targets and use the form of
9508 LD1RQ appropariate for the mode.
9510 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9512 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9513 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9516 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9519 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9520 check for operands that need to go through aarch64_sve_reload_be.
9522 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9524 PR tree-optimization/81661
9525 PR tree-optimization/84117
9526 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9527 * tree-eh.c: Include gimplify.h.
9528 (find_trapping_overflow, replace_trapping_overflow,
9529 rewrite_to_non_trapping_overflow): New functions.
9530 * tree-vect-loop.c: Include tree-eh.h.
9531 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9532 * tree-data-ref.c: Include tree-eh.h.
9533 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9535 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9537 PR rtl-optimization/84123
9538 * combine.c (change_zero_ext): Check if hard register satisfies
9539 can_change_dest_mode before calling gen_lowpart_SUBREG.
9541 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9544 * ira.c (ira_init_register_move_cost): Remove assert.
9546 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9548 PR rtl-optimization/84071
9549 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9550 * doc/tm.texi: Regenerate.
9552 2018-01-31 Richard Biener <rguenther@suse.de>
9554 PR tree-optimization/84132
9555 * tree-data-ref.c (analyze_miv_subscript): Properly
9556 check whether evolution_function_is_affine_multivariate_p
9557 before calling gcd_of_steps_may_divide_p.
9559 2018-01-31 Julia Koval <julia.koval@intel.com>
9562 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9563 * config/i386/i386.md (rdpid_rex64) New.
9564 (rdpid): Make 32bit only.
9566 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9569 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9570 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9572 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9575 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9577 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9579 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9581 PR rtl-optimization/84071
9582 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9583 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9585 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9587 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9588 (arc_attribute_table): Add 'aux' attribute.
9589 (arc_in_small_data_p): Consider aux like variables.
9590 (arc_is_aux_reg_p): New function.
9591 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9592 (arc_get_aux_arg): New function.
9593 (prepare_move_operands): Handle aux-register access.
9594 (arc_handle_aux_attribute): New function.
9595 * doc/extend.texi (ARC Variable attributes): Add subsection.
9597 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9599 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9600 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9601 (arc_attribute_table): Add 'uncached' attribute.
9602 (arc_print_operand): Print '.di' flag for uncached memory
9604 (arc_in_small_data_p): Do not consider for small data the uncached
9606 (arc_is_uncached_mem_p): New function.
9607 * config/arc/predicates.md (compact_store_memory_operand): Check
9608 for uncached memory accesses.
9609 (nonvol_nonimm_operand): Likewise.
9610 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9612 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9615 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9616 falign-loops=): Add Optimization flag.
9618 2018-01-30 Jeff Law <law@redhat.com>
9621 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9622 INT_REGISTERS_SAVED. Check it prior to calling
9623 get_scratch_register_on_entry.
9624 (ix86_adjust_stack_and_probe): Similarly.
9625 (ix86_emit_probe_stack_range): Similarly.
9626 (ix86_expand_prologue): Corresponding changes.
9628 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9631 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9632 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9634 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9637 * lra-constraints.c (curr_insn_transform): Process AND in the
9640 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9642 PR rtl-optimization/83986
9643 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9644 dependence against last_pending_memory_flush in addition to
9647 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9649 PR tree-optimization/81611
9650 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9653 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9656 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9659 2018-01-30 Richard Biener <rguenther@suse.de>
9660 Jakub Jelinek <jakub@redhat.com>
9662 PR tree-optimization/84111
9663 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9664 inner loops added during recursion, as they don't have up-to-date
9667 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9670 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9671 (can_inline_edge_by_limits_p): ... here.
9672 (can_early_inline_edge_p, check_callers,
9673 update_caller_keys, update_callee_keys, recursive_inlining,
9674 add_new_edges_to_heap, speculation_useful_p,
9675 inline_small_functions,
9676 inline_small_functions, flatten_function,
9677 inline_to_all_callers_1): Update.
9679 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9681 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9684 2018-01-30 Richard Biener <rguenther@suse.de>
9686 PR tree-optimization/83008
9687 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9688 invariant and constant vector uses in stmts when they need
9691 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9694 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9695 * configure: Regenerate.
9697 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9699 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9701 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9702 Use gen_rtx_REG rather than gen_lowpart.
9704 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9706 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9707 rather than 0 when creating partial subregs.
9709 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9711 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9714 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9717 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9718 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9719 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9720 flags. This restores the settings used before the 2017-07-24.
9721 Turning off pre increment/decrement/modify allows IVOPTS to
9722 optimize DF/SF loops where the index is an int.
9724 2018-01-29 Richard Biener <rguenther@suse.de>
9725 Kelvin Nilsen <kelvin@gcc.gnu.org>
9728 * tree-vect-stmts.c (vectorizable_call): Don't call
9729 targetm.vectorize_builtin_md_vectorized_function if callee is
9732 2018-01-22 Carl Love <cel@us.ibm.com>
9734 * doc/extend.tex: Fix typo in second arg in
9735 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9737 2018-01-29 Richard Biener <rguenther@suse.de>
9739 PR tree-optimization/84086
9740 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9741 (flush_ssaname_freelist): When SSA names were released reset
9742 the SCEV hash table.
9744 2018-01-29 Richard Biener <rguenther@suse.de>
9746 PR tree-optimization/84057
9747 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9748 removed paths when removing edges.
9750 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9752 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9753 -mfunction-return=@var{choice}.
9755 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9758 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9760 (layout::print_source_line): Likewise.
9761 (test_get_line_width_without_trailing_whitespace): Add test cases.
9763 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9766 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9769 2018-01-26 Jim Wilson <jimw@sifive.com>
9771 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9773 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9776 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9778 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9779 and CMP + SUB-immediate -> SUBS.
9781 2018-01-26 Martin Sebor <msebor@redhat.com>
9783 PR tree-optimization/83896
9784 * tree-ssa-strlen.c (get_string_len): Rename...
9785 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9786 Avoid assuming length is constant.
9787 (handle_char_store): Use HOST_WIDE_INT for string length.
9789 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9792 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9793 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9795 2018-01-26 Richard Biener <rguenther@suse.de>
9797 PR rtl-optimization/84003
9798 * dse.c (record_store): Only record redundant stores when
9799 the earlier store aliases at least all accesses the later one does.
9801 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9803 PR rtl-optimization/83985
9804 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9805 REG_CFA_RESTORE insns.
9806 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9807 REG_CFA_RESTORE insns here.
9810 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9811 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9813 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9815 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9816 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9817 (arc_init): Likewise.
9818 (arc_override_options): Likewise.
9819 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9821 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9823 * config/arc/arc.h (TARGET_DBNZ): Define.
9824 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9825 properly set the tune attribute.
9826 (dbnz): Use TARGET_DBNZ guard.
9827 * config/arc/arc.opt (mtune): Add core3 option.
9829 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9831 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9832 recognize new pic like addresses.
9833 (arc_delegitimize_address): Clean up.
9835 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9837 * config/arc/arc-arches.def: Option mrf16 valid for all
9839 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9840 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9841 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9842 * config/arc/arc-tables.opt: Regenerate.
9843 * config/arc/arc.c (arc_conditional_register_usage): Handle
9844 reduced register file case.
9845 (arc_file_start): Set must have build attributes.
9846 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9848 * config/arc/arc.opt (mrf16): Add new option.
9849 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9850 * config/arc/genmultilib.awk: Handle new mrf16 option.
9851 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9852 * config/arc/t-multilib: Regenerate.
9853 * doc/invoke.texi (ARC Options): Document mrf16 option.
9855 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9857 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9858 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9859 (arc_attribute_table): Add 'secure_call' attribute.
9860 (arc_print_operand): Print secure call operand.
9861 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9863 (arc_is_secure_call_p): New function. * config/arc/arc.md
9864 (call_i): Add support for sjli instruction.
9865 (call_value_i): Likewise.
9866 * config/arc/constraints.md (Csc): New constraint.
9868 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9869 John Eric Martin <John.Martin@emmicro-us.com>
9871 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9872 * config/arc/arc.c (_arc_jli_section): New struct.
9873 (arc_jli_section): New type.
9874 (rc_jli_sections): New static variable.
9875 (arc_handle_jli_attribute): New function.
9876 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9877 (arc_file_end): New function.
9878 (TARGET_ASM_FILE_END): Define.
9879 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9880 (arc_add_jli_section): New function.
9881 (jli_call_scan): Likewise.
9882 (arc_reorg): Call jli_call_scan.
9883 (arc_output_addsi): Remove 'S' from printing asm operand.
9884 (arc_is_jli_call_p): New function.
9885 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9887 (movhi_insn): Likewise.
9888 (movsi_insn): Likewise.
9889 (movsi_set_cc_insn): Likewise.
9890 (loadqi_update): Likewise.
9891 (load_zeroextendqisi_update): Likewise.
9892 (load_signextendqisi_update): Likewise.
9893 (loadhi_update): Likewise.
9894 (load_zeroextendhisi_update): Likewise.
9895 (load_signextendhisi_update): Likewise.
9896 (loadsi_update): Likewise.
9897 (loadsf_update): Likewise.
9898 (movsicc_insn): Likewise.
9899 (bset_insn): Likewise.
9900 (bxor_insn): Likewise.
9901 (bclr_insn): Likewise.
9902 (bmsk_insn): Likewise.
9903 (bicsi3_insn): Likewise.
9904 (cmpsi_cc_c_insn): Likewise.
9905 (movsi_ne): Likewise.
9906 (movsi_cond_exec): Likewise.
9907 (clrsbsi2): Likewise.
9918 (call_i): Remove 'S' asm letter, add jli instruction.
9919 (call_value_i): Likewise.
9920 * config/arc/arc.op (mjli-always): New option.
9921 * config/arc/constraints.md (Cji): New constraint.
9922 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9924 (subsf3_fpx): Likewise.
9925 (mulsf3_fpx): Likewise.
9926 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9928 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9930 * doc/invoke.texi (ARC): Document mjli-always option.
9932 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9934 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9935 avoid addition with 0 and use incw and decw where possible.
9937 2018-01-26 Richard Biener <rguenther@suse.de>
9939 PR tree-optimization/81082
9940 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9941 association if it requires casting to unsigned.
9942 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9943 from fold_plusminus_mult_expr to catch important cases late when
9944 range info is available.
9946 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9948 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9949 * configure.ac (hidden_linkonce): New test.
9950 * configure: Regenerate.
9951 * config.in: Regenerate.
9953 2018-01-26 Julia Koval <julia.koval@intel.com>
9955 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9956 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9957 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9958 _mm_mask_bitshuffle_epi64_mask): Fix type.
9959 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9960 USI_FTYPE_V4DI_V4DI_USI): Remove.
9961 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9962 __builtin_ia32_vpshufbitqmb256_mask,
9963 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9964 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9965 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9967 2018-01-26 Alan Modra <amodra@gmail.com>
9970 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9971 UNSPEC_VBPERMQ. Sort other unspecs.
9973 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9975 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9977 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9980 * predict.c (drop_profile): Do not push/pop cfun; update also
9982 (handle_missing_profiles): Fix logic looking for zero profiles.
9984 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9987 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9988 on functions with #pragma omp declare simd or functions with simd
9990 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9991 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9992 Remove trailing \n from warning_at calls.
9994 2018-01-25 Tom de Vries <tom@codesourcery.com>
9997 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9998 for neutered workers.
10000 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10003 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10004 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10006 2018-01-24 Jeff Law <law@redhat.com>
10009 * i386.c (get_probe_interval): Move to earlier point.
10010 (ix86_compute_frame_layout): If -fstack-clash-protection and
10011 the frame is larger than the probe interval, then use pushes
10012 to save registers rather than reg->mem moves.
10013 (ix86_expand_prologue): Remove conditional for int_registers_saved
10016 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10019 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10020 min/max for never referenced object.
10022 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10024 PR middle-end/83977
10025 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10027 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10028 attributes from DECL_ATTRIBUTES (decl) without affecting
10029 DECL_ATTRIBUTES (current_function_decl).
10030 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10031 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10033 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10035 PR tree-optimization/83979
10036 * fold-const.c (fold_comparison): Use constant_boolean_node
10037 instead of boolean_{true,false}_node.
10039 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10041 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10044 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10046 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10047 Simplify the clause that sets the length attribute.
10048 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10049 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10050 clause that sets the length attribute.
10051 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10053 2018-01-24 Tom de Vries <tom@codesourcery.com>
10056 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10057 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10058 Add strict parameter.
10059 (prevent_branch_around_nothing): Insert dummy insn between branch to
10060 label and label with no ptx insn inbetween.
10061 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10063 2018-01-24 Tom de Vries <tom@codesourcery.com>
10066 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10067 for neutered threads in warp.
10068 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10070 2018-01-24 Richard Biener <rguenther@suse.de>
10072 PR tree-optimization/83176
10073 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10076 2018-01-24 Richard Biener <rguenther@suse.de>
10078 PR tree-optimization/82819
10079 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10080 code generating pluses that are no-ops in the target precision.
10082 2018-01-24 Richard Biener <rguenther@suse.de>
10084 PR middle-end/84000
10085 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10087 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10089 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10090 to merge probabilities.
10091 * predict.c (probably_never_executed): Also mark as cold functions
10092 with global 0 profile and guessed local profile.
10093 * profile-count.c (profile_probability::combine_with_count): New
10095 * profile-count.h (profile_probability::operator*,
10096 profile_probability::operator*=, profile_probability::operator/,
10097 profile_probability::operator/=): Reduce precision to adjusted
10098 and set value to guessed on contradictory divisions.
10099 (profile_probability::combine_with_freq): Remove.
10100 (profile_probability::combine_wiht_count): Declare.
10101 (profile_count::force_nonzero):: Set to adjusted.
10102 (profile_count::probability_in):: Set quality to adjusted.
10103 * tree-ssa-tail-merge.c (replace_block_by): Use
10104 combine_with_count.
10106 2018-01-23 Andrew Waterman <andrew@sifive.com>
10107 Jim Wilson <jimw@sifive.com>
10109 * config/riscv/riscv.c (riscv_stack_boundary): New.
10110 (riscv_option_override): Set riscv_stack_boundary. Handle
10111 riscv_preferred_stack_boundary_arg.
10112 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10113 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10114 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10115 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10116 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10117 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10119 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10122 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10123 of struct ix86_frame.
10124 (ix86_expand_epilogue): Likewise. Add a local variable for
10125 the reg_save_offset field in struct ix86_frame.
10127 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10129 PR tree-optimization/82604
10130 * tree-loop-distribution.c (enum partition_kind): New enum item
10131 PKIND_PARTIAL_MEMSET.
10132 (partition_builtin_p): Support above new enum item.
10133 (generate_code_for_partition): Ditto.
10134 (compute_access_range): Differentiate cases that equality can be
10135 proven at all loops, the innermost loops or no loops.
10136 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10137 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10138 (finalize_partitions, distribute_loop): Don't fuse partition of
10139 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10140 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10141 parloop is enabled.
10143 2018-01-23 Martin Liska <mliska@suse.cz>
10145 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10146 order to ignore the predictor.
10147 (PRED_POLYMORPHIC_CALL): Likewise.
10148 (PRED_RECURSIVE_CALL): Likewise.
10150 2018-01-23 Martin Liska <mliska@suse.cz>
10152 * tree-profile.c (tree_profiling): Print function header to
10153 aware reader which function we are working on.
10154 * value-prof.c (gimple_find_values_to_profile): Do not print
10155 not interesting value histograms.
10157 2018-01-23 Martin Liska <mliska@suse.cz>
10159 * profile-count.h (enum profile_quality): Add
10160 profile_uninitialized as the first value. Do not number values
10161 as they are zero based.
10162 (profile_count::verify): Update sanity check.
10163 (profile_probability::verify): Likewise.
10165 2018-01-23 Nathan Sidwell <nathan@acm.org>
10167 * doc/invoke.texi (ffor-scope): Deprecate.
10169 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10171 PR tree-optimization/83510
10172 * domwalk.c (set_all_edges_as_executable): New function.
10173 (dom_walker::dom_walker): Convert bool param
10174 "skip_unreachable_blocks" to enum reachability. Move setup of
10175 edge flags to set_all_edges_as_executable and only do it when
10176 reachability is REACHABLE_BLOCKS.
10177 * domwalk.h (enum dom_walker::reachability): New enum.
10178 (dom_walker::dom_walker): Convert bool param
10179 "skip_unreachable_blocks" to enum reachability.
10180 (set_all_edges_as_executable): New decl.
10181 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10182 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10184 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10185 but converting true to REACHABLE_BLOCKS.
10186 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10188 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10189 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10190 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10192 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10193 if check_all_array_refs will be called.
10195 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10197 * tree.c (selftest::test_location_wrappers): Add more test
10200 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10202 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10203 (selftest::test_bit_in_range): Likewise.
10205 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10208 * doc/sourcebuild.texi (vect_float): Say that the selector
10209 only describes the situation when -funsafe-math-optimizations is on.
10210 (vect_float_strict): Document.
10212 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10214 PR tree-optimization/83965
10215 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10216 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10217 instead of checking only for a reduction.
10218 (vect_recog_widen_sum_pattern): Likewise.
10220 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10222 * predict.c (probably_never_executed): Only use precise profile info.
10223 (compute_function_frequency): Skip after inlining hack since we now
10224 have quality checking.
10226 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10228 * profile-count.h (profile_probability::very_unlikely,
10229 profile_probability::unlikely, profile_probability::even): Set
10230 precision to guessed.
10232 2018-01-23 Richard Biener <rguenther@suse.de>
10234 PR tree-optimization/83963
10235 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10236 Properly terminate dominator walk when crossing the exit edge not
10237 when visiting its source block.
10239 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10242 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10243 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10245 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10247 PR tree-optimization/83957
10248 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10249 semicolon after for body surrounded by braces.
10251 PR tree-optimization/83081
10252 * profile-count.h (profile_probability::split): New method.
10253 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10254 Use profile_probability::split.
10255 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10256 when splitting a single conditional jump into 2.
10258 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10260 PR tree-optimization/69452
10261 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10264 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10266 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10267 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10268 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10270 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10272 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10274 * config/rl78/rl78.md (movdi): New define_expand.
10275 * config/rl78/rl78.c (rl78_split_movdi): New function.
10277 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10280 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10282 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10283 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10284 128-bit to produce an UNSPEC move to get the double word with the
10285 signbit and then a shift directly to do signbit.
10286 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10287 implementation with a new version that just does either a direct
10288 move or a regular move. Move memory interface to separate insns.
10289 Move insns so they are next to the expander.
10290 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10291 with signbit move. Split big and little endian case.
10292 (signbit<mode>2_dm_mem_le): Likewise.
10293 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10294 (signbit<mode>2_dm2): Likewise.
10296 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10298 * config/rl78/rl78.md (anddi3): New define_expand.
10300 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10302 * config/rl78/rl78.md (umindi3): New define_expand.
10304 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10306 * config/rl78/rl78.md (smindi3): New define_expand.
10308 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10310 * config/rl78/rl78.md (smaxdi3): New define_expand.
10312 2018-01-22 Carl Love <cel@us.ibm.com>
10314 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10315 LVX_V1TI): Add macro expansion.
10316 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10317 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10318 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10319 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10320 Change check to determine if the instruction is a byte reversing
10321 entry. Fix typo in comment.
10322 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10323 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10324 Add def_builtin calls for new builtins.
10325 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10326 Add define_insn expansion.
10328 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10330 * config/rl78/rl78.md (umaxdi3): New define_expand.
10332 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10334 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10335 for non-QImode registers.
10337 2018-01-22 Richard Biener <rguenther@suse.de>
10339 PR tree-optimization/83963
10340 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10341 including the loop exit block.
10342 (scop_detection::merge_sese): Likewise.
10343 (scop_detection::add_scop): Do it here instead.
10345 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10347 * doc/sourcebuild.texi (arm_softfloat): Document.
10349 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10352 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10353 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10354 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10356 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10357 David Edelsohn <dje.gcc@gmail.com>
10360 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10361 Change "crset eq" to "crset 2".
10362 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10363 (*call_indirect_aix<mode>_nospec): Likewise.
10364 (*call_value_indirect_aix<mode>_nospec): Likewise.
10365 (*call_indirect_elfv2<mode>_nospec): Likewise.
10366 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10367 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10368 change assembly output from . to $.
10369 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10370 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10371 (*tablejump<mode>_internal1_nospec): Likewise.
10373 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10376 * config/sh/sh_optimize_sett_clrt.cc:
10377 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10379 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10381 PR tree-optimization/83940
10382 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10383 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10384 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10385 use it to pass back the definition type.
10386 (vect_check_store_rhs): Likewise rhs_dt_out.
10387 (vect_build_gather_load_calls): Add a mask_dt argument and use
10388 it instead of a call to vect_is_simple_use.
10389 (vectorizable_store): Update calls to vect_check_load_store_mask
10390 and vect_check_store_rhs. Use the dt returned by the latter instead
10391 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10392 instead of calls to vect_is_simple_use. Pass the scalar rather
10393 than the vector operand to vect_is_simple_use when handling
10394 second and subsequent copies of an rhs value.
