1 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5 Don't handle -mcpu=power8 if -mpower9-vector is also used.
7 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
11 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12 several redundant entries.
14 2018-06-06 David Malcolm <dmalcolm@redhat.com>
16 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
17 type from "rtx" to "rtx_insn *".
18 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
19 for local "call_insn", removing cast.
20 (ix86_expand_call): Likewise, introducing a "call_insn" local.
22 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
24 PR tree-optimization/86066
25 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
26 for BIT_INSERT_EXPR stores.
28 2018-06-06 Richard Biener <rguenther@suse.de>
30 PR tree-optimization/86062
31 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
33 of to be offsetted base.
35 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
37 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
38 to be static and remove check on interrupt attribute name.
40 gcc/testsuite/gcc.target/msp430/
41 * function-attributes-4.c: New test.
42 * static-interrupts.c: New test.
44 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
46 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
47 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
49 2018-06-05 Steve Ellcey <sellcey@cavium.com>
52 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
54 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
55 Remove second argument, change how error is called.
56 (aarch64_layout_arg): Remove second argument from
57 aarch64_err_no_fpadvsimd call.
58 (aarch64_init_cumulative_args): Ditto.
59 (aarch64_gimplify_va_arg_expr): Ditto.
60 * config/aarch64/aarch64.md (mov<mode>): Ditto.
62 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
64 * config/i386/i386.md (simple_return_indirect_internal): New expander.
65 (*simple_return_indirect_internal<mode>): Rename from
66 simple_return_indirect_internal. Use W mode iterator.
67 (rstorssp): New expander.
68 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
69 (clrssbsy): New expander.
70 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
72 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
74 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
76 (cmse_check_pointed_object): Likewise.
78 2018-06-05 Martin Liska <mliska@suse.cz>
81 * doc/invoke.texi: Document how -fprofile-dir format
84 2018-06-05 Richard Biener <rguenther@suse.de>
86 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
87 removal pretend DOM info isn't available so we do not update
88 it and only remove edges, not dominated blocks. Actually free
89 DOM info in case we removed something. Remove unreachable blocks.
90 (mfb_keep_latches): Work with either DOM info or marked backedges.
91 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
92 first. Mark backedges if DOM info isn't available.
93 (Re-)compute DOM info after cleanup_control_flow_pre.
95 2018-06-05 Richard Biener <rguenther@suse.de>
97 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
98 (locus_discrim_hasher::hash): Adjust.
99 (locus_discrim_hasher::equal): Likewise.
100 (next_discriminator_for_locus): Work on line directly.
101 (same_line_p): Pass in expanded locus1 as well.
102 (assign_discriminators): Avoid redundant location expansions.
104 2018-06-05 Richard Biener <rguenther@suse.de>
106 PR tree-optimization/86046
107 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
108 if required after clearing TREE_ADDRESSABLE.
110 2018-06-05 Richard Biener <rguenther@suse.de>
112 PR tree-optimization/86047
113 * tree-ssa-loop.c (for_each_index): Glob handling of all
114 decls and constants and really handle all of them.
116 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
119 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
120 qualifier_void_pointer and qualifier_const_void_pointer.
121 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
122 (arm_init_builtins): Handle the above.
123 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
124 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
125 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
128 2018-06-05 Martin Liska <mliska@suse.cz>
130 * auto-profile.c (read_autofdo_file): Do not use
131 gcov_ctr_summary struct.
132 (afdo_callsite_hot_enough_for_early_inline): Likewise.
133 * coverage.c (struct counts_entry): Likewise.
134 (read_counts_file): Read just single summary entry.
135 (get_coverage_counts): Use gcov_summary struct.
136 * coverage.h (get_coverage_counts): Likewise.
137 * gcov-dump.c (dump_working_sets): Likewise.
138 (tag_summary): Dump just single summary.
139 * gcov-io.c (gcov_write_summary): Write just histogram
141 (gcov_read_summary): Read just single summary.
142 (compute_working_sets): Use gcov_summary struct.
143 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
144 of GCOV_COUNTERS_SUMMABLE.
145 (GCOV_COUNTERS_SUMMABLE): Remove.
146 (GCOV_FIRST_VALUE_COUNTER): Replace with
147 GCOV_COUNTER_V_INTERVAL.
148 (struct gcov_ctr_summary): Remove.
149 (struct gcov_summary): Directly use fields of former
151 (compute_working_sets): Use gcov_summary struct.
152 * gcov.c (read_count_file): Do not use ctrs fields.
153 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
155 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
157 * profile.c: Likewise.
158 * profile.h: Likewise.
160 2018-06-05 Martin Liska <mliska@suse.cz>
162 PR gcov-profile/84846
163 * gcov.c (output_lines): Print working directory only
164 in intermediate format.
166 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
168 * config/s390/s390-builtin-types.def: Add void function type.
169 * config/s390/s390-builtins.def: Use the function type for the
172 2018-06-04 Jim Wilson <jimw@sifive.com>
174 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
176 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
177 and maybe_eh_return. Change regno to unsigned int. Use new args to
178 handle EH_RETURN_DATA_REGNO registers properly.
179 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
180 (riscv_expand_epilogue): Update comment. Change argument name and
181 type. Update code to use new name and type. Pass new args to
182 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
184 * config/riscv/riscv.md (NORMAL_RETURN): New.
185 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
186 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
187 (eh_return): Call gen_eh_return_internal and emit barrier.
188 (eh_return_internal): Call riscv_expand_epilogue.
190 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
192 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
193 bit_insertion field and declare can_be_merged_into method.
194 (merged_store_group::can_be_merged_into): New method.
195 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
196 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
197 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
199 2018-06-04 Richard Biener <rguenther@suse.de>
201 PR tree-optimization/85955
202 * builtins.c (fold_builtin_sincos): Convert pointers to
203 destination to appropriate type before dereferencing.
205 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
207 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
209 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
211 * expr.c (expand_expr_real_1): Force the operand into memory if
212 its TYPE_MODE is BLKmode and if there is no integer mode for
213 the number of bits being extracted.
215 2018-06-04 Jakub Jelinek <jakub@redhat.com>
219 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
220 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
222 2018-06-04 Richard Biener <rguenther@suse.de>
224 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
225 (cleanup_tree_cfg_noloop): ... single caller. Do
226 start_recording_case_labels later.
228 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
230 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
231 to _IMMINTRIN_H_INCLUDED.
232 * config/i386/pconfigintrin.h: Ditto.
233 * config/i386/waitpkgintrin.h: Ditto.
234 * config/i386/immintrin.h: Add includes for sgxintrin.h,
235 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
236 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
237 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
238 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
239 waitpkgintrin.h and cldemoteintrin.h.
241 2018-06-04 Richard Biener <rguenther@suse.de>
243 PR tree-optimization/86038
244 * tracer.c (find_best_successor): Check probability for
245 being initialized, bail out if not.
247 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
250 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
251 of bits to ignore when comparing architectures.
253 2018-06-04 Jakub Jelinek <jakub@redhat.com>
255 PR tree-optimization/69615
256 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
257 maximum or minimum of the type, try to merge it also as if
258 range1 is + [-, x - 1] or + [x + 1, -].
261 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
263 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
265 PR tree-optimization/86034
266 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
267 the unsigned bitfield type in a bit insertion sequence if it does not
268 have a larger precision than the bitfield size.
269 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
271 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
273 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
275 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
277 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
278 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
279 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
280 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
282 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
284 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
285 Disable -fdelete-null-pointer-checks for ELF toolchain.
287 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
288 Kito Cheng <kito.cheng@gmail.com>
290 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
291 (nds32le-*-*, nds32be-*-*): Integrate checking process.
292 (nds32*-*-*): Add glibc and uclibc conditions.
293 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
294 (TARGET_EXCEPT_UNWIND_INFO): Define.
295 * config/nds32/elf.h: New file.
296 * config/nds32/linux.h: New file.
297 * config/nds32/nds32-elf.opt: New file.
298 * config/nds32/nds32-linux.opt: New file.
299 * config/nds32/nds32-fp-as-gp.c
300 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
301 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
303 (nds32_asm_file_end): Ditto.
304 (nds32_print_operand): Ditto.
305 (nds32_insert_attributes): Ditto.
306 (nds32_init_libfuncs): New function.
307 (TARGET_HAVE_TLS): Define.
308 (TARGET_INIT_LIBFUNCS): Define.
309 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
311 (TARGET_ELF): Apply different mcmodel setting.
312 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
313 been migrated into elf.h and linux.h files.
314 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
315 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
316 (mcmodel): The content has been migrated into nds32-elf.opt and
317 nds32-linux.opt files.
318 * config/nds32/t-elf: New file.
319 * config/nds32/t-linux: New file.
321 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
322 Shiva Chen <shiva0217@gmail.com>
324 * config/nds32/constants.md (unspec_volatile_element): Add
325 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
326 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
328 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
329 (make_pass_nds32_fp_as_gp): Declare.
330 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
332 (nds32_asm_function_end_prologue): Remove unused asm output.
333 (nds32_asm_function_begin_epilogue): Remove unused asm output.
334 (nds32_asm_file_start): Output necessary fp_as_gp information.
335 (nds32_option_override): Adjust register usage.
336 (nds32_expand_prologue): Consider fp_as_gp situation.
337 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
338 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
341 (simple_return): Ditto.
342 (omit_fp_begin): Output special directive for fp_as_gp.
343 (omit_fp_end): Output special directive for fp_as_gp.
344 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
345 mforbid-fp-as-gp): New options.
347 2018-06-01 Mark Wielaard <mark@klomp.org>
349 * dwarf2out.c (dwarf2out_finish): Remove generation of
352 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
354 * gimple-ssa-store-merging.c: Include gimple-fold.h.
355 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
356 (struct merged_store_group): Add bit_insertion field.
357 (dump_char_array): Use standard hexadecimal format.
358 (merged_store_group::merged_store_group): Set bit_insertion to false.
359 (merged_store_group::apply_stores): Use optimal buffer size. Deal
360 with BIT_INSERT_EXPR stores. Move up code updating the mask and
361 also print the mask in the dump file.
362 (pass_store_merging::gate): Minor tweak.
363 (imm_store_chain_info::coalesce_immediate): Fix wrong association
364 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
365 stores with INTEGER_CST stores.
366 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
367 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
368 and use it throughout. Generate bit insertion sequences if need be.
369 (pass_store_merging::process_store): Remove redundant condition.
370 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
372 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
374 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
375 the 128-bit floating point types. Fix function comment.
377 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
379 * config/aarch64/aarch64-simd.md
380 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
382 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
385 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
387 PR tree-optimization/85989
388 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
390 (backprop::intersect_uses): Check it when deciding whether this
391 is a backedge reference.
392 (backprop::process_block): Add each phi to m_visited_phis
393 after visiting it, then clear it at the end.
395 2018-06-01 Richard Biener <rguenther@suse.de>
397 * tree-vectorizer.h (vect_dr_stmt): New function.
398 (vect_get_load_cost): Adjust.
399 (vect_get_store_cost): Likewise.
400 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
401 Use vect_dr_stmt instead of DR_SMTT.
402 (vect_record_base_alignments): Likewise.
403 (vect_calculate_target_alignment): Likewise.
404 (vect_compute_data_ref_alignment): Likewise and make static.
405 (vect_update_misalignment_for_peel): Likewise.
406 (vect_verify_datarefs_alignment): Likewise.
407 (vector_alignment_reachable_p): Likewise.
408 (vect_get_data_access_cost): Likewise. Pass down
409 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
410 (vect_get_peeling_costs_all_drs): Likewise.
411 (vect_peeling_hash_get_lowest_cost): Likewise.
412 (vect_enhance_data_refs_alignment): Likewise.
413 (vect_find_same_alignment_drs): Likewise.
414 (vect_analyze_data_refs_alignment): Likewise.
415 (vect_analyze_group_access_1): Likewise.
416 (vect_analyze_group_access): Likewise.
417 (vect_analyze_data_ref_access): Likewise.
418 (vect_analyze_data_ref_accesses): Likewise.
419 (vect_vfa_segment_size): Likewise.
420 (vect_small_gap_p): Likewise.
421 (vectorizable_with_step_bound_p): Likewise.
422 (vect_prune_runtime_alias_test_list): Likewise.
423 (vect_analyze_data_refs): Likewise.
424 (vect_supportable_dr_alignment): Likewise.
425 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
426 (vect_gen_prolog_loop_niters): Likewise.
427 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
428 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
430 (vect_recog_mask_conversion_pattern): Likewise.
431 (vect_try_gather_scatter_pattern): Likewise.
432 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
433 to vect_get_store_cost.
434 (vect_get_store_cost): Get stmt_info instead of DR.
435 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
436 (vect_get_load_cost): Get stmt_info instead of DR.
438 2018-06-01 Richard Biener <rguenther@suse.de>
441 * gimple-fold.c (var_decl_component_p): Also allow offsetted
442 vars wrapped in MEM_REFs.
444 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
446 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
447 Fix subreg tests so that we only return a choice between
448 GENERAL_REGS and FP_REGS if the original classes included both.
450 2018-06-01 Richard Biener <rguenther@suse.de>
453 * tree-ssa-structalias.c (get_function_part_constraint):
454 Handle NULL fi->decl.
455 (find_func_aliases_for_call): Properly handle indirect
457 (find_func_clobbers): Likewise.
458 (ipa_pta_execute): Likewise.
459 (create_variable_info_for): For functions that are ifunc_resolver
460 resolve to a varinfo that contains the result of the resolver call.
461 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
464 2018-05-31 Michael Collison <michael.collison@arm.com>
466 * config/aarch64/aarch64.md:
467 (*fix_to_zero_extenddfdi2): New pattern.
468 * gcc.target/aarch64/fix_extend1.c: New testcase.
470 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
474 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
475 and BUILT_IN_STRNCMP_EQ.
476 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
478 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
479 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
480 (gimple_fold_builtin): Likewise.
481 * tree-ssa-strlen.c (compute_string_length): New function.
482 (determine_min_obsize): New function.
483 (handle_builtin_string_cmp): New function to handle calls to
484 string compare functions.
485 (strlen_optimize_stmt): Add handling to builtin string compare
487 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
488 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
489 * tree.c (build_common_builtin_nodes): Add new defines of
490 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
492 2018-05-31 Jakub Jelinek <jakub@redhat.com>
495 * bb-reorder.c (pass_partition_blocks::gate): Return false for
496 functions with naked attribute.
498 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
500 * config/i386/sse.md (avx_vec_concat<mode>):
501 Substitute concat_tg_mode mode attribute with xtg_mode.
502 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
503 (concat_tg_mode): Remove mode attribute.
505 2018-05-31 Martin Sebor <msebor@redhat.com>
508 * calls.c (alloc_max_size): Correct a logic error/typo.
509 Treat excessive arguments as infinite. Warn for invalid arguments.
510 * doc/invoke.texi (-Walloc-size-larger-than): Update.
512 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
515 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
516 and movx for Haswell.
518 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
519 Cesar Philippidis <cesar@codesourcery.com>
522 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
523 when emitting error on private/firstprivate reductions.
524 * omp-low.c (lower_omp_target): Avoid reference-type processing
525 on pointers for firstprivate clause.
527 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
529 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
532 * config/aarch64/aarch64-simd.md
533 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
534 (aarch64_ld1_x3_<mode>): Likewise
535 (aarch64_st1x2<VALLDIF:mode>): Likewise
536 (aarch64_st1_x2_<mode>): Likewise
537 (aarch64_st1x3<VALLDIF:mode>): Likewise
538 (aarch64_st1_x3_<mode>): Likewise
539 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
540 (vld1_s8_x3): Likewise.
541 (vld1_u16_x3): Likewise.
542 (vld1_s16_x3): Likewise.
543 (vld1_u32_x3): Likewise.
544 (vld1_s32_x3): Likewise.
545 (vld1_u64_x3): Likewise.
546 (vld1_s64_x3): Likewise.
547 (vld1_f16_x3): Likewise.
548 (vld1_f32_x3): Likewise.
549 (vld1_f64_x3): Likewise.
550 (vld1_p8_x3): Likewise.
551 (vld1_p16_x3): Likewise.
552 (vld1_p64_x3): Likewise.
553 (vld1q_u8_x3): Likewise.
554 (vld1q_s8_x3): Likewise.
555 (vld1q_u16_x3): Likewise.
556 (vld1q_s16_x3): Likewise.
557 (vld1q_u32_x3): Likewise.
558 (vld1q_s32_x3): Likewise.
559 (vld1q_u64_x3): Likewise.
560 (vld1q_s64_x3): Likewise.
561 (vld1q_f16_x3): Likewise.
562 (vld1q_f32_x3): Likewise.
563 (vld1q_f64_x3): Likewise.
564 (vld1q_p8_x3): Likewise.
565 (vld1q_p16_x3): Likewise.
566 (vld1q_p64_x3): Likewise.
567 (vst1_s64_x2): Likewise.
568 (vst1_u64_x2): Likewise.
569 (vst1_f64_x2): Likewise.
570 (vst1_s8_x2): Likewise.
571 (vst1_p8_x2): Likewise.
572 (vst1_s16_x2): Likewise.
573 (vst1_p16_x2): Likewise.
574 (vst1_s32_x2): Likewise.
575 (vst1_u8_x2): Likewise.
576 (vst1_u16_x2): Likewise.
577 (vst1_u32_x2): Likewise.
578 (vst1_f16_x2): Likewise.
579 (vst1_f32_x2): Likewise.
580 (vst1_p64_x2): Likewise.
581 (vst1q_s8_x2): Likewise.
582 (vst1q_p8_x2): Likewise.
583 (vst1q_s16_x2): Likewise.
584 (vst1q_p16_x2): Likewise.
585 (vst1q_s32_x2): Likewise.
586 (vst1q_s64_x2): Likewise.
587 (vst1q_u8_x2): Likewise.
588 (vst1q_u16_x2): Likewise.
589 (vst1q_u32_x2): Likewise.
590 (vst1q_u64_x2): Likewise.
591 (vst1q_f16_x2): Likewise.
592 (vst1q_f32_x2): Likewise.
593 (vst1q_f64_x2): Likewise.
594 (vst1q_p64_x2): Likewise.
595 (vst1_s64_x3): Likewise.
596 (vst1_u64_x3): Likewise.
597 (vst1_f64_x3): Likewise.
598 (vst1_s8_x3): Likewise.
599 (vst1_p8_x3): Likewise.
600 (vst1_s16_x3): Likewise.
601 (vst1_p16_x3): Likewise.
602 (vst1_s32_x3): Likewise.
603 (vst1_u8_x3): Likewise.
604 (vst1_u16_x3): Likewise.
605 (vst1_u32_x3): Likewise.
606 (vst1_f16_x3): Likewise.
607 (vst1_f32_x3): Likewise.
608 (vst1_p64_x3): Likewise.
609 (vst1q_s8_x3): Likewise.
610 (vst1q_p8_x3): Likewise.
611 (vst1q_s16_x3): Likewise.
612 (vst1q_p16_x3): Likewise.
613 (vst1q_s32_x3): Likewise.
614 (vst1q_s64_x3): Likewise.
615 (vst1q_u8_x3): Likewise.
616 (vst1q_u16_x3): Likewise.
617 (vst1q_u32_x3): Likewise.
618 (vst1q_u64_x3): Likewise.
619 (vst1q_f16_x3): Likewise.
620 (vst1q_f32_x3): Likewise.
621 (vst1q_f64_x3): Likewise.
622 (vst1q_p64_x3): Likewise.
624 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
626 * config/msp430/msp430.c (msp430_output_labelref): Prepend
627 user_label_prefix to name.
629 * tree-core.h: Update comment about the format of NAME string
630 passed to handler in attribute_spec.
632 * config/msp430/msp430.md: Remove erroneous subreg expression from
633 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
634 zero_extend{q,h}isi2.
636 2018-05-30 Borislav Petkov <bp@suse.de>
638 * doc/extend.texi: Document some architecture specific
639 constraints and sort entries.
641 2018-05-30 Martin Sebor <msebor@redhat.com>
644 * builtins.c (expand_builtin_stpcpy_1): New function.
645 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
646 only if the former succeeds.
648 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
650 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
653 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
655 * doc/invoke.texi (-flinker-output): Document
657 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
659 * passes.c (ipa_write_summaries): Only modify statements if body
661 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
662 incrementally linking.
663 (ipa_passes): Likewise.
664 * lto-cgraph.c (lto_output_node): When incrementally linking do not
665 pass down resolution info.
666 * common.opt (flag_incremental_link): Update info.
667 * gcc.c (plugin specs): Turn flinker-output=* to
668 -plugin-opt=-linker-output-known
669 * toplev.c (compile_file): Also cut compilation when doing incremental
671 * flag-types. (enum lto_partition_model): Add
672 LTO_LINKER_OUTPUT_NOLTOREL.
673 (invoke.texi): Add -flinker-output docs.
674 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
675 link same way as WPA; do not stream in dead initializers.
677 * dwarf2out.c (dwarf2out_die_ref_for_decl,
678 darf2out_register_external_decl): Support incremental link.
680 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
682 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
684 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
686 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
687 it down to simple_object_copy_lto_debug_sections.
688 (run_gcc): Determine incremental LTO link time and configure
689 lto1 into non-wpa mode, disable renaming of debug sections.
691 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
693 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
694 descriptions of various incorrectly documented functions.
696 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
699 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
700 address check not strict.
702 2018-05-30 Richard Biener <rguenther@suse.de>
704 PR tree-optimization/85964
705 * tracer.c (better_p): Drop initialized count check, we only
706 call the function with initialized counts now.
707 (find_best_successor): Do find a best edge if one
708 has uninitialized count.
709 (find_best_predecessor): Likewise. Do BB frequency check only
710 if count is initialized.
712 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
714 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
715 (aarch64_ldrstr_offset_compare): New.
716 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
717 load/store orderings.
718 (aarch64_gen_adjusted_ldpstp): Likewise.
720 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
722 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
723 Check for subset of GENERAL_REGS and FP_REGS.
724 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
727 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
729 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
730 and wi::to_poly_offset. Add the current offset and then check
731 whether the sum fits, rather than using an unchecked addition of
732 a checked term. Check for a shwi rather than a uhwi.
733 * expr.c (get_bit_range): Use tree_to_poly_uint64.
734 (store_constructor): Use poly_int_tree_p.
735 (expand_expr_real_1): Likewise.
736 * function.c (assign_temp): Likewise.
737 * fold-const.c (const_binop): Use poly_int_tree_p and
739 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
741 * ipa-icf-gimple.c (func_checker::compare_operand): Use
742 to_poly_offset for MEM offsets.
743 * ipa-icf.c (sem_variable::equals): Likewise.
744 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
745 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
746 wi::to_poly_offset for BIT_FIELD_REF offsets.
747 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
749 * var-tracking.c (emit_note_insn_var_location): Use
752 2018-05-29 Jim Wilson <jimw@sifive.com>
754 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
756 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
759 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
760 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
762 (sse4_1_round<mode>2): Use nonimmediate_operand
763 for operand 1 predicate.
765 2018-05-29 Martin Sebor <msebor@redhat.com>
766 Richard Biener <rguenther@suse.de>
769 * calls.c (get_size_range): Call determine_value_range instead
771 * tree-vrp.h (determine_value_range): Declared new function.
772 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
774 2018-05-29 Richard Biener <rguenther@suse.de>
776 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
777 sure to use non-pattern stmts for get_earlier_stmt arguments.
778 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
779 called on pattern stmts.
780 (get_later_stmt): Likewise.
782 2018-05-29 Martin Liska <mliska@suse.cz>
784 PR gcov-profile/85759
785 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
788 2018-05-29 Jakub Jelinek <jakub@redhat.com>
790 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
792 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
796 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
797 VEC_PACK_FLOAT_EXPR): New tree codes.
798 * tree-pretty-print.c (op_code_prio): Handle
799 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
800 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
801 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
802 * tree-inline.c (estimate_operator_cost): Likewise.
803 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
804 * fold-const.c (const_binop): Likewise.
805 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
806 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
807 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
808 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
809 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
810 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
811 * expr.c (expand_expr_real_2): Likewise.
812 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
813 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
814 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
816 * optabs.c (expand_widen_pattern_expr): For
817 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
818 sign from result type rather than operand's type.
819 (expand_binop_directly): For vec_packu_float_optab and
820 vec_packs_float_optab allow result type to be different from operand's
822 * optabs-tree.c (optab_for_tree_code): Handle
823 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
824 VEC_PACK_FLOAT_EXPR. Formatting fixes.
825 * tree-vect-generic.c (expand_vector_operations_1): Handle
826 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
828 * tree-vect-stmts.c (supportable_widening_operation): Handle
830 (supportable_narrowing_operation): Handle FLOAT_EXPR.
831 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
832 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
833 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
834 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
836 (vec_pack<floatprefix>_float_<mode>): New expander.
837 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
839 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
840 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
841 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
842 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
843 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
845 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
846 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
847 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
848 VEC_PACK_FLOAT_EXPR): Document.
850 2018-05-29 Richard Biener <rguenther@suse.de>
852 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
854 (stmt_vec_info_vec): Make pointer.
855 (init_stmt_vec_info_vec): Remove.
856 (free_stmt_vec_info_vec): Likewise.
857 (set_stmt_vec_info_vec): New function.
858 (free_stmt_vec_infos): Likewise.
859 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
860 (set_vinfo_for_stmt): Likewise.
861 (get_earlier_stmt): Likewise.
862 (get_later_stmt): Likewise.
863 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
864 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
865 (vec_info::~vec_info): Free stmt_vec_infos.
866 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
867 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
868 (pass_slp_vectorize::execute): Likewise.
869 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
870 (free_stmt_vec_info_vec): Likewise.
871 (set_stmt_vec_info_vec): New function.
872 (free_stmt_vec_infos): Likewise.
873 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
874 the global stmt_vec_info_vec.
875 * tree-parloops.c (gather_scalar_reductions): Use
876 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
879 2018-05-29 Richard Biener <rguenther@suse.de>
881 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
883 2018-05-29 Martin Liska <mliska@suse.cz>
884 David Malcolm <dmalcolm@redhat.com>
886 * vec.c (test_reverse): New.
887 (vec_c_tests): Add new test.
888 * vec.h (vl_ptr>::reverse): New function.
890 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
892 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
894 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
897 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
899 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
901 2018-05-28 Richard Biener <rguenther@suse.de>
903 PR tree-optimization/85933
904 * tree-vect-data-refs.c (vect_record_base_alignments): Only
905 look at stmts marked as vectorizable.
907 2018-05-28 Richard Biener <rguenther@suse.de>
909 PR tree-optimization/85934
910 * tree-vect-generic.c (expand_vector_operations_1): Hoist
911 vector boolean check before scalar optimization.
913 2018-05-28 Jakub Jelinek <jakub@redhat.com>
915 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
918 2018-05-28 Mark Wielaard <mark@klomp.org>
920 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
921 if it is an expression containing a minus sign.
923 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
925 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
927 2018-05-27 Paul Koning <ni1d@arrl.net>
929 * config/pdp11/pdp11.md (truncsihi2): Remove.
931 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
932 Chung-Ju Wu <jasonwucj@gmail.com>
934 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
936 (unaligned_store_dw): Ditto.
937 * config/nds32/nds32-memory-manipulation.c
938 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
939 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
940 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
941 (emit_setmem_word_loop): Rename to ...
942 (emit_setmem_doubleword_loop): ... this.
943 (nds32_gen_dup_4_byte_to_word_value): New function.
944 (nds32_gen_dup_8_byte_to_double_word_value): New function.
945 (nds32_expand_setmem_loop): Refine implementation.
946 (nds32_expand_setmem_loop_v3m): Ditto.
947 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
950 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
952 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
954 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
956 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
957 (nds32_init_machine_status): Initialize machine->attr_naked_p and
958 machine->attr_no_prologue_p.
959 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
960 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
961 (nds32_expand_epilogue): Consider attr_naked_p.
962 (nds32_expand_epilogue_v3pop): Likewise.
963 (nds32_can_use_return_insn): Likewise.
964 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
965 attr_no_prologue_p fields.
966 * config/nds32/nds32.opt (mret-in-naked-func): New option.
968 2018-05-27 Jakub Jelinek <jakub@redhat.com>
971 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
974 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
976 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
978 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
980 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
982 (*<floatsuffix>floatv2div2sf2): Rename to ...
983 (*float<floatunssuffix>v2div2sf2): ... this.
984 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
985 (float<floatunssuffix>v2div2sf2_mask): ... this.
986 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
987 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
988 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
990 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
992 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
994 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
996 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
998 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1000 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1001 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1002 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1003 gen_ufix_truncv8dfv8si2.
1004 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1005 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1006 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1007 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1008 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1009 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1010 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1011 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1013 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1017 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1019 2018-05-25 Jim Wilson <jimw@sifive.com>
1021 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1022 * config/riscv/riscv.c (struct machine_function): Add
1023 interrupt_handler_p and attribute_checked_p fields.
1024 (riscv_attribute_table): Add interrupt.
1025 (riscv_interrupt_type_p): New.
1026 (riscv_save_reg_p): Save extra regs for interrupt handler.
1027 (riscv_use_save_libcall): Return false for interrupt handler.
1028 (riscv_first_stack_step): Add forward declaration.
1029 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1030 for interrupt handler with large frame. Use it for saved reg list.
1031 (riscv_expand_prologue): Move flag_stack_usage_info support to
1032 eliminate duplication.
1033 (riscv_expand_epilogue): Generate mret for interrupt handler.
1034 (riscv_epilogue_uses): New.
1035 (riscv_can_use_return_insn): Return false for interrupt handler.
1036 (riscv_function_ok_for_sibcall): Likewise.
1037 (riscv_set_current_function): Add interrupt handler support.
1038 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1039 * config/riscv/riscv.md (UNSPECV_MRET): New.
1041 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1043 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1045 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1047 PR tree-optimization/85712
1048 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1049 this candidate has already been replaced in-situ by a copy.
1051 2018-05-25 Jason Merrill <jason@redhat.com>
1053 PR c++/80485 - inline function non-zero address.
1054 * symtab.c (nonzero_address): Check DECL_COMDAT.
1056 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1059 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1060 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1061 extension of SImode operation. Use const123_operand predicate.
1062 (*saddsi_1): Remove.
1063 (*saddl_se_1): Ditto.
1065 (*ssubl_se_1): Ditto.
1066 * config/alpha/predicates.md (const123_operand): New predicate.
1067 * config/alpha/constraints.md (P): Use IN_RANGE.
1069 2018-05-25 Richard Biener <rguenther@suse.de>
1071 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1073 (ref_maybe_used_by_stmt_p): Likewise.
1074 (stmt_may_clobber_ref_p): Likewise.
1075 (stmt_may_clobber_ref_p_1): Likewise.
1076 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1078 (ref_maybe_used_by_stmt_p): Likewise.
1079 (stmt_may_clobber_ref_p): Likewise.
1080 (stmt_may_clobber_ref_p_1): Likewise.
1081 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1082 the alias oracle to disambiguate DRs with stmts DR analysis
1084 (vect_analyze_data_refs): Do not give up on not analyzable
1085 DRs for BB vectorization. Remove code truncating the dataref
1088 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1091 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1092 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1093 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1095 2018-05-25 Richard Biener <rguenther@suse.de>
1097 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1098 function, combining stmt data ref gathering and fatal analysis
1100 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1101 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1102 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1103 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1104 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1106 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1108 PR tree-optimization/85720
1109 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1110 SCC if all partitions are builtins.
1111 (version_loop_by_alias_check): New parameter. Generate cancelable
1112 runtime alias check if all partitions are builtins.
1113 (distribute_loop): Update call to above function.
1115 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1117 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1118 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1119 (parm_default_def_partition_arg): Ditto.
1120 (set_parm_default_def_partition): Ditto.
1121 (get_parm_default_def_partitions): Ditto and make it static.
1122 (get_undefined_value_partitions): Ditto and make it static.
1123 (remove_ssa_form): Refactor call to init_var_map here.
1124 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1125 computation for loop region.
1126 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1127 (register_default_def): Delete.
1128 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1129 (parm_default_def_partition_arg): Ditto.
1130 (set_parm_default_def_partition): Ditto.
1131 (get_parm_default_def_partitions): Ditto and make it static.
1132 (get_undefined_value_partitions): Ditto and make it static.
1133 (coalesce_with_default, coalesce_with_default): Update comment.
1134 (create_coalesce_list_for_region): New func factored out from
1135 create_outofssa_var_map.
1136 (populate_coalesce_list_for_outofssa): New func factored out from
1137 create_outofssa_var_map and coalesce_ssa_name.
1138 (create_outofssa_var_map): Delete.
1139 (coalesce_ssa_name): Refactor to support live range computation.
1140 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1141 (get_parm_default_def_partitions): Delete.
1142 (get_undefined_value_partitions): Ditto.
1143 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1144 computation for loop region.
1145 (new_tree_live_info, loe_visit_block): Ditto.
1146 (live_worklist, set_var_live_on_entry): Ditto.
1147 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1148 * tree-ssa-live.h (struct _var_map): New fields.
1149 (init_var_map): Change decl.
1150 (region_contains_p): New.
1152 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1154 * tree-ssa-live.h (live_merge_and_clear): Delete.
1156 2018-05-25 Richard Biener <rguenther@suse.de>
1159 * tree-dump.c (dequeue_and_dump): Remove access to removed
1160 operand 2 of a SWITCH_EXPR.
1162 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1164 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1165 multiplication and division.
1166 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1167 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1168 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1169 (cond_udiv_optab, cond_umod_optab): New optabs.
1170 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1171 (IFN_COND_RDIV): New internal functions.
1172 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1173 TRUNC_MOD_EXPR and RDIV_EXPR.
1174 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1175 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1177 (SVE_INT_BINARY): Include mult.
1178 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1179 (optab, sve_int_op): Handle mult.
1180 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1182 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1183 for SVE_INT_BINARY_SD.
1185 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1187 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1188 (optab, sve_int_op): Handle div and udiv.
1189 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1190 for SVE_INT_BINARY_SD.
1191 (*<optab><mode>3): New insn for the same.
1193 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1195 * tree-vect-patterns.c: Include predict.h.
1196 (vect_recog_divmod_pattern): Restrict check for division support
1197 to when optimizing for size.
1199 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1201 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1202 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1203 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1204 (gimple_match_op::set_op): Likewise.
1205 (gimple_resimplify4): Declare.
1206 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1207 (expr::gen_transform): Likewise.
1208 (decision_tree::gen): Generate a simplification routine for 4 operands.
1209 * gimple-match-head.c (gimple_simplify): Add an overload for
1210 4 operands. In the top-level function, handle up to 4 call
1211 arguments and call gimple_resimplify4.
1212 (gimple_resimplify4): New function.
1213 (build_call_internal): Pass a fourth operand.
1214 (maybe_push_to_seq): Likewise.
1215 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1216 Fold VEC_COND_EXPRs of an operation and a default value into
1217 an IFN_COND_* function if possible.
1218 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1220 (SVE_COND_FP_BINARY): Include them.
1221 (optab, sve_fp_op): Handle them.
1222 (SVE_INT_BINARY_REV): New code iterator.
1223 (SVE_COND_FP_BINARY_REV): New int iterator.
1224 (commutative): New int attribute.
1225 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1227 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1229 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1230 (*cond_<optab><mode>): New patterns for reversed operands.
1232 2018-05-25 Richard Biener <rguenther@suse.de>
1234 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1235 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1236 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1237 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1238 * tree-vect-data-refs.c (everywhere): Adjust users.
1239 * tree-vect-loop.c (everywhere): Likewise.
1240 * tree-vect-slp.c (everywhere): Likewise.
1241 * tree-vect-stmts.c (everywhere): Likewise.
1242 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1244 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1246 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1248 (gcc_cv_as_section_exclude): ... this.
1249 Try Solaris as #exclude syntax.
1250 * configure: Regenerate.
1251 * config.in: Regenerate.
1252 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1254 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1255 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1257 * varasm.c (default_elf_asm_named_section): Don't check if
1258 HAVE_GAS_SECTION_EXCLUDE is defined.
1260 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1262 * doc/md.texi: Update the documentation of the cond_* optabs
1263 to mention the new final operand. Fix GET_MODE_NUNITS call.
1264 Describe the scalar case too.
1265 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1266 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1268 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1269 (get_conditional_internal_fn): Update comment.
1270 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1271 accumulator value as a final argument to conditional functions.
1272 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1273 a define_expand and add an "else" operand. Assert for now that
1274 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1275 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1276 (*cond_<optab><mode>): New patterns.
1277 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1278 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1279 (UNSPEC_COND_EOR): Delete.
1280 (optab): Remove associated mappings.
1281 (SVE_INT_BINARY): New code iterator.
1282 (sve_int_op): Remove int attribute and add "minus" to the code
1284 (SVE_COND_INT_OP): Delete.
1285 (SVE_COND_FP_OP): Rename to...
1286 (SVE_COND_FP_BINARY): ...this.
1288 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1290 * optabs.c (can_reuse_operands_p): New function.
1291 (maybe_legitimize_operands): Try to reuse the results for
1294 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1296 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1297 Add {q} suffix to insn mnemonic.
1299 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1301 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1302 (msp430_warn_func_return): New.
1304 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1306 * fold-const.c (tree_nonzero_bits): New function.
1307 * fold-const.h (tree_nonzero_bits): Likewise.
1308 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1309 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1311 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1315 * varasm.c (assemble_alias): Check ifunc_resolver only on
1318 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1321 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1322 when memory input operand is handled.
1324 2018-05-24 Luis Machado <luis.machado@linaro.org>
1326 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1328 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1330 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1332 * match.pd: Delay FMA folds until after vectorization.
1334 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1337 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1338 address check not strict.
1340 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1342 * gimple-match.h (gimple_match_op): New class.
1343 (mprts_hook): Replace parameters with a gimple_match_op *.
1344 (maybe_build_generic_op): Likewise.
1345 (gimple_simplified_result_is_gimple_val): Replace parameters with
1346 a const gimple_match_op *.
1347 (gimple_simplify): Replace code_helper * and tree * parameters with
1348 a gimple_match_op * parameter.
1349 (gimple_resimplify1): Replace code_helper *, tree and tree *
1350 parameters with a gimple_match_op * parameter.
1351 (gimple_resimplify2): Likewise.
1352 (gimple_resimplify3): Likewise.
1353 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1354 parameters with a gimple_match_op * parameter.
1355 * gimple-match-head.c (gimple_simplify): Change prototypes of
1356 auto-generated functions to take a gimple_match_op * instead of
1357 separate code_helper * and tree * parameters. Make the same
1358 change in the top-level overload and update calls to the
1359 gimple_resimplify routines. Update calls to the auto-generated
1360 functions and to maybe_push_res_to_seq in the publicly-facing
1361 operation-specific gimple_simplify overloads.
1362 (gimple_match_op::MAX_NUM_OPS): Define.
1363 (gimple_resimplify1): Replace rcode and ops with a single res_op
1364 parameter. Update call to gimple_simplify.
1365 (gimple_resimplify2): Likewise.
1366 (gimple_resimplify3): Likewise.
1367 (mprts_hook): Replace parameters with a gimple_match_op *.
1368 (maybe_build_generic_op): Likewise.
1369 (build_call_internal): Replace type, nargs and ops with
1370 a gimple_match_op *.
1371 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1372 with a single gimple_match_op *. Update calls to mprts_hook,
1373 build_call_internal and gimple_simplified_result_is_gimple_val.
1374 Factor out code that is common to the tree_code and combined_fn cases.
1375 * genmatch.c (expr::gen_transform): Replace tem_code and
1376 tem_ops with a gimple_match_op called tem_op. Update calls
1377 to the gimple_resimplify functions and maybe_push_res_to_seq.
1378 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1379 res_ops. Update call to the gimple_resimplify functions.
1380 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1381 (decision_tree::gen): Make the functions take a gimple_match_op *
1382 called res_op instead of separate res_code and res_ops parameters.
1383 Update call accordingly.
1384 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1385 and ops with a single res_op parameter. Update calls to
1386 maybe_build_generic_op and maybe_push_res_to_seq.
1387 (fold_stmt_1): Update calls to gimple_simplify and
1388 replace_stmt_with_simplification.
1389 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1390 and gimple_simplified_result_is_gimple_val.
1391 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1393 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1394 with a gimple_match_op *.
1395 (vn_nary_build_or_lookup): Likewise. Update call to
1396 vn_nary_build_or_lookup_1.
1397 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1398 gimple_match_op *. Update calls to the gimple_resimplify routines
1399 and to gimple_simplified_result_is_gimple_val.
1400 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1401 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1402 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1403 (visit_nary_op): Likewise.
1404 (visit_reference_op_load): Likewise.
1406 2018-05-23 Luis Machado <luis.machado@linaro.org>
1408 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1409 modifier for printing the step amount.
1411 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1414 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1417 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1419 * doc/sourcebuild.texi (Endianness): New subsubsection.
1421 2018-05-23 Luis Machado <luis.machado@linaro.org>
1423 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1424 <prefetch_dynamic_strides>: New const bool field.
1425 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1426 prefetch_dynamic_strides.
1427 (exynosm1_prefetch_tune): Likewise.
1428 (thunderxt88_prefetch_tune): Likewise.
1429 (thunderx_prefetch_tune): Likewise.
1430 (thunderx2t99_prefetch_tune): Likewise.
1431 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1433 (aarch64_override_options_internal): Update to set
1434 PARAM_PREFETCH_DYNAMIC_STRIDES.
1435 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1436 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1437 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1438 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1439 prefetch-dynamic-strides setting.
1441 2018-05-23 Luis Machado <luis.machado@linaro.org>
1443 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1444 <minimum_stride>: New const int field.
1445 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1446 minimum_stride field defaulting to -1.
1447 (exynosm1_prefetch_tune): Likewise.
1448 (thunderxt88_prefetch_tune): Likewise.
1449 (thunderx_prefetch_tune): Likewise.
1450 (thunderx2t99_prefetch_tune): Likewise.
1451 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1452 <default_opt_level>: Set to 3.
1453 (aarch64_override_options_internal): Update to set
1454 PARAM_PREFETCH_MINIMUM_STRIDE.
1455 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1456 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1457 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1458 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1459 stride is constant and is below the minimum stride threshold.
1461 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1463 * config/arm/arm-cpus.in (mode26): Delete.
1464 (armv4): Delete mode26 reference.
1465 * config/arm/arm.c (arm_configure_build_target): Delete use of
1468 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1470 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1472 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1473 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1475 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1476 (floatunsdidf2): Ditto.
1478 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1480 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1481 (fixuns_trunc<mode>si2_avx512f): Ditto.
1482 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1483 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1484 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1486 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1488 PR rtl-optimization/79985
1489 * df-scan.c (df_insn_refs_collect): Remove special case for
1490 global registers and asm statements.
1492 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1494 * extend.texi (Global Register Variables): Rewrite the bullet list.
1495 Note that the register is available for allocation. Note that access
1496 via inline asm must use constraints. Add note about async-signal
1497 handlers. Remove paragraph about automagic register selection.
1499 2018-05-23 Richard Biener <rguenther@suse.de>
1501 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1502 of fixed offset from memset VN.
1504 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1506 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1508 (alloc_cand_and_find_basis): Initialize first_interp field.
1509 (slsr_process_mul): Modify first_interp field.
1510 (slsr_process_add): Likewise.
1511 (slsr_process_cast): Modify first_interp field for each new
1513 (slsr_process_copy): Likewise.
1514 (dump_candidate): Dump first_interp field.
1515 (replace_mult_candidate): Process all interpretations, not just
1517 (replace_rhs_if_not_dup): Likewise.
1518 (replace_one_candidate): Likewise.
1520 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1522 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1524 (aarch64_needs_frame_chain): New function.
1525 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1527 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1530 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1531 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1532 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1533 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1534 as true for strict-align.
1535 (aarch64_can_inline_p): Perform checks even when callee has no
1536 attributes to check for strict alignment.
1537 * doc/extend.texi (AArch64 Function Attributes): Document
1539 * doc/invoke.texi: (AArch64 Options): Likewise.
1541 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1543 PR tree-optimization/85853
1544 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1545 the handling of the root of the node to...
1546 (vect_slp_analyze_node_operations_1): ...this new function,
1547 and run the whole thing with the child nodes' def types
1548 set according to their SLP node's def type.
1550 2018-05-23 Richard Biener <rguenther@suse.de>
1553 * tree-data-ref.c (create_runtime_alias_checks): Defer
1554 and ignore overflow warnings.
1556 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1558 PR tree-optimization/85822
1559 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1562 2018-05-23 Richard Biener <rguenther@suse.de>
1564 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1565 memset constants via native_interpret_expr.
1567 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1570 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1572 (cgraph_node::create_alias): Likewise.
1573 (cgraph_node::get_availability): Check ifunc_resolver instead
1574 of looking up ifunc attribute.
1575 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1576 * varasm.c (do_assemble_alias): Likewise.
1577 (assemble_alias): Likewise.
1578 (default_binds_local_p_3): Likewise.
1579 * cgraph.h (cgraph_node): Add ifunc_resolver.
1580 (cgraph_node::only_called_directly_or_aliased_p): Return false
1582 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1584 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1585 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1586 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1587 instead of looking up ifunc attribute.
1589 2018-05-22 Luis Machado <luis.machado@linaro.org>
1591 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1593 2018-05-22 Martin Sebor <msebor@redhat.com>
1596 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1597 only when expasion succeeds.
1598 (expand_builtin_strcmp): Same.
1599 (expand_builtin_strncmp): Same.
1601 2018-05-22 Martin Sebor <msebor@redhat.com>
1603 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1605 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1606 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1608 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1609 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1611 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1612 Add check for legitimate_address.
1613 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1614 (aarch64_swap_ldrstr_operands): New.
1615 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1618 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1619 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1621 * config/aarch64/aarch64.md: New patterns to generate stp
1623 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1624 single words and double words.
1625 (load_pair_sw, load_pair_dw): Likewise.
1626 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1628 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1630 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1631 for different mode ldpstp and add peephole for merged zero stores.
1633 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1635 (aarch64_gen_store_pair): Rename calls to match new patterns.
1636 (aarch64_gen_load_pair): Rename calls to match new patterns.
1637 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1638 (load_pair<DREG:mode><DREG2:mode>): ... This.
1639 (store_pair<mode>): Rename to...
1640 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1641 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1643 (V_INT_EQUIV): Handle SImode.
1644 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1647 2018-05-22 Martin Sebor <msebor@redhat.com>
1650 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1651 or ajust the presumed bound on an operation to avoid unnecessary
1654 2018-05-22 Martin Sebor <msebor@redhat.com>
1656 PR tree-optimization/85826
1657 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1658 assuming that a DECL necesarily has a constant size.
1660 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1663 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1665 2018-05-22 Richard Biener <rguenther@suse.de>
1667 PR tree-optimization/85834
1668 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1669 non-constant and non-zero memset arguments.
1671 2018-05-22 Martin Liska <mliska@suse.cz>
1674 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1676 2018-05-22 Richard Biener <rguenther@suse.de>
1678 PR tree-optimization/85863
1679 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1680 comparisons when vectype is specified.
1681 (vectorizable_condition): Do not specify vectype for
1682 vect_is_simple_cond when SLP vectorizing.
1684 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1687 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1688 define __ibm128 as long double.
1689 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1690 as a distinct type when IEEE 128-bit support is enabled.
1691 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1692 128-bit types to use the correct functions.
1693 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1694 convert between 128-bit floating point types that have different
1695 modes but the same representation, instead of using gen_lowpart to
1697 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1699 (IFKF_reg): New attributes to give the register constraints for
1701 (extend<mode>tf2_internal): New insns to mark an explicit
1702 conversion between 128-bit floating point types that have a
1703 different mode but share the same representation.
1705 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1707 PR tree-optimization/85814
1708 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1709 a null return from get_strinfo when unsharing the next
1710 strinfo in the chain.
1712 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1715 * varasm.c (weak_finish): Clean up weak_decls.
1717 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1719 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1720 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1721 UNSPEC_UADALP values.
1722 * config/aarch64/iterators.md (ABAL): New int iterator.
1725 (sur): Add mappings for the above.
1726 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1728 (aarch64_<sur>abal<mode>_4): Likewise.
1729 (aarch64_<sur>adalp<mode>_3): Likewise.
1730 (<sur>sadv16qi): New define_expand.
1732 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1734 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1735 (*movdf_internal): Ditto.
1736 (*rcpsf2_sse): Ditto.
1737 (*rsqrtsf2_sse): Ditto.
1738 (*sqrt<mode>2_sse): Ditto.
1740 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1742 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1744 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1745 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1746 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1747 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1749 * config/aarch64/arm_neon.h: Likewise.
1750 * config/aarch64/iterators.md (VQ_I): New.
1752 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1754 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1756 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1759 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1761 * config/nds32/constraints.md (S): New constraint.
1762 * config/nds32/nds32.md (call_internal): Use constraint S.
1763 (call_value_internal): Likewise.
1764 (sibcall_internal): Likewise.
1765 (sibcall_value_internal): Likewise.
1767 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1768 Chung-Ju Wu <jasonwucj@gmail.com>
1770 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1773 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1774 Chung-Ju Wu <jasonwucj@gmail.com>
1776 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1777 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1778 (nds32_rtx_costs_impl): Simplify.
1779 (nds32_address_cost_impl): Simplify.
1780 (nds32_init_rtx_costs): New function.
1781 (nds32_rtx_costs_speed_prefer): Likewise.
1782 (nds32_rtx_costs_size_prefer): Likewise.
1783 (nds32_address_cost_speed_prefer): Likewise.
1784 (nds32_address_cost_speed_fwprop): Likewise.
1785 (nds32_address_cost_size_prefer): Likewise.
1786 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1787 * config/nds32/nds32.c (nds32_option_override): Use
1788 nds32_init_rtx_costs function.
1790 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1792 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1793 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1794 (TARGET_PIPELINE_N8): Likewise.
1795 (TARGET_PIPELINE_N10): Likewise.
1796 (TARGET_PIPELINE_N13): Likewise.
1797 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1799 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1801 * config/nds32/nds32-fpu.md: Update copyright year.
1803 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1805 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1807 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1809 * config/nds32/nds32.c
1810 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1811 * config/nds32/nds32.opt (minline-asm-r15): New option.
1813 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1815 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1817 * config/nds32/nds32.md (abssi2): New pattern.
1819 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1821 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1822 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1823 Merge insn pattern from sse_cvtsi2ss<round_name> and
1824 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1825 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1826 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1827 using SWI48 mode iterator.