10395 (vectorizable_load): Update calls to vect_check_load_store_mask
10396 and vect_build_gather_load_calls. Use the cached mask_dt and
10397 gs_info.offset_dt instead of calls to vect_is_simple_use.
10399 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10401 PR middle-end/83945
10402 * tree-emutls.c: Include gimplify.h.
10403 (lower_emutls_2): New function.
10404 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10405 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10406 it before further processing.
10409 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10410 UINTVAL (trueop1) instead of INTVAL (op1).
10412 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10416 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10417 INCOMING_FRAME_SP_OFFSET if not defined.
10418 (scan_trace): Add ENTRY argument. If true and
10419 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10420 emit a note to adjust the CFA offset.
10421 (create_cfi_notes): Adjust scan_trace callers.
10422 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10423 INCOMING_FRAME_SP_OFFSET in the CIE.
10424 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10425 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10427 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10428 * doc/tm.texi: Regenerated.
10430 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10432 PR rtl-optimization/83147
10433 * lra-constraints.c (remove_inheritance_pseudos): Use
10434 lra_substitute_pseudo_within_insn.
10436 2018-01-19 Tom de Vries <tom@codesourcery.com>
10437 Cesar Philippidis <cesar@codesourcery.com>
10440 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10442 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10445 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10446 spaces for function labels.
10448 2018-01-19 Martin Liska <mliska@suse.cz>
10450 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10451 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10452 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10453 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10454 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10455 (PRED_CONST_RETURN): Change from 69 to 65.
10456 (PRED_NULL_RETURN): Change from 91 to 71.
10457 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10458 (PRED_LOOP_GUARD): Change from 66 to 73.
10460 2018-01-19 Martin Liska <mliska@suse.cz>
10462 * predict.c (predict_insn_def): Add new assert.
10463 (struct branch_predictor): Change type to signed integer.
10464 (test_prediction_value_range): Amend test to cover
10465 PROB_UNINITIALIZED.
10466 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10467 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10468 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10469 (PRED_LOOP_IV_COMPARE): Likewise.
10470 * predict.h (PROB_UNINITIALIZED): Define new constant.
10472 2018-01-19 Martin Liska <mliska@suse.cz>
10474 * predict.c (dump_prediction): Add new format for
10475 analyze_brprob.py script which is enabled with -details
10477 * profile-count.h (precise_p): New function.
10479 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10481 PR tree-optimization/83922
10482 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10483 there are no statements that need masking.
10484 (vect_active_double_reduction_p): New function.
10485 (vect_analyze_loop_operations): Use it when handling phis that
10486 are not in the loop header.
10488 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10490 PR tree-optimization/83914
10491 * tree-vect-loop.c (vectorizable_induction): Don't convert
10492 init_expr or apply the peeling adjustment for inductions
10493 that are nested within the vectorized loop.
10495 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10497 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10500 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10504 * function.h (gimplify_parameters): Add gimple_seq * argument.
10505 * function.c: Include gimple.h and options.h.
10506 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10507 for the added local temporaries if needed.
10508 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10509 if there are any parameter cleanups, wrap whole body into a
10510 try/finally with the cleanups.
10512 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10515 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10516 Use GET_MODE_CLASS for scalar floating point.
10518 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10522 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10523 Fix call of call_cgraph_insertion_hooks.
10525 2018-01-18 Martin Sebor <msebor@redhat.com>
10527 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10529 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10532 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10535 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10538 * common.opt: (-ffile-prefix-map): New option.
10539 * opts.c (common_handle_option): Defer it.
10540 * opts-global.c (handle_common_deferred_options): Handle it.
10541 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10542 * file-prefix-map.h: New file.
10543 (remap_debug_filename, add_debug_prefix_map): ...here.
10544 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10545 * final.c (debug_prefix_map, add_debug_prefix_map
10546 remap_debug_filename): Move to...
10547 * file-prefix-map.c: New file.
10548 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10549 generalize, get rid of alloca(), use strrchr() instead of strchr().
10550 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10551 Implement in terms of add_prefix_map().
10552 (remap_macro_filename, remap_debug_filename): Implement in term of
10554 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10555 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10556 * dbxout.c: Include file-prefix-map.h.
10557 * varasm.c: Likewise.
10558 * vmsdbgout.c: Likewise.
10559 * xcoffout.c: Likewise.
10560 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10561 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10562 * doc/invoke.texi (-ffile-prefix-map): Document.
10563 (-fdebug-prefix-map): Update description.
10565 2018-01-18 Martin Liska <mliska@suse.cz>
10567 * config/i386/i386.c (indirect_thunk_name): Document that also
10569 (output_indirect_thunk): Document why both instructions
10570 (pause and lfence) are generated.
10572 2018-01-18 Richard Biener <rguenther@suse.de>
10574 PR tree-optimization/83887
10575 * graphite-scop-detection.c
10576 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10577 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10578 (scop_detection::merge_sese): Re-implement with a flood-fill
10579 algorithm that properly finds a SESE region if it exists.
10581 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10584 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10585 pointer_diff optimizations use view_convert instead of convert.
10587 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10589 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10590 Generate different code for -mno-speculate-indirect-jumps.
10591 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10592 (*call_indirect_aix<mode>): Disable for
10593 -mno-speculate-indirect-jumps.
10594 (*call_indirect_aix<mode>_nospec): New define_insn.
10595 (*call_value_indirect_aix<mode>): Disable for
10596 -mno-speculate-indirect-jumps.
10597 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10598 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10599 -mno-speculate-indirect-jumps.
10600 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10602 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10604 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10605 long double type, set the flags for noting the default long double
10606 type, even if we don't pass or return a long double type.
10608 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10611 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10614 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10616 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10617 support for merge[hl].
10618 (fold_mergehl_helper): New helper function.
10619 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10620 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10621 (altivec_vmrglw_direct): Add xxmrglw insn.
10623 2018-01-17 Andrew Waterman <andrew@sifive.com>
10625 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10626 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10628 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10631 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10632 call the lto_location_cache before reading the
10633 DECL_SOURCE_LOCATION of the types.
10635 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10636 Richard Sandiford <richard.sandiford@linaro.org>
10638 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10639 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10640 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10641 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10642 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10644 * config/aarch64/constraints.md (aarch64_movti_operand):
10646 * config/aarch64/predicates.md (Uti): Add new constraint.
10648 2018-01-17 Carl Love <cel@us.ibm.com>
10650 * config/rs6000/vsx.md (define_expand xl_len_r,
10651 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10652 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10654 (define_expand, define_insn): Move the shift left from the
10655 define_insn to the define_expand for lxvl and stxvl instructions.
10656 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10657 and XL_LEN_R definitions to PURE.
10659 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10661 * config/i386/i386.c (indirect_thunk_name): Declare regno
10662 as unsigned int. Compare regno with INVALID_REGNUM.
10663 (output_indirect_thunk): Ditto.
10664 (output_indirect_thunk_function): Ditto.
10665 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10666 in the call to output_indirect_thunk_function.
10668 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10670 PR middle-end/83884
10671 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10672 rather than the size of inner_type to determine the stack slot size
10673 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10675 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10678 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10681 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10683 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10684 endian Linux systems to optionally enable multilibs for selecting
10685 the long double type if the user configured an explicit type.
10686 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10687 have no long double multilibs if not defined.
10688 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10689 warn if the user used -mabi={ieee,ibm}longdouble and we built
10690 multilibs for long double.
10691 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10692 appropriate multilib option.
10693 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10695 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10696 for building long double multilibs.
10697 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10699 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10701 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10704 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10706 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10709 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10712 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10715 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10717 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10718 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10720 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10722 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10723 different rtl trees depending on TARGET_64BIT.
10724 (rs6000_gen_lvx): Likewise.
10726 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10728 * config/visium/visium.md (nop): Tweak comment.
10729 (hazard_nop): Likewise.
10731 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10733 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10734 -mspeculate-indirect-jumps.
10735 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10736 for -mno-speculate-indirect-jumps.
10737 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10738 (*call_value_indirect_elfv2<mode>): Disable for
10739 -mno-speculate-indirect-jumps.
10740 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10741 (indirect_jump): Emit different RTL for
10742 -mno-speculate-indirect-jumps.
10743 (*indirect_jump<mode>): Disable for
10744 -mno-speculate-indirect-jumps.
10745 (*indirect_jump<mode>_nospec): New define_insn.
10746 (tablejump): Emit different RTL for
10747 -mno-speculate-indirect-jumps.
10748 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10749 (tablejumpsi_nospec): New define_expand.
10750 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10751 (tablejumpdi_nospec): New define_expand.
10752 (*tablejump<mode>_internal1): Disable for
10753 -mno-speculate-indirect-jumps.
10754 (*tablejump<mode>_internal1_nospec): New define_insn.
10755 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10758 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10760 * caller-save.c (insert_save): Drop unnecessary parameter. All
10763 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10764 Richard Biener <rguenth@suse.de>
10767 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10768 return early, inline manually is_gimple_sizepos. Make sure if we
10769 call gimplify_expr we don't end up with a gimple constant.
10770 * tree.c (variably_modified_type_p): Don't return true for
10771 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10772 * gimplify.h (is_gimple_sizepos): Remove.
10774 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10776 PR tree-optimization/83857
10777 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10778 vectorizable_live_operation for pure SLP statements.
10779 (vectorizable_live_operation): Handle PHIs.
10781 2018-01-16 Richard Biener <rguenther@suse.de>
10783 PR tree-optimization/83867
10784 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10785 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10787 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10790 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10791 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10792 If off is not INTEGER_CST, issue a may not be aligned warning
10793 rather than isn't aligned. Use isn%'t rather than isn't.
10794 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10796 <case MULT_EXPR>: Improve the case when bottom and one of the
10797 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10798 operand, in that case check if the other operand is multiple of
10799 bottom divided by the INTEGER_CST operand.
10801 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10804 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10805 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10806 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10807 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10808 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10809 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10810 (pa_function_arg_size): New function.
10812 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10814 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10815 in a separate statement.
10817 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10819 PR tree-optimization/83847
10820 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10821 group gathers and scatters.
10823 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10825 PR rtl-optimization/86620
10826 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10828 PR rtl-optimization/83213
10829 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10830 to last if both are JUMP_INSNs.
10832 PR tree-optimization/83843
10833 * gimple-ssa-store-merging.c
10834 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10835 store_immediate_info for bswap/nop orig_stores.
10837 2018-01-15 Andrew Waterman <andrew@sifive.com>
10839 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10841 <UDIV>: Increase cost if !TARGET_DIV.
10843 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10845 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10846 (define_attr "cr_logical_3op"): New.
10847 (cceq_ior_compare): Adjust.
10848 (cceq_ior_compare_complement): Adjust.
10849 (*cceq_rev_compare): Adjust.
10850 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10851 (is_cracked_insn): Adjust.
10852 (insn_must_be_first_in_group): Adjust.
10853 * config/rs6000/40x.md: Adjust.
10854 * config/rs6000/440.md: Adjust.
10855 * config/rs6000/476.md: Adjust.
10856 * config/rs6000/601.md: Adjust.
10857 * config/rs6000/603.md: Adjust.
10858 * config/rs6000/6xx.md: Adjust.
10859 * config/rs6000/7450.md: Adjust.
10860 * config/rs6000/7xx.md: Adjust.
10861 * config/rs6000/8540.md: Adjust.
10862 * config/rs6000/cell.md: Adjust.
10863 * config/rs6000/e300c2c3.md: Adjust.
10864 * config/rs6000/e500mc.md: Adjust.
10865 * config/rs6000/e500mc64.md: Adjust.
10866 * config/rs6000/e5500.md: Adjust.
10867 * config/rs6000/e6500.md: Adjust.
10868 * config/rs6000/mpc.md: Adjust.
10869 * config/rs6000/power4.md: Adjust.
10870 * config/rs6000/power5.md: Adjust.
10871 * config/rs6000/power6.md: Adjust.
10872 * config/rs6000/power7.md: Adjust.
10873 * config/rs6000/power8.md: Adjust.
10874 * config/rs6000/power9.md: Adjust.
10875 * config/rs6000/rs64.md: Adjust.
10876 * config/rs6000/titan.md: Adjust.
10878 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10880 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10881 ix86_indirect_branch_register logic.
10883 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10885 * config/i386/constraints.md (Bs): Update
10886 ix86_indirect_branch_register check. Don't check
10887 ix86_indirect_branch_register with GOT_memory_operand.
10889 * config/i386/predicates.md (GOT_memory_operand): Don't check
10890 ix86_indirect_branch_register here.
10891 (GOT32_symbol_operand): Likewise.
10893 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10895 * config/i386/predicates.md (constant_call_address_operand):
10896 Rewrite ix86_indirect_branch_register logic.
10897 (sibcall_insn_operand): Likewise.
10899 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10901 * config/i386/constraints.md (Bs): Replace
10902 ix86_indirect_branch_thunk_register with
10903 ix86_indirect_branch_register.
10905 * config/i386/i386.md (indirect_jump): Likewise.
10906 (tablejump): Likewise.
10907 (*sibcall_memory): Likewise.
10908 (*sibcall_value_memory): Likewise.
10909 Peepholes of indirect call and jump via memory: Likewise.
10910 * config/i386/i386.opt: Likewise.
10911 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10912 (GOT_memory_operand): Likewise.
10913 (call_insn_operand): Likewise.
10914 (sibcall_insn_operand): Likewise.
10915 (GOT32_symbol_operand): Likewise.
10917 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10919 PR middle-end/83837
10920 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10921 type rather than type addr's type points to.
10922 (expand_omp_atomic_mutex): Likewise.
10923 (expand_omp_atomic): Likewise.
10925 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10928 * config/i386/i386.c (output_indirect_thunk_function): Use
10929 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10930 for __x86_return_thunk.
10932 2018-01-15 Richard Biener <rguenther@suse.de>
10934 PR middle-end/83850
10935 * expmed.c (extract_bit_field_1): Fix typo.
10937 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10940 * config/arm/iterators.md (VF): New mode iterator.
10941 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10942 Remove integer-related logic from pattern.
10943 (neon_vabd<mode>_3): Likewise.
10945 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10947 PR middle-end/82694
10948 * common.opt (fstrict-overflow): No longer an alias.
10949 (fwrapv-pointer): New option.
10950 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10951 also for pointer types based on flag_wrapv_pointer.
10952 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10953 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10954 opts->x_flag_wrapv got set.
10955 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10956 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10957 POINTER_TYPE_OVERFLOW_UNDEFINED.
10958 * match.pd: Likewise in address comparison pattern.
10959 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10961 2018-01-15 Richard Biener <rguenther@suse.de>
10964 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10965 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10966 Reset type names to their identifier if their TYPE_DECL doesn't
10967 have linkage (and thus is used for ODR and devirt).
10968 (save_debug_info_for_decl): Remove.
10969 (save_debug_info_for_type): Likewise.
10970 (add_tree_to_fld_list): Adjust.
10971 * tree-pretty-print.c (dump_generic_node): Make dumping of
10972 type names more robust.
10974 2018-01-15 Richard Biener <rguenther@suse.de>
10976 * BASE-VER: Bump to 8.0.1.
10978 2018-01-14 Martin Sebor <msebor@redhat.com>
10981 * builtins.c (check_access): Avoid warning when the no-warning bit
10984 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10986 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10987 * ira-color (allocno_hard_regs_compare): Likewise.
10989 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10992 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10993 Use .pushsection/.popsection.
10995 2018-01-14 Martin Sebor <msebor@redhat.com>
10998 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11000 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11002 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11003 entry from extra_headers.
11004 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11005 extra_headers, make the list bitwise identical to the i?86-*-* one.
11007 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11009 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11010 -mcmodel=large with -mindirect-branch=thunk,
11011 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11012 -mfunction-return=thunk-extern.
11013 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11014 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11015 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11017 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11019 * config/i386/i386.c (print_reg): Print the name of the full
11020 integer register without '%'.
11021 (ix86_print_operand): Handle 'V'.
11022 * doc/extend.texi: Document 'V' modifier.
11024 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11026 * config/i386/constraints.md (Bs): Disallow memory operand for
11027 -mindirect-branch-register.
11029 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11030 (GOT_memory_operand): Likewise.
11031 (call_insn_operand): Likewise.
11032 (sibcall_insn_operand): Likewise.
11033 (GOT32_symbol_operand): Likewise.
11034 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11035 for -mindirect-branch-register.
11036 (tablejump): Likewise.
11037 (*sibcall_memory): Likewise.
11038 (*sibcall_value_memory): Likewise.
11039 Disallow peepholes of indirect call and jump via memory for
11040 -mindirect-branch-register.
11041 (*call_pop): Replace m with Bw.
11042 (*call_value_pop): Likewise.
11043 (*sibcall_pop_memory): Replace m with Bs.
11044 * config/i386/i386.opt (mindirect-branch-register): New option.
11045 * doc/invoke.texi: Document -mindirect-branch-register option.
11047 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11049 * config/i386/i386-protos.h (ix86_output_function_return): New.
11050 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11051 set function_return_type.
11052 (indirect_thunk_name): Add ret_p to indicate thunk for function
11054 (output_indirect_thunk_function): Pass false to
11055 indirect_thunk_name.
11056 (ix86_output_indirect_branch_via_reg): Likewise.
11057 (ix86_output_indirect_branch_via_push): Likewise.
11058 (output_indirect_thunk_function): Create alias for function
11059 return thunk if regno < 0.
11060 (ix86_output_function_return): New function.
11061 (ix86_handle_fndecl_attribute): Handle function_return.
11062 (ix86_attribute_table): Add function_return.
11063 * config/i386/i386.h (machine_function): Add
11064 function_return_type.
11065 * config/i386/i386.md (simple_return_internal): Use
11066 ix86_output_function_return.
11067 (simple_return_internal_long): Likewise.
11068 * config/i386/i386.opt (mfunction-return=): New option.
11069 (indirect_branch): Mention -mfunction-return=.
11070 * doc/extend.texi: Document function_return function attribute.
11071 * doc/invoke.texi: Document -mfunction-return= option.
11073 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11075 * config/i386/i386-opts.h (indirect_branch): New.
11076 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11077 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11078 with local indirect jump when converting indirect call and jump.
11079 (ix86_set_indirect_branch_type): New.
11080 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11081 (indirectlabelno): New.
11082 (indirect_thunk_needed): Likewise.
11083 (indirect_thunk_bnd_needed): Likewise.
11084 (indirect_thunks_used): Likewise.
11085 (indirect_thunks_bnd_used): Likewise.
11086 (INDIRECT_LABEL): Likewise.
11087 (indirect_thunk_name): Likewise.
11088 (output_indirect_thunk): Likewise.
11089 (output_indirect_thunk_function): Likewise.
11090 (ix86_output_indirect_branch_via_reg): Likewise.
11091 (ix86_output_indirect_branch_via_push): Likewise.
11092 (ix86_output_indirect_branch): Likewise.
11093 (ix86_output_indirect_jmp): Likewise.
11094 (ix86_code_end): Call output_indirect_thunk_function if needed.
11095 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11097 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11098 (ix86_attribute_table): Add indirect_branch.
11099 * config/i386/i386.h (machine_function): Add indirect_branch_type
11100 and has_local_indirect_jump.
11101 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11103 (tablejump): Likewise.
11104 (*indirect_jump): Use ix86_output_indirect_jmp.
11105 (*tablejump_1): Likewise.
11106 (simple_return_indirect_internal): Likewise.
11107 * config/i386/i386.opt (mindirect-branch=): New option.
11108 (indirect_branch): New.
11111 (thunk-inline): Likewise.
11112 (thunk-extern): Likewise.
11113 * doc/extend.texi: Document indirect_branch function attribute.
11114 * doc/invoke.texi: Document -mindirect-branch= option.
11116 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11119 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11121 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11123 * ipa-inline.c (want_inline_small_function_p): Return false if
11124 inlining has already failed with CIF_FINAL_ERROR.
11125 (update_caller_keys): Call want_inline_small_function_p before
11127 (update_callee_keys): Likewise.
11129 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11131 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11133 (rs6000_quadword_masked_address_p): Likewise.
11134 (quad_aligned_load_p): Likewise.
11135 (quad_aligned_store_p): Likewise.
11136 (const_load_sequence_p): Add comment to describe the outer-most loop.
11137 (mimic_memory_attributes_and_flags): New function.
11138 (rs6000_gen_stvx): Likewise.
11139 (replace_swapped_aligned_store): Likewise.
11140 (rs6000_gen_lvx): Likewise.
11141 (replace_swapped_aligned_load): Likewise.
11142 (replace_swapped_load_constant): Capitalize argument name in
11143 comment describing this function.
11144 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11145 and stores that access quad-word aligned addresses and replace
11146 with stvx or lvx instructions when appropriate.
11147 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11148 New function prototype.
11149 (rs6000_quadword_masked_address_p): Likewise.
11150 (rs6000_gen_lvx): Likewise.
11151 (rs6000_gen_stvx): Likewise.
11152 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11153 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11154 when memory address is aligned.