1828 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1829 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1830 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1831 pattern from sse_cvttss2si<round_saeonly_name>
1832 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1833 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1834 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1835 using SWI48 mode iterator.
1836 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1837 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1838 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1839 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1840 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1841 using SWI48 mode iterator.
1842 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1843 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1844 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1845 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1846 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1847 SWI48 mode iterator.
1848 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1849 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1850 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1851 pattern from sse_cvttsd2si<round_saeonly_name>
1852 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1854 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1856 * config/nds32/nds32-md-auxiliary.c
1857 (nds32_valid_smw_lwm_base_p): Refine.
1858 (nds32_output_smw_single_word): Refine.
1859 (nds32_output_smw_double_word): New.
1860 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1862 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1864 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1865 (nds32_output_stack_pop): Refine.
1866 (nds32_expand_unaligned_load): Refine.
1867 (nds32_expand_unaligned_store): Refine.
1869 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1870 Chung-Ju Wu <jasonwucj@gmail.com>
1872 * config/nds32/constants.md: Add TP_REGNUM constant.
1873 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1874 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1876 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1877 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1878 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1879 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1880 and PIC code generation.
1881 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1883 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1885 * config/nds32/nds32.md: Support TLS and PIC.
1886 * config/nds32/nds32.c: Support TLS and PIC.
1887 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1888 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1891 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1893 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1894 mode with E_ prefix.
1896 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1897 Chung-Ju Wu <jasonwucj@gmail.com>
1899 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1900 * config/nds32/nds32-md-auxiliary.c
1901 (symbolic_reference_mentioned_p): New.
1902 (nds32_legitimize_ict_address): New.
1903 (nds32_expand_ict_move): New.
1904 (nds32_indirect_call_referenced_p): New.
1905 (nds32_symbol_binds_local_p): Delete.
1906 (nds32_long_call_p): Modify.
1907 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1908 * config/nds32/nds32-protos.h
1909 (symbolic_reference_mentioned_p): Declare.
1910 (nds32_legitimize_ict_address): Declare.
1911 (nds32_expand_ict_move): Declare.
1912 (nds32_indirect_call_referenced_p): Declare.
1913 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1914 (nds32_relax_group): Use nds32_ict_const_p as condition.
1915 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1916 (nds32_asm_file_start): Output ict_model directive in asm code.
1917 (nds32_legitimate_address_p): Consider indirect call.
1918 (nds32_print_operand): Consider indirect call.
1919 (nds32_print_operand_address): Consider indirect call.
1920 (nds32_insert_attributes): Handle "indirect_call" attribute.
1921 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1922 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1923 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1924 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1925 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1926 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1927 (TARGET_ICT_MODEL_SMALL): Define.
1928 (TARGET_ICT_MODEL_LARGE): Define.
1929 * config/nds32/nds32.md (movsi): Consider ict model.
1930 (call, call_value): Consider ict model.
1931 (sibcall, sibcall_value): Consider ict model.
1932 * config/nds32/nds32.opt (mict-model): New option.
1933 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1936 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1937 Monk Chiang <sh.chiang04@gmail.com>
1938 Jim Wilson <jimw@sifive.com>
1940 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1941 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1942 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1943 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1944 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1945 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1946 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1947 compute save_libcall_adjustment properly.
1948 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1949 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1950 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1951 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1952 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1953 (ABI_SPEC): Handle mabi=ilp32e.
1954 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1955 (RVE): Add RVE mask.
1956 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1957 <-march>: Add rv32e as an example.
1959 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1962 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1963 (intra_create_variable_infos): Handle C++ constructors.
1965 2018-05-18 Martin Liska <mliska@suse.cz>
1967 * passes.def: Remove a redundant pass.
1969 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1972 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1974 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1976 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1978 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1980 (armv2, armv2a, armv3, armv3m): Delete architectures.
1981 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1982 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1983 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1985 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1986 (*mulsidi3adddi): Likewise.
1987 (mulsidi3): Likewise.
1988 (*mulsidi3_nov6): Likewise.
1989 (umulsidi3): Likewise.
1990 (umulsidi3_nov6): Likewise.
1991 (umaddsidi4): Likewise.
1992 (*umulsidi3adddi): Likewise.
1993 (smulsi3_highpart): Likewise.
1994 (*smulsi3_highpart_nov6): Likewise.
1995 (umulsi3_highpart): Likewise.
1996 (*umulsi3_highpart_nov6): Likewise.
1997 * config/arm/arm.h (arm_arch3m): Delete.
1998 * config/arm/arm.c (arm_arch3m): Delete.
1999 (arm_option_override_internal): Update armv3-related comment.
2000 (arm_configure_build_target): Delete use of isa_bit_mode32.
2001 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2002 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2003 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2005 (mulusa3): Likewise.
2006 * config/arm/arm-protos.h (arm_arch3m): Delete.
2007 * config/arm/arm-tables.opt: Regenerate.
2008 * config/arm/arm-tune.md: Likewise.
2009 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2010 deleted architectures.
2012 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2014 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2015 (armv5t, armv5te): New features.
2016 (ARMv5, ARMv5e): Delete fgroups.
2017 (ARMv5t, ARMv5te): Adjust for above changes.
2019 (armv5, armv5e): Delete arches.
2020 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2022 (*call_reg_arm): Likewise.
2023 (*call_value_reg_armv5): Likewise.
2024 (*call_value_reg_arm): Likewise.
2025 (*call_symbol): Likewise.
2026 (*call_value_symbol): Likewise.
2027 (*sibcall_insn): Likewise.
2028 (*sibcall_value_insn): Likewise.
2030 (prefetch): Likewise.
2031 (define_split and define_peephole2 dependent on arm_arch5):
2033 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2035 (TARGET_ARM_QBIT): Likewise.
2036 (TARGET_DSP_MULTIPLY): Likewise.
2037 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2038 (arm_arch5, arm_arch5e): Delete.
2039 (arm_arch5t, arm_arch5te): Declare.
2040 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2041 (arm_arch5t): Declare.
2042 (arm_option_reconfigure_globals): Update for the above.
2043 (arm_options_perform_arch_sanity_checks): Update comment, replace
2044 use of arm_arch5 with arm_arch5t.
2045 (use_return_insn): Likewise.
2046 (arm_emit_call_insn): Likewise.
2047 (output_return_instruction): Likewise.
2048 (arm_final_prescan_insn): Likewise.
2049 (arm_coproc_builtin_available): Likewise.
2050 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2051 arm_arch5e with arm_arch5t and arm_arch5te.
2052 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2053 (arm_arch5t, arm_arch5te): Declare.
2054 * config/arm/arm-tables.opt: Regenerate.
2055 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2056 * config/arm/t-multilib: Likewise.
2057 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2058 instead of arm_arch5.
2059 (*call_reg_thumb1): Likewise.
2060 (*call_value_reg_thumb1_v5): Likewise.
2061 (*call_value_reg_thumb1): Likewise.
2062 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2064 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2066 2018-05-18 Martin Liska <mliska@suse.cz>
2068 PR gcov-profile/84846
2069 * doc/gcov.texi: Document -t option of gcov tool.
2071 2018-05-18 Martin Liska <mliska@suse.cz>
2073 PR gcov-profile/84846
2074 * gcov.c (print_usage): Add new -t option.
2075 (process_args): Handle the option.
2076 (generate_results): Use stdout as output when requested by
2079 2018-05-18 Martin Liska <mliska@suse.cz>
2081 PR gcov-profile/84846
2082 * coverage.c (coverage_init): Write PWD to .gcno file.
2083 * doc/gcov.texi: Document how working directory is printed.
2084 * gcov-dump.c (dump_gcov_file): Print PWD.
2085 * gcov.c (output_intermediate_file): Likewise.
2086 (read_graph_file): Read PWD string.
2087 (output_lines): Print PWD.
2089 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2092 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2093 for retval and return false if all args to phi are zero.
2095 2018-05-18 Richard Biener <rguenther@suse.de>
2097 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2099 (evrp_dom_walker::before_dom_children): Call it.
2101 2018-05-18 Richard Biener <rguenther@suse.de>
2103 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2104 results when processing array refs with variable index.
2106 2018-05-18 Toon Moene <toon@moene.org>
2108 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2109 directly after that of -floop-interchange. Indicate that both
2110 options are enabled by default when specifying -O3.
2112 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2114 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2115 iterator. Delete separate integer-mode vec_set<mode> expander.
2116 (aarch64_simd_vec_setv2di): Delete.
2117 (vec_setv2di): Delete.
2118 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2119 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2120 the "w, r" alternative.
2122 2018-05-18 Martin Liska <mliska@suse.cz>
2124 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2125 * tree-pass.h (make_pass_lower_switch_O0): New function.
2126 * tree-switch-conversion.c (node_has_low_bound): Remove.
2127 (node_has_high_bound): Likewise.
2128 (node_is_bounded): Likewise.
2129 (class pass_lower_switch): Make it a template type and create
2131 (pass_lower_switch::execute): Add template argument.
2132 (make_pass_lower_switch): New function.
2133 (make_pass_lower_switch_O0): New function.
2134 (do_jump_if_equal): Remove.
2135 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2136 all the hard work to tree optimization passes.
2138 2018-05-18 Martin Liska <mliska@suse.cz>
2140 * dbgcnt.c (limit_low): Renamed from limit.
2141 (limit_high): New variable.
2142 (dbg_cnt_is_enabled): Check for upper limit.
2143 (dbg_cnt): Adjust dumping.
2144 (dbg_cnt_set_limit_by_index): Add new argument for high
2146 (dbg_cnt_set_limit_by_name): Likewise.
2147 (dbg_cnt_process_single_pair): Parse new format.
2148 (dbg_cnt_process_opt): Use strtok.
2149 (dbg_cnt_list_all_counters): Remove 'value' and add
2151 * doc/invoke.texi: Document changes.
2153 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2155 * doc/sourcebuild.texi (scalar_all_fma): Document.
2156 * tree.def (FMA_EXPR): Delete.
2157 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2158 * internal-fn.c (ternary_direct): New macro.
2159 (expand_ternary_optab_fn): Likewise.
2160 (direct_ternary_optab_supported_p): Likewise.
2161 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2162 * builtins.c (fold_builtin_fma): Delete.
2163 (fold_builtin_3): Don't call it.
2164 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2165 * expr.c (expand_expr_real_2): Likewise.
2166 * fold-const.c (operand_equal_p): Likewise.
2167 (fold_ternary_loc): Likewise.
2168 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2169 * gimple.c (DEFTREECODE): Likewise.
2170 * gimplify.c (gimplify_expr): Likewise.
2171 * optabs-tree.c (optab_for_tree_code): Likewise.
2172 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2173 * tree-eh.c (operation_could_trap_p): Likewise.
2174 (stmt_could_throw_1_p): Likewise.
2175 * tree-inline.c (estimate_operator_cost): Likewise.
2176 * tree-pretty-print.c (dump_generic_node): Likewise.
2177 (op_code_prio): Likewise.
2178 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2179 * tree-ssa-operands.c (get_expr_operands): Likewise.
2180 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2181 * fold-const-call.h (fold_fma): Delete.
2182 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2183 CFN_FNMA and CFN_FNMS.
2185 * genmatch.c (combined_fn): New enum.
2186 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2187 (commutative_op): New function.
2188 (commutate): Use it. Handle more than 2 operands.
2189 (dt_operand::gen_gimple_expr): Use commutative_op.
2190 (parser::parse_expr): Allow :c to be used with non-binary
2191 operators if the commutative operand is known.
2192 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2193 CFN_FMS, CFN_FNMA and CFN_FNMS.
2194 (backprop::process_assign_use): Remove FMA_EXPR handling.
2195 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2196 (gen_hsa_fma): New function.
2197 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2198 IFN_FNMA and IFN_FNMS.
2199 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2200 * gimple-fold.h (follow_all_ssa_edges): Declare.
2201 * gimple-fold.c (follow_all_ssa_edges): New function.
2202 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2203 gimple_build interface and use follow_all_ssa_edges to fold the result.
2204 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2205 instead of checking for optabs directly.
2206 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2207 rather than FMA_EXPRs.
2208 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2209 call to IFN_FMA instead of an FMA_EXPR.
2211 2018-05-17 Jim Wilson <jimw@sifive.com>
2213 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2214 sign extended promoted subreg, and the range does not have the sign bit
2215 set, then do a sign extend.
2217 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2218 test, check for sign extended subreg and/or constant operands, and
2219 do a sign extend in that case.
2221 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2223 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2224 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2226 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2227 Change logics_shift_reg to logics_shift_imm.
2228 (thunderx2t99_fp_loadpair_basic): Delete.
2229 (thunderx2t99_fp_storepair_basic): Delete.
2230 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2231 (thunderx2t99_asimd_polynomial): Delete.
2232 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2233 and neon_fp_mul_d_scalar_q.
2234 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2235 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2236 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2237 (thunderx2t99_asimd_lut): Add missing tbl types.
2238 (thunderx2t99_asimd_ext): Delete.
2239 (thunderx2t99_asimd_load1_1_mult): Delete.
2240 (thunderx2t99_asimd_load1_2_mult): Delete.
2241 (thunderx2t99_asimd_load1_ldp): New.
2242 (thunderx2t99_asimd_load1): New.
2243 (thunderx2t99_asimd_load2): Add missing *load2* types.
2244 (thunderx2t99_asimd_load3): New.
2245 (thunderx2t99_asimd_load4): New.
2246 (thunderx2t99_asimd_store1_1_mult): Delete.
2247 (thunderx2t99_asimd_store1_2_mult): Delete.
2248 (thunderx2t99_asimd_store2_mult): Delete.
2249 (thunderx2t99_asimd_store2_onelane): Delete.
2250 (thunderx2t99_asimd_store_stp): New.
2251 (thunderx2t99_asimd_store1): New.
2252 (thunderx2t99_asimd_store2): New.
2253 (thunderx2t99_asimd_store3): New.
2254 (thunderx2t99_asimd_store4): New.
2256 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2258 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2259 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2261 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2262 Segher Boessenkool <segher@kernel.crashing.org>
2265 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2268 2018-05-17 Richard Biener <rguenther@suse.de>
2270 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2271 for pruning loop and prune defs feeding only already visited PHIs.
2273 2018-05-17 Richard Biener <rguenther@suse.de>
2275 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2277 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2278 Richard Biener <rguenther@suse.de>
2280 PR tree-optimization/85793
2281 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2282 for VMAT_ELEMENTWISE.
2284 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2286 * internal-fn.h (lookup_internal_fn): Declare
2287 * internal-fn.c (lookup_internal_fn): New function.
2288 * gimple.c (gimple_build_call_from_tree): Handle calls to
2290 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2291 internal function names.
2292 * tree-pretty-print.c (dump_generic_node): Likewise.
2293 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2295 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2297 * gimple-fold.h (gimple_build): Make the function forms take
2298 combined_fn rather than built_in_function.
2299 (gimple_simplify): Likewise.
2300 * gimple-match-head.c (gimple_simplify): Likewise.
2301 * gimple-fold.c (gimple_build): Likewise.
2302 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2303 rather than gimple_build_call_internal.
2304 (get_initial_defs_for_reduction): Likewise.
2305 (vect_create_epilog_for_reduction): Likewise.
2306 (vectorizable_live_operation): Likewise.
2308 2018-05-17 Martin Liska <mliska@suse.cz>
2310 * gimple-ssa-sprintf.c (format_directive): Do not use
2311 space in between 'G_' and '('.
2313 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2316 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2317 even if the mask is not all ones.
2320 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2322 (ix86_gimple_fold_builtin): Likewise.
2325 * config/i386/i386.c: Include tree-vector-builder.h.
2326 (ix86_vector_shift_count): New function.
2327 (ix86_fold_builtin): Fold shift builtins by scalar count.
2328 (ix86_gimple_fold_builtin): Likewise.
2330 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2331 _mm512_setzero): New intrinsics.
2333 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2334 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2336 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2337 code generation for cases where splatting a value is not useful.
2338 * simplify-rtx.c (simplify_ternary_operation): Simplify
2339 vec_merge across a vec_duplicate and a paradoxical subreg forming
2340 a vector mode to a vec_concat.
2342 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2344 * config.gcc: Support "goldmont-plus".
2345 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2347 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2348 PROCESSOR_GOLDMONT_PLUS.
2349 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2350 (processor_target_table): Add "goldmont-plus".
2351 (PTA_GOLDMONT_PLUS): Define.
2352 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2353 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2354 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2355 (fold_builtin_cpu): Add "goldmont-plus".
2356 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2357 (ix86_option_override_internal): Add "goldmont-plus".
2358 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2359 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2360 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2361 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2363 2018-05-17 Richard Biener <rguenther@suse.de>
2365 PR tree-optimization/85757
2366 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2367 remove defs that only feed that PHI from further processing.
2369 2018-05-16 Jim Wilson <jimw@sifive.com>
2371 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2373 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2375 2018-05-16 Mark Wielaard <mark@klomp.org>
2377 * dwarf2out.c (count_index_strings): New function.
2378 (output_indirect_strings): Call count_index_strings and generate
2379 header for dwarf_version >= 5.
2381 2018-05-16 Mark Wielaard <mark@klomp.org>
2383 * dwarf2out.c (dwarf_FORM): New function.
2384 (set_indirect_string): Use dwarf_FORM.
2385 (reset_indirect_string): Likewise.
2386 (size_of_die): Likewise.
2387 (value_format): Likewise.
2388 (output_die): Likewise.
2389 (add_skeleton_AT_string): Likewise.
2390 (output_macinfo_op): Likewise.
2391 (index_string): Likewise.
2392 (output_index_string_offset): Likewise.
2393 (output_index_string): Likewise.
2394 (count_index_strings): Likewise.
2396 2018-05-16 Carl Love <cel@us.ibm.com>
2398 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2399 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2401 2018-05-16 Martin Jambor <mjambor@suse.cz>
2403 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2404 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2406 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2408 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2409 (fnma<mode>4): Likewise.
2410 (fms<mode>4): Likewise.
2411 (fnms<mode>4): Likewise.
2412 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2413 (aarch64_fnma<mode>4): Likewise.
2414 (aarch64_fms<mode>4): Likewise.
2415 (aarch64_fnms<mode>4): Likewise.
2416 (aarch64_fnmadd<mode>4): Likewise.
2418 2018-05-16 Jason Merrill <jason@redhat.com>
2420 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2422 2018-05-16 Richard Biener <rguenther@suse.de>
2424 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2425 (dump_stmt_cost): Declare.
2426 (add_stmt_cost): Dump cost we add.
2427 (add_stmt_costs): New function.
2428 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2430 (vect_analyze_stmt): Adjust prototype.
2431 (vectorizable_condition): Likewise.
2432 (vectorizable_live_operation): Likewise.
2433 (vectorizable_reduction): Likewise.
2434 (vectorizable_induction): Likewise.
2435 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2436 cost vector to pass to vectorizable_ and record afterwards.
2437 (vect_model_reduction_cost): Take cost vector argument and adjust.
2438 (vect_model_induction_cost): Likewise.
2439 (vectorizable_reduction): Likewise.
2440 (vectorizable_induction): Likewise.
2441 (vectorizable_live_operation): Likewise.
2442 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2443 SLP_TREE_NUMBER_OF_VEC_STMTS.
2444 (vect_analyze_slp_cost_1): Remove.
2445 (vect_analyze_slp_cost): Likewise.
2446 (vect_slp_analyze_node_operations): Take visited args and
2447 a target cost vector. Avoid processing already visited stmt sets.
2448 (vect_slp_analyze_operations): Use a local cost vector to gather
2449 costs and register those of non-discarded instances.
2450 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2451 (vect_schedule_slp_instance): Remove copying of
2452 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2454 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2455 adding cost. Record cost entry location.
2456 (vect_prologue_cost_for_slp_op): Function to compute cost of
2457 a constant or invariant generated for SLP vect in the prologue,
2458 split out from vect_analyze_slp_cost_1.
2459 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2460 (vect_model_promotion_demotion_cost): Likewise.
2461 (vect_model_store_cost): Likewise, make static.
2462 (vect_model_load_cost): Likewise.
2463 (vectorizable_bswap): Add cost vector arg and adjust.
2464 (vectorizable_call): Likewise.
2465 (vectorizable_simd_clone_call): Likewise.
2466 (vectorizable_conversion): Likewise.
2467 (vectorizable_assignment): Likewise.
2468 (vectorizable_shift): Likewise.
2469 (vectorizable_operation): Likewise.
2470 (vectorizable_store): Likewise.
2471 (vectorizable_load): Likewise.
2472 (vectorizable_condition): Likewise.
2473 (vectorizable_comparison): Likewise.
2474 (can_vectorize_live_stmts): Likewise.
2475 (vect_analyze_stmt): Likewise.
2476 (vect_transform_stmt): Adjust calls to vectorizable_*.
2477 * tree-vectorizer.c: Include gimple-pretty-print.h.
2478 (dump_stmt_cost): New function.
2480 2018-05-16 Richard Biener <rguenther@suse.de>
2482 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2483 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2484 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2485 (check_name): New callback.
2486 (dse_classify_store): Track cycles via a visited bitmap of PHI
2487 defs and simplify handling of in-loop and across loop dead stores
2488 and properly fail for loop-variant refs. Handle byte-tracking with
2489 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2492 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2494 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2495 (vect_get_mask_type_for_stmt): Likewise.
2496 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2498 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2499 to determine the statement's vector type and the vector type that
2500 should be used for calculating nunits. Deal with cases in which
2501 the type has to be deferred.
2502 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2503 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2504 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2505 (vect_determine_vf_for_stmt): New functions, split out from...
2506 (vect_determine_vectorization_factor): ...here.
2507 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2508 (vect_get_mask_type_for_stmt): New functions, split out from
2509 vect_determine_vectorization_factor.
2511 2018-05-16 Richard Biener <rguenther@suse.de>
2513 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2514 verify the [VEC_]COND_EXPR embedded comparison.
2516 2018-05-15 Martin Sebor <msebor@redhat.com>
2518 PR tree-optimization/85753
2519 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2520 RECORD_TYPE in addition to ARRAY_TYPE.
2522 2018-05-15 Martin Sebor <msebor@redhat.com>
2525 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2527 2018-05-15 Richard Biener <rguenther@suse.de>
2529 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2530 add by_clobber_p one. Change algorithm to collect all defs
2531 representing uses we need to walk and try reducing them to
2532 a single one before failing.
2533 (dse_dom_walker::dse_optimize_stmt): Adjust.
2535 2018-05-13 Mark Wielaard <mark@klomp.org>
2537 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2538 (size_of_loc_descr): Likewise.
2539 (output_loc_operands): Likewise.
2540 (output_loc_operands_raw): Likewise.
2541 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2542 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2543 (hash_loc_operands): Likewise.
2544 (compare_loc_operands): Likewise.
2546 2018-05-14 Mark Wielaard <mark@klomp.org>
2548 * dwarf2out.c (count_index_addrs): New function.
2549 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2551 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2553 PR tree-optimization/83648
2554 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2555 return value as malloc candidate.
2557 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2560 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2561 param as true in call to suggest_attribute.
2563 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2565 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2566 -mreadonly-in-sdata.
2568 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2570 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2572 (aarch64_crypto_aesd_fused): Likewise.
2574 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2576 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2577 (movsi_aarch64): Likewise.
2578 (load_pairsi): Likewise.
2579 (load_pairdi): Likewise.
2580 (store_pairsi): Likewise.
2581 (store_pairdi): Likewise.
2582 (load_pairsf): Likewise.
2583 (load_pairdf): Likewise.
2584 (store_pairsf): Likewise.
2585 (store_pairdf): Likewise.
2586 (zero_extend): Likewise.
2587 (trunc): Swap alternatives.
2588 (fcvt_target): Add '?' to prefer w over r.
2590 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2593 * config/i386/i386.md: Disallow non-commutative arithmetics in
2594 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2595 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2596 in the peephole2 before it.
2598 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2600 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2601 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2602 (ix86_handle_option): Handle -mcldemote.
2603 * config.gcc: New header.
2604 * config/i386/cldemoteintrin.h: New file.
2605 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2606 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2608 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2609 OPTION_MASK_ISA_CLDEMOTE.
2610 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2611 (ix86_valid_target_attribute_inner_p): Ditto.
2612 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2613 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2614 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2615 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2616 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2618 * config/i386/i386.opt: Add -mcldemote.
2619 * config/i386/x86intrin.h: New header.
2620 * doc/invoke.texi: Add -mcldemote.
2622 2018-05-14 Richard Biener <rguenther@suse.de>
2624 * doc/match-and-simplify.texi: Adjust :s documentation.
2626 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2628 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2629 intended memcpy size.
2630 (REORDER_45): Likewise.
2632 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2634 * sort.cc: New file.
2635 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2636 * vec.c (qsort_chk): Use gcc_qsort.
2637 * Makefile.in (OBJS-libcommon): Add sort.o.
2638 (build/sort.o): New target. Use it...
2639 (BUILD_RTL): ... here, and...
2640 (build/gencfn-macros): ... here, and...
2641 (build/genmatch): ... here.
2643 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2644 Chung-Ju Wu <jasonwucj@gmail.com>
2646 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2647 * config/nds32/nds32-graywolf.md: New file.
2648 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2649 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2651 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2652 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2653 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2654 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2655 * config/nds32/pipelines.md: Include n15 settings.
2657 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2658 Chung-Ju Wu <jasonwucj@gmail.com>
2660 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2661 * config/nds32/nds32-n13.md: New file.