11155 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11156 this split to select lvx instruction when memory address is aligned.
11157 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11158 instruction when memory address is aligned.
11159 (*vsx_le_perm_load_v16qi): Likewise.
11160 (four unnamed splitters): Modify to select the stvx instruction
11161 when memory is aligned.
11163 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11165 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11166 which appears in the queue multiple times.
11168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11169 Alan Hayward <alan.hayward@arm.com>
11170 David Sherwood <david.sherwood@arm.com>
11172 * tree-vectorizer.h (vec_lower_bound): New structure.
11173 (_loop_vec_info): Add check_nonzero and lower_bounds.
11174 (LOOP_VINFO_CHECK_NONZERO): New macro.
11175 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11176 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11177 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11178 fields. Make seg_len the distance travelled, not including the
11180 (dr_direction_indicator): Declare.
11181 (dr_zero_step_indicator): Likewise.
11182 (dr_known_forward_stride_p): Likewise.
11183 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11185 (runtime_alias_check_p): Allow runtime alias checks with
11187 (operator ==): Compare access_size and align.
11188 (prune_runtime_alias_test_list): Rework for new distinction between
11189 the access_size and seg_len.
11190 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11192 (get_segment_min_max): New function.
11193 (create_intersect_range_checks): Use it.
11194 (dr_step_indicator): New function.
11195 (dr_direction_indicator): Likewise.
11196 (dr_zero_step_indicator): Likewise.
11197 (dr_known_forward_stride_p): Likewise.
11198 * tree-loop-distribution.c (data_ref_segment_size): Return
11199 DR_STEP * (niters - 1).
11200 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11202 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11203 (vect_preserves_scalar_order_p): New function, split out from...
11204 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11205 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11206 (vect_vfa_access_size): New function.
11207 (vect_vfa_align): Likewise.
11208 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11209 (dump_lower_bound): New function.
11210 (vect_check_lower_bound): Likewise.
11211 (vect_small_gap_p): Likewise.
11212 (vectorizable_with_step_bound_p): Likewise.
11213 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11214 depencies if the vectorization factor is 1. Convert the checks
11215 for nonzero steps into checks on the bounds of DR_STEP. Try using
11216 a bunds check for variable steps if the minimum required step is
11217 relatively small. Update calls to the dr_with_seg_len
11218 constructor and to vect_compile_time_alias.
11219 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11221 (vect_loop_versioning): Call it.
11222 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11224 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11227 Alan Hayward <alan.hayward@arm.com>
11228 David Sherwood <david.sherwood@arm.com>
11230 * doc/sourcebuild.texi (vect_scatter_store): Document.
11231 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11233 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11235 * genopinit.c (main): Add supports_vec_scatter_store and
11236 supports_vec_scatter_store_cached to target_optabs.
11237 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11238 IFN_MASK_SCATTER_STORE.
11239 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11241 * internal-fn.h (internal_store_fn_p): Declare.
11242 (internal_fn_stored_value_index): Likewise.
11243 * internal-fn.c (scatter_store_direct): New macro.
11244 (expand_scatter_store_optab_fn): New function.
11245 (direct_scatter_store_optab_supported_p): New macro.
11246 (internal_store_fn_p): New function.
11247 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11248 IFN_MASK_SCATTER_STORE.
11249 (internal_fn_mask_index): Likewise.
11250 (internal_fn_stored_value_index): New function.
11251 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11252 for scatter stores.
11253 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11254 * optabs-query.c (supports_vec_scatter_store_p): New function.
11255 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11256 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11257 true for scatter stores.
11258 (vect_gather_scatter_fn_p): Handle scatter stores too.
11259 (vect_check_gather_scatter): Consider using scatter stores if
11260 supports_vec_scatter_store_p.
11261 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11262 scatter stores too.
11263 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11264 internal_fn_stored_value_index.
11265 (check_load_store_masking): Handle scatter stores too.
11266 (vect_get_store_rhs): Make public.
11267 (vectorizable_call): Use internal_store_fn_p.
11268 (vectorizable_store): Handle scatter store internal functions.
11269 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11270 when deciding whether the end of the group has been reached.
11271 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11272 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11273 (mask_scatter_store<mode>): New insns.
11275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11276 Alan Hayward <alan.hayward@arm.com>
11277 David Sherwood <david.sherwood@arm.com>
11279 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11280 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11281 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11283 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11284 Use vect_truncate_gather_scatter_offset if we can't treat the
11285 operation as a normal gather load or scatter store.
11286 (get_group_load_store_type): Take the gather_scatter_info
11287 as argument. Try using a gather load or scatter store for
11288 single-element groups.
11289 (get_load_store_type): Update calls to get_group_load_store_type
11290 and vect_use_strided_gather_scatters_p.
11292 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11293 Alan Hayward <alan.hayward@arm.com>
11294 David Sherwood <david.sherwood@arm.com>
11296 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11297 optional tree argument.
11298 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11300 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11301 but continue to use the current value as a fallback.
11302 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11303 to compare the updates.
11304 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11305 (get_load_store_type): Use it when handling a strided access.
11306 (vect_get_strided_load_store_ops): New function.
11307 (vect_get_data_ptr_increment): Likewise.
11308 (vectorizable_load): Handle strided gather loads. Always pass
11309 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11312 Alan Hayward <alan.hayward@arm.com>
11313 David Sherwood <david.sherwood@arm.com>
11315 * doc/md.texi (gather_load@var{m}): Document.
11316 (mask_gather_load@var{m}): Likewise.
11317 * genopinit.c (main): Add supports_vec_gather_load and
11318 supports_vec_gather_load_cached to target_optabs.
11319 * optabs-tree.c (init_tree_optimization_optabs): Use
11320 ggc_cleared_alloc to allocate target_optabs.
11321 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11322 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11324 * internal-fn.h (internal_load_fn_p): Declare.
11325 (internal_gather_scatter_fn_p): Likewise.
11326 (internal_fn_mask_index): Likewise.
11327 (internal_gather_scatter_fn_supported_p): Likewise.
11328 * internal-fn.c (gather_load_direct): New macro.
11329 (expand_gather_load_optab_fn): New function.
11330 (direct_gather_load_optab_supported_p): New macro.
11331 (direct_internal_fn_optab): New function.
11332 (internal_load_fn_p): Likewise.
11333 (internal_gather_scatter_fn_p): Likewise.
11334 (internal_fn_mask_index): Likewise.
11335 (internal_gather_scatter_fn_supported_p): Likewise.
11336 * optabs-query.c (supports_at_least_one_mode_p): New function.
11337 (supports_vec_gather_load_p): Likewise.
11338 * optabs-query.h (supports_vec_gather_load_p): Declare.
11339 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11340 and memory_type field.
11341 (NUM_PATTERNS): Bump to 15.
11342 * tree-vect-data-refs.c: Include internal-fn.h.
11343 (vect_gather_scatter_fn_p): New function.
11344 (vect_describe_gather_scatter_call): Likewise.
11345 (vect_check_gather_scatter): Try using internal functions for
11346 gather loads. Recognize existing calls to a gather load function.
11347 (vect_analyze_data_refs): Consider using gather loads if
11348 supports_vec_gather_load_p.
11349 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11350 (vect_get_gather_scatter_offset_type): Likewise.
11351 (vect_convert_mask_for_vectype): Likewise.
11352 (vect_add_conversion_to_patterm): Likewise.
11353 (vect_try_gather_scatter_pattern): Likewise.
11354 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11355 (vect_vect_recog_func_ptrs): Add it.
11356 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11357 internal_fn_mask_index and internal_gather_scatter_fn_p.
11358 (check_load_store_masking): Take the gather_scatter_info as an
11359 argument and handle gather loads.
11360 (vect_get_gather_scatter_ops): New function.
11361 (vectorizable_call): Check internal_load_fn_p.
11362 (vectorizable_load): Likewise. Handle gather load internal
11364 (vectorizable_store): Update call to check_load_store_masking.
11365 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11366 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11367 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11368 (aarch64_gather_scale_operand_d): New predicates.
11369 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11370 (mask_gather_load<mode>): New insns.
11372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11373 Alan Hayward <alan.hayward@arm.com>
11374 David Sherwood <david.sherwood@arm.com>
11376 * optabs.def (fold_left_plus_optab): New optab.
11377 * doc/md.texi (fold_left_plus_@var{m}): Document.
11378 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11379 * internal-fn.c (fold_left_direct): Define.
11380 (expand_fold_left_optab_fn): Likewise.
11381 (direct_fold_left_optab_supported_p): Likewise.
11382 * fold-const-call.c (fold_const_fold_left): New function.
11383 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11384 * tree-parloops.c (valid_reduction_p): New function.
11385 (gather_scalar_reductions): Use it.
11386 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11387 (vect_finish_replace_stmt): Declare.
11388 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11389 (needs_fold_left_reduction_p): New function, split out from...
11390 (vect_is_simple_reduction): ...here. Accept reductions that
11391 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11392 (vect_force_simple_reduction): Also store the reduction type in
11393 the assignment's STMT_VINFO_REDUC_TYPE.
11394 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11395 (merge_with_identity): New function.
11396 (vect_expand_fold_left): Likewise.
11397 (vectorize_fold_left_reduction): Likewise.
11398 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11399 scalar phi in place for it. Check for target support and reject
11400 cases that would reassociate the operation. Defer the transform
11401 phase to vectorize_fold_left_reduction.
11402 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11403 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11404 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11408 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11409 call to ifc_temp_var.
11411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11412 Alan Hayward <alan.hayward@arm.com>
11413 David Sherwood <david.sherwood@arm.com>
11415 * target.def (legitimize_address_displacement): Take the original
11416 offset as a poly_int.
11417 * targhooks.h (default_legitimize_address_displacement): Update
11419 * targhooks.c (default_legitimize_address_displacement): Likewise.
11420 * doc/tm.texi: Regenerate.
11421 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11422 as an argument, moving assert of ad->disp == ad->disp_term to...
11423 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11424 Try calling targetm.legitimize_address_displacement before expanding
11425 the address rather than afterwards, and adjust for the new interface.
11426 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11427 Match the new hook interface. Handle SVE addresses.
11428 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11429 new hook interface.
11431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11433 * Makefile.in (OBJS): Add early-remat.o.
11434 * target.def (select_early_remat_modes): New hook.
11435 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11436 * doc/tm.texi: Regenerate.
11437 * targhooks.h (default_select_early_remat_modes): Declare.
11438 * targhooks.c (default_select_early_remat_modes): New function.
11439 * timevar.def (TV_EARLY_REMAT): New timevar.
11440 * passes.def (pass_early_remat): New pass.
11441 * tree-pass.h (make_pass_early_remat): Declare.
11442 * early-remat.c: New file.
11443 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11445 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11447 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11448 Alan Hayward <alan.hayward@arm.com>
11449 David Sherwood <david.sherwood@arm.com>
11451 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11452 vfm1 with a bound_epilog parameter.
11453 (vect_do_peeling): Update calls accordingly, and move the prologue
11454 call earlier in the function. Treat the base bound_epilog as 0 for
11455 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11456 this base when peeling for gaps.
11457 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11458 with fully-masked loops.
11459 (vect_estimate_min_profitable_iters): Handle the single peeled
11460 iteration in that case.
11462 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11463 Alan Hayward <alan.hayward@arm.com>
11464 David Sherwood <david.sherwood@arm.com>
11466 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11467 single-element interleaving even if the size is not a power of 2.
11468 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11469 accesses for single-element interleaving if the group size is
11472 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11473 Alan Hayward <alan.hayward@arm.com>
11474 David Sherwood <david.sherwood@arm.com>
11476 * doc/md.texi (fold_extract_last_@var{m}): Document.
11477 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11478 * optabs.def (fold_extract_last_optab): New optab.
11479 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11480 * internal-fn.c (fold_extract_direct): New macro.
11481 (expand_fold_extract_optab_fn): Likewise.
11482 (direct_fold_extract_optab_supported_p): Likewise.
11483 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11484 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11485 EXTRACT_LAST_REDUCTION.
11486 (get_initial_def_for_reduction): Do not create an initial vector
11487 for EXTRACT_LAST_REDUCTION reductions.
11488 (vectorizable_reduction): Leave the scalar phi in place for
11489 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11490 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11491 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11492 transform phase to vectorizable_condition.
11493 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11495 (vect_finish_stmt_generation): ...here.
11496 (vect_finish_replace_stmt): New function.
11497 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11498 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11500 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11503 Alan Hayward <alan.hayward@arm.com>
11504 David Sherwood <david.sherwood@arm.com>
11506 * doc/md.texi (extract_last_@var{m}): Document.
11507 * optabs.def (extract_last_optab): New optab.
11508 * internal-fn.def (EXTRACT_LAST): New internal function.
11509 * internal-fn.c (cond_unary_direct): New macro.
11510 (expand_cond_unary_optab_fn): Likewise.
11511 (direct_cond_unary_optab_supported_p): Likewise.
11512 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11513 loops using EXTRACT_LAST.
11514 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11515 (extract_last_<mode>): ...this optab.
11516 (vec_extract<mode><Vel>): Update accordingly.
11518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11519 Alan Hayward <alan.hayward@arm.com>
11520 David Sherwood <david.sherwood@arm.com>
11522 * target.def (empty_mask_is_expensive): New hook.
11523 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11524 * doc/tm.texi: Regenerate.
11525 * targhooks.h (default_empty_mask_is_expensive): Declare.
11526 * targhooks.c (default_empty_mask_is_expensive): New function.
11527 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11528 if the target says that empty masks are expensive.
11529 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11531 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11534 Alan Hayward <alan.hayward@arm.com>
11535 David Sherwood <david.sherwood@arm.com>
11537 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11538 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11539 (vect_use_loop_mask_for_alignment_p): New function.
11540 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11541 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11542 niters_skip argument. Make sure that the first niters_skip elements
11543 of the first iteration are inactive.
11544 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11545 Update call to vect_set_loop_masks_directly.
11546 (get_misalign_in_elems): New function, split out from...
11547 (vect_gen_prolog_loop_niters): ...here.
11548 (vect_update_init_of_dr): Take a code argument that specifies whether
11549 the adjustment should be added or subtracted.
11550 (vect_update_init_of_drs): Likewise.
11551 (vect_prepare_for_masked_peels): New function.
11552 (vect_do_peeling): Skip prologue peeling if we're using a mask
11553 instead. Update call to vect_update_inits_of_drs.
11554 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11556 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11557 alignment. Do not include the number of peeled iterations in
11558 the minimum threshold in that case.
11559 (vectorizable_induction): Adjust the start value down by
11560 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11561 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11562 Take the number of skipped iterations into account when calculating
11564 * tree-vect-stmts.c (vect_gen_while_not): New function.
11566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11567 Alan Hayward <alan.hayward@arm.com>
11568 David Sherwood <david.sherwood@arm.com>
11570 * doc/sourcebuild.texi (vect_fully_masked): Document.
11571 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11572 default value to 0.
11573 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11575 (vect_analyze_loop_2): ...here. Don't check the vectorization
11576 factor against the number of loop iterations if the loop is
11579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11580 Alan Hayward <alan.hayward@arm.com>
11581 David Sherwood <david.sherwood@arm.com>
11583 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11584 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11585 (dump_groups): Update accordingly.
11586 (iv_use::mem_type): New member variable.
11587 (address_p): New function.
11588 (record_use): Add a mem_type argument and initialize the new
11590 (record_group_use): Add a mem_type argument. Use address_p.
11591 Remove obsolete null checks of base_object. Update call to record_use.
11592 (find_interesting_uses_op): Update call to record_group_use.
11593 (find_interesting_uses_cond): Likewise.
11594 (find_interesting_uses_address): Likewise.
11595 (get_mem_type_for_internal_fn): New function.
11596 (find_address_like_use): Likewise.
11597 (find_interesting_uses_stmt): Try find_address_like_use before
11598 calling find_interesting_uses_op.
11599 (addr_offset_valid_p): Use the iv mem_type field as the type
11600 of the addressed memory.
11601 (add_autoinc_candidates): Likewise.
11602 (get_address_cost): Likewise.
11603 (split_small_address_groups_p): Use address_p.
11604 (split_address_groups): Likewise.
11605 (add_iv_candidate_for_use): Likewise.
11606 (autoinc_possible_for_pair): Likewise.
11607 (rewrite_groups): Likewise.
11608 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11609 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11610 (get_alias_ptr_type_for_ptr_address): New function.
11611 (rewrite_use_address): Rewrite address uses in calls that were
11612 identified by find_address_like_use.
11614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11615 Alan Hayward <alan.hayward@arm.com>
11616 David Sherwood <david.sherwood@arm.com>
11618 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11620 * gimple-expr.h (is_gimple_addressable: Likewise.
11621 * gimple-expr.c (is_gimple_address): Likewise.
11622 * internal-fn.c (expand_call_mem_ref): New function.
11623 (expand_mask_load_optab_fn): Use it.
11624 (expand_mask_store_optab_fn): Likewise.
11626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11627 Alan Hayward <alan.hayward@arm.com>
11628 David Sherwood <david.sherwood@arm.com>
11630 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11631 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11632 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11633 (cond_umax@var{mode}): Document.
11634 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11635 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11636 (cond_umin_optab, cond_umax_optab): New optabs.
11637 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11638 (COND_IOR, COND_XOR): New internal functions.
11639 * internal-fn.h (get_conditional_internal_fn): Declare.
11640 * internal-fn.c (cond_binary_direct): New macro.
11641 (expand_cond_binary_optab_fn): Likewise.
11642 (direct_cond_binary_optab_supported_p): Likewise.
11643 (get_conditional_internal_fn): New function.
11644 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11645 Cope with reduction statements that are vectorized as calls rather
11647 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11648 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11649 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11650 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11651 (UNSPEC_COND_EOR): New unspecs.
11652 (optab): Add mappings for them.
11653 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11654 (sve_int_op, sve_fp_op): New int attributes.
11656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11657 Alan Hayward <alan.hayward@arm.com>
11658 David Sherwood <david.sherwood@arm.com>
11660 * optabs.def (while_ult_optab): New optab.
11661 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11662 * internal-fn.def (WHILE_ULT): New internal function.
11663 * internal-fn.h (direct_internal_fn_supported_p): New override
11664 that takes two types as argument.
11665 * internal-fn.c (while_direct): New macro.
11666 (expand_while_optab_fn): New function.
11667 (convert_optab_supported_p): Likewise.
11668 (direct_while_optab_supported_p): New macro.
11669 * wide-int.h (wi::udiv_ceil): New function.
11670 * tree-vectorizer.h (rgroup_masks): New structure.
11671 (vec_loop_masks): New typedef.
11672 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11673 and fully_masked_p.
11674 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11675 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11676 (vect_max_vf): New function.
11677 (slpeel_make_loop_iterate_ntimes): Delete.
11678 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11679 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11680 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11681 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11682 internal-fn.h, stor-layout.h and optabs-query.h.
11683 (vect_set_loop_mask): New function.
11684 (add_preheader_seq): Likewise.
11685 (add_header_seq): Likewise.
11686 (interleave_supported_p): Likewise.
11687 (vect_maybe_permute_loop_masks): Likewise.
11688 (vect_set_loop_masks_directly): Likewise.
11689 (vect_set_loop_condition_masked): Likewise.
11690 (vect_set_loop_condition_unmasked): New function, split out from
11691 slpeel_make_loop_iterate_ntimes.
11692 (slpeel_make_loop_iterate_ntimes): Rename to..
11693 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11694 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11695 (vect_do_peeling): Update call accordingly.
11696 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11698 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11699 mask_compare_type, can_fully_mask_p and fully_masked_p.
11700 (release_vec_loop_masks): New function.
11701 (_loop_vec_info): Use it to free the loop masks.
11702 (can_produce_all_loop_masks_p): New function.
11703 (vect_get_max_nscalars_per_iter): Likewise.
11704 (vect_verify_full_masking): Likewise.
11705 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11706 retries, and free the mask rgroups before retrying. Check loop-wide
11707 reasons for disallowing fully-masked loops. Make the final decision
11708 about whether use a fully-masked loop or not.
11709 (vect_estimate_min_profitable_iters): Do not assume that peeling
11710 for the number of iterations will be needed for fully-masked loops.
11711 (vectorizable_reduction): Disable fully-masked loops.
11712 (vectorizable_live_operation): Likewise.
11713 (vect_halve_mask_nunits): New function.
11714 (vect_double_mask_nunits): Likewise.
11715 (vect_record_loop_mask): Likewise.
11716 (vect_get_loop_mask): Likewise.
11717 (vect_transform_loop): Handle the case in which the final loop
11718 iteration might handle a partial vector. Call vect_set_loop_condition
11719 instead of slpeel_make_loop_iterate_ntimes.
11720 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11721 (check_load_store_masking): New function.
11722 (prepare_load_store_mask): Likewise.
11723 (vectorizable_store): Handle fully-masked loops.
11724 (vectorizable_load): Likewise.
11725 (supportable_widening_operation): Use vect_halve_mask_nunits for
11727 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11728 (vect_gen_while): New function.
11729 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11730 (aarch64_uqdec<mode>): New insn.
11732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11733 Alan Hayward <alan.hayward@arm.com>
11734 David Sherwood <david.sherwood@arm.com>
11736 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11737 (reduc_xor_scal_optab): New optabs.
11738 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11739 (reduc_xor_scal_@var{m}): Document.
11740 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11741 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11742 internal functions.
11743 * fold-const-call.c (fold_const_call): Handle them.
11744 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11745 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11746 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11747 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11748 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11749 (UNSPEC_XORV): New unspecs.
11750 (optab): Add entries for them.
11751 (BITWISEV): New int iterator.
11752 (bit_reduc_op): New int attributes.
11754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11755 Alan Hayward <alan.hayward@arm.com>
11756 David Sherwood <david.sherwood@arm.com>
11758 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11759 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11760 * optabs.def (vec_shl_insert_optab): New optab.
11761 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11762 (duplicate_and_interleave): Likewise.
11763 * tree-vect-loop.c: Include internal-fn.h.
11764 (neutral_op_for_slp_reduction): New function, split out from
11765 get_initial_defs_for_reduction.
11766 (get_initial_def_for_reduction): Handle option 2 for variable-length
11767 vectors by loading the neutral value into a vector and then shifting
11768 the initial value into element 0.
11769 (get_initial_defs_for_reduction): Replace the code argument with
11770 the neutral value calculated by neutral_op_for_slp_reduction.
11771 Use gimple_build_vector for constant-length vectors.
11772 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11773 but the first group_size elements have a neutral value.
11774 Use duplicate_and_interleave otherwise.
11775 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11776 Update call to get_initial_defs_for_reduction. Handle SLP
11777 reductions for variable-length vectors by creating one vector
11778 result for each scalar result, with the elements associated
11779 with other scalar results stubbed out with the neutral value.
11780 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11781 Require IFN_VEC_SHL_INSERT for double reductions on
11782 variable-length vectors, or SLP reductions that have
11783 a neutral value. Require can_duplicate_and_interleave_p
11784 support for variable-length unchained SLP reductions if there
11785 is no neutral value, such as for MIN/MAX reductions. Also require
11786 the number of vector elements to be a multiple of the number of
11787 SLP statements when doing variable-length unchained SLP reductions.
11788 Update call to vect_create_epilog_for_reduction.
11789 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11790 and remove initial values.
11791 (duplicate_and_interleave): Make public.
11792 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11793 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11796 Alan Hayward <alan.hayward@arm.com>
11797 David Sherwood <david.sherwood@arm.com>
11799 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11800 (can_duplicate_and_interleave_p): New function.
11801 (vect_get_and_check_slp_defs): Take the vector of statements
11802 rather than just the current one. Remove excess parentheses.
11803 Restriction rejectinon of vect_constant_def and vect_external_def
11804 for variable-length vectors to boolean types, or types for which
11805 can_duplicate_and_interleave_p is false.
11806 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11807 (duplicate_and_interleave): New function.
11808 (vect_get_constant_vectors): Use gimple_build_vector for
11809 constant-length vectors and suitable variable-length constant
11810 vectors. Use duplicate_and_interleave for other variable-length
11811 vectors. Don't defer the update when inserting new statements.
11813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11814 Alan Hayward <alan.hayward@arm.com>
11815 David Sherwood <david.sherwood@arm.com>
11817 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11818 min_profitable_iters doesn't go negative.
11820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11821 Alan Hayward <alan.hayward@arm.com>
11822 David Sherwood <david.sherwood@arm.com>
11824 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11825 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11826 * optabs.def (vec_mask_load_lanes_optab): New optab.
11827 (vec_mask_store_lanes_optab): Likewise.
11828 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11829 (MASK_STORE_LANES): Likewise.
11830 * internal-fn.c (mask_load_lanes_direct): New macro.
11831 (mask_store_lanes_direct): Likewise.
11832 (expand_mask_load_optab_fn): Handle masked operations.
11833 (expand_mask_load_lanes_optab_fn): New macro.
11834 (expand_mask_store_optab_fn): Handle masked operations.
11835 (expand_mask_store_lanes_optab_fn): New macro.
11836 (direct_mask_load_lanes_optab_supported_p): Likewise.
11837 (direct_mask_store_lanes_optab_supported_p): Likewise.
11838 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11840 (vect_load_lanes_supported): Likewise.
11841 * tree-vect-data-refs.c (strip_conversion): New function.
11842 (can_group_stmts_p): Likewise.
11843 (vect_analyze_data_ref_accesses): Use it instead of checking
11844 for a pair of assignments.
11845 (vect_store_lanes_supported): Take a masked_p parameter.
11846 (vect_load_lanes_supported): Likewise.
11847 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11848 vect_store_lanes_supported and vect_load_lanes_supported.
11849 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11850 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11851 parameter. Don't allow gaps for masked accesses.
11852 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11853 and vect_load_lanes_supported.
11854 (get_load_store_type): Take a masked_p parameter and update
11855 call to get_group_load_store_type.
11856 (vectorizable_store): Update call to get_load_store_type.
11857 Handle IFN_MASK_STORE_LANES.
11858 (vectorizable_load): Update call to get_load_store_type.
11859 Handle IFN_MASK_LOAD_LANES.
11861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11862 Alan Hayward <alan.hayward@arm.com>
11863 David Sherwood <david.sherwood@arm.com>
11865 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11867 * config/aarch64/aarch64-protos.h
11868 (aarch64_sve_struct_memory_operand_p): Declare.
11869 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11870 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11871 (VPRED, vpred): Handle SVE structure modes.
11872 * config/aarch64/constraints.md (Utx): New constraint.
11873 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11874 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11875 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11876 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11877 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11878 structure modes. Split into pieces after RA.
11879 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11880 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11882 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11883 SVE structure modes.
11884 (aarch64_classify_address): Likewise.
11885 (sizetochar): Move earlier in file.
11886 (aarch64_print_operand): Handle SVE register lists.
11887 (aarch64_array_mode): New function.
11888 (aarch64_sve_struct_memory_operand_p): Likewise.
11889 (TARGET_ARRAY_MODE): Redefine.
11891 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11892 Alan Hayward <alan.hayward@arm.com>
11893 David Sherwood <david.sherwood@arm.com>
11895 * target.def (array_mode): New target hook.
11896 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11897 * doc/tm.texi: Regenerate.
11898 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11899 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11900 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11901 targetm.array_mode.
11902 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11906 Alan Hayward <alan.hayward@arm.com>
11907 David Sherwood <david.sherwood@arm.com>
11909 * fold-const.c (fold_binary_loc): Check the argument types
11910 rather than the result type when testing for a vector operation.
11912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11914 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11915 * doc/tm.texi: Regenerate.
11917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11918 Alan Hayward <alan.hayward@arm.com>
11919 David Sherwood <david.sherwood@arm.com>
11921 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11922 (sve): Document new AArch64 extension.
11923 * doc/md.texi (w): Extend the description of the AArch64
11924 constraint to include SVE vectors.
11925 (Upl, Upa): Document new AArch64 predicate constraints.
11926 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11928 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11929 (msve-vector-bits=): New option.
11930 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11931 SVE when these are disabled.
11932 (sve): New extension.
11933 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11934 modes. Adjust their number of units based on aarch64_sve_vg.
11935 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11936 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11937 aarch64_addr_query_type.
11938 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11939 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11940 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11941 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11942 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11943 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11944 (aarch64_simd_imm_zero_p): Delete.
11945 (aarch64_check_zero_based_sve_index_immediate): Declare.
11946 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11947 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11948 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11949 (aarch64_sve_float_mul_immediate_p): Likewise.
11950 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11951 rather than an rtx.
11952 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11953 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11954 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11955 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11956 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11957 (aarch64_regmode_natural_size): Likewise.
11958 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11959 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11961 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11962 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11963 for VG and the SVE predicate registers.
11964 (V_ALIASES): Add a "z"-prefixed alias.
11965 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11966 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11967 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11968 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11969 (REG_CLASS_NAMES): Add entries for them.
11970 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11971 and the predicate registers.
11972 (aarch64_sve_vg): Declare.
11973 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11974 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11975 (REGMODE_NATURAL_SIZE): Define.
11976 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11978 * config/aarch64/aarch64.c: Include cfgrtl.h.
11979 (simd_immediate_info): Add a constructor for series vectors,
11980 and an associated step field.
11981 (aarch64_sve_vg): New variable.
11982 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11983 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11984 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11985 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11986 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11987 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11988 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11989 (aarch64_get_mask_mode): New functions.
11990 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11991 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11992 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11993 predicate modes and predicate registers. Explicitly restrict
11994 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11995 to store a vector mode if it is recognized by
11996 aarch64_classify_vector_mode.
11997 (aarch64_regmode_natural_size): New function.
11998 (aarch64_hard_regno_caller_save_mode): Return the original mode
12000 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12001 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12002 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12003 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12005 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12006 does not overlap dest if the function is frame-related. Handle
12008 (aarch64_split_add_offset): New function.
12009 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12010 them aarch64_add_offset.
12011 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12012 and update call to aarch64_sub_sp.
12013 (aarch64_add_cfa_expression): New function.
12014 (aarch64_expand_prologue): Pass extra temporary registers to the
12015 functions above. Handle the case in which we need to emit new
12016 DW_CFA_expressions for registers that were originally saved
12017 relative to the stack pointer, but now have to be expressed
12018 relative to the frame pointer.
12019 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12021 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12022 IP0 and IP1 values for SVE frames.
12023 (aarch64_expand_vec_series): New function.
12024 (aarch64_expand_sve_widened_duplicate): Likewise.
12025 (aarch64_expand_sve_const_vector): Likewise.
12026 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12027 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12028 into the register, rather than emitting a SET directly.
12029 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12030 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12031 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12032 (offset_9bit_signed_scaled_p): New functions.
12033 (aarch64_replicate_bitmask_imm): New function.
12034 (aarch64_bitmask_imm): Use it.
12035 (aarch64_cannot_force_const_mem): Reject expressions involving
12036 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12037 (aarch64_classify_index): Handle SVE indices, by requiring
12038 a plain register index with a scale that matches the element size.
12039 (aarch64_classify_address): Handle SVE addresses. Assert that
12040 the mode of the address is VOIDmode or an integer mode.
12041 Update call to aarch64_classify_symbol.
12042 (aarch64_classify_symbolic_expression): Update call to
12043 aarch64_classify_symbol.
12044 (aarch64_const_vec_all_in_range_p): New function.
12045 (aarch64_print_vector_float_operand): Likewise.
12046 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12047 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12048 and the FP immediates 1.0 and 0.5.
12049 (aarch64_print_address_internal): Handle SVE addresses.
12050 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12051 (aarch64_regno_regclass): Handle predicate registers.
12052 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12054 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12055 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12056 (aarch64_convert_sve_vector_bits): New function.
12057 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12058 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12059 rather than an rtx.
12060 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12061 Handle SVE vector and predicate modes. Accept VL-based constants
12062 that need only one temporary register, and VL offsets that require
12063 no temporary registers.
12064 (aarch64_conditional_register_usage): Mark the predicate registers
12065 as fixed if SVE isn't available.
12066 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12067 Return true for SVE vector and predicate modes.
12068 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12069 rather than an unsigned int. Handle SVE modes.
12070 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12072 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12074 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12075 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12076 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12077 (aarch64_sve_float_mul_immediate_p): New functions.
12078 (aarch64_sve_valid_immediate): New function.
12079 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12080 Explicitly reject structure modes. Check for INDEX constants.
12081 Handle PTRUE and PFALSE constants.
12082 (aarch64_check_zero_based_sve_index_immediate): New function.
12083 (aarch64_simd_imm_zero_p): Delete.
12084 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12085 vector modes. Accept constants in the range of CNT[BHWD].
12086 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12087 ask for an Advanced SIMD mode.
12088 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12089 (aarch64_simd_vector_alignment): Handle SVE predicates.
12090 (aarch64_vectorize_preferred_vector_alignment): New function.
12091 (aarch64_simd_vector_alignment_reachable): Use it instead of
12093 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12094 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12096 (MAX_VECT_LEN): Delete.
12097 (expand_vec_perm_d): Add a vec_flags field.
12098 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12099 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12100 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12102 (aarch64_evpc_rev): Rename to...
12103 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12104 (aarch64_evpc_rev_global): New function.
12105 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12106 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12108 (aarch64_evpc_sve_tbl): New function.
12109 (aarch64_expand_vec_perm_const_1): Update after rename of
12110 aarch64_evpc_rev. Handle SVE permutes too, trying
12111 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12112 than aarch64_evpc_tbl.
12113 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12114 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12115 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12116 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12117 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12118 (aarch64_expand_sve_vcond): New functions.
12119 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12120 of aarch64_vector_mode_p.
12121 (aarch64_dwarf_poly_indeterminate_value): New function.
12122 (aarch64_compute_pressure_classes): Likewise.
12123 (aarch64_can_change_mode_class): Likewise.
12124 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12125 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12126 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12127 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12128 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12129 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12130 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12131 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12133 (Dn, Dl, Dr): Accept const as well as const_vector.
12134 (Dz): Likewise. Compare against CONST0_RTX.
12135 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12136 of "vector" where appropriate.
12137 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12138 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12139 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12140 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12141 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12142 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12143 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12144 (v_int_equiv): Extend to SVE modes.
12145 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12147 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12148 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12149 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12150 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12151 (SVE_COND_FP_CMP): New int iterators.
12152 (perm_hilo): Handle the new unpack unspecs.
12153 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12155 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12156 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12157 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12158 (aarch64_equality_operator, aarch64_constant_vector_operand)
12159 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12160 (aarch64_sve_nonimmediate_operand): Likewise.
12161 (aarch64_sve_general_operand): Likewise.
12162 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12163 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12164 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12165 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12166 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12167 (aarch64_sve_float_arith_immediate): Likewise.
12168 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12169 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12170 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12171 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12172 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12173 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12174 (aarch64_sve_float_arith_operand): Likewise.
12175 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12176 (aarch64_sve_float_mul_operand): Likewise.
12177 (aarch64_sve_vec_perm_operand): Likewise.
12178 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12179 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12180 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12181 as well as const_vector.
12182 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12183 in file. Use CONST0_RTX and CONSTM1_RTX.
12184 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12185 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12186 Use aarch64_simd_imm_zero.
12187 * config/aarch64/aarch64-sve.md: New file.
12188 * config/aarch64/aarch64.md: Include it.
12189 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12190 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12191 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12192 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12193 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12194 (sve): New attribute.
12195 (enabled): Disable instructions with the sve attribute unless
12197 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12198 aarch64_expand_mov_immediate.
12199 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12200 CNT[BHSD] immediates.
12201 (movti): Split CONST_POLY_INT moves into two halves.
12202 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12203 Split additions that need a temporary here if the destination
12204 is the stack pointer.
12205 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12206 (*add<mode>3_poly_1): New instruction.
12207 (set_clobber_cc): New expander.
12209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12211 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12212 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12213 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12214 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12215 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12216 Change innermode from fixed_mode_size to machine_mode.
12217 (simplify_subreg): Update call accordingly. Handle a constant-sized
12218 subreg of a variable-length CONST_VECTOR.
12220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12221 Alan Hayward <alan.hayward@arm.com>
12222 David Sherwood <david.sherwood@arm.com>
12224 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12225 (add_offset_to_base): New function, split out from...
12226 (create_mem_ref): ...here. When handling a scale other than 1,
12227 check first whether the address is valid without the offset.
12228 Add it into the base if so, leaving the index and scale as-is.
12230 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12233 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12234 fold_for_warn before checking if arg2 is INTEGER_CST.
12236 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12238 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12239 (store_multiple_operation): Delete.
12240 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12241 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12242 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12243 guarded by TARGET_STRING.
12244 (rs6000_output_load_multiple): Delete.
12245 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12246 OPTION_MASK_STRING / TARGET_STRING handling.
12247 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12248 (const rs6000_opt_masks) <"string">: Change mask to 0.
12249 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12250 (MASK_STRING): Delete.
12251 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12253 (load_multiple): Delete.
12260 (store_multiple): Delete.
12267 (movmemsi_8reg): Delete.
12268 (corresponding unnamed define_insn): Delete.
12269 (movmemsi_6reg): Delete.
12270 (corresponding unnamed define_insn): Delete.
12271 (movmemsi_4reg): Delete.
12272 (corresponding unnamed define_insn): Delete.
12273 (movmemsi_2reg): Delete.
12274 (corresponding unnamed define_insn): Delete.
12275 (movmemsi_1reg): Delete.
12276 (corresponding unnamed define_insn): Delete.
12277 * config/rs6000/rs6000.opt (mno-string): New.
12278 (mstring): Replace by deprecation warning stub.
12279 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12281 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12283 * regrename.c (regrename_do_replace): If replacing the same
12284 reg multiple times, try to reuse last created gen_raw_REG.
12287 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12288 main to workaround a bug in GDB.
12290 2018-01-12 Tom de Vries <tom@codesourcery.com>
12293 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12295 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12297 PR rtl-optimization/80481
12298 * ira-color.c (get_cap_member): New function.
12299 (allocnos_conflict_by_live_ranges_p): Use it.
12300 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12301 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12303 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12306 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12307 (*saddl_se_1): Ditto.
12308 (*ssubsi_1): Ditto.
12309 (*ssubl_se_1): Ditto.
12311 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12313 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12314 rather than wi::to_widest for DR_INITs.
12315 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12316 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12317 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12319 (vect_analyze_group_access_1): Note that here.
12321 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12323 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12324 polynomial type sizes.
12326 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12328 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12329 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12330 (gimple_add_tmp_var): Likewise.
12332 2018-01-12 Martin Liska <mliska@suse.cz>
12334 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12335 (gimple_alloc_sizes): Likewise.
12336 (dump_gimple_statistics): Use PRIu64 in printf format.
12337 * gimple.h: Change uint64_t to int.
12339 2018-01-12 Martin Liska <mliska@suse.cz>
12341 * tree-core.h: Use uint64_t instead of int.
12342 * tree.c (tree_node_counts): Likewise.
12343 (tree_node_sizes): Likewise.
12344 (dump_tree_statistics): Use PRIu64 in printf format.
12346 2018-01-12 Martin Liska <mliska@suse.cz>
12348 * Makefile.in: As qsort_chk is implemented in vec.c, add
12349 vec.o to linkage of gencfn-macros.
12350 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12351 passing the info to record_node_allocation_statistics.
12352 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12354 * ggc-common.c (struct ggc_usage): Add operator== and use
12355 it in operator< and compare function.
12356 * mem-stats.h (struct mem_usage): Likewise.
12357 * vec.c (struct vec_usage): Remove operator< and compare
12358 function. Can be simply inherited.
12360 2018-01-12 Martin Jambor <mjambor@suse.cz>
12363 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12364 * tree-ssa-math-opts.c: Include domwalk.h.
12365 (convert_mult_to_fma_1): New function.
12366 (fma_transformation_info): New type.
12367 (fma_deferring_state): Likewise.
12368 (cancel_fma_deferring): New function.
12369 (result_of_phi): Likewise.
12370 (last_fma_candidate_feeds_initial_phi): Likewise.
12371 (convert_mult_to_fma): Added deferring logic, split actual
12372 transformation to convert_mult_to_fma_1.
12373 (math_opts_dom_walker): New type.
12374 (math_opts_dom_walker::after_dom_children): New method, body moved
12375 here from pass_optimize_widening_mul::execute, added deferring logic
12377 (pass_optimize_widening_mul::execute): Moved most of code to
12378 math_opts_dom_walker::after_dom_children.
12379 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12380 * config/i386/i386.c (ix86_option_override_internal): Added
12381 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12383 2018-01-12 Richard Biener <rguenther@suse.de>
12386 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12387 inline instance vars.
12389 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12392 * config/rx/rx.c (rx_is_restricted_memory_address):
12393 Handle SUBREG case.
12395 2018-01-12 Richard Biener <rguenther@suse.de>
12397 PR tree-optimization/80846
12398 * target.def (split_reduction): New target hook.
12399 * targhooks.c (default_split_reduction): New function.
12400 * targhooks.h (default_split_reduction): Declare.
12401 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12402 target requests first reduce vectors by combining low and high
12404 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12405 (get_vectype_for_scalar_type_and_size): Export.
12406 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12407 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12408 * doc/tm.texi: Regenerate.
12409 * config/i386/i386.c (ix86_split_reduction): Implement
12410 TARGET_VECTORIZE_SPLIT_REDUCTION.
12412 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12415 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12416 in PIC mode except for TARGET_VXWORKS_RTP.
12417 * config/sparc/sparc.c: Include cfgrtl.h.
12418 (TARGET_INIT_PIC_REG): Define.
12419 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12420 (sparc_pic_register_p): New predicate.
12421 (sparc_legitimate_address_p): Use it.
12422 (sparc_legitimize_pic_address): Likewise.
12423 (sparc_delegitimize_address): Likewise.
12424 (sparc_mode_dependent_address_p): Likewise.
12425 (gen_load_pcrel_sym): Remove 4th parameter.