2662 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2663 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2665 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2666 * config/nds32/nds32.md (pipeline_model): Add n13.
2667 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2668 * config/nds32/pipelines.md: Include n13 settings.
2670 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2671 Chung-Ju Wu <jasonwucj@gmail.com>
2673 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2674 * config/nds32/nds32-n10.md: New file.
2675 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2676 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2678 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2679 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2680 * config/nds32/nds32.md (pipeline_model): Add n10.
2681 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2682 * config/nds32/pipelines.md: Include n10 settings.
2684 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2685 Kito Cheng <kito.cheng@gmail.com>
2686 Chung-Ju Wu <jasonwucj@gmail.com>
2688 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2689 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2690 Add enum values for DSP extension instructions.
2691 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2693 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2694 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2696 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2697 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2698 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2699 * config/nds32/nds32-intrinsic.md: Likewise.
2700 * config/nds32/nds32_intrinsic.h: Likewise.
2701 * config/nds32/nds32-md-auxiliary.c: Likewise.
2702 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2703 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2704 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2705 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2706 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2707 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2708 TYPE_DMAC in switch statement.
2709 * config/nds32/nds32.c: New checking and implementation for DSP
2710 extension instructions.
2711 * config/nds32/nds32.h: Likewise.
2712 * config/nds32/nds32.md: Likewise.
2713 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2714 * config/nds32/predicates.md: Implement new predicates for DSP
2717 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2719 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2720 Reformat alternatives and attributes so it is easier to identify
2721 which constraints/attributes go with which instruction.
2722 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2723 (mov<mode>_softfloat32, FMOVE64): Likewise.
2724 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2725 (mov<mode>_softfloat64, FMOVE64): Likewise.
2727 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2729 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2731 (Basic PowerPC Built-in Functions): The new name of the
2732 subsection previously known as "PowerPC Built-in Functions".
2733 (Basic PowerPC Built-in Functions Available on all Configurations):
2735 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2736 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2737 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2738 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2740 2018-05-11 Martin Jambor <mjambor@suse.cz>
2743 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2746 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2749 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2751 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2753 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2754 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2755 (ix86_handle_option): Handle -mwaitpkg.
2756 * config.gcc: New header.
2757 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2758 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2759 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2761 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2762 OPTION_MASK_ISA_WAITPKG.
2763 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2764 (ix86_option_override_internal): Add PTA_WAITPKG.
2765 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2766 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2767 IX86_BUILTIN_TPAUSE.
2768 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2769 __builtin_ia32_umwait and __builtin_ia32_tpause.
2770 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2771 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2772 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2773 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2774 UNSPECV_TPAUSE): New.
2775 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2776 * config/i386/i386.opt: Add -mwaitpkg.
2777 * config/i386/waitpkgintrin.h: New file.
2778 * config/i386/x86intrin.h: New header.
2779 * doc/invoke.texi: Add -mwaitpkg.
2781 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2784 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2786 (cortex-m0): Use armv6s-m isa.
2787 (cortex-m0plus): Likewise.
2788 (cortex-m1): Likewise.
2789 (cortex-m0.small-multiply): Likewise.
2790 (cortex-m0plus.small-multiply): Likewise.
2791 (cortex-m1.small-multiply): Likewise.
2793 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2794 Jakub Jelinek <jakub@redhat.com>
2796 PR tree-optimization/85692
2797 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2798 source permute as well.
2800 2018-05-11 Martin Liska <mliska@suse.cz>
2803 * doc/extend.texi: Document LLVM style format for no_sanitize
2806 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2808 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2809 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2810 (mode_supports_vsx_dform_quad): Likewise.
2811 (mode_supports_vmx_dform): Move these functions to be next to the
2812 other mode_supports functions.
2813 (mode_supports_dq_form): Likewise.
2814 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2815 mode_supports_dq_form.
2816 (reg_offset_addressing_ok_p): Likewise.
2817 (offsettable_ok_by_alignment): Likewise.
2818 (rs6000_legitimate_offset_address_p): Likewise.
2819 (legitimate_lo_sum_address_p): Likewise.
2820 (rs6000_legitimize_address): Likewise.
2821 (rs6000_legitimize_reload_address): Likewise.
2822 (rs6000_secondary_reload_inner): Likewise.
2823 (rs6000_preferred_reload_class): Likewise.
2824 (rs6000_output_move_128bit): Likewise.
2826 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2828 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2829 Generate SImode target register for null target.
2830 <case IX86_BUILTIN_XGETBV>: Ditto.
2831 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2832 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2834 2018-05-10 Carl Love <cel@us.ibm.com>
2836 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2837 dcbtt and dcbtstt if operands[2] is 0.
2839 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2842 * config/i386/sse.md (usadv64qi): New expander.
2844 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2846 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2847 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2848 -maltivec=be support.
2849 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2850 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2851 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2852 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2853 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2854 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2855 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2856 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2857 altivec_vsumsws): Adjust.
2858 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2859 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2860 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2862 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2863 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2864 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2865 (altivec_lve<VI_char>x): Delete expand.
2866 (*altivec_lve<VI_char>x_internal): Rename to...
2867 (altivec_lve<VI_char>x): ... this.
2868 (altivec_lvxl_<mode>): Delete expand.
2869 (*altivec_lvxl_<mode>_internal): Rename to ...
2870 (altivec_lvxl_<mode>): ... this.
2871 (altivec_stvxl_<mode>): Delete expand.
2872 (*altivec_stvxl_<mode>_internal): Rename to ...
2873 (altivec_stvxl_<mode>): ... this.
2874 (altivec_stve<VI_char>x): Delete expand.
2875 (*altivec_stve<VI_char>x_internal): Rename to ...
2876 (altivec_stve<VI_char>x): ... this.
2877 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2878 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2879 reduc_plus_scal_<mode>): Adjust.
2880 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2882 (rs6000_cpu_cpp_builtins): Adjust.
2883 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2884 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2885 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2886 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2887 -maltivec=be support.
2888 (rs6000_split_vec_extract_var): Adjust.
2889 (rs6000_split_v4si_init): Adjust.
2890 (swap_selector_for_mode): Delete.
2891 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2892 altivec_expand_stvex_be): Delete.
2893 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2894 -maltivec=be support.
2895 (rs6000_gimple_fold_builtin): Ditto.
2896 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2898 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2899 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2900 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2901 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2902 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2903 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2904 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2905 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2906 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2907 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2908 anonymous split): Adjust.
2909 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2910 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2912 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2914 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2915 when --with-gxx-include-dir is also specified.
2916 * configure: Regenerate.
2918 2018-05-09 Jim Wilson <jimw@sifive.com>
2921 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2922 * config/riscv/t-withmultilib: New.
2923 * config/riscv/withmultilib.h: New.
2924 * doc/install.texi: Document RISC-V --with-multilib-list support.
2926 2018-05-09 Richard Biener <rguenther@suse.de>
2928 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2930 (vect_bb_vectorization_profitable_p): Adjust. Compute
2931 actual scalar cost using the cost vector and the add_stmt_cost
2934 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2936 PR rtl-optimization/85645
2937 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2938 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2940 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2942 PR rtl-optimization/85645
2943 * shrink-wrap.c (spread_components): Return a boolean saying if
2944 anything was changed.
2945 (try_shrink_wrapping_separate): Iterate spread_components until
2946 nothing changes anymore.
2948 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2950 PR rtl-optimization/85645
2951 * regrename.c (build_def_use): Also kill the chains that include the
2952 destination of a REG_CFA_REGISTER note.
2954 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2956 PR rtl-optimization/85645
2957 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2958 insn that has a REG_CFA_REGISTER note.
2960 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2962 * cfgexpand.c (expand_clobber): New function.
2963 (expand_gimple_stmt_1): Use it.
2964 * tree-vect-stmts.c (vect_clobber_variable): New function,
2966 (vectorizable_simd_clone_call): ...here.
2967 (vectorizable_store): Emit a clobber either side of an
2968 IFN_STORE_LANES sequence.
2969 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2971 2018-05-09 Tom de Vries <tom@codesourcery.com>
2974 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2975 (define_insn "trap_if_false"): Add exit after trap.
2977 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2979 PR rtl-optimization/85638
2980 * bb-reorder.c: Include common/common-target.h.
2981 (create_forwarder_block): New function extracted from...
2982 (fix_up_crossing_landing_pad): ...here. Rename into...
2983 (dw2_fix_up_crossing_landing_pad): ...this.
2984 (sjlj_fix_up_crossing_landing_pad): New function.
2985 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2986 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2987 from both partitions and exit the loop after one iteration.
2989 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2992 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2994 (Basic PowerPC Built-in Functions): The new name of the
2995 subsection previously known as "PowerPC Built-in Functions".
2996 (Basic PowerPC Built-in Functions Available on all Configurations):
2998 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3000 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3002 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3004 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3007 2018-05-08 Jim Wilson <jimw@sifive.com>
3009 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3010 (LD_EMUL_SUFFIX): New.
3011 (LINK_SPEC): Use it.
3013 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3015 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3017 (Basic PowerPC Built-in Functions): The new name of the
3018 subsection previously known as "PowerPC Built-in Functions".
3019 (Basic PowerPC Built-in Functions Available on all Configurations):
3021 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3023 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3025 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3027 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3030 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3033 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3034 after cmpelim optimization.
3036 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3038 * config.gcc: Support "goldmont".
3039 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3040 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3042 * config/i386/i386.c (m_GOLDMONT): Define.
3043 (processor_target_table): Add "goldmont".
3044 (PTA_GOLDMONT): Define.
3045 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3046 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3047 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3048 (fold_builtin_cpu): Add "goldmont".
3049 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3050 (ix86_option_override_internal): Add "goldmont".
3051 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3052 (processor_type): Add PROCESSOR_GOLDMONT.
3053 * config/i386/i386.md: Add CPU "glm".
3054 * config/i386/glm.md: New file.
3055 * config/i386/x86-tune.def: Add m_GOLDMONT.
3056 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3058 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3061 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3063 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3064 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3065 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3068 * config/i386/i386.c (ix86_fold_builtin): Handle
3069 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3072 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3073 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3075 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3078 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3079 (check_arch): Likewise.
3080 (check_fpu): Return the result rather than printing it.
3081 (end arch): Fix operator precedence.
3082 (end cpu): Likewise.
3083 (END): Print the result from check_fpu.
3085 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3086 Alan Hayward <alan.hayward@arm.com>
3087 David Sherwood <david.sherwood@arm.com>
3089 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3090 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3091 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3092 (*fcmuo<mode>_and): New patterns.
3094 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3096 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3097 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3098 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3099 (cmp_op, sve_imm_con): New code attributes.
3100 (SVE_COND_INT_CMP, imm_con): Delete.
3101 (cmp_op): Remove above unspecs from int attribute.
3102 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3104 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3105 comparison-specific unspecs.
3106 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3107 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3108 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3109 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3110 (*vec_fcm<cmp_op><mode>): Rename to...
3111 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3112 (*vec_fcmuo<mode>): Rename to...
3113 (*fcmuo<mode>): ...this and adjust likewise.
3114 (*pred_fcm<cmp_op><mode>): New pattern.
3115 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3116 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3118 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3120 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3121 (aarch64_emit_sve_predicated_cond): New function.
3122 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3123 (aarch64_emit_unspec_cond_or): Replace with...
3124 (aarch64_emit_sve_or_conds): ...this new function. Use
3125 aarch64_emit_sve_ptrue_op for the individual comparisons and
3126 aarch64_emit_binop to OR them together.
3127 (aarch64_emit_inverted_unspec_cond): Replace with...
3128 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3129 aarch64_emit_sve_ptrue_op for the comparison and
3130 aarch64_emit_unop to invert the result.
3131 (aarch64_expand_sve_vec_cmp_float): Update after the above
3132 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3134 2018-05-07 Nathan Sidwell <nathan@acm.org>
3136 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3137 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3138 (Backwards Compatibility): Likewise.
3140 2018-05-07 Luis Machado <luis.machado@linaro.org>
3144 2018-05-07 Luis Machado <luis.machado@linaro.org>
3146 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3147 <prefetch_dynamic_strides>: New const bool field.
3148 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3149 prefetch_dynamic_strides.
3150 (exynosm1_prefetch_tune): Likewise.
3151 (thunderxt88_prefetch_tune): Likewise.
3152 (thunderx_prefetch_tune): Likewise.
3153 (thunderx2t99_prefetch_tune): Likewise.
3154 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3156 (aarch64_override_options_internal): Update to set
3157 PARAM_PREFETCH_DYNAMIC_STRIDES.
3158 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3159 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3160 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3161 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3162 prefetch-dynamic-strides setting.
3164 2018-05-07 Luis Machado <luis.machado@linaro.org>
3166 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3167 <minimum_stride>: New const int field.
3168 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3169 minimum_stride field.
3170 (exynosm1_prefetch_tune): Likewise.
3171 (thunderxt88_prefetch_tune): Likewise.
3172 (thunderx_prefetch_tune): Likewise.
3173 (thunderx2t99_prefetch_tune): Likewise.
3174 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3175 (aarch64_override_options_internal): Update to set
3176 PARAM_PREFETCH_MINIMUM_STRIDE.
3177 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3178 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3179 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3180 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3181 stride is constant and is below the minimum stride threshold.
3183 2018-05-07 Luis Machado <luis.machado@linaro.org>
3185 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3188 2018-05-07 Luis Machado <luis.machado@linaro.org>
3190 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3191 <prefetch_dynamic_strides>: New const bool field.
3192 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3193 prefetch_dynamic_strides.
3194 (exynosm1_prefetch_tune): Likewise.
3195 (thunderxt88_prefetch_tune): Likewise.
3196 (thunderx_prefetch_tune): Likewise.
3197 (thunderx2t99_prefetch_tune): Likewise.
3198 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3200 (aarch64_override_options_internal): Update to set
3201 PARAM_PREFETCH_DYNAMIC_STRIDES.
3202 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3203 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3204 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3205 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3206 prefetch-dynamic-strides setting.
3208 2018-05-07 Luis Machado <luis.machado@linaro.org>
3210 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3211 <minimum_stride>: New const int field.
3212 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3213 minimum_stride field.
3214 (exynosm1_prefetch_tune): Likewise.
3215 (thunderxt88_prefetch_tune): Likewise.
3216 (thunderx_prefetch_tune): Likewise.
3217 (thunderx2t99_prefetch_tune): Likewise.
3218 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3219 (aarch64_override_options_internal): Update to set
3220 PARAM_PREFETCH_MINIMUM_STRIDE.
3221 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3222 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3223 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3224 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3225 stride is constant and is below the minimum stride threshold.
3227 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3230 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3231 the type is addressable. Don't force op into register if it has
3234 2018-05-05 Roland McGrath <mcgrathr@google.com>
3237 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3238 any section for which we don't know a specific type it should have,
3239 regardless of name. Previously this was done only for the exact
3240 names ".init_array", ".fini_array", and ".preinit_array".
3241 (default_elf_asm_named_section): Add comment about
3242 relationship with default_section_type_flags and SECTION_NOTYPE.
3243 (get_section): Don't consider it a type conflict if one side has
3244 SECTION_NOTYPE and the other doesn't, as long as neither has the
3245 SECTION_BSS et al used in the default_section_type_flags logic.
3247 2018-05-05 Tom de Vries <tom@codesourcery.com>
3250 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3251 (workaround_barsyncs): New function.
3252 (nvptx_reorg): Use workaround_barsyncs.
3253 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3254 (define_expand "nvptx_membar_cta"): New define_expand.
3255 (define_insn "*nvptx_membar_cta"): New insn.
3257 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3259 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3260 To improve optimization opportunities.
3261 * builtin-types.def: The new needed builtin types for the above.
3263 2018-05-04 Richard Biener <rguenther@suse.de>
3265 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3266 * gimple-ssa-store-merging.c
3267 (imm_store_chain_info::output_merged_store): Remove redundant create,
3268 release split_store vector contents on failure.
3269 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3270 scalar stmt vector on cache hit.
3272 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3274 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3276 * config.gcc (powerpc-xilinx-eabi*): Remove.
3277 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3279 (fusion_addis_mem_combo_load): Ditto.
3280 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3282 (rs6000_cpu_cpp_builtins): Ditto.
3283 * config/rs6000/rs6000-linux.c
3284 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3285 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3286 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3288 (rs6000_setup_reg_addr_masks): Ditto.
3289 (rs6000_init_hard_regno_mode_ok): Ditto.
3290 (rs6000_option_override_internal): Ditto.
3291 (legitimate_lo_sum_address_p): Ditto.
3292 (rs6000_legitimize_address): Ditto.
3293 (rs6000_legitimize_reload_address): Ditto.
3294 (rs6000_legitimate_address_p): Ditto.
3295 (abi_v4_pass_in_fpr): Ditto.
3296 (setup_incoming_varargs): Ditto.
3297 (rs6000_gimplify_va_arg): Ditto.
3298 (rs6000_split_multireg_move): Ditto.
3299 (rs6000_savres_strategy): Ditto.
3300 (rs6000_emit_prologue_components): Ditto.
3301 (rs6000_emit_epilogue_components): Ditto.
3302 (rs6000_emit_prologue): Ditto.
3303 (rs6000_emit_epilogue): Ditto.
3304 (rs6000_elf_file_end): Ditto.
3305 (rs6000_function_value): Ditto.
3306 (rs6000_libcall_value): Ditto.
3307 * config/rs6000/rs6000.h: Ditto.
3308 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3309 (TARGET_MINMAX): ... this. New.
3310 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3311 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3312 (*movsi_internal1_single): Delete.
3313 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3314 mfpu=, mxilinx-fpu): Delete.
3315 * config/rs6000/singlefp.h: Delete.
3316 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3317 * config/rs6000/t-rs6000: Ditto.
3318 * config/rs6000/t-xilinx: Delete.
3319 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3320 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3321 (VStype_simple): Delete.
3322 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3323 * config/rs6000/xfpu.h: Delete.
3324 * config/rs6000/xfpu.md: Delete.
3325 * config/rs6000/xilinx.h: Delete.
3326 * config/rs6000/xilinx.opt: Delete.
3327 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3328 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3330 2018-05-04 Tom de Vries <tom@codesourcery.com>
3333 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3336 2018-05-04 Richard Biener <rguenther@suse.de>
3339 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3340 (expand_complex_div_wide): Likewise.
3341 (expand_complex_operations_1): Likewise.
3342 (expand_complex_libcall): Preserve EH info of the original stmt.
3343 (tree_lower_complex): Handle removed blocks.
3344 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3345 on complex multiplication and division libcall builtins.
3347 2018-05-04 Richard Biener <rguenther@suse.de>
3350 * fold-const.c (negate_expr_p): Restrict negation of operand
3351 zero of a division to when we know that can happen without
3353 (fold_negate_expr_1): Likewise.
3355 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3358 * real.h (real_nextafter): Declare.
3359 * real.c (real_nextafter): New function.
3360 * fold-const-call.c (fold_const_nextafter): New function.
3361 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3362 CASE_CFN_NEXTTOWARD.
3363 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3364 even when arg1_mode is different from arg0_mode.
3366 2018-05-03 Nathan Sidwell <nathan@acm.org>
3368 * doc/extend.texi (Deprecated Features): Remove
3370 (Backwards Compatibility): Likewise.
3371 * doc/invoke.texi (C++ Language Options): Likewise.
3372 (C++ Dialect Options): Likewise.
3374 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3377 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3378 _mm512_mask_mullox_epi64): New intrinsics.
3380 2018-05-03 Tom de Vries <tom@codesourcery.com>
3383 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3384 dump files): Add offload-tree.
3386 2018-05-03 Richard Biener <rguenther@suse.de>
3388 PR tree-optimization/85615
3389 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3390 to loops not nested in BBs loop father to avoid creating multi-entry
3393 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3395 PR tree-optimization/70291
3396 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3397 arguments. Change return type to tree. Emit libcall as a new
3398 statement rather than replacing existing one when inplace_p is true.
3399 (expand_complex_multiplication_components): New function.
3400 (expand_complex_multiplication): Expand floating-point complex
3401 multiplication using the above.
3402 (expand_complex_division): Rename inner_type parameter to type.
3403 Update expand_complex_libcall call-site.
3404 (expand_complex_operations_1): Update expand_complex_multiplication
3405 and expand_complex_division call-sites.
3407 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3410 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3411 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3412 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3413 the highest significant bit of the shift count mask is clear. In
3414 check whether and[sq]i3 is needed verify that all significant bits
3415 of the shift count other than the highest are set.
3417 2018-05-02 Tom de Vries <tom@codesourcery.com>
3420 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3421 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3422 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3423 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3424 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3425 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3426 __builtin_goacc_parlevel_size.
3428 2018-05-02 Richard Biener <rguenther@suse.de>
3430 PR tree-optimization/85597
3431 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3432 do not use split vect_get_vec_defs call but call vect_get_slp_defs
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 ltrans-tree.
3441 2018-05-02 Tom de Vries <tom@codesourcery.com>
3444 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3445 dump files): Add wpa-ipa.
3447 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3449 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3450 powerpc*-*-linux*paired* target.
3451 * config/rs6000/750cl.h: Delete.
3452 * config/rs6000/paired.h: Delete.
3453 * config/rs6000/paired.md: Delete.
3454 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3456 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3457 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3458 comment. Remove paired float support.
3459 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3460 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3462 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3463 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3465 * config/rs6000/rs6000.c: Remove paired float support.
3466 (paired_expand_vector_init, paired_expand_vector_move,
3467 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3468 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3469 paired_expand_builtin, paired_expand_predicate_builtin,
3470 paired_init_builtins): Delete.
3471 * config/rs6000/rs6000.h: Remove paired float support.
3472 * config/rs6000/rs6000.md: Remove paired float support.
3473 (move_from_CR_ov_bit): Delete.
3474 * config/rs6000/rs6000.opt (mpaired): Delete.
3475 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3476 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3478 2018-05-02 Richard Biener <rguenther@suse.de>
3481 * gimplify.c (gimplify_save_expr): When in SSA form allow
3482 SAVE_EXPRs to compute to SSA vars.
3484 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3487 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3488 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3489 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3490 clobber operands[2], instead use a new pseudo. Formatting fixes.
3492 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3494 PR tree-optimization/85586
3495 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3496 exit early for statements in the same group if the accesses are
3499 2018-05-02 Tom de Vries <tom@codesourcery.com>
3502 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3505 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3507 PR tree-optimization/85143
3508 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3510 2018-05-01 Tom de Vries <tom@codesourcery.com>
3513 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3514 not found" error message.
3516 2018-05-01 Tom de Vries <tom@codesourcery.com>
3519 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3520 * vec.c (test_ordered_remove_if): New function.
3521 (vec_c_tests): Call test_ordered_remove_if.
3522 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3523 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3524 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3525 VEC_ORDERED_REMOVE_IF.
3527 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3529 PR tree-optimization/82665
3530 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3531 pointer subtraction where arguments come from a memchr call.
3533 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3535 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3536 --push-state --as-needed and --pop-state instead of --as-needed and
3537 --no-as-needed if ld supports it.
3538 * configure: Regenerated.
3541 * doc/install.texi2html: Replace _002d with - and _002a with * in
3542 generated html files using sed.
3544 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3547 * gcc-rich-location.c (blank_line_before_p): New function.
3548 (use_new_line): New function.
3549 (gcc_rich_location::add_fixit_insert_formatted): New function.
3550 * gcc-rich-location.h
3551 (gcc_rich_location::add_fixit_insert_formatted): New function.
3553 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3555 * selftest.c (assert_streq): Rename "expected" and "actual" to
3556 "val1" and "val2". Extend NULL-handling to cover both inputs
3557 symmetrically, while still requiring both to be non-NULL for a pass.
3558 * selftest.h (assert_streq): Rename "expected" and "actual" to
3560 (ASSERT_EQ): Likewise.
3561 (ASSERT_EQ_AT): Likewise.
3562 (ASSERT_KNOWN_EQ): Likewise.
3563 (ASSERT_KNOWN_EQ_AT): Likewise.
3564 (ASSERT_NE): Likewise.
3565 (ASSERT_MAYBE_NE): Likewise.
3566 (ASSERT_MAYBE_NE_AT): Likewise.
3567 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3568 the assertion to pass.
3569 (ASSERT_STREQ_AT): Likewise.
3571 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3573 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3574 interaction with -pie.
3576 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3578 * selftest.h: Fix alphabetization of per-source-file selftest
3581 2018-04-30 Jason Merrill <jason@redhat.com>
3583 PR c++/61982 - dead stores to destroyed objects.
3584 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3587 2018-04-30 Jason Merrill <jason@redhat.com>
3589 * tree.c (build_clobber): New.
3590 * tree.h: Declare it.
3591 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3593 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3595 * diagnostic-show-locus.c (layout::layout): Update for
3596 location_get_source_line returning a char_span.
3597 (struct char_span): Move to input.h.
3598 (struct correction): Update for fields in char_span becoming
3600 (struct source_line): Update for location_get_source_line
3601 returning a char_span.
3602 (layout::print_line): Likewise.
3603 * edit-context.c (edited_file::print_content): Likewise.
3604 (edited_file::print_diff_hunk): Likewise.
3605 (edited_file::print_run_of_changed_lines): Likewise.
3606 (edited_file::get_num_lines): Likewise.
3607 (edited_line::edited_line): Likewise.
3608 * final.c (asm_show_source): Likewise.
3609 * input.c (location_get_source_line): Convert return type
3610 from const char * to char_span, losing the final "line_len"
3612 (dump_location_info): Update for the above.