12426 (load_got_register): Adjust call to above. Remove obsolete stuff.
12427 (sparc_expand_prologue): Do not call load_got_register here.
12428 (sparc_flat_expand_prologue): Likewise.
12429 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12430 (sparc_use_pseudo_pic_reg): New function.
12431 (sparc_init_pic_reg): Likewise.
12432 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12433 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12435 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12437 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12438 Add item for branch_cost.
12440 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12442 PR rtl-optimization/83565
12443 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12444 not extend the result to a larger mode for rotate operations.
12445 (num_sign_bit_copies1): Likewise.
12447 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12450 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12452 Use values-Xc.o for -pedantic.
12453 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12455 2018-01-12 Martin Liska <mliska@suse.cz>
12458 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12460 (possible_polymorphic_call_targets): Use it.
12461 (ipa_devirt): Likewise.
12463 2018-01-12 Martin Liska <mliska@suse.cz>
12465 * profile-count.h (enum profile_quality): Use 0 as invalid
12466 enum value of profile_quality.
12468 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12470 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12471 -mext-string options.
12473 2018-01-12 Richard Biener <rguenther@suse.de>
12475 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12476 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12477 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12479 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12481 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12483 * configure.ac (--with-long-double-format): Add support for the
12484 configuration option to change the default long double format on
12486 * config.gcc (powerpc*-linux*-*): Likewise.
12487 * configure: Regenerate.
12488 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12489 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12490 used without modification.
12492 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12494 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12495 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12496 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12497 MISC_BUILTIN_SPEC_BARRIER.
12498 (rs6000_init_builtins): Likewise.
12499 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12501 (speculation_barrier): New define_insn.
12502 * doc/extend.texi: Document __builtin_speculation_barrier.
12504 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12507 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12508 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12509 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12511 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12512 integral modes instead of "ss" and "sd".
12513 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12514 vectors with 32-bit and 64-bit elements.
12515 (vecdupssescalarmodesuffix): New mode attribute.
12516 (vec_dup<mode>): Use it.
12518 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12521 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12522 frame if argument is passed on stack.
12524 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12527 * ree.c (combine_reaching_defs): Optimize also
12528 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12529 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12531 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12533 PR middle-end/83189
12534 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12536 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12538 PR middle-end/83718
12539 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12540 after they are computed.
12542 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12544 PR tree-optimization/83695
12545 * gimple-loop-linterchange.cc
12546 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12547 reset cached scev information after interchange.
12548 (pass_linterchange::execute): Remove call to scev_reset_htab.
12550 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12552 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12553 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12554 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12555 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12556 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12557 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12558 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12559 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12560 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12561 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12562 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12563 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12564 (V_lane_reg): Likewise.
12565 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12567 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12568 (vfmal_lane_low<mode>_intrinsic,
12569 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12570 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12571 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12572 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12573 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12574 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12576 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12578 * config/arm/arm-cpus.in (fp16fml): New feature.
12579 (ALL_SIMD): Add fp16fml.
12580 (armv8.2-a): Add fp16fml as an option.
12581 (armv8.3-a): Likewise.
12582 (armv8.4-a): Add fp16fml as part of fp16.
12583 * config/arm/arm.h (TARGET_FP16FML): Define.
12584 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12586 * config/arm/arm-modes.def (V2HF): Define.
12587 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12588 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12589 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12590 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12591 vfmsl_low, vfmsl_high): New set of builtins.
12592 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12593 (vfml_op): New code attribute.
12594 (VFMLHALVES): New int iterator.
12595 (VFML, VFMLSEL): New mode attributes.
12596 (V_reg): Define mapping for V2HF.
12597 (V_hi, V_lo): New mode attributes.
12598 (VF_constraint): Likewise.
12599 (vfml_half, vfml_half_selector): New int attributes.
12600 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12602 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12603 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12605 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12606 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12607 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12608 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12610 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12611 Document new effective target and option set.
12613 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12615 * config/arm/arm-cpus.in (armv8_4): New feature.
12616 (ARMv8_4a): New fgroup.
12617 (armv8.4-a): New arch.
12618 * config/arm/arm-tables.opt: Regenerate.
12619 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12620 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12621 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12622 Add matching rules for -march=armv8.4-a and extensions.
12623 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12625 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12628 * config/rx/rx.md (BW): New mode attribute.
12629 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12631 2018-01-11 Richard Biener <rguenther@suse.de>
12633 PR tree-optimization/83435
12634 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12635 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12636 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12638 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12639 Alan Hayward <alan.hayward@arm.com>
12640 David Sherwood <david.sherwood@arm.com>
12642 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12644 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12645 (aarch64_print_address_internal): Use it to check for a zero offset.
12647 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12648 Alan Hayward <alan.hayward@arm.com>
12649 David Sherwood <david.sherwood@arm.com>
12651 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12652 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12653 Return a poly_int64 rather than a HOST_WIDE_INT.
12654 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12655 rather than a HOST_WIDE_INT.
12656 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12657 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12658 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12659 final_offset from HOST_WIDE_INT to poly_int64.
12660 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12661 to_constant when getting the number of units in an Advanced SIMD
12663 (aarch64_builtin_vectorized_function): Check for a constant number
12665 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12667 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12668 attribute instead of GET_MODE_NUNITS.
12669 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12670 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12671 GET_MODE_SIZE for fixed-size registers.
12672 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12673 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12674 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12675 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12676 (aarch64_print_operand, aarch64_print_address_internal)
12677 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12678 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12679 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12680 Handle polynomial GET_MODE_SIZE.
12681 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12682 wider than SImode without modification.
12683 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12684 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12685 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12686 passing and returning SVE modes.
12687 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12688 rather than GEN_INT.
12689 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12690 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12691 (aarch64_allocate_and_probe_stack_space): Likewise.
12692 (aarch64_layout_frame): Cope with polynomial offsets.
12693 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12694 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12695 polynomial offsets.
12696 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12697 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12698 poly_int64 rather than a HOST_WIDE_INT.
12699 (aarch64_get_separate_components, aarch64_process_components)
12700 (aarch64_expand_prologue, aarch64_expand_epilogue)
12701 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12702 (aarch64_anchor_offset): New function, split out from...
12703 (aarch64_legitimize_address): ...here.
12704 (aarch64_builtin_vectorization_cost): Handle polynomial
12705 TYPE_VECTOR_SUBPARTS.
12706 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12708 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12709 number of elements from the PARALLEL rather than the mode.
12710 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12711 rather than GET_MODE_BITSIZE.
12712 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12713 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12714 (aarch64_expand_vec_perm_const_1): Handle polynomial
12715 d->perm.length () and d->perm elements.
12716 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12717 Apply to_constant to d->perm elements.
12718 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12719 polynomial CONST_VECTOR_NUNITS.
12720 (aarch64_move_pointer): Take amount as a poly_int64 rather
12722 (aarch64_progress_pointer): Avoid temporary variable.
12723 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12724 the mode attribute instead of GET_MODE.
12726 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12727 Alan Hayward <alan.hayward@arm.com>
12728 David Sherwood <david.sherwood@arm.com>
12730 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12731 x exists before using it.
12732 (aarch64_add_constant_internal): Rename to...
12733 (aarch64_add_offset_1): ...this. Replace regnum with separate
12734 src and dest rtxes. Handle the case in which they're different,
12735 including when the offset is zero. Replace scratchreg with an rtx.
12736 Use 2 additions if there is no spare register into which we can
12737 move a 16-bit constant.
12738 (aarch64_add_constant): Delete.
12739 (aarch64_add_offset): Replace reg with separate src and dest
12740 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12741 Use aarch64_add_offset_1.
12742 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12743 an rtx rather than an int. Take the delta as a poly_int64
12744 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12745 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12746 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12747 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12748 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12749 and aarch64_add_sp.
12750 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12751 aarch64_add_constant.
12753 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12755 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12756 Use scalar_float_mode.
12758 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12760 * config/aarch64/aarch64-simd.md
12761 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12762 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12763 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12764 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12765 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12766 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12767 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12768 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12769 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12770 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12772 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12775 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12776 targ_options->x_arm_arch_string is non NULL.
12778 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12780 * config/aarch64/aarch64.h
12781 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12783 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12786 * expmed.c (emit_store_flag_force): Swap if const op0
12787 and change VOIDmode to mode of op0.
12789 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12791 PR rtl-optimization/83761
12792 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12793 than bytes to mode_for_size.
12795 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12797 PR middle-end/83189
12798 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12799 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12802 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12804 PR middle-end/83575
12805 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12806 when in layout mode.
12807 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12808 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12811 2018-01-10 Michael Collison <michael.collison@arm.com>
12813 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12814 * config/aarch64/aarch64-option-extension.def: Add
12815 AARCH64_OPT_EXTENSION of 'fp16fml'.
12816 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12817 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12818 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12819 * config/aarch64/constraints.md (Ui7): New constraint.
12820 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12821 (VFMLA_SEL_W): Ditto.
12824 (VFMLA16_LOW): New int iterator.
12825 (VFMLA16_HIGH): Ditto.
12826 (UNSPEC_FMLAL): New unspec.
12827 (UNSPEC_FMLSL): Ditto.
12828 (UNSPEC_FMLAL2): Ditto.
12829 (UNSPEC_FMLSL2): Ditto.
12830 (f16mac): New code attribute.
12831 * config/aarch64/aarch64-simd-builtins.def
12832 (aarch64_fmlal_lowv2sf): Ditto.
12833 (aarch64_fmlsl_lowv2sf): Ditto.
12834 (aarch64_fmlalq_lowv4sf): Ditto.
12835 (aarch64_fmlslq_lowv4sf): Ditto.
12836 (aarch64_fmlal_highv2sf): Ditto.
12837 (aarch64_fmlsl_highv2sf): Ditto.
12838 (aarch64_fmlalq_highv4sf): Ditto.
12839 (aarch64_fmlslq_highv4sf): Ditto.
12840 (aarch64_fmlal_lane_lowv2sf): Ditto.
12841 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12842 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12843 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12844 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12845 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12846 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12847 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12848 (aarch64_fmlal_lane_highv2sf): Ditto.
12849 (aarch64_fmlsl_lane_highv2sf): Ditto.
12850 (aarch64_fmlal_laneq_highv2sf): Ditto.
12851 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12852 (aarch64_fmlalq_lane_highv4sf): Ditto.
12853 (aarch64_fmlsl_lane_highv4sf): Ditto.
12854 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12855 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12856 * config/aarch64/aarch64-simd.md:
12857 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12858 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12859 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12860 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12861 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12862 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12863 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12864 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12865 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12866 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12867 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12868 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12869 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12870 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12871 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12872 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12873 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12874 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12875 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12876 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12877 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12878 (vfmlsl_low_u32): Ditto.
12879 (vfmlalq_low_u32): Ditto.
12880 (vfmlslq_low_u32): Ditto.
12881 (vfmlal_high_u32): Ditto.
12882 (vfmlsl_high_u32): Ditto.
12883 (vfmlalq_high_u32): Ditto.
12884 (vfmlslq_high_u32): Ditto.
12885 (vfmlal_lane_low_u32): Ditto.
12886 (vfmlsl_lane_low_u32): Ditto.
12887 (vfmlal_laneq_low_u32): Ditto.
12888 (vfmlsl_laneq_low_u32): Ditto.
12889 (vfmlalq_lane_low_u32): Ditto.
12890 (vfmlslq_lane_low_u32): Ditto.
12891 (vfmlalq_laneq_low_u32): Ditto.
12892 (vfmlslq_laneq_low_u32): Ditto.
12893 (vfmlal_lane_high_u32): Ditto.
12894 (vfmlsl_lane_high_u32): Ditto.
12895 (vfmlal_laneq_high_u32): Ditto.
12896 (vfmlsl_laneq_high_u32): Ditto.
12897 (vfmlalq_lane_high_u32): Ditto.
12898 (vfmlslq_lane_high_u32): Ditto.
12899 (vfmlalq_laneq_high_u32): Ditto.
12900 (vfmlslq_laneq_high_u32): Ditto.
12901 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12902 (AARCH64_FL_FOR_ARCH8_4): New.
12903 (AARCH64_ISA_F16FML): New ISA flag.
12904 (TARGET_F16FML): New feature flag for fp16fml.
12905 (doc/invoke.texi): Document new fp16fml option.
12907 2018-01-10 Michael Collison <michael.collison@arm.com>
12909 * config/aarch64/aarch64-builtins.c:
12910 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12911 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12912 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12913 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12914 (AARCH64_ISA_SHA3): New ISA flag.
12915 (TARGET_SHA3): New feature flag for sha3.
12916 * config/aarch64/iterators.md (sha512_op): New int attribute.
12917 (CRYPTO_SHA512): New int iterator.
12918 (UNSPEC_SHA512H): New unspec.
12919 (UNSPEC_SHA512H2): Ditto.
12920 (UNSPEC_SHA512SU0): Ditto.
12921 (UNSPEC_SHA512SU1): Ditto.
12922 * config/aarch64/aarch64-simd-builtins.def
12923 (aarch64_crypto_sha512hqv2di): New builtin.
12924 (aarch64_crypto_sha512h2qv2di): Ditto.
12925 (aarch64_crypto_sha512su0qv2di): Ditto.
12926 (aarch64_crypto_sha512su1qv2di): Ditto.
12927 (aarch64_eor3qv8hi): Ditto.
12928 (aarch64_rax1qv2di): Ditto.
12929 (aarch64_xarqv2di): Ditto.
12930 (aarch64_bcaxqv8hi): Ditto.
12931 * config/aarch64/aarch64-simd.md:
12932 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12933 (aarch64_crypto_sha512su0qv2di): Ditto.
12934 (aarch64_crypto_sha512su1qv2di): Ditto.
12935 (aarch64_eor3qv8hi): Ditto.
12936 (aarch64_rax1qv2di): Ditto.
12937 (aarch64_xarqv2di): Ditto.
12938 (aarch64_bcaxqv8hi): Ditto.
12939 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12940 (vsha512h2q_u64): Ditto.
12941 (vsha512su0q_u64): Ditto.
12942 (vsha512su1q_u64): Ditto.
12943 (veor3q_u16): Ditto.
12944 (vrax1q_u64): Ditto.
12945 (vxarq_u64): Ditto.
12946 (vbcaxq_u16): Ditto.
12947 * config/arm/types.md (crypto_sha512): New type attribute.
12948 (crypto_sha3): Ditto.
12949 (doc/invoke.texi): Document new sha3 option.
12951 2018-01-10 Michael Collison <michael.collison@arm.com>
12953 * config/aarch64/aarch64-builtins.c:
12954 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12955 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12956 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12957 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12958 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12959 (AARCH64_ISA_SM4): New ISA flag.
12960 (TARGET_SM4): New feature flag for sm4.
12961 * config/aarch64/aarch64-simd-builtins.def
12962 (aarch64_sm3ss1qv4si): Ditto.
12963 (aarch64_sm3tt1aq4si): Ditto.
12964 (aarch64_sm3tt1bq4si): Ditto.
12965 (aarch64_sm3tt2aq4si): Ditto.
12966 (aarch64_sm3tt2bq4si): Ditto.
12967 (aarch64_sm3partw1qv4si): Ditto.
12968 (aarch64_sm3partw2qv4si): Ditto.
12969 (aarch64_sm4eqv4si): Ditto.
12970 (aarch64_sm4ekeyqv4si): Ditto.
12971 * config/aarch64/aarch64-simd.md:
12972 (aarch64_sm3ss1qv4si): Ditto.
12973 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12974 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12975 (aarch64_sm4eqv4si): Ditto.
12976 (aarch64_sm4ekeyqv4si): Ditto.
12977 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12978 (sm3part_op): Ditto.
12979 (CRYPTO_SM3TT): Ditto.
12980 (CRYPTO_SM3PART): Ditto.
12981 (UNSPEC_SM3SS1): New unspec.
12982 (UNSPEC_SM3TT1A): Ditto.
12983 (UNSPEC_SM3TT1B): Ditto.
12984 (UNSPEC_SM3TT2A): Ditto.
12985 (UNSPEC_SM3TT2B): Ditto.
12986 (UNSPEC_SM3PARTW1): Ditto.
12987 (UNSPEC_SM3PARTW2): Ditto.
12988 (UNSPEC_SM4E): Ditto.
12989 (UNSPEC_SM4EKEY): Ditto.
12990 * config/aarch64/constraints.md (Ui2): New constraint.
12991 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12992 * config/arm/types.md (crypto_sm3): New type attribute.
12993 (crypto_sm4): Ditto.
12994 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12995 (vsm3tt1aq_u32): Ditto.
12996 (vsm3tt1bq_u32): Ditto.
12997 (vsm3tt2aq_u32): Ditto.
12998 (vsm3tt2bq_u32): Ditto.
12999 (vsm3partw1q_u32): Ditto.
13000 (vsm3partw2q_u32): Ditto.
13001 (vsm4eq_u32): Ditto.
13002 (vsm4ekeyq_u32): Ditto.
13003 (doc/invoke.texi): Document new sm4 option.
13005 2018-01-10 Michael Collison <michael.collison@arm.com>
13007 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13008 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13009 (AARCH64_FL_FOR_ARCH8_4): New.
13010 (AARCH64_FL_V8_4): New flag.
13011 (doc/invoke.texi): Document new armv8.4-a option.
13013 2018-01-10 Michael Collison <michael.collison@arm.com>
13015 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13016 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13017 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13018 * config/aarch64/aarch64-option-extension.def: Add
13019 AARCH64_OPT_EXTENSION of 'sha2'.
13020 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13021 (crypto): Disable sha2 and aes if crypto disabled.
13022 (crypto): Enable aes and sha2 if enabled.
13023 (simd): Disable sha2 and aes if simd disabled.
13024 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13026 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13027 (TARGET_SHA2): New feature flag for sha2.
13028 (TARGET_AES): New feature flag for aes.
13029 * config/aarch64/aarch64-simd.md:
13030 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13031 conditional on TARGET_AES.
13032 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13033 (aarch64_crypto_sha1hsi): Make pattern conditional
13035 (aarch64_crypto_sha1hv4si): Ditto.
13036 (aarch64_be_crypto_sha1hv4si): Ditto.
13037 (aarch64_crypto_sha1su1v4si): Ditto.
13038 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13039 (aarch64_crypto_sha1su0v4si): Ditto.
13040 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13041 (aarch64_crypto_sha256su0v4si): Ditto.
13042 (aarch64_crypto_sha256su1v4si): Ditto.
13043 (doc/invoke.texi): Document new aes and sha2 options.
13045 2018-01-10 Martin Sebor <msebor@redhat.com>
13047 PR tree-optimization/83781
13048 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13051 2018-01-11 Martin Sebor <msebor@gmail.com>
13052 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13054 PR tree-optimization/83501
13055 PR tree-optimization/81703
13057 * tree-ssa-strlen.c (get_string_cst): Rename...
13058 (get_string_len): ...to this. Handle global constants.
13059 (handle_char_store): Adjust.
13061 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13062 Jim Wilson <jimw@sifive.com>
13064 * config/riscv/riscv-protos.h (riscv_output_return): New.
13065 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13066 (riscv_attribute_table, riscv_output_return),
13067 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13068 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13069 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13071 (riscv_expand_prologue): Add early return for naked function.
13072 (riscv_expand_epilogue): Likewise.
13073 (riscv_function_ok_for_sibcall): Return false for naked function.
13074 (riscv_set_current_function): New.
13075 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13076 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13077 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13078 * doc/extend.texi (RISC-V Function Attributes): New.
13080 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13082 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13083 check for 128-bit long double before checking TCmode.
13084 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13085 128-bit long doubles before checking TFmode or TCmode.
13086 (FLOAT128_IBM_P): Likewise.
13088 2018-01-10 Martin Sebor <msebor@redhat.com>
13090 PR tree-optimization/83671
13091 * builtins.c (c_strlen): Unconditionally return zero for the empty
13093 Use -Warray-bounds for warnings.
13094 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13095 for non-constant array indices with COMPONENT_REF, arrays of
13096 arrays, and pointers to arrays.
13097 (gimple_fold_builtin_strlen): Determine and set length range for
13098 non-constant character arrays.
13100 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13102 PR middle-end/81897
13103 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13106 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13108 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13110 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13113 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13114 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13115 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13116 indexed_or_indirect_operand predicate.
13117 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13118 (*vsx_le_perm_load_v8hi): Likewise.
13119 (*vsx_le_perm_load_v16qi): Likewise.
13120 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13121 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13122 (*vsx_le_perm_store_v8hi): Likewise.
13123 (*vsx_le_perm_store_v16qi): Likewise.
13124 (eight unnamed splitters): Likewise.
13126 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13128 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13129 * config/rs6000/emmintrin.h: Likewise.
13130 * config/rs6000/mmintrin.h: Likewise.
13131 * config/rs6000/xmmintrin.h: Likewise.
13133 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13136 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13138 * tree.c (tree_nop_conversion): Return true for location wrapper
13140 (maybe_wrap_with_location): New function.
13141 (selftest::check_strip_nops): New function.
13142 (selftest::test_location_wrappers): New function.
13143 (selftest::tree_c_tests): Call it.
13144 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13145 (maybe_wrap_with_location): New decl.
13146 (EXPR_LOCATION_WRAPPER_P): New macro.
13147 (location_wrapper_p): New inline function.
13148 (tree_strip_any_location_wrapper): New inline function.
13150 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13153 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13154 stack_realign_offset for the largest alignment of stack slot
13156 (ix86_find_max_used_stack_alignment): New function.
13157 (ix86_finalize_stack_frame_flags): Use it. Set
13158 max_used_stack_alignment if we don't realign stack.
13159 * config/i386/i386.h (machine_function): Add
13160 max_used_stack_alignment.
13162 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13164 * config/arm/arm.opt (-mbranch-cost): New option.
13165 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13168 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13171 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13172 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13174 2018-01-10 Richard Biener <rguenther@suse.de>
13177 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13178 early out so it also covers the case where we have a non-NULL
13181 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13183 PR tree-optimization/83753
13184 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13185 for non-strided grouped accesses if the number of elements is 1.
13187 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13190 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13191 * i386.h (TARGET_USE_GATHER): Define.
13192 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13194 2018-01-10 Martin Liska <mliska@suse.cz>
13197 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13198 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13200 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13201 CLEANUP_NO_PARTITIONING is not set.
13203 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13205 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13206 for vectors, as a partial revert of r254296.
13207 * rtl.h (const_vec_p): Delete.
13208 (const_vec_duplicate_p): Don't test for vector CONSTs.
13209 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13210 * expmed.c (make_tree): Likewise.
13213 * common.md (E, F): Use CONSTANT_P instead of checking for
13215 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13216 checking for CONST_VECTOR.
13218 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13220 PR middle-end/83575
13221 * predict.c (force_edge_cold): Handle in more sane way edges
13222 with no prediction.
13224 2018-01-09 Carl Love <cel@us.ibm.com>
13226 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13228 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13229 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13230 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13231 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13232 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13233 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13234 * config/rs6000/rs6000-protos.h: Add extern defition for
13235 rs6000_generate_float2_double_code.
13236 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13238 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13239 (float2_v2df): Add define_expand.
13241 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13244 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13245 op_mode in the force_to_mode call.
13247 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13249 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13250 instead of checking each element individually.
13251 (aarch64_evpc_uzp): Likewise.
13252 (aarch64_evpc_zip): Likewise.
13253 (aarch64_evpc_ext): Likewise.
13254 (aarch64_evpc_rev): Likewise.
13255 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13256 instead of checking each element individually. Return true without
13258 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13259 whether all selected elements come from the same input, instead of
13260 checking each element individually. Remove calls to gen_rtx_REG,
13261 start_sequence and end_sequence and instead assert that no rtl is
13264 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13266 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13267 order of HIGH and CONST checks.
13269 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13271 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13272 if the destination isn't an SSA_NAME.
13274 2018-01-09 Richard Biener <rguenther@suse.de>
13276 PR tree-optimization/83668
13277 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13279 (canonicalize_loop_form): ... here, renamed from ...
13280 (canonicalize_loop_closed_ssa_form): ... this and amended to
13281 swap successor edges for loop exit blocks to make us use
13282 the RPO order we need for initial schedule generation.
13284 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13286 PR tree-optimization/64811
13287 * match.pd: When optimizing comparisons with Inf, avoid
13288 introducing or losing exceptions from comparisons with NaN.
13290 2018-01-09 Martin Liska <mliska@suse.cz>
13293 * asan.c (shadow_mem_size): Add gcc_assert.
13295 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13297 Don't save registers in main().
13300 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13301 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13302 * config/avr/avr.c (avr_set_current_function): Don't error if
13303 naked, OS_task or OS_main are specified at the same time.
13304 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13306 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13308 * common/config/avr/avr-common.c (avr_option_optimization_table):
13309 Switch on -mmain-is-OS_task for optimizing compilations.
13311 2018-01-09 Richard Biener <rguenther@suse.de>
13313 PR tree-optimization/83572
13314 * graphite.c: Include cfganal.h.
13315 (graphite_transform_loops): Connect infinite loops to exit
13316 and remove fake edges at the end.
13318 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13320 * ipa-inline.c (edge_badness): Revert accidental checkin.
13322 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13325 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13326 symbols; not inline clones.
13328 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13331 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13332 hard registers. Formatting fixes.
13334 PR preprocessor/83722
13335 * gcc.c (try_generate_repro): Pass
13336 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13337 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13340 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13341 Kito Cheng <kito.cheng@gmail.com>
13343 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13344 (riscv_leaf_function_p): Delete.
13345 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13347 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13349 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13351 (do_ifelse): New function.
13352 (do_isel): New function.
13353 (do_sub3): New function.
13354 (do_add3): New function.
13355 (do_load_mask_compare): New function.
13356 (do_overlap_load_compare): New function.
13357 (expand_compare_loop): New function.
13358 (expand_block_compare): Call expand_compare_loop() when appropriate.
13359 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13360 option description.
13361 (-mblock-compare-inline-loop-limit): New option.
13363 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13366 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13367 Reverse order of second and third operands in first alternative.
13368 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13369 of first and second elements in UNSPEC_VPERMR vector.
13370 (altivec_expand_vec_perm_le): Likewise.
13372 2018-01-08 Jeff Law <law@redhat.com>
13374 PR rtl-optimizatin/81308
13375 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13376 (process_switch): If group_case_labels makes a change, then set
13378 (pass_convert_switch::execute): If a switch is converted, then
13379 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13381 PR rtl-optimization/81308
13382 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13385 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13387 PR target/83663 - Revert r255946
13388 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13389 generation for cases where splatting a value is not useful.
13390 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13391 across a vec_duplicate and a paradoxical subreg forming a vector
13392 mode to a vec_concat.
13394 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13396 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13397 -march=armv8.3-a variants.
13398 * config/arm/t-multilib: Likewise.
13399 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13401 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13403 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13405 (cceq_ior_compare_complement): Give it a name so I can use it, and
13406 change boolean_or_operator predicate to boolean_operator so it can
13407 be used to generate a crand.
13408 (eqne): New code iterator.
13409 (bd/bd_neg): New code_attrs.
13410 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13411 a single define_insn.
13412 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13413 decrement (bdnzt/bdnzf/bdzt/bdzf).
13414 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13415 with the new names of the branch decrement patterns, and added the
13416 names of the branch decrement conditional patterns.
13418 2018-01-08 Richard Biener <rguenther@suse.de>
13420 PR tree-optimization/83563
13421 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13424 2018-01-08 Richard Biener <rguenther@suse.de>
13426 PR middle-end/83713
13427 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13429 2018-01-08 Richard Biener <rguenther@suse.de>
13431 PR tree-optimization/83685
13432 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13433 references to abnormals.
13435 2018-01-08 Richard Biener <rguenther@suse.de>
13438 * dwarf2out.c (output_indirect_strings): Handle empty
13439 skeleton_debug_str_hash.
13440 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13442 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13444 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13445 (emit_store_direct): Likewise.
13446 (arc_trampoline_adjust_address): Likewise.
13447 (arc_asm_trampoline_template): New function.
13448 (arc_initialize_trampoline): Use asm_trampoline_template.
13449 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13450 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13451 * config/arc/arc.md (flush_icache): Delete pattern.
13453 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13455 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13456 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13459 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13462 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13463 by not USED_FOR_TARGET.
13464 (make_pass_resolve_sw_modes): Likewise.
13466 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13468 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13471 2018-01-08 Richard Biener <rguenther@suse.de>
13473 PR middle-end/83580
13474 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13476 2018-01-08 Richard Biener <rguenther@suse.de>
13478 PR middle-end/83517
13479 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13481 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13483 PR middle-end/81897
13484 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13485 basic blocks with a small number of successors.
13486 (convert_control_dep_chain_into_preds): Improve handling of
13488 (dump_predicates): Split apart into...
13489 (dump_pred_chain): ...here...
13490 (dump_pred_info): ...and here.
13491 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13492 (can_chain_union_be_invalidated_p): Improve check for invalidation
13494 (uninit_uses_cannot_happen): Avoid unnecessary if
13495 convert_control_dep_chain_into_preds yielded nothing.
13497 2018-01-06 Martin Sebor <msebor@redhat.com>
13499 PR tree-optimization/83640
13500 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13501 subtracting negative offset from size.
13502 (builtin_access::overlap): Adjust offset bounds of the access to fall
13503 within the size of the object if possible.
13505 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13507 PR rtl-optimization/83699
13508 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13509 extract_bit_field_as_subreg to cases in which the extracted
13510 value is also a vector.
13512 * lra-constraints.c (process_alt_operands): Test for the equivalence
13513 substitutions when detecting a possible reload cycle.
13515 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13518 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13519 by default if flag_selective_schedling{,2}. Formatting fixes.
13521 PR rtl-optimization/83682
13522 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13523 if it has non-VECTOR_MODE element mode.
13524 (vec_duplicate_p): Likewise.
13526 PR middle-end/83694
13527 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13528 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13530 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13533 * config/i386/i386-builtin.def
13534 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13535 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13536 Require also OPTION_MASK_ISA_AVX512F in addition to
13537 OPTION_MASK_ISA_GFNI.
13538 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13539 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13540 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13541 to OPTION_MASK_ISA_GFNI.
13542 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13543 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13544 OPTION_MASK_ISA_AVX512BW.
13545 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13546 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13547 addition to OPTION_MASK_ISA_GFNI.
13548 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13549 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13550 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13551 to OPTION_MASK_ISA_GFNI.
13552 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13553 a requirement for all ISAs rather than any of them with a few
13555 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13557 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13558 bitmasks to be enabled with 3 exceptions, instead of requiring any
13559 enabled ISA with lots of exceptions.
13560 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13561 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13562 Change avx512bw in isa attribute to avx512f.
13563 * config/i386/sgxintrin.h: Add license boilerplate.
13564 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13565 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13566 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13567 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13569 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13570 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13571 temporarily sse2 rather than sse if not enabled already.
13574 * config/i386/sse.md (VI248_VLBW): Rename to ...
13575 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13576 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13577 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13578 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13579 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13580 mode iterator instead of VI248_VLBW.
13582 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13584 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13585 (record_modified): Skip clobbers; add debug output.
13586 (param_change_prob): Use sreal frequencies.
13588 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13590 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13591 punt for user-aligned variables.
13593 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13595 * tree-chrec.c (chrec_contains_symbols): Return true for
13598 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13601 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13602 of (x|y) == x for BICS pattern.
13604 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13606 PR tree-optimization/83605
13607 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13608 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13611 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13613 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13614 * config/epiphany/rtems.h: New file.
13616 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13617 Uros Bizjak <ubizjak@gmail.com>
13620 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13621 QIreg_operand instead of register_operand predicate.
13622 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13623 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13624 comments instead of -fmitigate[-_]rop.
13626 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13629 * cgraphunit.c (symbol_table::compile): Switch to text_section
13630 before calling assembly_start debug hook.
13631 * run-rtl-passes.c (run_rtl_passes): Likewise.
13634 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13636 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13637 range_int_cst_p rather than !symbolic_range_p before calling
13638 extract_range_from_multiplicative_op_1.
13640 2018-01-04 Jeff Law <law@redhat.com>
13642 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13643 redundant test in assertion.
13645 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13647 * doc/rtl.texi: Document machine_mode wrapper classes.
13649 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13651 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13652 using tree_to_uhwi.
13654 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13656 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13657 the VEC_PERM_EXPR fold to fail.
13659 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13662 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13663 to switched_sections.
13665 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13668 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13669 test for d.testing.
13671 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13674 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13675 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13677 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13680 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13681 is BLKmode and bitpos not zero or mode change is needed.
13683 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13686 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13689 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13692 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13693 instead of MULT rtx. Update all corresponding splitters.
13694 (*saddl_se): Ditto.
13695 (*ssub<modesuffix>): Ditto.
13696 (*ssubl_se): Ditto.
13697 (*cmp_sadd_di): Update split patterns.
13698 (*cmp_sadd_si): Ditto.
13699 (*cmp_sadd_sidi): Ditto.
13700 (*cmp_ssub_di): Ditto.
13701 (*cmp_ssub_si): Ditto.
13702 (*cmp_ssub_sidi): Ditto.
13703 * config/alpha/predicates.md (const23_operand): New predicate.
13704 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13705 Look for ASHIFT, not MULT inner operand.
13706 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13708 2018-01-04 Martin Liska <mliska@suse.cz>
13710 PR gcov-profile/83669
13711 * gcov.c (output_intermediate_file): Add version to intermediate
13713 * doc/gcov.texi: Document new field 'version' in intermediate
13714 file format. Fix location of '-k' option of gcov command.
13716 2018-01-04 Martin Liska <mliska@suse.cz>
13719 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13721 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13723 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13725 2018-01-03 Martin Sebor <msebor@redhat.com>
13727 PR tree-optimization/83655
13728 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13729 checking calls with invalid arguments.
13731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13733 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13734 (vectorizable_mask_load_store): Delete.
13735 (vectorizable_call): Return false for masked loads and stores.
13736 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13737 instead of gimple_assign_rhs1.
13738 (vectorizable_load): Handle IFN_MASK_LOAD.
13739 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13743 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13745 (vectorizable_mask_load_store): ...here.
13746 (vectorizable_load): ...and here.
13748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13750 * tree-vect-stmts.c (vect_build_all_ones_mask)
13751 (vect_build_zero_merge_argument): New functions, split out from...
13752 (vectorizable_load): ...here.
13754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13756 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13758 (vectorizable_mask_load_store): ...here.
13759 (vectorizable_store): ...and here.
13761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13763 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13765 (vectorizable_mask_load_store): ...here.
13767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13769 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13770 (vect_model_store_cost): Take a vec_load_store_type instead of a
13772 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13773 (vect_model_store_cost): Take a vec_load_store_type instead of a
13775 (vectorizable_mask_load_store): Update accordingly.
13776 (vectorizable_store): Likewise.
13777 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13781 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13782 IFN_MASK_LOAD calls here rather than...
13783 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
13790 fall back to extract_bit_field_as_subreg if vec_extract
13793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13794 Alan Hayward <alan.hayward@arm.com>
13795 David Sherwood <david.sherwood@arm.com>
13797 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13798 they are variable or constant sized.
13799 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13800 slots for constant-sized data.
13802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13803 Alan Hayward <alan.hayward@arm.com>
13804 David Sherwood <david.sherwood@arm.com>
13806 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13807 handling COND_EXPRs with boolean comparisons, try to find a better
13808 basis for the mask type than the boolean itself.
13810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13812 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13813 is calculated and how it can be overridden.
13814 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13815 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13817 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13821 Alan Hayward <alan.hayward@arm.com>
13822 David Sherwood <david.sherwood@arm.com>
13824 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13825 Remove the mode argument.
13826 (aarch64_simd_valid_immediate): Remove the mode and inverse
13828 * config/aarch64/iterators.md (bitsize): New iterator.
13829 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13830 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13831 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13832 aarch64_simd_valid_immediate.
13833 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13834 (aarch64_reg_or_bic_imm): Likewise.
13835 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13836 with an insn_type enum and msl with a modifier_type enum.
13837 Replace element_width with a scalar_mode. Change the shift
13838 to unsigned int. Add constructors for scalar_float_mode and
13839 scalar_int_mode elements.
13840 (aarch64_vect_float_const_representable_p): Delete.
13841 (aarch64_can_const_movi_rtx_p)
13842 (aarch64_simd_scalar_immediate_valid_for_move)
13843 (aarch64_simd_make_constant): Update call to
13844 aarch64_simd_valid_immediate.
13845 (aarch64_advsimd_valid_immediate_hs): New function.
13846 (aarch64_advsimd_valid_immediate): Likewise.
13847 (aarch64_simd_valid_immediate): Remove mode and inverse
13848 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13849 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13850 and aarch64_float_const_representable_p on the result.
13851 (aarch64_output_simd_mov_immediate): Remove mode argument.
13852 Update call to aarch64_simd_valid_immediate and use of
13853 simd_immediate_info.
13854 (aarch64_output_scalar_simd_mov_immediate): Update call
13857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13858 Alan Hayward <alan.hayward@arm.com>
13859 David Sherwood <david.sherwood@arm.com>
13861 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13862 (mode_nunits): Likewise CONST_MODE_NUNITS.
13863 * machmode.def (ADJUST_NUNITS): Document.
13864 * genmodes.c (mode_data::need_nunits_adj): New field.
13865 (blank_mode): Update accordingly.
13866 (adj_nunits): New variable.
13867 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13869 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13870 listed in adj_nunits.
13871 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13872 listed in adj_nunits. Don't emit case statements for such modes.
13873 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13874 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13875 nothing if adj_nunits is nonnull.
13876 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13877 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13878 (emit_mode_fbit): Update use of print_maybe_const_decl.
13879 (emit_move_size): Likewise. Treat the array as non-const
13881 (emit_mode_adjustments): Handle adj_nunits.
13883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13885 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13886 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13887 (VECTOR_MODES): Use it.
13888 (make_vector_modes): Take the prefix as an argument.
13890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13891 Alan Hayward <alan.hayward@arm.com>
13892 David Sherwood <david.sherwood@arm.com>
13894 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13895 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13896 for MODE_VECTOR_BOOL.
13897 * machmode.def (VECTOR_BOOL_MODE): Document.
13898 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13899 (make_vector_bool_mode): New function.
13900 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13902 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13903 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13905 * stor-layout.c (int_mode_for_mode): Likewise.
13906 * tree.c (build_vector_type_for_mode): Likewise.
13907 * varasm.c (output_constant_pool_2): Likewise.
13908 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13909 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13910 for MODE_VECTOR_BOOL.
13911 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13912 of mode class checks.
13913 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13914 instead of a list of mode class checks.
13915 (expand_vector_scalar_condition): Likewise.
13916 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13919 Alan Hayward <alan.hayward@arm.com>
13920 David Sherwood <david.sherwood@arm.com>
13922 * machmode.h (mode_size): Change from unsigned short to
13924 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13925 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13926 or if measurement_type is not polynomial.
13927 (fixed_size_mode::includes_p): Check for constant-sized modes.
13928 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13929 return a poly_uint16 rather than an unsigned short.
13930 (emit_mode_size): Change the type of mode_size from unsigned short
13931 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13932 (emit_mode_adjustments): Cope with polynomial vector sizes.
13933 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13935 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13937 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13938 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13939 * caller-save.c (setup_save_areas): Likewise.
13940 (replace_reg_with_saved_mem): Likewise.
13941 * calls.c (emit_library_call_value_1): Likewise.
13942 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13943 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13944 (gen_lowpart_for_combine): Likewise.
13945 * convert.c (convert_to_integer_1): Likewise.
13946 * cse.c (equiv_constant, cse_insn): Likewise.
13947 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13948 (cselib_subst_to_values): Likewise.
13949 * dce.c (word_dce_process_block): Likewise.
13950 * df-problems.c (df_word_lr_mark_ref): Likewise.
13951 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13952 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13953 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13954 (rtl_for_decl_location): Likewise.
13955 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13956 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13957 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13958 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13959 (expand_expr_real_1): Likewise.
13960 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13961 (pad_below): Likewise.
13962 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13963 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13964 * ira.c (get_subreg_tracking_sizes): Likewise.
13965 * ira-build.c (ira_create_allocno_objects): Likewise.
13966 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13967 (ira_sort_regnos_for_alter_reg): Likewise.
13968 * ira-costs.c (record_operand_costs): Likewise.
13969 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13970 (resolve_simple_move): Likewise.
13971 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13972 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13973 (lra_constraints): Likewise.
13974 (CONST_POOL_OK_P): Reject variable-sized modes.
13975 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13976 (add_pseudo_to_slot, lra_spill): Likewise.
13977 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13978 * optabs-query.c (get_best_extraction_insn): Likewise.
13979 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13980 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13981 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13982 * recog.c (offsettable_address_addr_space_p): Likewise.
13983 * regcprop.c (maybe_mode_change): Likewise.
13984 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13985 * regrename.c (build_def_use): Likewise.
13986 * regstat.c (dump_reg_info): Likewise.
13987 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13988 (find_reloads, find_reloads_subreg_address): Likewise.
13989 * reload1.c (eliminate_regs_1): Likewise.
13990 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13991 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13992 (simplify_binary_operation_1, simplify_subreg): Likewise.
13993 * targhooks.c (default_function_arg_padding): Likewise.
13994 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13995 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13996 (verify_gimple_assign_ternary): Likewise.
13997 * tree-inline.c (estimate_move_cost): Likewise.
13998 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13999 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14000 (get_address_cost_ainc): Likewise.
14001 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14002 (vect_supportable_dr_alignment): Likewise.
14003 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14004 (vectorizable_reduction): Likewise.
14005 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14006 (vectorizable_operation, vectorizable_load): Likewise.
14007 * tree.c (build_same_sized_truth_vector_type): Likewise.
14008 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14009 * var-tracking.c (emit_note_insn_var_location): Likewise.
14010 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14011 (ADDR_VEC_ALIGN): Likewise.