3613 (get_substring_ranges_for_loc): Likewise. Use a char_span
3614 when handling the literal within the line.
3615 (test_reading_source_line): Update for location_get_source_line
3616 returning a char_span.
3617 * input.h (class char_span): Move here from
3618 diagnostic-show-locus.c, converting from a struct to a class.
3619 Make data members private.
3620 (char_span::operator bool): New.
3621 (char_span::length): New.
3622 (char_span::get_buffer): New.
3623 (char_span::operator[]): New.
3624 (char_span::subspan): Make const.
3625 (char_span::xstrdup): New.
3626 (location_get_source_line): Convert return type from const char *
3627 to char_span, losing the final "line_size" param.
3629 2018-04-30 Jan Hubicka <jh@suse.cz>
3631 * lto-wrapper.c (ltrans_priorities): New static var.
3632 (cmp_priority): New.
3633 (run_gcc): Read priorities and if doing parallel build order
3634 the Makefile by them.
3636 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3638 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3640 2018-04-30 Richard Biener <rguenther@suse.de>
3642 * tree-cfg.c (verify_address): Remove base argument, add
3643 flag whether to check TREE_ADDRESSABLE and do that.
3644 (verify_expr): Remove.
3645 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3646 (verify_gimple_assign_single): Likewise.
3647 (verify_gimple_switch): Likewise.
3648 (verify_expr_location_1): Dereference tp once. Add (disabled)
3649 piece from verify_expr.
3650 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3652 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3654 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3656 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3658 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3659 (small_data_pattern): Likewise.
3660 (arc_rewrite_small_data): Likewise.
3661 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3662 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3663 (get_symbol_alignment): New function.
3664 (legitimate_small_data_address_p): Likewise.
3665 (legitimate_scaled_address): Update, call
3666 legitimate_small_data_address_p.
3667 (output_sdata): New static variable.
3668 (arc_print_operand): Update how we handle small data operands.
3669 (arc_print_operand_address): Likewise.
3670 (arc_legitimate_address_p): Update, use
3671 legitimate_small_data_address_p.
3672 (arc_rewrite_small_data_p): Remove.
3673 (arc_rewrite_small_data_1): Likewise.
3674 (arc_rewrite_small_data): Likewise.
3675 (small_data_pattern): Likewise.
3676 (compact_sda_memory_operand): Update to use
3677 legitimate_small_data_address_p and get_symbol_alignment.
3678 (prepare_move_operands): Don't rewite sdata pattern.
3679 (prepare_extend_operands): Remove.
3680 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3682 (zero_extendqisi2): Likewise.
3683 (zero_extendhisi2): Likewise.
3684 (extendqihi2): Likewise.
3685 (extendqisi2): Likewise.
3686 (extendhisi2): Likewise.
3690 * config/arc/constraints.md (Usd): Change it to memory constraint.
3692 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3694 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3695 as source of std instructions.
3696 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3697 allow 6-bit constants as source for store instructions.
3698 (movdi_insn): Update instruction pattern to allow 6-bit constants
3699 as source for store instructions.
3701 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3703 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3705 2018-04-30 Nathan Sidwell <nathan@acm.org>
3706 Sandra Loosemore <sandra@codesourcery.com>
3708 * dumpfile.c (dump_open): Allow '-' for stdout.
3709 * doc/invoke.texi (Developer Options): Document dump filename
3710 determination early. Document stdin/stdout selection.
3712 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3714 Microblaze Target: PIC data text relative
3716 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3717 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3719 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3720 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3721 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3722 New addressing mode for data-text relative position indepenedent code.
3723 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3724 'ADDRESS_SYMBOLIC_TXT_REL'.
3725 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3726 (microblaze_legitimate_pic_operand): Exclude function calls from
3727 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3728 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3730 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3731 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3732 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3733 for 'address + offset'.
3734 (microblaze_expand_prologue): Add new function prologue call for
3736 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3737 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3738 table in case of TARGET_PIC_DATA_TEXT_REL.
3739 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3740 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3741 Add new macros 'UNSPEC_TEXT',
3742 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3743 + exclude function calls from 'UNSPEC_PLT' in case of data text
3745 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3746 new target hook for generating address diff vector tables in case of
3748 * doc/tm.texi : Regenerate.
3749 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3750 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3751 of addr diff vector generation.
3752 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3753 target hook definition.
3754 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3755 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3756 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3757 option for data text relative.
3759 2018-04-30 Richard Biener <rguenther@suse.de>
3761 * tree-chrec.h (evolution_function_is_constant_p): Remove
3763 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3765 2018-04-30 Richard Biener <rguenther@suse.de>
3768 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3770 2018-04-30 Richard Biener <rguenther@suse.de>
3772 PR tree-optimization/28364
3773 PR tree-optimization/85275
3774 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3775 copying first exit test.
3777 2018-04-28 Mark Wielaard <mark@klomp.org>
3779 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3781 (dwarf_AT): Handle DW_AT_addr_base.
3782 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3784 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3787 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3788 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3789 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3790 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3792 2018-04-28 Richard Biener <rguenther@suse.de>
3794 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3795 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3796 to reflect use. Only add interesting stmts.
3798 2018-04-27 Martin Jambor <mjambor@suse.cz>
3801 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3802 the jump function allows for passing through aggregate values.
3804 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3806 * input.h (in_system_header_at): Convert from macro to inline
3808 (from_macro_expansion_at): Likewise.
3809 (from_macro_definition_at): Likewise.
3811 2018-04-27 Jeff Law <law@redhat.com>
3813 * config.gcc: Mark tile* targets as deprecated/obsolete.
3815 2018-04-27 Richard Biener <rguenther@suse.de>
3817 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3820 2018-04-27 Richard Biener <rguenther@suse.de>
3822 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3824 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3826 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3827 with Yd constraint. Set "preferred_for_speed" attribute from
3828 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3830 (*movdi_internal): Ditto.
3831 (movti_interunit splitters): Remove
3832 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3833 (movdi_interunit splitters): Ditto.
3834 * config/i386/constraints.md (Ye): Remove.
3835 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3837 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3840 * config/aarch64/constraints.md (Usg): Limit to 31.
3843 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3845 PR tree-optimization/85529
3846 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3847 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3848 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3849 zero extension or masking of the MSB bit.
3850 (optimize_range_tests): Add FIRST_BB argument, pass it through
3851 to optimize_range_tests_var_bound.
3852 (maybe_optimize_range_tests, reassociate_bb): Adjust
3853 optimize_range_tests callers.
3855 2018-04-26 Richard Biener <rguenther@suse.de>
3856 Jakub Jelinek <jakub@redhat.com>
3858 * cgraph.h (symbol_table): Just declare debug method here.
3859 * symtab.c (symbol_table::debug): Define.
3861 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3863 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3865 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3867 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3868 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3869 (*movdi_internal): Substitute Yi and Yj constraint with x
3870 and Ym and Yn constraint with y constraint. Update "isa"
3871 attribute and set "preferred_for_speed" attribute from
3872 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3873 (*movsi_internal): Ditto.
3874 (*movdf_internal): Ditto.
3875 (*movsf_internal): Ditto.
3876 (*zero_extendsidi2): Ditto.
3877 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3878 (sse2_loadld): Ditto.
3879 (*vec_extract<ssevecmodelower>_0): Ditto.
3880 (*vec_extractv4si_0_zext_sse4): Ditto.
3881 (vec_concatv2di): Ditto.
3882 (*vec_dup<mode>): Ditto.
3883 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3884 * config/i386/constraints.md (Yi): Remove.
3889 2018-04-26 Nathan Sidwell <nathan@acm.org>
3891 * dumpfile.c (dump_open): New.
3892 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3893 (dump_finish): Detect stdio/stderr by value not name.
3895 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3897 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3899 2018-04-26 Tom de Vries <tom@codesourcery.com>
3902 * config/nvptx/nvptx.c (verify_neutering_jumps)
3903 (verify_neutering_labels): New function
3904 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3906 2018-04-26 Tom de Vries <tom@codesourcery.com>
3909 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3910 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3913 2018-04-26 Richard Biener <rguenther@suse.de>
3914 Tom de Vries <tom@codesourcery.com>
3917 * lto-streamer-out.c (output_function): Fixup loops if required to match
3918 discovery done in the reader.
3920 2018-04-26 Richard Biener <rguenther@suse.de>
3922 PR tree-optimization/85116
3923 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3924 have a loop exit from the single latch predecessor. Remove
3925 case of header with just condition.
3926 (ch_base::copy_headers): Exclude infinite loops from any
3928 (pass_ch::execute): Record exits.
3930 2018-04-26 Richard Biener <rguenther@suse.de>
3932 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3933 prologue cost vector and pass it to vect_get_load_cost.
3934 (vect_get_peeling_costs_all_drs): Likewise.
3935 (vect_peeling_hash_get_lowest_cost): Likewise.
3936 (vect_enhance_data_refs_alignment): Likewise.
3938 2018-04-26 Richard Biener <rguenther@suse.de>
3941 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3942 checking of integer<->pointer conversions.
3943 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3944 sign-/zero-extending pointer types.
3945 (expand_omp_for_static_chunk): Likewise.
3947 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3948 Jean Lee <xiaoyur347@gmail.com>
3950 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3951 (TARGET_ASAN_SHADOW_OFFSET): Define.
3952 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3953 true for -fsanitize=address.
3955 2018-04-25 Mark Wielaard <mark@klomp.org>
3957 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3960 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3962 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3963 than "alu", remove explicit "memory" and "imm_disp" attributes.
3964 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3967 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3968 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3969 gen_lowpart_no_emit.
3971 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3974 * config/i386/i386.c (ix86_expand_builtin): Change memory
3975 operand to XI, extend p0 to Pmode.
3976 * config/i386/i386.md: Change unspec volatile and operand
3977 1 mode to XI, change operand 0 mode to P.
3979 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3981 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3982 GET_MODE_MASK before any checking.
3983 (nds32_can_use_bset_p): Likewise.
3984 (nds32_can_use_btgl_p): Likewise.
3986 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3988 * config/nds32/nds32-doubleword.md: New define_split pattern for
3989 illegal register number.
3991 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3993 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3995 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3997 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3999 2018-04-25 Richard Biener <rguenther@suse.de>
4001 * lto-streamer.h (LTO_major_version): Bump to 8.
4003 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4005 * BASE-VER: Set to 9.0.0.
4007 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4009 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4010 in __abskf2 and __powikf2.
4012 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4015 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4016 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4017 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4018 Use the above on operand 2. Reindent.
4019 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4021 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4024 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4026 * config/i386/i386.opt (mcet): Removed.
4027 * doc/install.texi: Remove -mcet documentation.
4028 * doc/invoke.texi: Likewise.
4030 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4033 * doc/install.texi: Remove -mcet from bootstrap-cet.
4035 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4038 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4039 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4043 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4044 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4045 containing a CONST_VECTOR.
4047 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4049 * doc/install.texi: Update newlib dependency for nvptx.
4051 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4054 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4055 instead of INTVAL when shifting x left.
4057 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4059 PR tree-optimization/85478
4060 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4061 vect_grouped_store_supported for single element vectors.
4063 2018-04-24 Richard Biener <rguenther@suse.de>
4066 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4067 load cost increase to the case of non-constant step.
4069 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4072 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4073 destination if any_malformed_asm.
4075 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4078 * expr.c (store_field): In the bitfield case, if the value comes from
4079 a function call and is returned in registers by means of a PARALLEL,
4080 do not change the mode of the temporary unless BLKmode and VOIDmode.
4082 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4084 PR rtl-optimization/85423
4085 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4086 dependencies to debug insns when the previous insn is non-debug.
4088 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4090 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4091 enums into a single definition.
4092 (fls): Fix predicates and printing.
4095 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4097 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4098 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4099 and short u6 immediate.
4100 (check_if_valid_sleep_operand): Remove.
4101 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4103 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4105 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4106 flag_always_save_lp condition.
4107 * config/nds32/nds32.opt (malways-save-lp): New option.
4109 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4111 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4112 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4113 * config/nds32/nds32.h
4114 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4115 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4117 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4119 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4120 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4122 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4123 Chung-Ju Wu <jasonwucj@gmail.com>
4125 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4127 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4128 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4130 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4132 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4134 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4136 * config/nds32/nds32-protos.h (nds32_data_alignment,
4137 nds32_local_alignment): Declare.
4138 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4139 nds32_local_alignment): New functions.
4140 (TARGET_CONSTANT_ALIGNMENT): Define.
4141 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4143 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4145 * config/nds32/nds32.c
4146 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4147 (TARGET_MODES_TIEABLE_P): Likewise.
4149 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4151 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4154 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4155 Chung-Ju Wu <jasonwucj@gmail.com>
4157 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4158 for unaligned access.
4159 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4161 * config/nds32/nds32-intrinsic.md: Likewise.
4162 * config/nds32/nds32_intrinsic.h: Likewise.
4163 * config/nds32/nds32.h (nds32_builtins): Likewise.
4164 * config/nds32/nds32.opt (munaligned-access): New option.
4165 * config/nds32/nds32.c (nds32_asm_file_start): Display
4166 flag_unaligned_access status.
4168 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4170 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4171 -mno-relax is present.
4172 * config/riscv/linux.h (LINK_SPEC): Ditto.
4174 2018-04-20 Martin Sebor <msebor@redhat.com>
4177 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4179 (gimple_fold_builtin_stxcpy_chk): Same.
4180 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4182 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4185 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4186 __powikf2 when long double is IEEE 128-bit.
4188 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4190 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4191 step to make sure stack always aligned.
4193 2018-04-20 Carl Love <cel@us.ibm.com>
4196 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4197 size check for arg0.
4199 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4200 Tom de Vries <tom@codesourcery.com>
4203 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4204 Emit insns for calls too.
4205 (nvptx_find_par): Always look for worker-level predecessor insn.
4206 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4208 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4209 (nvptx_process_pars): Propagate frames for calls.
4211 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4214 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4216 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4217 (ix86_handle_option): Don't handle OPT_mibt.
4218 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4220 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4222 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4223 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4224 (ix86_target_macros): Define __CET__ with flag_cf_protection
4225 for -fcf-protection.
4226 * config/i386/i386.c (isa2_opts): Remove -mibt.
4227 * config/i386/i386.h (TARGET_IBT): Removed.
4228 (TARGET_IBT_P): Likewise.
4229 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4230 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4231 * config/i386/i386.opt (mcet): Update help message.
4234 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4235 -mcet as an alias for -mshstk.
4237 2018-04-20 Richard Biener <rguenther@suse.de>
4240 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4241 complexity by forcing a single use of the multiply operand.
4243 2018-04-20 Martin Jambor <mjambor@suse.cz>
4246 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4247 recursion dependency to only apply to non-clones.
4249 2018-04-20 Martin Jambor <mjambor@suse.cz>
4252 * ipa-cp.c (create_specialized_node): Check that clones of
4253 self-recursive edges exist during IPA-CP.
4255 2018-04-19 Toon Moene <toon@moene.org>
4257 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4260 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4262 PR tree-optimization/85467
4263 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4264 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4265 VECTOR_CST element to type.
4267 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4270 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4271 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4272 (builtin_longjmp): Likewise.
4273 (save_stack_nonlocal): New pattern.
4274 (restore_stack_nonlocal): Likewise.
4276 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4279 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4280 Replace ASM_OUTPUT_LABEL with fprintf.
4282 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4285 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4286 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4287 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4288 define __IBT__ and __SHSTK__ for -fcf-protection.
4289 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4291 (ix86_trampoline_init): Likewise.
4292 (x86_output_mi_thunk): Likewise.
4293 (ix86_notrack_prefixed_insn_p): Likewise.
4294 (ix86_option_override_internal): Don't disallow -fcf-protection.
4295 * config/i386/i386.md (rdssp<mode>): Also enable for
4297 (incssp<mode>): Likewise.
4298 (nop_endbr): Likewise.
4299 * config/i386/i386.opt (mcet): Change help message to built-in
4303 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4304 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4305 enable CET built-in functions.
4307 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4309 * common/config/i386/i386-common.c
4310 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4311 OPTION_MASK_ISA_MOVDIRI_UNSET,
4312 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4313 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4314 * config.gcc (movdirintrin.h): New header.
4315 * config/i386/cpuid.h (bit_MOVDIRI,
4316 bit_MOVDIR64B): New bits.
4317 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4319 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4320 (VOID, PVOID, PCVOID)): New function types.
4321 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4322 __builtin_ia32_directstoreu_u64,
4323 __builtin_ia32_movdir64b): New builtins.
4324 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4325 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4327 (ix86_valid_target_attribute_inner_p): Ditto.
4328 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4329 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4330 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4331 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4332 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4333 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4334 (movdiri<mode>, movdir64b_<mode>): New.
4335 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4336 * config/i386/immintrin.h: Include movdirintrin.h.
4337 * config/i386/movdirintrin.h: New file.
4338 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4340 2018-04-19 Richard Biener <rguenther@suse.de>
4343 * cfg.c (clear_bb_flags): When loop state says we have
4344 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4346 2018-04-19 Richard Biener <rguenther@suse.de>
4348 PR tree-optimization/84737
4349 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4350 copying restrict info.
4351 (vect_setup_realignment): Use it.
4352 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4353 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4354 the first DR to all generated stores.
4355 (vectorizable_load): Likewise for loads.
4357 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4359 PR tree-optimization/85446
4360 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4361 the integral and pointer types to have the same precision.
4363 * doc/install.texi: Document --disable-cet being the default and
4366 2018-04-18 Martin Liska <mliska@suse.cz>
4368 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4371 2018-04-18 Martin Liska <mliska@suse.cz>
4374 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4377 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4378 arguments if they are comparable.
4380 2018-04-18 Martin Liska <mliska@suse.cz>
4383 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4386 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4389 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4392 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4393 ENDBR after calling __morestack.
4395 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4398 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4399 by using gcc_base_ver to generate a gcc_driver_version, and use
4400 it when generating GCC_DRIVER_NAME.
4401 * configure: Regenerate.
4403 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4406 * config.gcc: Obsolete powerpc*-*-*spe*.
4408 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4411 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4412 (stabstr_D): Change type of unum from unsigned int to
4413 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4416 2018-04-17 Jim Wilson <jimw@sifive.com>
4419 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4420 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4421 Set arg_pointer_offset after using pretend_args_size.
4423 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4425 PR rtl-optimization/85431
4426 * dse.c (record_store): Ignore zero width stores.
4429 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4430 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4431 __builtin_stack_restore rather than after it.
4432 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4433 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4434 argument instead of virtual_dynamic_stack_rtx.
4436 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4438 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4440 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4441 Add note to error message to explain internal mapping of overloaded
4442 built-in function name to non-overloaded built-in function name.
4443 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4446 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4449 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4450 where the inputs overlap with the output.
4452 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4454 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4455 (=v, v) alternative and explicit "memory" attribute.
4456 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4457 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4459 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4460 "sselog1" type instead of "sselog".
4461 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4462 "sselog". Remove explicit "memory" attribute.
4463 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4464 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4466 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4467 "sselog1" type instead of "sselog". Remove explicit "memory"
4469 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4470 use "sselog1" type instead of "sselog". Remove explicit "memory"
4472 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4473 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4475 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4476 "sselog1" type instead of "sselog". Remove explicit "memory"
4478 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4479 use "sselog1" type instead of "sselog". Remove explicit "memory"
4483 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4486 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4489 2018-04-17 Martin Jambor <mjambor@suse.cz>
4492 * ipa-cp.c (create_specialized_node): Call
4493 expand_all_artificial_thunks if necessary.
4495 2018-04-17 Martin Liska <mliska@suse.cz>
4498 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4499 in message, remote space in between '_G' and '('.
4501 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4504 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4505 avx512f_vmcmp<mode>3<round_saeonly_name>,
4506 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4507 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4508 avx512f_rndscale<mode><round_saeonly_name>,
4509 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4510 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4511 Use %<iptr>2 instead of %2 for -masm=intel.
4512 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4513 avx512f_vcvttss2usi<round_saeonly_name>,
4514 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4516 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4517 avx512f_vcvttsd2usi<round_saeonly_name>,
4518 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4519 Use %q1 instead of %1 for -masm=intel.
4520 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4521 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4522 of %3 for -masm=intel.
4523 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4525 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4527 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4529 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4530 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4532 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4534 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4535 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4536 %g1 and one with %0 and %1.
4537 (avx512er_vmrcp28<mode><round_saeonly_name>,
4538 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4540 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4541 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4542 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4543 of %0 and %{%4%} for -masm=intel.
4544 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4545 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4546 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4547 order of %0 and %{%5%}%{z%} for -masm=intel.
4549 2018-04-17 Jan Hubicka <jh@suse.cz>
4552 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4554 2018-04-17 Martin Liska <mliska@suse.cz>
4557 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4558 for target_clone error message. Make default implementation
4559 clone to be a local declaration.
4560 (separate_attrs): Add new argument and check for an empty
4562 (expand_target_clones): Handle it.
4563 (ipa_target_clone): Make redirection just for target_clones
4566 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4567 Tom de Vries <tom@codesourcery.com>
4570 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4571 tiled basic blocks without omp continue statements.
4573 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4576 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4577 vec_extract expression as having side effects to make sure it gets
4580 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4583 * config/i386/i386.c (get_builtin_code_for_version): Check
4586 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4589 * gcc/config.gcc: Support "skylake".
4590 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4592 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4593 (processor_target_table): Add "skylake".
4594 (ix86_option_override_internal): Add "skylake".
4595 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4596 PROCESSOR_CANNONLAKE.
4597 (get_builtin_code_for_version): Fix priority for
4598 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4599 PROCESSOR_SKYLAKE-AVX512.
4600 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4601 (processor_type): Add PROCESSOR_SKYLAKE.
4603 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4604 Jason Merrill <jason@redhat.com>
4607 * convert.c (convert_to_integer_1): Use direct recursion for
4608 enumeral types and types with a precision less than the number
4609 of bits in their mode.
4611 2018-04-16 Julia Koval <julia.koval@intel.com>
4614 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4615 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4617 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4620 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4621 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4622 and -mno-direct-move.
4624 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4627 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4628 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4629 Ensure negative shifts result in {0}.
4631 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4633 PR rtl-optimization/79916
4634 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4635 regs (if any) to define how to gnerate SD moves when LRA is in
4638 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4640 PR rtl-optimization/85393
4641 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4642 * except.c (expand_dw2_landing_pad_for_region): Make static.
4643 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4644 a label and unconditional jump to old_bb, rather than
4645 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4648 PR rtl-optimization/85376
4649 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4650 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4651 instead of a specific value.
4653 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4654 Bin Cheng <bin.cheng@arm.com>
4656 PR tree-optimization/82965
4657 PR tree-optimization/83991
4658 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4659 by_profile_only parameter.
4660 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4661 information if the loop was predicted to iterate too many times.
4662 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4664 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4667 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4670 2018-04-13 Martin Liska <mliska@suse.cz>
4671 Jakub Jelinek <jakub@redhat.com>
4674 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4675 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4676 * builtins.c (expand_builtin_memory_copy_args): Use
4677 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4678 handle dest_addr == pc_rtx.
4680 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4683 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4684 asked to not generate direct moves.
4685 (fix_trunc<mode>si2_stfiwx): Similar.
4686 (fix_trunc<mode>si2_internal): Similar.
4688 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4691 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4692 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4693 lookup if dest in some wider mode is known to be const0_rtx and
4694 if so, record permanent equivalence for it to be ZERO_EXTEND of
4695 the narrower mode destination.
4697 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4699 * lto-streamer-out.c (output_function): Revert 259346.
4700 * omp-expand.c (expand_oacc_for): Likewise.
4702 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4704 PR rtl-optimization/85354
4705 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4706 * sel-sched.c (sel_global_init): ... here.
4708 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4711 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4712 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4713 mode for PE-COFF targets.
4714 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4715 (i386_pe_asm_lto_end): Likewise.
4716 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4717 (TARGET_ASM_LTO_END): Likewise.
4718 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4719 (i386_pe_asm_lto_start): New function.
4720 (i386_pe_asm_lto_end): Likewise.
4722 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4723 Richard Biener <rguenther@suse.de>
4726 * lto-streamer-out.c (output_function): Fix CFG loop state before
4728 * omp-expand.c (expand_oacc_for): Handle calls to internal
4729 functions like regular functions.
4731 2018-04-12 Richard Biener <rguenther@suse.de>
4734 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4735 for the early LTO debug to properly generate references to it
4736 during DIE emission. Do not re-use that for the skeleton for
4738 (dwarf2out_early_finish): Likewise.
4740 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4743 * config/i386/sse.md
4744 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4745 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4746 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4747 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4748 and output is a reg, avoid creating invalid lowpart subreg, but
4749 instead split into a 512-bit move. Don't split if not AVX512VL,
4750 input is xmm16+ reg and output is a mem.
4751 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4752 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4753 xmm16+ reg and output is a mem.
4755 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4757 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4758 also for flag_dwarf2_cfi_asm.
4760 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4762 PR rtl-optimization/85342
4763 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4764 a bool scalar var inside of the loop instead. Don't try to update
4765 recog_data.operand after failed apply_change_group.
4767 2018-04-12 Tom de Vries <tom@codesourcery.com>
4770 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4771 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4772 array with flexible array member as array without given dimension.
4773 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4774 argument for undefined param to true.
4776 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4779 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4780 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4781 from PowerPC section.
4782 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4783 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4784 help text that is too long.
4785 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4786 help text that is too long.
4787 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4788 help text that is too long.
4790 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4792 * config/alpha/alpha.md (stack_probe_internal): Rename
4793 from "probe_stack". Update all callers.
4795 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4797 PR rtl-optimization/84566
4798 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4799 sched_macro_fuse_insns.
4801 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4804 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4805 (compute_block_dependences): ... from here.
4807 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4809 PR tree-optimization/85331
4810 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4811 from int to HOST_WIDE_INT.
4813 2018-04-11 Martin Jambor <mjambor@suse.cz>
4816 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4817 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4818 not the same as the source val.
4819 (cgraph_edge_brings_value_p): New parameter.
4820 (gather_edges_for_value): Pass destination value to
4821 cgraph_edge_brings_value_p.
4822 (perhaps_add_new_callers): Likewise.
4823 (get_info_about_necessary_edges): Likewise and exclude values brought
4824 only by self-recursive edges.
4825 (create_specialized_node): Redirect only clones of self-calling edges.
4826 (+self_recursive_pass_through_p): New function.
4827 (find_more_scalar_values_for_callers_subset): Use it.
4828 (find_aggregate_values_for_callers_subset): Likewise.
4829 (known_aggs_to_agg_replacement_list): Removed.
4830 (decide_whether_version_node): Re-calculate known constants for all
4831 remaining context clones.
4833 2018-04-11 Richard Biener <rguenther@suse.de>
4836 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4837 from early DWARF output.
4838 (dwarf2out_early_finish): Output line info unconditionally into
4839 early DWARF and add reference to it.
4841 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4844 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4845 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4846 other than V2DFmode using iptr mode attribute.
4847 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4849 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4851 PR rtl-optimization/84659
4852 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4854 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4857 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4859 (output_loc_list): Pass address of a dummy size variable even in the
4860 locview handling loop.
4861 (index_location_lists): Add comment on why skip_loc_list_entry can't
4864 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4867 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4870 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4873 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4874 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4875 and -mstring-compare-inline-limit.
4877 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4880 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4881 for stack clash protection in a register whenever we need it to be in
4884 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4886 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4887 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4889 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4892 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4894 (mlong-double-): Ditto.
4895 * config/rs6000/sysv4.opt (msdata=): Ditto.
4896 (mtls-size=): Ditto.
4898 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4900 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4901 erroneous entries for
4902 "vector int vec_ldl (int, long int *)", and
4903 "vector unsigned int vec_ldl (int, unsigned long int *)".
4904 Add comments and entries for
4905 "vector bool char vec_ldl (int, bool char *)",
4906 "vector bool short vec_ldl (int, bool short *)",
4907 "vector bool int vec_ldl (int, bool int *)",
4908 "vector bool long long vec_ldl (int, bool long long *)",
4909 "vector pixel vec_ldl (int, pixel *)",
4910 "vector long long vec_ldl (int, long long *)",
4911 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4912 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4913 type tree bool_long_long_type_node and correct definition of
4914 bool_V2DI_type_node to make reference to this new type tree.
4915 (rs6000_mangle_type): Replace erroneous reference to
4916 bool_long_type_node with bool_long_long_type_node.
4917 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4918 comments to emphasize sign distinctions for char and int types and
4919 replace RS6000_BTI_bool_long constant with
4920 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4921 use of RS6000_BTI_pixel.
4922 (bool_long_type_node): Remove this macro definition.
4923 (bool_long_long_type_node): New macro definition
4925 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4927 PR rtl-optimization/85300
4928 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4929 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4930 simplify_unary_operation fails.
4932 2018-04-10 Martin Liska <mliska@suse.cz>
4934 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4935 cgraph_edge and ipa_ref.
4937 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4941 * config/i386/sse.md
4942 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4943 computation of the VEC_MERGE selector from mask.
4944 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4945 Fix decoding of the VEC_MERGE selector into mask.
4947 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4949 PR tree-optimization/85286
4950 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4952 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4954 * final.c (final_1): Set insn_last_address as well as
4955 insn_current_address.
4957 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4960 * explow.c (emit_stack_probe): Call validize_mem on memory location
4961 before passing it to gen_probe_stack. Create address operand and
4962 legitimize it for the probe_stack_address case.
4964 2018-04-09 Jan Hubicka <jh@suse.cz>
4967 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4968 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4969 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4971 (free_lang_data): Rebuild type inheritance graph.
4973 2018-04-09 Martin Sebor <msebor@redhat.com>
4975 * invoke.texi (-finline-small-functions): Mention other optimization
4977 (-findirect-inlining, -fpartial-inlining): Same.
4978 (-finline-functions-called-once): Same.
4979 (-freorder-blocks-and-partition): Same.
4981 2018-04-09 Jan Hubicka <jh@suse.cz>
4984 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4985 jumps; choose last target that matches the criteria (i.e.
4986 no partition changes for non-crossing jumps).
4987 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4988 support for redirecting crossing jumps to non-crossing.
4990 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4992 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4993 also for naked functions.
4995 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4997 * config/arc/arc.md (add_shift): New pattern.
4998 (add_shift2): Likewise.
4999 (sub_shift): Likewise.
5000 (sub_shift_cmp0_noout): Likewise.
5001 (compare_si_ashiftsi): Likewise.
5002 (xbfu_cmp0_noout): New combine pattern.
5003 (xbfu_cmp0"): Likewise.
5004 (movsi_set_cc_insn): Place the predicable variant first.
5005 (commutative_binary_cmp0_noout): Remove clobber.
5006 (commutative_binary_cmp0): New pattern.
5007 (noncommutative_binary_cmp0): Likewise.
5008 (noncommutative_binary_cmp0_noout): Likewise.
5009 (noncommutative_binary_comparison_result_used): Removed.
5010 (rsub_cmp0): New pattern.
5011 (rsub_cmp0_noout): Likewise.
5012 (extzvsi): Changed, keep only meaningful variants.
5013 (SQH, SEZ): New iterators.
5014 (SQH_postfix): New mode attribute.
5015 (SEZ_prefix): New code attribute.
5016 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5017 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5018 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5020 (noncommutative_operator): Check the availability of barrel
5023 2018-04-09 Richard Biener <rguenther@suse.de>
5025 PR tree-optimization/85284
5026 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5027 Only use the niter constraining form of simple_iv when the exit
5030 2018-04-09 Tom de Vries <tom@codesourcery.com>
5033 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5034 (define_expand "*memory_barrier"): New define_expand.
5035 (define_insn "memory_barrier"): New insn.
5037 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5039 PR rtl-optimization/80463
5040 PR rtl-optimization/83972
5041 PR rtl-optimization/83480
5043 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5044 correct producer for the insn.
5045 (tidy_control_flow): Fixup seqnos in case of debug insns.
5047 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5049 PR rtl-optimization/83913
5051 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5052 different sched-times when merging exprs.
5054 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5056 PR rtl-optimization/83962
5058 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5059 tidy_fallthru_edge and tidy_control_flow.
5061 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5063 PR rtl-optimization/83530
5065 * sel-sched.c (force_next_insn): New global variable.
5066 (remove_insn_for_debug): When force_next_insn is true, also leave only
5067 next insn in the ready list.
5068 (sel_sched_region): When the region wasn't scheduled, make another pass
5069 over it with force_next_insn set to 1.
5071 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5073 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5075 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5076 for interrupt control.
5077 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5078 functions for interrupt control.
5079 * config/nds32/nds32-intrinsic.md: Likewise.
5080 * config/nds32/nds32_intrinsic.h: Likewise.
5081 * config/nds32/nds32.h (nds32_builtins): Likewise.
5083 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5085 * config/nds32/nds32.c (nds32_init_machine_status,
5086 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5087 strict_aligned_p field.
5088 (nds32_expand_to_rtl_hook): New function.
5089 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5090 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5092 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5093 Chung-Ju Wu <jasonwucj@gmail.com>
5095 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5096 * config/nds32/nds32-n7.md: New file.
5097 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5098 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5100 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5101 * config/nds32/nds32.md (pipeline_model): Add n7.
5102 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5103 * config/nds32/pipelines.md: Include n7 settings.
5105 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5106 Chung-Ju Wu <jasonwucj@gmail.com>
5108 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5109 * config/nds32/nds32-e8.md: New file.
5110 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5111 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5113 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5114 * config/nds32/nds32.md (pipeline_model): Add e8.
5115 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5116 * config/nds32/pipelines.md: Include e8 settings.
5118 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5119 Chung-Ju Wu <jasonwucj@gmail.com>
5121 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5122 * config/nds32/nds32-n8.md: New file.
5123 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5124 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5126 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5127 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5128 * config/nds32/nds32.md (pipeline_model): Add n8.
5129 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5130 * config/nds32/pipelines.md: Include n8 settings.
5132 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5133 Chung-Ju Wu <jasonwucj@gmail.com>
5135 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5136 * config/nds32/nds32-n9-2r1w.md: New file.
5137 * config/nds32/nds32-n9-3r2w.md: New file.
5138 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5139 nds32_register_ports): New or modify for cpu n9.
5140 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5142 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5143 * config/nds32/nds32-utils.c: New file.
5144 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5145 TARGET_MUL_SLOW): Define.
5146 * config/nds32/nds32.md (pipeline_model): New attribute.
5147 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5148 New options that support cpu n9.
5149 * config/nds32/pipelines.md: Include n9 settings.
5150 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5152 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5154 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5155 information if necessary.
5156 (output_cond_branch_compare_zero): Likewise.
5157 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5158 (nds32_target_alignment): Refine for alignment.
5159 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5160 (FUNCTION_BOUNDARY): Modify.
5161 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5163 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5165 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5167 * config/nds32/constants.md (unspec_volatile_element): Add values for
5168 TLB operation and data prefetch.
5169 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5170 functions for TLB operation and data prefetch.
5171 * config/nds32/nds32-intrinsic.md: Likewise.
5172 * config/nds32/nds32_intrinsic.h: Likewise.
5173 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5174 (nds32_print_operand): Likewise.
5175 * config/nds32/nds32.h (nds32_builtins): Likewise.
5177 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5178 Andrew Pinski <pinsika@gcc.gnu.org>
5181 * match.pd: Use constant_boolean_node of correct type instead of
5182 boolean_true_node or boolean_false_node for simplifying
5183 pointer comparisons to zero.
5185 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5187 PR tree-optimization/80021
5188 * tree.c (verify_type_variant): Make error call in verify_variant_match
5189 translatable and remove final full stop.
5191 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5193 * config/nds32/constants.md (unspec_volatile_element): Add
5194 UNSPEC_VOLATILE_EH_RETURN.
5195 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5196 nds32_output_stack_pop): Support dwarf exception handling process.
5197 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5198 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5199 exception handling process.
5200 (nds32_compute_stack_frame): Likewise.
5201 (nds32_return_addr_rtx): Likewise.
5202 (nds32_initial_elimination_offset): Likewise.
5203 (nds32_expand_prologue): Likewise.
5204 (nds32_expand_epilogue): Likewise.
5205 (nds32_dynamic_chain_address): New function.
5206 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5208 (DYNAMIC_CHAIN_ADDRESS): Define.
5209 (EH_RETURN_DATA_REGNO): Define.
5210 (EH_RETURN_STACKADJ_RTX): Define.
5211 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5212 patterns for dwarf exception handling.
5214 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5216 * config/nds32/nds32.h: Clean up obsolete macros.
5218 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5220 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5221 Add enum values for particular instructions.
5222 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5223 particular intrinsic functions.
5224 * config/nds32/nds32-intrinsic.md: Likewise.
5225 * config/nds32/nds32_intrinsic.h: Likewise.
5226 * config/nds32/nds32.h (nds32_builtins): Likewise.
5227 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5228 (btst, ave): New patterns for particular instructions.
5230 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5232 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5233 Add enum values for atomic load/store and memory sync.
5234 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5236 * config/nds32/nds32-intrinsic.md: Likewise.
5237 * config/nds32/nds32_intrinsic.h: Likewise.
5238 * config/nds32/nds32.h (nds32_builtins): Likewise.
5240 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5242 PR tree-optimization/85257
5243 * fold-const.c (native_encode_vector): If not all elts could fit
5244 and off is -1, return 0 rather than offset.
5245 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5246 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5247 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5248 adjust buffer in native_interpret_expr call.
5250 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5252 * config/nds32/constants.md (unspec_volatile_element): Add cache
5253 control enum values.
5254 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5255 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5256 * config/nds32/nds32.c (nds32_cctl_names): New.
5257 (nds32_print_operand): Handle cache control register names.
5258 * config/nds32/nds32.h (nds32_builtins): New enum values.
5259 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5261 * config/nds32/nds32.md (type): Add mmu.
5262 * config/nds32/pipelines.md (simple_insn): Add mmu.
5264 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5266 * config/nds32/nds32.md (type): Remove call.
5267 * config/nds32/pipelines.md (simple_insn): Likewise.
5269 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5271 * config/nds32/constants.md (unspec_volatile_element): Add
5272 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5273 UNSPEC_VOLATILE_FMFCFG.
5274 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5275 description for fmfcfg and fmfcsr.
5276 (bdesc_1arg): Add fmtcsr.
5277 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5278 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5279 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5280 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5281 unspec_fmfcfg): New patterns.
5282 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5283 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5284 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5285 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5286 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5287 __nds32__fmfcfg): Define.
5289 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5291 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5292 intrinsic register names.
5293 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5294 intrinsic register enum values and macros.
5296 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5298 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5299 for load/store addressing form.
5300 (nds32_print_operand_address): Likewise.
5302 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5305 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5306 based on LABEL_REF. Remove useless assertion.
5307 (pic_address_needs_scratch): Fix formatting.
5308 (sparc_legitimize_pic_address): Minor tweaks.
5309 (sparc_delegitimize_address): Adjust assertion accordingly.
5310 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5311 into symbolic_operand.
5312 (movsi_high_pic_label_ref): Likewise.
5313 (movsi_lo_sum_pic_label_ref): Likewise.
5314 (movdi_pic_label_ref): Likewise.
5315 (movdi_high_pic_label_ref): Likewise.
5316 (movdi_lo_sum_pic_label_ref): Likewise.
5318 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5320 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5321 custom LIB_SPEC setup.
5323 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5324 Kito Cheng <kito.cheng@gmail.com>
5326 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5327 * config/riscv/freebsd.h: New.
5329 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5331 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5332 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5335 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5336 Kito Cheng <kito.cheng@gmail.com>
5338 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5339 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5340 * config/nds32/nds32-protos.h (nds32_output_call,
5341 nds32_output_return): Declare.
5342 * config/nds32/nds32.md: Refine all the call and return patterns.
5344 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5347 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5348 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5350 PR rtl-optimization/84872
5351 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5352 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5355 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5357 * expr.c (copy_blkmode_to_reg): Revert 254862.
5358 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5360 2018-04-06 Richard Biener <rguenther@suse.de>
5363 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5364 after seeing a component reference with an adjacent field. Treat
5365 refs to arrays at struct end of external decls similar to
5366 refs to unconstrained commons.
5368 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5371 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5372 look through SAVE_EXPRs with non-side-effects argument. Adjust
5374 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5377 2018-04-06 Richard Biener <rguenther@suse.de>
5380 * alias.c (find_base_term): New wrapper around find_base_term
5381 unwinding CSELIB_VAL_PTR changes.
5382 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5385 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5387 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5389 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5390 constant definitions.
5391 ("nop"): lr 0,0 -> nopr r0
5392 ("nop_lr0", "nop_lr1"): New insn definitions.
5394 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5396 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5397 NDS32_V3PUSH_AVAILABLE_P macro.
5399 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5400 Chung-Ju Wu <jasonwucj@gmail.com>
5402 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5403 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5404 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5405 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5406 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5407 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5408 * config/nds32/constraints.md: New constraints and checking for hard
5409 float configuration.
5410 * config/nds32/iterators.md: New mode iterator and attribute for hard
5411 float configuration.
5412 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5414 * config/nds32/nds32-fpu.md: New file.
5415 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5416 deal with hard float code generation.
5417 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5419 (abi_type, float_reg_number): New enum type.
5420 * config/nds32/nds32-predicates.c: New predicates for hard float.
5421 * config/nds32/nds32-protos.h: Declare functions for hard float.
5422 * config/nds32/nds32.c: Implementation for hard float configuration.
5423 * config/nds32/nds32.h: Definitions for hard float configuration.
5424 * config/nds32/nds32.md: Include hard float machine description and
5425 modify patterns for hard float configuration.
5426 * config/nds32/nds32.opt: New options for hard float configuration.
5427 * config/nds32/predicates.md: New predicates for hard float
5430 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5432 * common/config/nds32/nds32-common.c
5433 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5435 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5438 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5439 CONSTRUCTOR_ELT (ctor, ...)->value.
5441 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5444 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5446 2018-04-05 Tom de Vries <tom@codesourcery.com>
5449 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5452 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5453 Kito Cheng <kito.cheng@gmail.com>
5455 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5456 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5457 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5458 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5460 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5461 Kito Cheng <kito.cheng@gmail.com>
5463 * config/nds32/constraints.md (Ufe): New memory constraint.
5464 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5465 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5466 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5468 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5469 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5471 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5473 * config/nds32/nds32.md: Use optimize_size in the condition for
5474 alu-shift instructions.
5476 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5478 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5480 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5482 * config/nds32/nds32.md (negsi2): Refine pattern.
5484 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5485 Chung-Ju Wu <jasonwucj@gmail.com>
5487 * config/nds32/iterators.md (shift_rotate): New code iterator.
5488 (shift): New code attribute.
5489 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5490 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5491 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5492 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5493 bit-wise operations.
5494 (andsi3, *andsi3): Ditto.
5495 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5496 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5497 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5498 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5499 nds32_ior_operand, nds32_xor_operand): New predicates.
5501 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5503 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5504 (addsi3, subsi3): ... this.
5506 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5508 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5510 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5512 * config/nds32/nds32.md: Adjust indention.
5514 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5516 * config/nds32/nds32.md (feature): New attribute.
5518 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5520 * config/nds32/nds32.md (subtype): New attribute.
5522 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5525 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5526 expansion to perform a bitwise AND of the argument followed by a
5527 boolean negation of the result.
5529 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5531 PR rtl-optimization/84878
5532 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5533 the basic block. Assert the use reference is not artificial and that
5534 it has an associated insn.
5536 2018-04-04 Michael Matz <matz@suse.de>
5538 * builtins.c (compute_objsize): Pass correct operand
5539 to array_at_struct_end_p.
5541 2018-04-04 Richard Biener <rguenther@suse.de>
5544 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5545 from contexts for DINFO_LEVEL_TERSE and below.
5547 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5549 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5550 resiter_operand condition.
5551 * config/nds32/nds32.md (*move<mode>): Ditto.
5553 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5554 Monk Chiang <sh.chiang04@gmail.com>
5556 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5558 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5560 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5562 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5563 Kito Cheng <kito.cheng@gmail.com>
5565 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5566 nds32_cond_code_str, output_cond_branch,
5567 output_cond_branch_compare_zero, 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: New functions.
5573 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5574 nds32_expand_cstore, nds32_expand_movcc,
5575 nds32_output_cbranchsi4_equality_zero,
5576 nds32_output_cbranchsi4_equality_reg,
5577 nds32_output_cbranchsi4_equality_reg_or_const_int,
5578 nds32_output_cbranchsi4_greater_less_zero): Declare.
5579 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5580 nds32_rimm11s_operand): New predicates.
5581 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5582 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5585 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5587 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5588 * config/nds32/nds32.md: Ditto.
5589 * config/nds32/pipelines.md: Ditto.
5591 2018-04-04 Richard Biener <rguenther@suse.de>
5593 PR tree-optimization/85168
5594 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5595 propagating abnormals.
5597 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5599 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5601 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5602 Kito Cheng <kito.cheng@gmail.com>
5604 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5605 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5606 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5607 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5608 * config/nds32/nds32.md (sibcall_internal): New.
5609 (sibcall_register): Remove.
5610 (sibcall_immediate): Remove.
5611 (sibcall_value_internal): New.
5612 (sibcall_value_register): Remove.
5613 (sibcall_value_immediate): Remove.
5614 * config/nds32/predicates.md (nds32_general_register_operand): New.
5615 (nds32_call_address_operand): New.
5617 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5619 PR rtl-optimization/85167
5620 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5621 bb_defs if *split_p, instead preinitialize it to NULL.
5623 PR tree-optimization/85156
5624 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5625 evaluating the argument multiple times.
5627 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5629 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5631 (_mm_cvtpd_ps): Likewise.
5632 (_mm_cvttpd_epi32): Likewise.
5633 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5634 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5635 vector, pixel, and bool following altivec.h include.
5637 2018-04-03 Martin Sebor <msebor@redhat.com>
5639 * doc/extend.texi (Common Function Attributes): Clarify.
5640 (const attribute): Likewise.
5641 (pure attribute): Likewise.
5643 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5646 * config/i386/i386.c (ix86_expand_vector_set): Use
5647 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5649 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5651 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5652 instructions when changing rounding bits to preserve precision bits
5653 in the x87 control word.
5655 2018-04-03 Martin Liska <mliska@suse.cz>
5657 PR tree-optimization/82491
5658 * rtl.h (strip_offset_and_add): Replace += suboffset with
5659 poly_uint64 () + suboffset.
5661 2018-03-29 Martin Liska <mliska@suse.cz>
5662 Martin Jambor <mjambor@suse.cz>
5665 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5666 param_type is not an integral or pointer type.
5668 2018-04-03 Richard Biener <rguenther@suse.de>
5670 * sese.h (recompute_all_dominators): Remove.
5672 2018-04-02 Martin Sebor <msebor@redhat.com>
5674 * doc/invoke.texi (-Wrestrict): Fix typos.
5676 2018-04-02 Jim Wilson <jimw@sifive.com>
5678 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5679 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5680 (<optab>di3, <optab>si3_extend): Likewise.
5681 (<optab>si3_mask, <optab>si3_mask_1): New.
5682 (<optab>di3_mask, <optab>di3_mask_1): New.
5683 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5684 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5685 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5687 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5689 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5692 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5694 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5695 (nds32_canonicalize_comparison): New function.
5697 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5698 Kito Cheng <kito.cheng@gmail.com>
5699 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5701 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5702 * config/nds32/constants.md (unspec_volatile_element): Add
5703 UNSPEC_VOLATILE_RELAX_GROUP.
5704 * config/nds32/nds32-relax-opt.c: New file.
5705 * config/nds32/nds32-predicates.c
5706 (nds32_symbol_load_store_p): New function.
5707 * config/nds32/nds32-protos.h
5708 (nds32_symbol_load_store_p): Declare function.
5709 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5710 * config/nds32/nds32.c
5711 (nds32_register_pass): New function to register pass.
5712 (nds32_register_passes): New function to register passes.
5713 * config/nds32/nds32.md (relax_group): New pattern.
5714 * config/nds32/nds32.opt (mrelax-hint): New option.
5715 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5717 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5719 * config/nds32/t-nds32: Modify files dependency.
5721 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5723 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5724 (PROFILE_HOOK): Define its implementation.
5726 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5728 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5729 type and 32-bit size.
5731 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5734 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5735 (V_128_256): New mode iterator.
5736 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5737 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5738 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5740 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5741 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5743 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5746 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5747 NaN inputs correctly.
5749 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5752 * config/rs6000/vsx.md (??r): New mode attribute.
5753 (*vsx_mov<mode>_64bit): Use it.
5754 (*vsx_mov<mode>_32bit): Likewise.
5756 2018-03-30 Martin Sebor <msebor@redhat.com>
5758 PR tree-optimization/84818
5759 * builtins.c (check_access): Use warning_n.
5761 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5764 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5766 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5769 2018-03-30 Julia Koval <julia.koval@intel.com>
5772 * x86-tune.def (movx, partial_reg_dependency): Enable for
5775 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5778 * lra-constraints.c (process_alt_operands): Move setting
5779 this_alternative_matches below.
5781 2018-03-29 Martin Liska <mliska@suse.cz>
5784 * doc/invoke.texi: Document how LTO works with debug info.
5785 Describe auto-load support of binutils. Mention 'x86-64'
5786 as valid option value of -march option.
5788 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5790 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5793 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5794 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5795 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5798 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5801 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5802 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5803 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5804 for RS6000_BTM_POWERPC64.
5805 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5806 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5807 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5812 2018-03-28 Carl Love <cel@us.ibm.com>
5815 2017-09-27 Carl Love <cel@us.ibm.com>
5817 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5818 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5819 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5822 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5824 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5825 instead of __vector bool.
5826 (_mm_max_pu8): Likewise.
5827 (_mm_min_pi16): Likewise.
5829 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5832 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5833 (DIVWEUO): Likewise.
5835 (DIVDEUO): Likewise.
5836 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5837 DIVWEUO and DIVDEUO.
5838 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5839 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5840 (div_extend): Likewise.
5841 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5843 (__builtin_divweuo): Likewise.
5844 (__builtin_divdeo): Likewise.
5845 (__builtin_divdeuo): Likewise.
5847 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5850 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5851 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5853 PR tree-optimization/82004
5854 * gimple-match-head.c (optimize_pow_to_exp): New function.
5855 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5856 Don't fold to exp if optimize_pow_to_exp is false.
5858 2018-03-28 Martin Liska <mliska@suse.cz>
5861 * calls.c (initialize_argument_information): Fix trailing space.