14013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14014 Alan Hayward <alan.hayward@arm.com>
14015 David Sherwood <david.sherwood@arm.com>
14017 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14019 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14020 or if measurement_type is polynomial.
14021 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14022 * combine.c (make_extraction): Likewise.
14023 * dse.c (find_shift_sequence): Likewise.
14024 * dwarf2out.c (mem_loc_descriptor): Likewise.
14025 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14026 (extract_bit_field, extract_low_bits): Likewise.
14027 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14028 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14029 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14030 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14031 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14032 * reload.c (find_reloads): Likewise.
14033 * reload1.c (alter_reg): Likewise.
14034 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14035 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14036 * tree-if-conv.c (predicate_mem_writes): Likewise.
14037 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14038 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14039 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14040 * valtrack.c (dead_debug_insert_temp): Likewise.
14041 * varasm.c (mergeable_constant_section): Likewise.
14042 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14045 Alan Hayward <alan.hayward@arm.com>
14046 David Sherwood <david.sherwood@arm.com>
14048 * expr.c (expand_assignment): Cope with polynomial mode sizes
14049 when assigning to a CONCAT.
14051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14052 Alan Hayward <alan.hayward@arm.com>
14053 David Sherwood <david.sherwood@arm.com>
14055 * machmode.h (mode_precision): Change from unsigned short to
14057 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14059 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14060 or if measurement_type is not polynomial.
14061 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14062 in which the mode is already known to be a scalar_int_mode.
14063 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14064 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14066 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14067 for GET_MODE_PRECISION.
14068 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14069 for GET_MODE_PRECISION.
14070 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14072 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14073 (expand_field_assignment, make_extraction): Likewise.
14074 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14075 (get_last_value): Likewise.
14076 * convert.c (convert_to_integer_1): Likewise.
14077 * cse.c (cse_insn): Likewise.
14078 * expr.c (expand_expr_real_1): Likewise.
14079 * lra-constraints.c (simplify_operand_subreg): Likewise.
14080 * optabs-query.c (can_atomic_load_p): Likewise.
14081 * optabs.c (expand_atomic_load): Likewise.
14082 (expand_atomic_store): Likewise.
14083 * ree.c (combine_reaching_defs): Likewise.
14084 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14085 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14086 * tree.h (type_has_mode_precision_p): Likewise.
14087 * ubsan.c (instrument_si_overflow): Likewise.
14089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14090 Alan Hayward <alan.hayward@arm.com>
14091 David Sherwood <david.sherwood@arm.com>
14093 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14094 polynomial numbers of units.
14095 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14096 (valid_vector_subparts_p): New function.
14097 (build_vector_type): Remove temporary shim and take the number
14098 of units as a poly_uint64 rather than an int.
14099 (build_opaque_vector_type): Take the number of units as a
14100 poly_uint64 rather than an int.
14101 * tree.c (build_vector_from_ctor): Handle polynomial
14102 TYPE_VECTOR_SUBPARTS.
14103 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14104 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14105 (build_vector_from_val): If the number of units is variable,
14106 use build_vec_duplicate_cst for constant operands and
14107 VEC_DUPLICATE_EXPR otherwise.
14108 (make_vector_type): Remove temporary is_constant ().
14109 (build_vector_type, build_opaque_vector_type): Take the number of
14110 units as a poly_uint64 rather than an int.
14111 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14113 * cfgexpand.c (expand_debug_expr): Likewise.
14114 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14115 (store_constructor, expand_expr_real_1): Likewise.
14116 (const_scalar_mask_from_tree): Likewise.
14117 * fold-const-call.c (fold_const_reduction): Likewise.
14118 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14119 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14120 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14121 (fold_relational_const): Likewise.
14122 (native_interpret_vector): Likewise. Change the size from an
14123 int to an unsigned int.
14124 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14125 TYPE_VECTOR_SUBPARTS.
14126 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14127 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14128 duplicating a non-constant operand into a variable-length vector.
14129 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14130 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14131 * ipa-icf.c (sem_variable::equals): Likewise.
14132 * match.pd: Likewise.
14133 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14134 * print-tree.c (print_node): Likewise.
14135 * stor-layout.c (layout_type): Likewise.
14136 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14137 * tree-cfg.c (verify_gimple_comparison): Likewise.
14138 (verify_gimple_assign_binary): Likewise.
14139 (verify_gimple_assign_ternary): Likewise.
14140 (verify_gimple_assign_single): Likewise.
14141 * tree-pretty-print.c (dump_generic_node): Likewise.
14142 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14143 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14144 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14145 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14146 (vect_shift_permute_load_chain): Likewise.
14147 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14148 (expand_vector_condition, optimize_vector_constructor): Likewise.
14149 (lower_vec_perm, get_compute_type): Likewise.
14150 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14151 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14152 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14153 (vect_recog_mask_conversion_pattern): Likewise.
14154 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14155 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14156 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14157 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14158 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14159 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14160 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14161 (supportable_widening_operation): Likewise.
14162 (supportable_narrowing_operation): Likewise.
14163 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14165 * varasm.c (output_constant): Likewise.
14167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14168 Alan Hayward <alan.hayward@arm.com>
14169 David Sherwood <david.sherwood@arm.com>
14171 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14172 so that both the length == 3 and length != 3 cases set up their
14173 own permute vectors. Add comments explaining why we know the
14174 number of elements is constant.
14175 (vect_permute_load_chain): Likewise.
14177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14178 Alan Hayward <alan.hayward@arm.com>
14179 David Sherwood <david.sherwood@arm.com>
14181 * machmode.h (mode_nunits): Change from unsigned char to
14183 (ONLY_FIXED_SIZE_MODES): New macro.
14184 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14185 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14186 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14188 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14189 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14190 or if measurement_type is not polynomial.
14191 * genmodes.c (ZERO_COEFFS): New macro.
14192 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14194 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14195 Use ZERO_COEFFS when emitting initializers.
14196 * data-streamer.h (bp_pack_poly_value): New function.
14197 (bp_unpack_poly_value): Likewise.
14198 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14199 for GET_MODE_NUNITS.
14200 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14201 for GET_MODE_NUNITS.
14202 * tree.c (make_vector_type): Remove temporary shim and make
14203 the real function take the number of units as a poly_uint64
14204 rather than an int.
14205 (build_vector_type_for_mode): Handle polynomial nunits.
14206 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14207 * emit-rtl.c (const_vec_series_p_1): Likewise.
14208 (gen_rtx_CONST_VECTOR): Likewise.
14209 * fold-const.c (test_vec_duplicate_folding): Likewise.
14210 * genrecog.c (validate_pattern): Likewise.
14211 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14212 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14213 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14214 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14215 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14216 * rtlanal.c (subreg_get_info): Likewise.
14217 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14218 (vect_grouped_load_supported): Likewise.
14219 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14220 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14221 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14222 (simplify_const_unary_operation, simplify_binary_operation_1)
14223 (simplify_const_binary_operation, simplify_ternary_operation)
14224 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14225 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14226 instead of CONST_VECTOR_NUNITS.
14227 * varasm.c (output_constant_pool_2): Likewise.
14228 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14229 explicit-encoded elements in the XVEC for variable-length vectors.
14231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14233 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14236 Alan Hayward <alan.hayward@arm.com>
14237 David Sherwood <david.sherwood@arm.com>
14239 * coretypes.h (fixed_size_mode): Declare.
14240 (fixed_size_mode_pod): New typedef.
14241 * builtins.h (target_builtins::x_apply_args_mode)
14242 (target_builtins::x_apply_result_mode): Change type to
14243 fixed_size_mode_pod.
14244 * builtins.c (apply_args_size, apply_result_size, result_vector)
14245 (expand_builtin_apply_args_1, expand_builtin_apply)
14246 (expand_builtin_return): Update accordingly.
14248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14250 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14251 * cselib.c (cselib_hash_rtx): Likewise.
14252 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14253 CONST_VECTOR encoding.
14255 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14256 Jeff Law <law@redhat.com>
14259 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14260 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14261 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14262 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14265 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14266 explicitly probe *sp in a noreturn function if there were any callee
14267 register saves or frame pointer is needed.
14269 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14272 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14273 BLKmode for ternary, binary or unary expressions.
14276 * var-tracking.c (delete_vta_debug_insn): New inline function.
14277 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14278 insns from get_insns () to NULL instead of each bb separately.
14279 Use delete_vta_debug_insn. No longer static.
14280 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14281 delete_vta_debug_insns callers.
14282 * rtl.h (delete_vta_debug_insns): Declare.
14283 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14284 instead of variable_tracking_main.
14286 2018-01-03 Martin Sebor <msebor@redhat.com>
14288 PR tree-optimization/83603
14289 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14290 arguments past the endof the argument list in functions declared
14291 without a prototype.
14292 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14293 Avoid checking when arguments are null.
14295 2018-01-03 Martin Sebor <msebor@redhat.com>
14298 * doc/extend.texi (attribute const): Fix a typo.
14299 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14300 issuing -Wsuggest-attribute for void functions.
14302 2018-01-03 Martin Sebor <msebor@redhat.com>
14304 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14305 offset_int::from instead of wide_int::to_shwi.
14306 (maybe_diag_overlap): Remove assertion.
14307 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14308 * gimple-ssa-sprintf.c (format_directive): Same.
14309 (parse_directive): Same.
14310 (sprintf_dom_walker::compute_format_length): Same.
14311 (try_substitute_return_value): Same.
14313 2018-01-03 Jeff Law <law@redhat.com>
14315 PR middle-end/83654
14316 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14317 non-constant residual for zero at runtime and avoid probing in
14318 that case. Reorganize code for trailing problem to mirror handling
14321 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14323 PR tree-optimization/83501
14324 * tree-ssa-strlen.c (get_string_cst): New.
14325 (handle_char_store): Call get_string_cst.
14327 2018-01-03 Martin Liska <mliska@suse.cz>
14329 PR tree-optimization/83593
14330 * tree-ssa-strlen.c: Include tree-cfg.h.
14331 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14332 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14333 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14335 (strlen_dom_walker::before_dom_children): Call
14336 gimple_purge_dead_eh_edges. Dump tranformation with details
14338 (strlen_dom_walker::before_dom_children): Update call by adding
14339 new argument cleanup_eh.
14340 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14342 2018-01-03 Martin Liska <mliska@suse.cz>
14345 * cif-code.def (VARIADIC_THUNK): New enum value.
14346 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14349 2018-01-03 Jan Beulich <jbeulich@suse.com>
14351 * sse.md (mov<mode>_internal): Tighten condition for when to use
14352 vmovdqu<ssescalarsize> for TI and OI modes.
14354 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14356 Update copyright years.
14358 2018-01-03 Martin Liska <mliska@suse.cz>
14361 * ipa-visibility.c (function_and_variable_visibility): Skip
14362 functions with noipa attribure.
14364 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14366 * gcc.c (process_command): Update copyright notice dates.
14367 * gcov-dump.c (print_version): Ditto.
14368 * gcov.c (print_version): Ditto.
14369 * gcov-tool.c (print_version): Ditto.
14370 * gengtype.c (create_file): Ditto.
14371 * doc/cpp.texi: Bump @copying's copyright year.
14372 * doc/cppinternals.texi: Ditto.
14373 * doc/gcc.texi: Ditto.
14374 * doc/gccint.texi: Ditto.
14375 * doc/gcov.texi: Ditto.
14376 * doc/install.texi: Ditto.
14377 * doc/invoke.texi: Ditto.
14379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14381 * vector-builder.h (vector_builder::m_full_nelts): Change from
14382 unsigned int to poly_uint64.
14383 (vector_builder::full_nelts): Update prototype accordingly.
14384 (vector_builder::new_vector): Likewise.
14385 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14386 (vector_builder::operator ==): Likewise.
14387 (vector_builder::finalize): Likewise.
14388 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14389 Take the number of elements as a poly_uint64 rather than an
14391 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14392 from unsigned int to poly_uint64.
14393 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14394 (vec_perm_indices::new_vector): Likewise.
14395 (vec_perm_indices::length): Likewise.
14396 (vec_perm_indices::nelts_per_input): Likewise.
14397 (vec_perm_indices::input_nelts): Likewise.
14398 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14399 number of elements per input as a poly_uint64 rather than an
14400 unsigned int. Use the original encoding for variable-length
14401 vectors, rather than clamping each individual element.
14402 For the second and subsequent elements in each pattern,
14403 clamp the step and base before clamping their sum.
14404 (vec_perm_indices::series_p): Handle polynomial element counts.
14405 (vec_perm_indices::all_in_range_p): Likewise.
14406 (vec_perm_indices_to_tree): Likewise.
14407 (vec_perm_indices_to_rtx): Likewise.
14408 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14409 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14410 (tree_vector_builder::new_binary_operation): Handle polynomial
14411 element counts. Return false if we need to know the number
14412 of elements at compile time.
14413 * fold-const.c (fold_vec_perm): Punt if the number of elements
14414 isn't known at compile time.
14416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14418 * vec-perm-indices.h (vec_perm_builder): Change element type
14419 from HOST_WIDE_INT to poly_int64.
14420 (vec_perm_indices::element_type): Update accordingly.
14421 (vec_perm_indices::clamp): Handle polynomial element_types.
14422 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14423 (vec_perm_indices::all_in_range_p): Likewise.
14424 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14426 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14427 polynomial vec_perm_indices element types.
14428 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14429 * fold-const.c (fold_vec_perm): Likewise.
14430 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14431 * tree-vect-generic.c (lower_vec_perm): Likewise.
14432 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14433 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14434 element type to HOST_WIDE_INT.
14436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14437 Alan Hayward <alan.hayward@arm.com>
14438 David Sherwood <david.sherwood@arm.com>
14440 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14441 rather than an int. Use plus_constant.
14442 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14443 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14446 Alan Hayward <alan.hayward@arm.com>
14447 David Sherwood <david.sherwood@arm.com>
14449 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14450 a HOST_WIDE_INT to a poly_int64.
14452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14453 Alan Hayward <alan.hayward@arm.com>
14454 David Sherwood <david.sherwood@arm.com>
14456 * calls.c (load_register_parameters): Cope with polynomial
14457 mode sizes. Require a constant size for BLKmode parameters
14458 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14459 forces a parameter to be padded at the lsb end in order to
14460 fill a complete number of words, require the parameter size
14461 to be ordered wrt UNITS_PER_WORD.
14463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14464 Alan Hayward <alan.hayward@arm.com>
14465 David Sherwood <david.sherwood@arm.com>
14467 * reload1.c (spill_stack_slot_width): Change element type
14468 from unsigned int to poly_uint64_pod.
14469 (alter_reg): Treat mode sizes as polynomial.
14471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14472 Alan Hayward <alan.hayward@arm.com>
14473 David Sherwood <david.sherwood@arm.com>
14475 * reload.c (complex_word_subreg_p): New function.
14476 (reload_inner_reg_of_subreg, push_reload): Use it.
14478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14479 Alan Hayward <alan.hayward@arm.com>
14480 David Sherwood <david.sherwood@arm.com>
14482 * lra-constraints.c (process_alt_operands): Reject matched
14483 operands whose sizes aren't ordered.
14484 (match_reload): Refer to this check here.
14486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14487 Alan Hayward <alan.hayward@arm.com>
14488 David Sherwood <david.sherwood@arm.com>
14490 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14491 that the mode size is in the set {1, 2, 4, 8, 16}.
14493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14494 Alan Hayward <alan.hayward@arm.com>
14495 David Sherwood <david.sherwood@arm.com>
14497 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14498 Use plus_constant instead of gen_rtx_PLUS.
14500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14501 Alan Hayward <alan.hayward@arm.com>
14502 David Sherwood <david.sherwood@arm.com>
14504 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14505 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14506 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14507 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14508 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14509 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14510 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14511 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14512 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14513 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14515 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14516 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14517 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14518 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14519 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14520 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14521 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14522 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14523 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14524 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14526 * expr.c (emit_move_resolve_push): Treat the input and result
14527 of PUSH_ROUNDING as a poly_int64.
14528 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14529 (emit_push_insn): Likewise.
14530 * lra-eliminations.c (mark_not_eliminable): Likewise.
14531 * recog.c (push_operand): Likewise.
14532 * reload1.c (elimination_effects): Likewise.
14533 * rtlanal.c (nonzero_bits1): Likewise.
14534 * calls.c (store_one_arg): Likewise. Require the padding to be
14535 known at compile time.
14537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14538 Alan Hayward <alan.hayward@arm.com>
14539 David Sherwood <david.sherwood@arm.com>
14541 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14542 Use plus_constant instead of gen_rtx_PLUS.
14544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14545 Alan Hayward <alan.hayward@arm.com>
14546 David Sherwood <david.sherwood@arm.com>
14548 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14549 rather than an int.
14551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14552 Alan Hayward <alan.hayward@arm.com>
14553 David Sherwood <david.sherwood@arm.com>
14555 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14556 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14557 via stack temporaries. Treat the mode size as polynomial too.
14559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14560 Alan Hayward <alan.hayward@arm.com>
14561 David Sherwood <david.sherwood@arm.com>
14563 * expr.c (expand_expr_real_2): When handling conversions involving
14564 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14565 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14566 as a poly_uint64 too.
14568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14569 Alan Hayward <alan.hayward@arm.com>
14570 David Sherwood <david.sherwood@arm.com>
14572 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14575 Alan Hayward <alan.hayward@arm.com>
14576 David Sherwood <david.sherwood@arm.com>
14578 * combine.c (can_change_dest_mode): Handle polynomial
14579 REGMODE_NATURAL_SIZE.
14580 * expmed.c (store_bit_field_1): Likewise.
14581 * expr.c (store_constructor): Likewise.
14582 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14583 and polynomial REGMODE_NATURAL_SIZE.
14584 (gen_lowpart_common): Likewise.
14585 * reginfo.c (record_subregs_of_mode): Likewise.
14586 * rtlanal.c (read_modify_subreg_p): Likewise.
14588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14589 Alan Hayward <alan.hayward@arm.com>
14590 David Sherwood <david.sherwood@arm.com>
14592 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14593 numbers of elements.
14595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14596 Alan Hayward <alan.hayward@arm.com>
14597 David Sherwood <david.sherwood@arm.com>
14599 * match.pd: Cope with polynomial numbers of vector elements.
14601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14602 Alan Hayward <alan.hayward@arm.com>
14603 David Sherwood <david.sherwood@arm.com>
14605 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14606 in a POINTER_PLUS_EXPR.
14608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14609 Alan Hayward <alan.hayward@arm.com>
14610 David Sherwood <david.sherwood@arm.com>
14612 * omp-simd-clone.c (simd_clone_subparts): New function.
14613 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14614 (ipa_simd_modify_function_body): Likewise.
14616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14617 Alan Hayward <alan.hayward@arm.com>
14618 David Sherwood <david.sherwood@arm.com>
14620 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14621 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14622 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14623 (expand_vector_condition, vector_element): Likewise.
14624 (subparts_gt): New function.
14625 (get_compute_type): Use subparts_gt.
14626 (count_type_subparts): Delete.
14627 (expand_vector_operations_1): Use subparts_gt instead of
14628 count_type_subparts.
14630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14631 Alan Hayward <alan.hayward@arm.com>
14632 David Sherwood <david.sherwood@arm.com>
14634 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14635 (vect_compile_time_alias): ...this new function. Do the calculation
14636 on poly_ints rather than trees.
14637 (vect_prune_runtime_alias_test_list): Update call accordingly.
14639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14640 Alan Hayward <alan.hayward@arm.com>
14641 David Sherwood <david.sherwood@arm.com>
14643 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14645 (vect_schedule_slp_instance): Likewise.
14647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14648 Alan Hayward <alan.hayward@arm.com>
14649 David Sherwood <david.sherwood@arm.com>
14651 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14652 constant and extern definitions for variable-length vectors.
14653 (vect_get_constant_vectors): Note that the number of units
14654 is known to be constant.
14656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14657 Alan Hayward <alan.hayward@arm.com>
14658 David Sherwood <david.sherwood@arm.com>
14660 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14661 of units as polynomial. Choose between WIDE and NARROW based
14664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14665 Alan Hayward <alan.hayward@arm.com>
14666 David Sherwood <david.sherwood@arm.com>
14668 * tree-vect-stmts.c (simd_clone_subparts): New function.
14669 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14672 Alan Hayward <alan.hayward@arm.com>
14673 David Sherwood <david.sherwood@arm.com>
14675 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14676 vectors as polynomial. Use build_index_vector for
14677 IFN_GOMP_SIMD_LANE.
14679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14680 Alan Hayward <alan.hayward@arm.com>
14681 David Sherwood <david.sherwood@arm.com>
14683 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14684 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14685 for variable-length vectors.
14686 (vectorizable_mask_load_store): Treat the number of units as
14687 polynomial, asserting that it is constant if the condition has
14688 already been enforced.
14689 (vectorizable_store, vectorizable_load): Likewise.
14691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14692 Alan Hayward <alan.hayward@arm.com>
14693 David Sherwood <david.sherwood@arm.com>
14695 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14696 of units as polynomial. Punt if we can't tell at compile time
14697 which vector contains the final result.