5862 * common.opt: Fix typo and provide better explanation for
5863 -fsanitize-coverage option.
5864 * config/i386/i386.opt: Fix typo.
5866 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5867 Martin Liska <mliska@suse.cz>
5870 * gimplify.c (asan_poison_variable): Don't do the check for
5871 gimplify_omp_ctxp here.
5872 (gimplify_decl_expr): Do it here.
5873 (gimplify_target_expr): Likewise.
5875 2018-03-28 Martin Liska <mliska@suse.cz>
5878 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5879 chkp_type_bounds_count if MPX is not enabled.
5881 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5883 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5885 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5888 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5889 function to create the function decl for complex long double
5890 multiply and divide for -mabi=ieeelongdouble.
5891 (init_float128_ieee): Call it.
5893 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5896 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5897 -fcf-protection=branch -mibt.
5898 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5900 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5903 * config/arm/arm.c (arm_valid_symbolic_address): Handle
5904 arm_word_relocations.
5906 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5909 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5910 extern array declarations.
5912 2018-03-27 Richard Biener <rguenther@suse.de>
5915 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5916 explicit single_use checks.
5918 2018-03-27 Richard Biener <rguenther@suse.de>
5920 PR tree-optimization/85082
5921 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5924 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5926 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5927 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5928 Turn on fasynchronous-unwind-tables and funwind-tables.
5930 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5933 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5934 (*bmi_blsr_<mode>_ccz): Ditto.
5936 2018-03-26 Tom de Vries <tom@codesourcery.com>
5938 PR tree-optimization/85063
5939 * omp-general.c (offloading_function_p): New function. Factor out
5941 * omp-offload.c (pass_omp_target_link::gate): ... here.
5942 * omp-general.h (offloading_function_p): Declare.
5943 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5944 with attribute omp declare target for offloading functions.
5946 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5948 PR tree-optimization/84005
5949 * tree-data-ref.h (get_base_for_alignment): Declare.
5950 * tree-data-ref.c (get_base_for_alignment_1): New function.
5951 (get_base_for_alignment): Likewise.
5952 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5953 get_base_for_alignment to find a suitable base object, instead
5954 of always using drb->base_address.
5956 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5959 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5960 known size by default.
5962 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5965 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5966 and non BLKmode operands.
5968 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5971 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5972 Clean up attributes.
5974 2018-03-23 Richard Biener <rguenther@suse.de>
5977 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5978 we are going to emit early debug for LTO.
5980 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5983 * function.c (match_asm_constraints_1): Don't optimize if input
5984 doesn't satisfy general_operand predicate for output's mode.
5987 * alias.c (write_dependence_p): Don't require for x_canonicalized
5988 non-VOIDmode if x has VOIDmode.
5991 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5992 just don't try to optimize it rather than assert it never happens.
5994 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5996 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5997 macro expansions for definition of ST_INTERNAL_<mode> and
5998 LD_INTERNAL_<mode> builtins.
5999 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6001 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6003 (altivec_expand_st_builtin): Likewise.
6004 (altivec_expand_builtin): Remove calls to deleted functions.
6005 (rs6000_address_for_altivec): Delete this function.
6006 * config/rs6000/vector.md: Remove expands for
6007 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6009 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6012 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6013 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6014 re-computing once computed.
6015 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6016 (arm_init_machine_status): Initialize
6017 machine->static_chain_stack_bytes.
6019 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6022 * doc/extend.texi: Add four new prototypes for vec_ld.
6023 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6024 definitions for more logical presentation.
6025 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6026 entries for V1TI variants of __builtin_altivec_ld builtin.
6027 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6028 handling of V1TI variant of LVX icode pattern.
6029 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6030 (rs6000_gimple_fold_builtin): Likewise.
6031 (altivec_init_builtins): Add code to define
6032 __builtin_altivec_lvx_v1ti function.
6034 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6037 * function.c (match_asm_constraints_1): Don't do the optimization
6038 if input isn't a REG, SUBREG, MEM or constant.
6040 2018-03-22 Tom de Vries <tom@codesourcery.com>
6042 PR tree-optimization/84956
6043 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6044 bb_has_abnormal_pred.
6046 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6049 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6050 DECL_INITIAL (decl) to decl at the end.
6051 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6054 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6056 * doc/extend.texi (__builtin_tgmath): Document when complex
6057 integer types are treated as _Complex _Float64.
6059 2018-03-21 Tom de Vries <tom@codesourcery.com>
6061 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6063 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6065 PR tree-optimization/84960
6066 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6067 if it is ENTRY block, move them into single succ of ENTRY in that case.
6069 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6071 PR tree-optimization/84811
6072 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6073 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6074 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6075 (known_subrange_p): Update accordingly. Cast each value involved
6076 in the size comparison, rather than casting the result of the
6079 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6081 PR tree-optimization/84982
6082 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6083 by flipping the least significant bit rather than all bits from
6084 bitpos to bitpos + bitsize - 1.
6086 2018-03-21 Nathan Sidwell <nathan@acm.org>
6088 * doc/extend.texi (Deprecated Features): Remove mention of
6089 long-deleted deprecations.
6091 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6094 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6095 * configure: Regenerate.
6097 2018-03-21 Tom de Vries <tom@codesourcery.com>
6099 PR tree-optimization/83126
6100 * tree-parloops.c (num_phis): New function.
6101 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6103 2018-03-21 Nathan Sidwell <nathan@acm.org>
6105 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6106 mention anon-struct/union members and trailing attributes.
6108 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6110 PR tree-optimization/84969
6111 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6112 builtin memset partitions if they set different rhs values.
6114 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6116 PR rtl-optimization/84989
6117 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6118 VEC_DUPLICATE with scalar result mode.
6120 2018-03-21 Martin Liska <mliska@suse.cz>
6123 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6124 not intended return statement.
6126 2018-03-21 Martin Liska <mliska@suse.cz>
6129 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6130 (chkp_find_bound_slots_1): Limit number of iterations.
6132 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6135 * Minor grammar fixes for x86 options.
6137 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6140 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6141 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6143 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6146 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6147 (altivec_lvx_<mode>_1op): Likewise.
6148 (altivec_stvx_<mode>_2op): Likewise.
6149 (altivec_stvx_<mode>_1op): Likewise.
6150 (altivec_lvx_<VM2:mode>): New define_expand.
6151 (altivec_stvx_<VM2:mode>): Likewise.
6152 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6153 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6154 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6155 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6156 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6157 (rs6000_gen_lvx): Likewise.
6158 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6159 (altivec_expand_stv_builtin): Likewise.
6160 (altivec_expand_builtin): Likewise.
6161 * config/rs6000/vector.md: Likewise.
6163 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6166 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6169 2018-03-20 Richard Biener <rguenther@suse.de>
6172 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6173 sign-conversions as zero, fall back to standard scalar_stmt
6176 2018-03-20 Martin Liska <mliska@suse.cz>
6179 * predict.c (rebuild_frequencies): Handle case when we have
6180 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6182 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6185 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6186 flag_section_anchors.
6187 * varasm.c (use_blocks_for_decl_p): Remove hack for
6188 dw2_force_const_mem.
6191 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6193 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6194 be created, use lowpart_subreg of operands[0] rather than operands[0]
6196 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6197 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6198 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6199 and n constraint instead of aarch64_shift_imm_di and Usd.
6200 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6201 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6203 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6206 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6207 to favor GPR over NEON registers.
6208 (<shift>di3_neon): Likewise.
6210 2018-03-20 Tom de Vries <tom@codesourcery.com>
6213 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6214 (nvptx_process_pars): Emit bar.sync asap and alap.
6216 2018-03-20 Tom de Vries <tom@codesourcery.com>
6219 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6220 seen_label if seen_label is already set.
6222 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6225 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6226 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6227 Use 1U instead of 1. Formatting fixes.
6230 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6231 instead of TREE_TYPE (s1) for the return value.
6233 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6235 PR tree-optimization/84946
6236 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6237 bitsize + bitsize in poly_uint64 rather than poly_int64.
6240 * dwarf2asm.c: Include fold-const.c.
6241 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6242 of decl rather than decl itself.
6244 PR rtl-optimization/84643
6245 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6247 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6250 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6251 calling assemble_variable.
6253 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6256 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6257 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6259 2018-03-19 Jim Wilson <jimw@sifive.com>
6262 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6263 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6264 (riscv_first_stack_step): Likewise.
6265 (riscv_option_override): Use STACK_BOUNDARY instead of
6267 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6269 (BIGGEST_ALIGNMENT): Set to 128.
6270 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6271 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6274 2018-03-19 Richard Biener <rguenther@suse.de>
6276 PR tree-optimization/84933
6277 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6278 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6280 2018-03-19 Richard Biener <rguenther@suse.de>
6282 PR tree-optimization/84859
6283 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6284 (cond_if_else_store_replacement): Perform sinking operation on
6285 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6286 Generalize what a BB with a single eligible store is.
6288 2018-03-19 Richard Biener <rguenther@suse.de>
6290 PR tree-optimization/84929
6291 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6292 chrec_is_positive against non-chrec arg.
6294 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6297 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6299 2018-03-18 Martin Liska <mliska@suse.cz>
6301 PR rtl-optimization/84635
6302 * regrename.c (build_def_use): Use matches_mode only when
6305 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6307 PR tree-optimization/84913
6308 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6309 vectorize chains of COND_EXPRs.
6311 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6313 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6315 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6317 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6319 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6321 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6323 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6324 Kito Cheng <kito.cheng@gmail.com>
6326 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6327 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6328 (nds32_adjust_reg_alloc_order): New function.
6329 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6331 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6333 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6334 nds32_print_operand, nds32_print_operand_address): Use
6335 HOST_WIDE_INT_PRINT_DEC instead.
6337 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6339 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6341 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6344 * config/i386/i386.c (initial_ix86_tune_features,
6345 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6347 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6348 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6349 rather than 1u << ix86_tune. Formatting fix.
6350 (ix86_option_override_internal): Change ix86_arch_mask from
6351 unsigned int to unsigned HOST_WIDE_INT, initialize to
6352 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6353 (ix86_function_specific_restore): Likewise.
6355 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6358 * postreload.c (reload_combine_recognize_pattern): Perform
6359 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6360 truncate_int_for_mode the result for the destination's mode.
6363 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6364 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6366 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6369 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6372 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6373 regno_allocno_class_array and sorted_pseudos.
6374 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6375 insns where regno is used.
6377 2018-03-16 Martin Liska <mliska@suse.cz>
6380 * multiple_target.c (create_dispatcher_calls): Redirect
6381 reference in the symbol table.
6383 2018-03-16 Martin Liska <mliska@suse.cz>
6386 * multiple_target.c (create_dispatcher_calls): Redirect also
6389 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6393 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6394 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6395 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6397 2018-03-16 Julia Koval <julia.koval@intel.com>
6399 * doc/invoke.texi (Skylake Server): Add CLWB.
6400 Cannonlake): Remove CLWB.
6402 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6404 PR tree-optimization/84841
6405 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6407 (FLOAT_ONE_CONST_TYPE): Define.
6408 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6409 (sort_by_operand_rank): Put entries with higher constant_type last
6410 rather than first to match comments.
6412 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6414 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6417 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6420 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6421 check and use address of target always.
6423 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6426 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6427 (indirect_thunk_bnd_needed): Likewise.
6428 (indirect_thunks_used): Likewise.
6429 (indirect_thunks_bnd_used): Likewise.
6430 (indirect_return_needed): New.
6431 (indirect_return_bnd_needed): Likewise.
6432 (output_indirect_thunk_function): Add a bool argument for
6434 (output_indirect_thunk_function): Don't generate alias for
6435 function return thunk.
6436 (ix86_code_end): Call output_indirect_thunk_function to generate
6437 function return thunks.
6438 (ix86_output_function_return): Set indirect_return_bnd_needed
6439 and indirect_return_needed instead of indirect_thunk_bnd_needed
6440 and indirect_thunk_needed.
6442 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6444 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6445 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6446 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6448 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6449 Paul Hua <paul.hua.gm@gmail.com>
6452 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6454 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6456 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6457 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6460 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6463 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6464 instead of GET_MODE_SIZE when comparing Units.
6466 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6469 * varasm.c (hash_section): Return an unchangeble hash value
6470 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6471 Return !aarch64_can_use_per_function_literal_pools_p ().
6473 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6476 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6477 rather than address of cmode as last argument to prepare_cmp_insn.
6479 2018-03-15 Julia Koval <julia.koval@intel.com>
6481 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6482 F_AVX512VNNI, F_AVX512BITALG): New.
6484 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6487 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6488 insn for floating-point loads and stores.
6490 2018-03-14 Carl Love <cel@us.ibm.com>
6492 * config/rs6000/rs6000-c.c: Add macro definitions for
6493 ALTIVEC_BUILTIN_VEC_PERMXOR.
6494 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6495 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6496 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6497 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6499 * config/doc/extend.texi: Add prototypes for vec_permxor.
6501 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6504 * diagnostic-show-locus.c (class layout_point): Convert m_line
6505 from int to linenum_type.
6506 (line_span::comparator): Use linenum "compare" function when
6507 comparing line numbers.
6508 (test_line_span): New function.
6509 (layout_range::contains_point): Convert param "row" from int to
6511 (layout_range::intersects_line_p): Likewise.
6512 (layout::will_show_line_p): Likewise.
6513 (layout::print_source_line): Likewise.
6514 (layout::should_print_annotation_line_p): Likewise.
6515 (layout::print_annotation_line): Likewise.
6516 (layout::print_leading_fixits): Likewise.
6517 (layout::annotation_line_showed_range_p): Likewise.
6518 (struct line_corrections): Likewise for field m_row.
6519 (line_corrections::line_corrections): Likewise for param "row".
6520 (layout::print_trailing_fixits): Likewise.
6521 (layout::get_state_at_point): Likewise.
6522 (layout::get_x_bound_for_row): Likewise.
6523 (layout::print_line): Likewise.
6524 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6525 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6526 * input.c (selftest::test_linenum_comparisons): New function.
6527 (selftest::input_c_tests): Call it.
6528 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6529 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6530 * selftest.h (ASSERT_GT): New macro.
6531 (ASSERT_GT_AT): New macro.
6532 (ASSERT_LT): New macro.
6533 (ASSERT_LT_AT): New macro.
6535 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6537 PR rtl-optimization/84780
6538 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6540 2018-03-14 Martin Liska <mliska@suse.cz>
6542 * tree.c (record_node_allocation_statistics): Use
6543 get_stats_node_kind.
6544 (get_stats_node_kind): New function extracted from
6545 record_node_allocation_statistics.
6546 (free_node): Use get_stats_node_kind.
6548 2018-03-14 Richard Biener <rguenther@suse.de>
6550 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6551 that the value-set of ANTIC_IN doesn't grow.
6554 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6556 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6557 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6559 2018-03-14 Julia Koval <julia.koval@intel.com>
6561 * config.gcc (icelake-client, icelake-server): New.
6563 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6564 (initial_ix86_arch_features): Ditto.
6565 (PTA_SKYLAKE): Add SGX.
6566 (PTA_ICELAKE): Remove.
6567 (PTA_ICELAKE_CLIENT): New.
6568 (PTA_ICELAKE_SERVER): New.
6569 (ix86_option_override_internal): Split up icelake on icelake client and
6571 (get_builtin_code_for_version): Ditto.
6572 (fold_builtin_cpu): Ditto.
6573 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6574 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6575 * config/i386/i386.h (processor_type): Ditto.
6576 * doc/invoke.texi: Ditto.
6578 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6581 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6582 INTEGER_CST offset, add it together with bitpos / 8 and
6583 sign extend based on POINTER_SIZE.
6587 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6590 * config/i386/constraints.md (Yc): New register constraint.
6591 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6592 Use Yc constraint for alternative 2 of operand 0. Remove
6593 preferred_for_speed attribute.
6595 2018-03-14 Richard Biener <rguenther@suse.de>
6597 PR tree-optimization/84830
6598 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6599 with the old one to avoid oscillations.
6601 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6604 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6606 (assign_by_spills): Return a flag of reload assignment failure.
6607 Do not process the reload assignment failures. Do not spill other
6608 reload pseudos if they has the same reg class. Update n if
6610 (lra_assign): Add a return arg. Set up from the result of
6611 assign_by_spills call.
6612 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6613 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6614 usage_insns if it is not NULL.
6615 (spill_hard_reg_in_range): New function.
6616 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6617 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6618 function prototypes.
6619 (lra_assign): Change prototype.
6620 * lra.c (lra): Add code to deal with fails by splitting hard reg
6623 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6625 * config/riscv/riscv.opt (mrelax): New option.
6626 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6627 "norelax" when riscv_mrelax is disabled.
6628 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6630 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6633 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6634 reassociation for int modes.
6636 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6638 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6639 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6641 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6642 * config/aarch64/aarch64-sve.md
6643 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6644 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6645 (*extend<mode><Vwide>2): Rename to...
6646 (aarch64_sve_extend<mode><Vwide>2): ...this.
6647 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6648 renaming the old pattern to...
6649 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6651 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6652 define_expand, renaming the old pattern to...
6653 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6654 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6655 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6656 account when deciding which SVE instruction the optab should use.
6657 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6659 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6661 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6662 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6663 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6664 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6665 (tlsdesc_small_<mode>): Turn a define_expand and use
6666 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6667 (tlsdesc_small_advsimd_<mode>): ...this.
6668 (tlsdesc_small_sve_<mode>): New pattern.
6670 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6672 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6673 (UNSPEC_UMUL_HIGHPART): New constants.
6674 (MUL_HIGHPART): New int iteraor.
6675 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6676 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6678 (*<su>mul<mode>3_highpart): New define_insn.
6680 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6683 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6686 2018-03-13 Martin Liska <mliska@suse.cz>
6689 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6691 (sem_item_optimizer::~sem_item_optimizer): Release it.
6692 (sem_item_optimizer::merge_classes): Register variable aliases.
6693 (sem_item_optimizer::fixup_pt_set): New function.
6694 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6695 * ipa-icf.h: Declare new variables and functions.
6697 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6700 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6701 integer_pow2p@2 and test integer_pow2p in condition.
6702 (A < 0 ? C : 0): Similarly for @1.
6705 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6706 characters starting at p contain '\0' character, don't look beyond
6710 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6711 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6714 * reg-stack.c (change_stack): Change update_end var from int to
6715 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6716 also call set_block_for_insn on the newly added insns and rescan.
6719 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6720 on the last operand.
6723 * tree.c (stabilize_reference_1): Return save_expr (e) for
6724 STATEMENT_LIST even if it doesn't have side-effects.
6726 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6728 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6730 2018-03-12 Renlin Li <renlin.li@arm.com>
6732 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6733 aarch64_output_scalar_simd_mov_immediate.
6735 2018-03-12 Martin Sebor <msebor@redhat.com>
6737 PR tree-optimization/83456
6738 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6739 for perfectly overlapping calls to memcpy.
6740 (gimple_fold_builtin_memory_chk): Same.
6741 (gimple_fold_builtin_strcpy): Handle no-warning.
6742 (gimple_fold_builtin_stxcpy_chk): Same.
6743 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6745 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6747 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6748 parameter. Use it for SFmode.
6749 (rs6000_function_arg_advance_1): Adjust.
6750 (rs6000_function_arg): Adjust.
6751 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6753 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6755 PR rtl-optimization/84169
6756 PR rtl-optimization/84780
6757 * combine.c (can_combine_p): Check for a 2-insn combination whether
6758 the destination register is used between the two insns, too.
6760 2018-03-12 Richard Biener <rguenther@suse.de>
6762 PR tree-optimization/84803
6763 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6764 for refs DR analysis didn't process.
6766 2018-03-12 Richard Biener <rguenther@suse.de>
6768 PR tree-optimization/84777
6769 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6770 force-vectorize loops ignore whether we are optimizing for size.
6772 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6774 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6775 (TARGET_MD_ASM_ADJUST): Define.
6777 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6778 Kito Cheng <kito.cheng@gmail.com>
6779 Chung-Ju Wu <jasonwucj@gmail.com>
6781 * config/nds32/nds32.c (nds32_compute_stack_frame,
6782 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6783 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6784 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6785 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6786 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6787 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6788 * config/nds32/nds32.md (prologue, epilogue): Use macro
6789 NDS32_V3PUSH_AVAILABLE_P to do checking.
6791 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6794 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6795 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6796 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6797 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6798 addition of most attributes on !orig_type_die or the attribute not
6799 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6801 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6802 Chung-Ju Wu <jasonwucj@gmail.com>
6804 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6806 * config/nds32/nds32.opt (mvh): New option.
6808 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6809 Chung-Ju Wu <jasonwucj@gmail.com>
6811 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6813 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6814 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6817 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6818 Chung-Ju Wu <jasonwucj@gmail.com>
6820 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6822 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6823 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6825 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6826 Kito Cheng <kito.cheng@gmail.com>
6827 Chung-Ju Wu <jasonwucj@gmail.com>
6829 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6830 UNSPEC_FFMISM and UNSPEC_FLMISM.
6831 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6832 for ffb, ffmism and flmism.
6833 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6834 (unspec_ffmism): Ditto.
6835 (unspec_flmism): Ditto.
6836 (nds32_expand_builtin_impl): Check if string extension is available.
6837 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6838 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6840 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6843 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6846 * lra-assigns.c (assign_by_spills): Return a flag of reload
6847 assignment failure. Do not process the reload assignment
6848 failures. Do not spill other reload pseudos if they has the same
6850 (lra_assign): Add a return arg. Set up from the result of
6851 assign_by_spills call.
6852 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6853 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6854 usage_insns if it is not NULL.
6855 (spill_hard_reg_in_range): New function.
6856 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6857 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6858 function prototypes.
6859 (lra_assign): Change prototype.
6860 * lra.c (lra): Add code to deal with fails by splitting hard reg
6863 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6866 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6868 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6871 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6872 (dw_val_node): Add val_symbolic_view.
6873 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6874 (symview_upper_bound): New.
6875 (new_line_info_table): Initialize symviews_since_reset.
6876 (dwarf2out_source_line): Count symviews_since_reset and set
6877 symview_upper_bound.
6878 (dw_val_equal_p): Handle symview.
6879 (add_AT_symview): New.
6880 (print_dw_val): Handle symview.
6881 (attr_checksum, attr_checksum_ordered): Likewise.
6882 (same_dw_val_p, size_of_die): Likewise.
6883 (value_format, output_die): Likewise.
6884 (add_high_low_attributes): Use add_AT_symview for entry_view.
6885 (dwarf2out_finish): Reset symview_upper_bound, clear
6888 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6891 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6892 Add strict argument and use it.
6893 (rs6000_split_multireg_move): Update for new strict argument.
6894 (mem_operand_gpr): Disallow all non-offsettable addresses.
6895 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6897 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6900 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6901 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6902 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6905 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6906 decl, use remap_type if we want to use the type.
6908 2018-03-09 Martin Sebor <msebor@redhat.com>
6910 PR tree-optimization/84526
6911 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6913 (builtin_access::generic_overlap): Be prepared to handle non-array
6916 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6918 PR rtl-optimization/84682
6919 * lra-constraints.c (process_address_1): Check is_address flag
6920 for address constraints.
6921 (process_alt_operands): Likewise.
6922 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6923 preprocess_constraints.
6924 * recog.h (preprocess_constraints): Add oploc parameter.
6926 * recog.c (preprocess_constraints): Test address_operand for
6927 CT_ADDRESS constraints.
6929 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6932 * lra-assigns.c (assign_by_spills): Return a flag of reload
6933 assignment failure. Do not process the reload assignment
6934 failures. Do not spill other reload pseudos if they has the same
6936 (lra_assign): Add a return arg. Set up from the result of
6937 assign_by_spills call.
6938 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6939 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6940 usage_insns if it is not NULL.
6941 (spill_hard_reg_in_range): New function.
6942 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6943 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6944 function prototypes.
6945 (lra_assign): Change prototype.
6946 * lra.c (lra): Add code to deal with fails by splitting hard reg
6949 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6952 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6953 Accept complain bool parameter. Only emit errors if it is true.
6954 (arm_parse_cpu_option_name): Likewise.
6955 (arm_target_thumb_only): Adjust callers of the above.
6956 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6957 prototype to take a default true bool parameter.
6958 (arm_parse_arch_option_name): Likewise.
6960 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6961 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6965 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6966 * configure: Regenerate.
6967 * configure.ac ("linker --version-script option"): New.
6968 ("linker soname option"): New.
6970 2018-03-09 Richard Biener <rguenther@suse.de>
6972 PR tree-optimization/84775
6973 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6974 immediate uses of predicate stmts and mark them modified.
6977 PR tree-optimization/84178
6978 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6980 (version_loop_for_if_conversion): Delay update_ssa call.
6981 (tree_if_conversion): Delay update_ssa until after predicate
6984 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6987 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6988 when the function accesses prior frames.
6990 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6993 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6994 gen_llsym, otherwise call maybe_gen_llsym.
6997 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6998 has ',' character inside of it.
7000 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7003 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7004 as clobbering CC_REGNUM.
7006 2018-03-08 Richard Biener <rguenther@suse.de>
7009 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7010 (follow_copies_to_constant): Do not follow SSA names registered
7013 2018-03-08 Richard Biener <rguenther@suse.de>
7015 PR tree-optimization/84178
7016 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7018 (version_loop_for_if_conversion): Delay update_ssa call.