14699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14700 Alan Hayward <alan.hayward@arm.com>
14701 David Sherwood <david.sherwood@arm.com>
14703 * tree-vect-loop.c (vectorizable_induction): Treat the number
14704 of units as polynomial. Punt on SLP inductions. Use an integer
14705 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14706 cast of such a series for variable-length floating-point
14709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14710 Alan Hayward <alan.hayward@arm.com>
14711 David Sherwood <david.sherwood@arm.com>
14713 * tree.h (build_index_vector): Declare.
14714 * tree.c (build_index_vector): New function.
14715 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14716 of units as polynomial, forcibly converting it to a constant if
14717 vectorizable_reduction has already enforced the condition.
14718 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14719 to create a {1,2,3,...} vector.
14720 (vectorizable_reduction): Treat the number of units as polynomial.
14721 Choose vectype_in based on the largest scalar element size rather
14722 than the smallest number of units. Enforce the restrictions
14725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14726 Alan Hayward <alan.hayward@arm.com>
14727 David Sherwood <david.sherwood@arm.com>
14729 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14730 number of units as polynomial.
14732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14733 Alan Hayward <alan.hayward@arm.com>
14734 David Sherwood <david.sherwood@arm.com>
14736 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14737 * target.def (autovectorize_vector_sizes): Return the vector sizes
14738 by pointer, using vector_sizes rather than a bitmask.
14739 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14740 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14741 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14743 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14744 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14745 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14746 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14747 * omp-general.c (omp_max_vf): Likewise.
14748 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14749 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14750 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14751 * tree-vect-slp.c (vect_slp_bb): Likewise.
14752 * doc/tm.texi: Regenerate.
14753 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14755 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14756 the vector size as a poly_uint64 rather than an unsigned int.
14757 (current_vector_size): Change from an unsigned int to a poly_uint64.
14758 (get_vectype_for_scalar_type): Update accordingly.
14759 * tree.h (build_truth_vector_type): Take the size and number of
14760 units as a poly_uint64 rather than an unsigned int.
14761 (build_vector_type): Add a temporary overload that takes
14762 the number of units as a poly_uint64 rather than an unsigned int.
14763 * tree.c (make_vector_type): Likewise.
14764 (build_truth_vector_type): Take the number of units as a poly_uint64
14765 rather than an unsigned int.
14767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14768 Alan Hayward <alan.hayward@arm.com>
14769 David Sherwood <david.sherwood@arm.com>
14771 * target.def (get_mask_mode): Take the number of units and length
14772 as poly_uint64s rather than unsigned ints.
14773 * targhooks.h (default_get_mask_mode): Update accordingly.
14774 * targhooks.c (default_get_mask_mode): Likewise.
14775 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14776 * doc/tm.texi: Regenerate.
14778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14779 Alan Hayward <alan.hayward@arm.com>
14780 David Sherwood <david.sherwood@arm.com>
14782 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14783 * omp-general.c (omp_max_vf): Likewise.
14784 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14785 (expand_omp_simd): Handle polynomial safelen.
14786 * omp-low.c (omplow_simd_context): Add a default constructor.
14787 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14788 (lower_rec_simd_input_clauses): Update accordingly.
14789 (lower_rec_input_clauses): Likewise.
14791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14792 Alan Hayward <alan.hayward@arm.com>
14793 David Sherwood <david.sherwood@arm.com>
14795 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14796 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14797 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14798 (vect_analyze_slp_cost): Likewise.
14799 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14800 (vect_model_load_cost): Likewise.
14802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14803 Alan Hayward <alan.hayward@arm.com>
14804 David Sherwood <david.sherwood@arm.com>
14806 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14807 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14808 from an unsigned int * to a poly_uint64_pod *.
14809 (calculate_unrolling_factor): New function.
14810 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14813 Alan Hayward <alan.hayward@arm.com>
14814 David Sherwood <david.sherwood@arm.com>
14816 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14817 from an unsigned int to a poly_uint64.
14818 (_loop_vec_info::slp_unrolling_factor): Likewise.
14819 (_loop_vec_info::vectorization_factor): Change from an int
14821 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14822 (vect_get_num_vectors): New function.
14823 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14824 (vect_get_num_copies): Use vect_get_num_vectors.
14825 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14826 to an unsigned int *.
14827 (vect_analyze_data_refs): Change min_vf from an int * to a
14829 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14830 than an unsigned HOST_WIDE_INT.
14831 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14832 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14833 to an unsigned int *.
14834 (vect_analyze_data_ref_dependences): Likewise.
14835 (vect_compute_data_ref_alignment): Handle polynomial vf.
14836 (vect_enhance_data_refs_alignment): Likewise.
14837 (vect_prune_runtime_alias_test_list): Likewise.
14838 (vect_shift_permute_load_chain): Likewise.
14839 (vect_supportable_dr_alignment): Likewise.
14840 (dependence_distance_ge_vf): Take the vectorization factor as a
14841 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14842 (vect_analyze_data_refs): Change min_vf from an int * to a
14844 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14845 vfm1 as a poly_uint64 rather than an int. Make the same change
14846 for the returned bound_scalar.
14847 (vect_gen_vector_loop_niters): Handle polynomial vf.
14848 (vect_do_peeling): Likewise. Update call to
14849 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14850 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14852 * tree-vect-loop.c (vect_determine_vectorization_factor)
14853 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14854 (vect_get_known_peeling_cost): Likewise.
14855 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14856 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14857 (vect_transform_loop): Likewise. Use the lowest possible VF when
14858 updating the upper bounds of the loop.
14859 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14860 rather than an int.
14861 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14862 polynomial unroll factors.
14863 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14864 (vect_make_slp_decision): Likewise.
14865 (vect_supported_load_permutation_p): Likewise, and polynomial
14867 (vect_analyze_slp_cost): Handle polynomial vf.
14868 (vect_slp_analyze_node_operations): Likewise.
14869 (vect_slp_analyze_bb_1): Likewise.
14870 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14871 than an unsigned HOST_WIDE_INT.
14872 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14873 (vectorizable_load): Handle polynomial vf.
14874 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14876 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14879 Alan Hayward <alan.hayward@arm.com>
14880 David Sherwood <david.sherwood@arm.com>
14882 * match.pd: Handle bit operations involving three constants
14883 and try to fold one pair.
14885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14887 * tree-vect-loop-manip.c: Include gimple-fold.h.
14888 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14889 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14890 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14891 Add a path that uses a step of VF instead of 1, but disable it
14893 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14894 and niters_no_overflow parameters. Update calls to
14895 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14896 Create a new SSA name if the latter choses to use a ste other
14897 than zero, and return it via niters_vector_mult_vf_var.
14898 * tree-vect-loop.c (vect_transform_loop): Update calls to
14899 vect_do_peeling, vect_gen_vector_loop_niters and
14900 slpeel_make_loop_iterate_ntimes.
14901 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14902 (vect_gen_vector_loop_niters): Update declarations after above changes.
14904 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14906 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14907 128-bit round to integer instructions.
14908 (ceil<mode>2): Likewise.
14909 (btrunc<mode>2): Likewise.
14910 (round<mode>2): Likewise.
14912 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14914 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14915 unaligned VSX load/store on P8/P9.
14916 (expand_block_clear): Allow the use of unaligned VSX
14917 load/store on P8/P9.
14919 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14921 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14923 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14924 swap associated with both a load and a store.
14926 2018-01-02 Andrew Waterman <andrew@sifive.com>
14928 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14929 * config/riscv/riscv.md (clear_cache): Use it.
14931 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14933 * web.c: Remove out-of-date comment.
14935 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14937 * expr.c (fixup_args_size_notes): Check that any existing
14938 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14939 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14940 (emit_single_push_insn): ...here.
14942 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14944 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14945 (const_vector_encoded_nelts): New function.
14946 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14947 (const_vector_int_elt, const_vector_elt): Declare.
14948 * emit-rtl.c (const_vector_int_elt_1): New function.
14949 (const_vector_elt): Likewise.
14950 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14951 of CONST_VECTOR_ELT.
14953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14955 * expr.c: Include rtx-vector-builder.h.
14956 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14957 directly on the tree encoding.
14958 (const_vector_from_tree): Likewise.
14959 * optabs.c: Include rtx-vector-builder.h.
14960 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14961 sequence of "u" values.
14962 * vec-perm-indices.c: Include rtx-vector-builder.h.
14963 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14964 directly on the vec_perm_indices encoding.
14966 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14968 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14969 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14970 * rtx-vector-builder.h: New file.
14971 * rtx-vector-builder.c: Likewise.
14972 * rtl.h (rtx_def::u2): Add a const_vector field.
14973 (CONST_VECTOR_NPATTERNS): New macro.
14974 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14975 (CONST_VECTOR_DUPLICATE_P): Likewise.
14976 (CONST_VECTOR_STEPPED_P): Likewise.
14977 (CONST_VECTOR_ENCODED_ELT): Likewise.
14978 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14979 (unwrap_const_vec_duplicate): Likewise.
14980 (const_vec_series_p): Check for a non-duplicated vector encoding.
14981 Say that the function only returns true for integer vectors.
14982 * emit-rtl.c: Include rtx-vector-builder.h.
14983 (gen_const_vec_duplicate_1): Delete.
14984 (gen_const_vector): Call gen_const_vec_duplicate instead of
14985 gen_const_vec_duplicate_1.
14986 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14987 (gen_const_vec_duplicate): Use rtx_vector_builder.
14988 (gen_const_vec_series): Likewise.
14989 (gen_rtx_CONST_VECTOR): Likewise.
14990 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14991 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14992 Build a new vector rather than modifying a CONST_VECTOR in-place.
14993 (handle_special_swappables): Update call accordingly.
14994 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14995 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14996 Build a new vector rather than modifying a CONST_VECTOR in-place.
14997 (handle_special_swappables): Update call accordingly.
14999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15001 * simplify-rtx.c (simplify_const_binary_operation): Use
15002 CONST_VECTOR_ELT instead of XVECEXP.
15004 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15006 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15007 the selector elements to be different from the data elements
15008 if the selector is a VECTOR_CST.
15009 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15010 ssizetype for the selector.
15012 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15014 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15015 before testing each element individually.
15016 * tree-vect-generic.c (lower_vec_perm): Likewise.
15018 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15020 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15021 * selftest-run-tests.c (selftest::run_tests): Call it.
15022 * vector-builder.h (vector_builder::operator ==): New function.
15023 (vector_builder::operator !=): Likewise.
15024 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15025 (vec_perm_indices::all_from_input_p): New function.
15026 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15027 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15028 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15029 instead of reading the VECTOR_CST directly. Detect whether both
15030 vector inputs are the same before constructing the vec_perm_indices,
15031 and update the number of inputs argument accordingly. Use the
15032 utility functions added above. Only construct sel2 if we need to.
15034 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15036 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15037 the broadcast of the low byte.
15038 (expand_mult_highpart): Use an explicit encoding for the permutes.
15039 * optabs-query.c (can_mult_highpart_p): Likewise.
15040 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15041 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15042 (vectorizable_bswap): Likewise.
15043 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15044 explicit encoding for the power-of-2 permutes.
15045 (vect_permute_store_chain): Likewise.
15046 (vect_grouped_load_supported): Likewise.
15047 (vect_permute_load_chain): Likewise.
15049 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15051 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15052 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15053 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15054 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15055 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15056 (vect_gen_perm_mask_any): Likewise.
15058 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15060 * int-vector-builder.h: New file.
15061 * vec-perm-indices.h: Include int-vector-builder.h.
15062 (vec_perm_indices): Redefine as an int_vector_builder.
15063 (auto_vec_perm_indices): Delete.
15064 (vec_perm_builder): Redefine as a stand-alone class.
15065 (vec_perm_indices::vec_perm_indices): New function.
15066 (vec_perm_indices::clamp): Likewise.
15067 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15068 (vec_perm_indices::new_vector): New function.
15069 (vec_perm_indices::new_expanded_vector): Update for new
15070 vec_perm_indices class.
15071 (vec_perm_indices::rotate_inputs): New function.
15072 (vec_perm_indices::all_in_range_p): Operate directly on the
15073 encoded form, without computing elided elements.
15074 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15075 encoding. Update for new vec_perm_indices class.
15076 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15077 the given vec_perm_builder.
15078 (expand_vec_perm_var): Update vec_perm_builder constructor.
15079 (expand_mult_highpart): Use vec_perm_builder instead of
15080 auto_vec_perm_indices.
15081 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15082 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15083 or double series encoding as appropriate.
15084 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15085 vec_perm_indices instead of auto_vec_perm_indices.
15086 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15087 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15088 (vect_permute_store_chain): Likewise.
15089 (vect_grouped_load_supported): Likewise.
15090 (vect_permute_load_chain): Likewise.
15091 (vect_shift_permute_load_chain): Likewise.
15092 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15093 (vect_transform_slp_perm_load): Likewise.
15094 (vect_schedule_slp_instance): Likewise.
15095 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15096 (vectorizable_mask_load_store): Likewise.
15097 (vectorizable_bswap): Likewise.
15098 (vectorizable_store): Likewise.
15099 (vectorizable_load): Likewise.
15100 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15101 vec_perm_indices instead of auto_vec_perm_indices. Use
15102 tree_to_vec_perm_builder to read the vector from a tree.
15103 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15104 vec_perm_builder instead of a vec_perm_indices.
15105 (have_whole_vector_shift): Use vec_perm_builder and
15106 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15107 truncation to calc_vec_perm_mask_for_shift.
15108 (vect_create_epilog_for_reduction): Likewise.
15109 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15110 from auto_vec_perm_indices to vec_perm_indices.
15111 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15112 instead of changing individual elements.
15113 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15114 the vector in d.perm.
15115 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15116 from auto_vec_perm_indices to vec_perm_indices.
15117 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15118 instead of changing individual elements.
15119 (arm_vectorize_vec_perm_const): Use new_vector to install
15120 the vector in d.perm.
15121 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15122 Update vec_perm_builder constructor.
15123 (rs6000_expand_interleave): Likewise.
15124 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15125 (rs6000_expand_interleave): Likewise.
15127 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15129 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15130 to qimode could truncate the indices.
15131 * optabs.c (expand_vec_perm_var): Likewise.
15133 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15135 * Makefile.in (OBJS): Add vec-perm-indices.o.
15136 * vec-perm-indices.h: New file.
15137 * vec-perm-indices.c: Likewise.
15138 * target.h (vec_perm_indices): Replace with a forward class
15140 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15141 * optabs.h: Include vec-perm-indices.h.
15142 (expand_vec_perm): Delete.
15143 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15144 (expand_vec_perm_const): Declare.
15145 * target.def (vec_perm_const_ok): Replace with...
15146 (vec_perm_const): ...this new hook.
15147 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15148 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15149 * doc/tm.texi: Regenerate.
15150 * optabs.def (vec_perm_const): Delete.
15151 * doc/md.texi (vec_perm_const): Likewise.
15152 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15153 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15154 expand_vec_perm for constant permutation vectors. Assert that
15155 the mode of variable permutation vectors is the integer equivalent
15156 of the mode that is being permuted.
15157 * optabs-query.h (selector_fits_mode_p): Declare.
15158 * optabs-query.c: Include vec-perm-indices.h.
15159 (selector_fits_mode_p): New function.
15160 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15161 is defined, instead of checking whether the vec_perm_const_optab
15162 exists. Use targetm.vectorize.vec_perm_const instead of
15163 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15164 fit in the vector mode before using a variable permute.
15165 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15166 vec_perm_indices instead of an rtx.
15167 (expand_vec_perm): Replace with...
15168 (expand_vec_perm_const): ...this new function. Take the selector
15169 as a vec_perm_indices rather than an rtx. Also take the mode of
15170 the selector. Update call to shift_amt_for_vec_perm_mask.
15171 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15172 Use vec_perm_indices::new_expanded_vector to expand the original
15173 selector into bytes. Check whether the indices fit in the vector
15174 mode before using a variable permute.
15175 (expand_vec_perm_var): Make global.
15176 (expand_mult_highpart): Use expand_vec_perm_const.
15177 * fold-const.c: Includes vec-perm-indices.h.
15178 * tree-ssa-forwprop.c: Likewise.
15179 * tree-vect-data-refs.c: Likewise.
15180 * tree-vect-generic.c: Likewise.
15181 * tree-vect-loop.c: Likewise.
15182 * tree-vect-slp.c: Likewise.
15183 * tree-vect-stmts.c: Likewise.
15184 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15186 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15187 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15188 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15189 (aarch64_vectorize_vec_perm_const): ...this new function.
15190 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15191 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15192 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15193 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15194 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15195 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15196 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15198 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15199 check for NEON modes.
15200 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15201 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15202 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15203 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15205 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15206 the old VEC_PERM_CONST conditions.
15207 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15208 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15209 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15210 (ia64_vectorize_vec_perm_const_ok): Merge into...
15211 (ia64_vectorize_vec_perm_const): ...this new function.
15212 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15213 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15214 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15215 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15216 * config/mips/mips.c (mips_expand_vec_perm_const)
15217 (mips_vectorize_vec_perm_const_ok): Merge into...
15218 (mips_vectorize_vec_perm_const): ...this new function.
15219 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15220 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15221 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15222 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15223 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15224 (rs6000_expand_vec_perm_const): Delete.
15225 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15227 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15228 (altivec_expand_vec_perm_const_le): Take each operand individually.
15229 Operate on constant selectors rather than rtxes.
15230 (altivec_expand_vec_perm_const): Likewise. Update call to
15231 altivec_expand_vec_perm_const_le.
15232 (rs6000_expand_vec_perm_const): Delete.
15233 (rs6000_vectorize_vec_perm_const_ok): Delete.
15234 (rs6000_vectorize_vec_perm_const): New function.
15235 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15236 an element count and rtx array.
15237 (rs6000_expand_extract_even): Update call accordingly.
15238 (rs6000_expand_interleave): Likewise.
15239 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15240 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15241 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15242 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15243 (rs6000_expand_vec_perm_const): Delete.
15244 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15245 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15246 (altivec_expand_vec_perm_const_le): Take each operand individually.
15247 Operate on constant selectors rather than rtxes.
15248 (altivec_expand_vec_perm_const): Likewise. Update call to
15249 altivec_expand_vec_perm_const_le.
15250 (rs6000_expand_vec_perm_const): Delete.
15251 (rs6000_vectorize_vec_perm_const_ok): Delete.
15252 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15253 reference to the SPE evmerge intructions.
15254 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15255 an element count and rtx array.
15256 (rs6000_expand_extract_even): Update call accordingly.
15257 (rs6000_expand_interleave): Likewise.
15258 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15259 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15261 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15265 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15266 vector mode and that that mode matches the mode of the data
15268 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15269 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15270 directly using expand_vec_perm_1 when forcing selectors into
15272 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15274 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15276 * optabs-query.h (can_vec_perm_p): Delete.
15277 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15278 * optabs-query.c (can_vec_perm_p): Split into...
15279 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15280 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15281 particular selector is valid.
15282 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15283 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15284 (vect_grouped_load_supported): Likewise.
15285 (vect_shift_permute_load_chain): Likewise.
15286 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15287 (vect_transform_slp_perm_load): Likewise.
15288 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15289 (vectorizable_bswap): Likewise.
15290 (vect_gen_perm_mask_checked): Likewise.
15291 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15292 implementations of variable permutation vectors into account
15293 when deciding which selector to use.
15294 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15295 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15296 with a false third argument.
15297 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15298 to test whether the constant selector is valid and can_vec_perm_var_p
15299 to test whether a variable selector is valid.
15301 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15303 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15304 * optabs-query.c (can_vec_perm_p): Likewise.
15305 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15306 instead of vec_perm_indices.
15307 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15308 (vect_gen_perm_mask_checked): Likewise,
15309 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15310 (vect_gen_perm_mask_checked): Likewise,
15312 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15314 * optabs-query.h (qimode_for_vec_perm): Declare.
15315 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15316 (qimode_for_vec_perm): ...this new function.
15317 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15319 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15321 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15322 does not have a conditional at the top.
15324 2018-01-02 Richard Biener <rguenther@suse.de>
15326 * ipa-inline.c (big_speedup_p): Fix expression.
15328 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15331 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15334 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15338 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15339 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15340 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15341 cond_taken_branch_cost 3->4.
15343 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15345 PR tree-optimization/83581
15346 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15347 TODO_cleanup_cfg if any changes have been made.
15349 PR middle-end/83608
15350 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15351 convert_modes if target mode has the right side, but different mode
15354 PR middle-end/83609
15355 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15356 last argument when extracting from CONCAT. If either from_real or
15357 from_imag is NULL, use expansion through memory. If result is not
15358 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15359 the parts directly to inner mode, if even that fails, use expansion
15362 PR middle-end/83623
15363 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15364 check for bswap in mode rather than HImode and use that in expand_unop
15367 Copyright (C) 2018 Free Software Foundation, Inc.
15369 Copying and distribution of this file, with or without modification,
15370 are permitted in any medium without royalty provided the copyright
15371 notice and this notice are preserved.