7019 (tree_if_conversion): Delay update_ssa until after predicate
7022 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7024 PR tree-optimization/84178
7025 * tree-if-conv.c (release_bb_predicate): Remove the
7026 the assertion that the stmts have NULL use_ops.
7027 Discard the statements, asserting that they haven't
7028 yet been added to a BB.
7030 2018-03-08 Richard Biener <rguenther@suse.de>
7032 PR tree-optimization/84746
7033 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7034 (phi_translate): Pass in destination ANTIC_OUT set.
7035 (phi_translate_1): Likewise. For a simplified result lookup
7036 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7037 (phi_translate_set): Adjust.
7038 (do_pre_regular_insertion): Likewise.
7039 (do_pre_partial_partial_insertion): Likewise.
7041 2018-03-08 Martin Liska <mliska@suse.cz>
7043 PR gcov-profile/84735
7044 * doc/gcov.texi: Document usage of profile files.
7045 * gcov-io.h: Document changes in the format.
7047 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7051 * dwarf2out.c (struct dw_line_info_table): Update comments for
7053 (FORCE_RESET_NEXT_VIEW): New.
7054 (FORCE_RESETTING_VIEW_P): New.
7055 (RESETTING_VIEW_P): Check for -1 too.
7056 (ZERO_VIEW_P): Likewise.
7057 (new_line_info_table): Force-reset next view.
7058 (dwarf2out_begin_function): Likewise.
7059 (dwarf2out_source_line): Simplify zero_view_p initialization.
7060 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7061 view directly. Omit view when omitting .loc at line 0.
7063 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7065 PR tree-optimization/84740
7066 * tree-switch-conversion.c (process_switch): Call build_constructors
7067 only if info.phi_count is non-zero.
7069 PR tree-optimization/84739
7070 * tree-tailcall.c (find_tail_calls): Check call arguments against
7071 DECL_ARGUMENTS (current_function_decl) rather than
7072 DECL_ARGUMENTS (func) when checking for tail recursion.
7074 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7076 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7077 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7078 Volker Reichelt's entry and add entries for people that perform
7079 GCC fuzzy testing and report numerous bugs.
7081 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7084 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7085 readonly data in sdata, if that is disabled.
7086 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7087 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7088 -mreadonly-in-sdata option.
7090 2018-03-07 Martin Sebor <msebor@redhat.com>
7092 PR tree-optimization/84468
7093 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7094 basic block when looking for nul assignment.
7096 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7099 * except.h (output_function_exception_table): Adjust prototype.
7100 * except.c (output_function_exception_table): Remove FNNAME parameter
7101 and add SECTION parameter. Ouput one part of the table at a time.
7102 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7103 the first part of the exception table and emit unwind directives.
7104 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7105 (i386_pe_seh_cold_init): Likewise.
7106 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7107 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7108 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7109 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7110 * config/i386/winnt.c: Include except.h.
7111 (struct seh_frame_state): Add reg_offset, after_prologue and
7112 in_cold_section fields.
7113 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7114 (i386_pe_seh_cold_init): New function.
7115 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7116 to seh->in_cold_section.
7117 (seh_emit_push): Record the offset of the push.
7118 (seh_emit_save): Record the offet of the save.
7119 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7120 Test seh->after_prologue to disregard the epilogue.
7121 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7122 (i386_pe_end_cold_function): New function.
7124 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7127 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7128 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7131 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7133 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7134 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7137 * multiple_target.c: Include tree-inline.h and intl.h.
7138 (expand_target_clones): Diagnose and fail if node->definition and
7139 !tree_versionable_function_p (node->decl).
7141 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7143 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7145 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7146 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7147 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7149 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7152 * combine.c (try_combine): Use reg_or_subregno instead of handling
7153 just paradoxical SUBREGs and REGs.
7155 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7157 * config/arc/arc.c (arc_finalize_pic): Remove function.
7158 (arc_must_save_register): We use single base PIC register, remove
7159 checks to save/restore the PIC register.
7160 (arc_expand_prologue): Likewise.
7161 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7163 (arc_verify_short): Likewise.
7164 (arc_attr_type): Likewise.
7165 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7166 (walk_stores): Likewise.
7167 (arc_address_cost): Make it static.
7168 (arc_verify_short): Likewise.
7169 (branch_dest): Likewise.
7170 (arc_attr_type): Likewise.
7171 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7172 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7173 (arc_final_prescan_insn): Remove inserting the nops due to
7174 hardware hazards. It is done in reorg step.
7175 (insn_length_variant_t): Remove.
7176 (insn_length_parameters_t): Likewise.
7177 (arc_insn_length_parameters): Likewise.
7178 (arc_get_insn_variants): Likewise.
7179 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7181 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7184 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7187 PR tree-optimization/84687
7188 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7190 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7192 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7194 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7195 Rename to ppc_speculation_barrier.
7196 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7197 __builtin_ppc_speculation_barrier.
7199 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7202 * combine.c (combine_simplify_rtx): Don't try to simplify if
7203 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7206 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7208 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7209 to 32 bytes when compiling for POWER9.
7211 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7214 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7215 regparm >= 3 with no arg reg available also for calls with
7216 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7219 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7221 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7223 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7226 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7228 2018-03-05 Richard Biener <rguenther@suse.de>
7230 PR tree-optimization/84486
7231 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7232 When inserting a __builtin_assume_aligned call set the LHS
7233 SSA name alignment info accordingly.
7235 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7237 PR tree-optimization/84114
7238 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7239 Avoid reassociation of FLOAT_MODE addition.
7241 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7243 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7244 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7245 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7246 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7247 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7248 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7249 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7251 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7252 __builtin_ia32_wbinvd): New builtins.
7253 (SPECIAL_ARGS2): New.
7254 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7255 (SPECIAL_ARGS2): New.
7256 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7257 (ix86_valid_target_attribute_inner_p): Ditto.
7258 (ix86_init_mmx_sse_builtins): Add special_args2.
7259 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7260 TARGET_WBNOINVD_P): New.
7261 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7262 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7263 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7264 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7265 * config/i386/pconfigintrin.h: New file.
7266 * config/i386/wbnoinvdintrin.h: Ditto.
7267 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7269 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7271 2018-03-05 Richard Biener <rguenther@suse.de>
7273 PR tree-optimization/84670
7274 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7276 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7277 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7278 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7279 doesn't grow if all successors (recursively) were visited at least
7282 2018-03-05 Richard Biener <rguenther@suse.de>
7284 PR tree-optimization/84650
7285 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7286 if executed in the loop pipeline.
7288 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7290 * doc/configfiles.texi (Configuration Files): Move info about
7291 conditionalizing $target-protos.h to...
7292 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7293 differs from $target-protos.h.
7295 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7296 Chung-Ju Wu <jasonwucj@gmail.com>
7298 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7299 * config/nds32/nds32-multiple.md (setmemsi): Define.
7300 * config/nds32/nds32-memory-manipulation.c
7301 (nds32_gen_dup_4_byte_to_word_value): New.
7302 (emit_setmem_word_loop): New.
7303 (emit_setmem_byte_loop): New.
7304 (nds32_expand_setmem_loop): New.
7305 (nds32_expand_setmem_loop_v3m): New.
7306 (nds32_expand_setmem_unroll): New.
7307 (nds32_expand_setmem): New.
7309 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7310 Chung-Ju Wu <jasonwucj@gmail.com>
7312 * config/nds32/nds32-memory-manipulation.c
7313 (nds32_emit_load_store): New.
7314 (nds32_emit_post_inc_load_store): New.
7315 (nds32_emit_mem_move): New.
7316 (nds32_emit_mem_move_block): New.
7317 (nds32_expand_movmemsi_loop_unknown_size): New.
7318 (nds32_expand_movmemsi_loop_known_size): New.
7319 (nds32_expand_movmemsi_loop): New.
7320 (nds32_expand_movmemsi_unroll): New.
7321 (nds32_expand_movmemqi): Rename ...
7322 (nds32_expand_movmemsi): ... to this.
7323 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7324 (movmemsi): ... to this.
7325 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7326 (nds32_expand_movmemsi): ... to this.
7328 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7329 Monk Chiang <sh.chiang04@gmail.com>
7330 Chung-Ju Wu <jasonwucj@gmail.com>
7332 * config/nds32/nds32-protos.h
7333 (nds32_expand_load_multiple): New arguments.
7334 (nds32_expand_store_multiple): Ditto.
7335 (nds32_valid_multiple_load_store): Rename ...
7336 (nds32_valid_multiple_load_store_p): ... to this.
7337 * config/nds32/nds32-memory-manipulation.c
7338 (nds32_expand_load_multiple): Refine implementation.
7339 (nds32_expand_store_multiple): Ditto.
7340 * config/nds32/nds32-multiple.md
7341 (load_multiple): Update nds32_expand_load_multiple interface.
7342 (store_multiple): Update nds32_expand_store_multiple interface.
7343 * config/nds32/nds32-predicates.c
7344 (nds32_valid_multiple_load_store): Rename ...
7345 (nds32_valid_multiple_load_store_p): ... to this and refine
7347 * config/nds32/predicates.md
7348 (nds32_load_multiple_and_update_address_operation): New predicate.
7349 (nds32_store_multiple_and_update_address_operation): New predicate.
7351 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7352 Chung-Ju Wu <jasonwucj@gmail.com>
7354 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7355 (combo): New attribute.
7356 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7358 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7360 * config/nds32/nds32.opt: Change -mcmodel= default value.
7362 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7363 Monk Chiang <sh.chiang04@gmail.com>
7364 Chung-Ju Wu <jasonwucj@gmail.com>
7366 * config/nds32/constants.md (unspec_element): New enum.
7367 * config/nds32/constraints.md (Umw): New constraint.
7368 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7369 * config/nds32/nds32-intrinsic.md: Likewise.
7370 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7371 (nds32_valid_smw_lwm_base_p): New.
7372 (nds32_output_smw_single_word): New.
7373 (nds32_output_lmw_single_word): New.
7374 (nds32_expand_unaligned_load): New.
7375 (nds32_expand_unaligned_store): New.
7376 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7377 (nds32_output_smw_single_word): Declare.
7378 (nds32_output_lmw_single_word): Declare.
7379 (nds32_expand_unaligned_load): Declare.
7380 (nds32_expand_unaligned_store): Declare.
7381 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7382 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7383 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7384 NDS32_BUILTIN_UASTORE_DW.
7385 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7388 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7389 Kito Cheng <kito.cheng@gmail.com>
7390 Chung-Ju Wu <jasonwucj@gmail.com>
7392 * config/nds32/nds32-intrinsic.c
7393 (nds32_expand_builtin_null_ftype_reg): Delete.
7394 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7395 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7396 (nds32_read_argument): New.
7397 (nds32_legitimize_target): Ditto.
7398 (nds32_legitimize_argument): Ditto.
7399 (nds32_check_constant_argument): Ditto.
7400 (nds32_expand_unop_builtin): Ditto.
7401 (nds32_expand_unopimm_builtin): Ditto.
7402 (nds32_expand_binop_builtin): Ditto.
7403 (nds32_builtin_decl_impl): Ditto.
7404 (builtin_description): Ditto.
7405 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7406 (nds32_init_builtins_impl): Ditto.
7407 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7408 (nds32_builtin_decl): New.
7409 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7410 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7412 2018-03-02 Jeff Law <law@redhat.com>
7414 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7415 (redundant_insn, fill_simple_delay_slots): Likewise.
7416 (fill_slots_from_thread): Likewise.
7417 * resource.c (mark_referenced_resources): Likewise.
7418 (mark_set_resources, find_dead_or_set_registers): Likewise.
7420 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7422 * substring-locations.h (format_warning_va): Formatting fix for
7424 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7426 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7427 * substring-locations.c: Include intl.h.
7428 (format_warning_va): Turned into small wrapper around
7429 format_warning_n_va, renamed to ...
7430 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7431 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7433 (format_warning_at_substring_n): New function.
7434 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7435 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7436 format_warning_at_substring with just a shorter name instead of
7437 const function pointer.
7438 (fmtwarn_n): New function.
7439 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7440 appropriate, get rid of all the fmtstr temporaries, move conditionals
7441 with G_() wrapped string literals directly into fmtwarn arguments,
7442 cast dir.len to (int), formatting fixes.
7444 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7446 * doc/invoke.texi: Remove "Cilk Plus" references.
7448 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7449 Richard Biener <rguenther@suse.de>
7452 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7453 for error or warning attributes if CALL_FROM_THUNK_P is set.
7456 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7459 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7460 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7463 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7464 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7466 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7467 instead of -1U in last predictors element's probability member.
7469 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7472 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7473 arguments if they are comparable.
7475 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7477 PR tree-optimization/84634
7478 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7479 masks and masked_loop_p with a single loop_masks, making sure it's
7480 null for bb vectorization.
7482 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7484 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7485 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7486 loop->force_vectorize to check whether there is no alias.
7488 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7491 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7493 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7495 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7496 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7497 instead of a loop around prev_real_insn.
7498 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7502 * config/i386/i386.c (ix86_print_operand): Use conditional
7503 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7506 2018-03-02 Richard Biener <rguenther@suse.de>
7508 PR tree-optimization/84427
7509 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7510 (bitmap_set_subtract_values): Rewrite to handle multiple
7513 (prune_clobbered_mems): Likewise.
7514 (phi_translate): Take edge instead of pred/phiblock.
7515 (phi_translate_1): Likewise.
7516 (phi_translate_set): Likewise. Insert all translated
7517 exprs for a value into the set, keeping possibly multiple
7518 expressions per value.
7519 (compute_antic_aux): Adjust for phi_translate changes.
7520 When intersecting union the expressions and prune those
7521 not in the final value set, keeping possibly multiple
7522 expressions per value. Do not use value-insertion
7523 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7524 all expressions. Add verification that the value-sets
7525 only shrink during iteration.
7526 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7527 (do_pre_regular_insertion): Likewise.
7528 (do_pre_partial_partial_insertion): Likewise.
7530 2018-03-02 Richard Biener <rguenther@suse.de>
7533 * config/darwin.c (saved_debug_info_level): New static global.
7534 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7535 (darwin_asm_lto_end): Restore debug info generation settings.
7537 2018-03-01 Martin Liska <mliska@suse.cz>
7540 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7543 2018-03-01 Richard Biener <rguenther@suse.de>
7546 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7547 type annotation with LTO when debug was disabled at compile-time.
7549 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7551 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7553 (mips_final_postscan_insn): Likewise.
7555 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7557 PR rtl-optimization/84528
7558 * alias.c (init_alias_target): Add commentary.
7559 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7560 a unique base value if the frame pointer is not eliminated
7561 to the stack pointer.
7563 2018-03-01 Tom de Vries <tom@codesourcery.com>
7565 PR rtl-optimization/83327
7566 * lra-int.h (hard_regs_spilled_into): Declare.
7567 * lra.c (hard_regs_spilled_into): Define.
7568 (init_reg_info): Init hard_regs_spilled_into.
7569 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7570 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7571 (process_bb_lives): Handle hard_regs_spilled_into.
7572 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7573 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7575 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7577 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7578 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7579 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7580 * config/rs6000/aix72.h: New file.
7582 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7584 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7585 instead of warning_at with conditional singular and plural messages
7589 * stor-layout.c (update_alignment_for_field): For
7590 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7591 && !DECL_PACKED (field), do the alignment update, just use
7592 only desired_align instead of MAX (type_align, desired_align)
7594 (place_field): Don't do known_align < desired_align handling
7595 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7596 is non-NULL, instead do it after rli->prev_field handling and
7597 only if not within a bitfield word. For DECL_PACKED (field)
7598 use type_align of BITS_PER_UNIT.
7600 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7602 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7603 superfluous parentheses and trailing spaces.
7605 2018-02-28 Richard Biener <rguenther@suse.de>
7607 PR tree-optimization/84584
7608 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7609 SCoPs with fake exit edge.
7611 2018-02-28 Martin Liska <mliska@suse.cz>
7614 * timevar.c (timer::print): Fix format to properly print 100%
7617 2018-02-28 Richard Biener <rguenther@suse.de>
7620 * genmatch.c (capture_info::walk_match): Do not mark
7621 captured expressions without operands as expr_p given
7622 they act more like predicates and should be subject to
7623 "lost tail" side-effect preserving.
7625 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7627 PR rtl-optimization/81611
7628 * auto-inc-dec.c (attempt_change): Move dead note from
7629 mem_insn if it's the next use of regno
7630 (find_address): Take address use of reg holding
7631 non-incremented value. Add parm to limit search to the named
7633 (merge_in_block): Attempt to use a mem insn that is the next
7634 use of the original regno.
7636 2018-02-27 Martin Sebor <msebor@redhat.com>
7639 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7640 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7642 2018-02-27 Martin Sebor <msebor@redhat.com>
7644 PR translation/84207
7645 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7646 n argument to unsigned HOST_WIDE_INT.
7647 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7648 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7649 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7650 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7652 2018-02-27 Richard Biener <rguenther@suse.de>
7654 PR tree-optimization/84512
7655 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7656 Do not use the estimate returned from record_stmt_cost for
7657 the scalar iteration cost but sum properly using add_stmt_cost.
7659 2018-02-27 Richard Biener <rguenther@suse.de>
7661 PR tree-optimization/84466
7662 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7663 Adjust last change to less strictly validate use operands.
7665 2018-02-27 Martin Liska <mliska@suse.cz>
7667 PR gcov-profile/84548
7668 * gcov.c (process_file): Allow partial overlap and consider it
7669 also as group functions.
7670 (output_lines): Properly calculate range of lines for a group.
7672 2018-02-27 Martin Liska <mliska@suse.cz>
7674 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7675 'ggc' suffixes. Change first column width.
7676 (timer::print): Fix formatting of the column.
7678 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7680 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7681 preserve inline entry blocks for the sake of debug inline
7682 entry point markers alone.
7683 (remove_unused_locals): Suggest in comments a better place to
7684 force the preservation of inline entry blocks that are
7685 otherwise unused, but do not preserve them.
7687 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7689 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7691 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7694 * config/i386/constraints.md (Bs): Replace
7695 ix86_indirect_branch_register with
7696 TARGET_INDIRECT_BRANCH_REGISTER.
7698 * config/i386/i386.md (indirect_jump): Likewise.
7699 (tablejump): Likewise.
7700 (*sibcall_memory): Likewise.
7701 (*sibcall_value_memory): Likewise.
7702 Peepholes of indirect call and jump via memory: Likewise.
7703 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7704 (*sibcall_value_GOT_32): Likewise.
7705 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7706 (GOT_memory_operand): Likewise.
7707 (call_insn_operand): Likewise.
7708 (sibcall_insn_operand): Likewise.
7709 (GOT32_symbol_operand): Likewise.
7710 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7712 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7714 PR rtl-optimization/83496
7715 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7716 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7717 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7718 redundant insn, if any.
7719 (relax_delay_slots): Likewise.
7720 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7722 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7724 PR tree-optimization/83965
7725 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7726 that grouped statements are part of a reduction chain. Return
7727 true if the statement is not marked as a reduction itself but
7729 (vect_recog_dot_prod_pattern): Don't check whether the statement
7730 is part of a group here.
7731 (vect_recog_sad_pattern): Likewise.
7732 (vect_recog_widen_sum_pattern): Likewise.
7734 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7737 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7739 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7742 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7744 (ix86_output_indirect_function_return): New prototype.
7745 (ix86_split_simple_return_pop_internal): Likewise.
7746 * config/i386/i386.c (indirect_return_via_cx): New.
7747 (indirect_return_via_cx_bnd): Likewise.
7748 (indirect_thunk_name): Handle return va CX_REG.
7749 (output_indirect_thunk_function): Create alias for
7750 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7751 (ix86_output_indirect_jmp): Remove the bool argument.
7752 (ix86_output_indirect_function_return): New function.
7753 (ix86_split_simple_return_pop_internal): Likewise.
7754 * config/i386/i386.md (*indirect_jump): Don't pass false
7755 to ix86_output_indirect_jmp.
7756 (*tablejump_1): Likewise.
7757 (simple_return_pop_internal): Change it to define_insn_and_split.
7758 Call ix86_split_simple_return_pop_internal to split it for
7760 (simple_return_indirect_internal): Call
7761 ix86_output_indirect_function_return instead of
7762 ix86_output_indirect_jmp.
7764 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7767 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7768 memset and value initialization afterwards.
7770 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7772 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7774 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7777 * common/config/aarch64/aarch64-common.c
7778 (aarch_option_optimization_table[]): Switch
7779 off fomit-frame-pointer
7781 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7782 Chung-Ju Wu <jasonwucj@gmail.com>
7784 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7786 (store_multiple): Ditto.
7788 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7790 * config.gcc: Add --with-cpu support for nds32 target.
7791 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7792 * config/nds32/nds32.opt: Add -mcpu= option.
7794 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7796 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7797 isel=yes): Warn for these deprecated options.
7799 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7801 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7802 ISA_2_5_MASKS_EMBEDDED.
7804 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7806 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7807 p->max as pointers rather than using iterative_hash_expr.
7809 2018-02-23 Carl Love <cel@us.ibm.com>
7811 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7812 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7814 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7815 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7816 P8V_BUILTIN_VEC_VUNSIGNED2.
7818 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7821 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7822 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7824 * lra-constraints.c (curr_insn_transform): Set up
7825 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7827 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7828 * lra-eliminations.c (spill_pseudos): Ditto.
7829 (process_insn_for_elimination): Ditto.
7830 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7831 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7832 LRA_NON_CLOBBERED_ALT.
7834 2018-02-22 Martin Sebor <msebor@redhat.com>
7836 PR tree-optimization/84480
7837 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7838 to maybe_diag_stxncpy_trunc. Call it.
7839 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7840 from gimple_fold_builtin_strcpy. Print inlining stack.
7841 (handle_builtin_stxncpy): Print inlining stack.
7842 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7844 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7847 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7848 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7849 and -fcheck-pointer-bounds are used together.
7850 (indirect_thunk_prefix): New enum.
7851 (indirect_thunk_need_prefix): New function.
7852 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7853 "_nt" instead of "_bnd" for NOTRACK prefix.
7854 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7855 (output_indirect_thunk_function): Likewise.
7857 (ix86_code_end): Update output_indirect_thunk_function calls.
7858 (ix86_output_indirect_branch_via_reg): Replace
7859 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7860 (ix86_output_indirect_branch_via_push): Likewise.
7861 (ix86_output_function_return): Likewise.
7862 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7863 incompatible with -fcf-protection=branch and
7864 -fcheck-pointer-bounds.
7866 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7869 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7870 Change gcc_assert call to output_operand_lossage.
7872 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7874 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7876 2018-02-22 DJ Delorie <dj@redhat.com>
7877 Sebastian Perta <sebastian.perta@renesas.com>
7878 Oleg Endo <olegendo@gcc.gnu.org>
7880 * config/rx/rx.c (rx_rtx_costs): New function.
7881 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7883 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7885 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7887 2018-02-22 Martin Liska <mliska@suse.cz>
7890 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7891 Add "native" as a possible value.
7893 2018-02-22 Martin Liska <mliska@suse.cz>
7896 * config/i386/i386.c (ix86_option_override_internal):
7897 Add "native" as a possible value for -march and -mtune.
7899 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7902 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7903 to all type variants.
7905 PR tree-optimization/84503
7906 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7907 width as info->bitpos + info->bitsize - start.
7908 (merged_store_group::merge_overlapping): Simplify width computation.
7909 (check_no_overlap): New function.
7910 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7911 start + width and last_order of the group, fail if check_no_overlap
7913 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7914 to group if check_no_overlap fails.
7916 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7918 * config/rs6000/altivec.md: Delete contraint arguments to
7919 define_expand, define_split, and define_peephole2, and in
7920 define_insn_and_split if always unused.
7921 * config/rs6000/darwin.md: Ditto.
7922 * config/rs6000/dfp.md: Ditto.
7923 * config/rs6000/rs6000.md: Ditto.
7924 * config/rs6000/sync.md: Ditto.
7925 * config/rs6000/vector.md: Ditto.
7926 * config/rs6000/vsx.md: Ditto.
7928 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7930 * config/rs6000/altivec.md: Write output control strings as braced
7931 blocks instead of double-quoted strings.
7932 * config/rs6000/darwin.md: Ditto.
7933 * config/rs6000/rs6000.md: Ditto.
7934 * config/rs6000/vector.md: Ditto.
7935 * config/rs6000/vsx.md: Ditto.
7937 2018-02-21 Jason Merrill <jason@redhat.com>
7939 PR c++/84314 - ICE with templates and fastcall attribute.
7940 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7942 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7944 * ipa-cp.c (determine_versionability): Fix comment typos.
7946 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7949 * ipa-cp.c (determine_versionability): Do not version functions caling
7952 2018-02-21 Martin Liska <mliska@suse.cz>
7955 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7956 Add "native" as a possible value.
7957 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7958 the macro when native cpu detection is available.
7960 2018-02-21 Martin Liska <mliska@suse.cz>
7963 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7964 Add "native" as a possible value.
7965 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7966 when native cpu detection is available.
7968 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7969 Martin Sebor <msebor@redhat.com>
7971 PR tree-optimization/84478
7972 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7974 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7975 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7976 support which is conservatively correct, for 2 only stay conservative
7977 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7978 argument to the 2 argument get_range_strlen, adjust 6 arg
7979 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7981 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7982 (gimple_fold_builtin_strlen): Pass true as last argument to
7985 2018-02-20 Martin Sebor <msebor@redhat.com>
7988 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7989 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7990 (builtin_memref::builtin_memref): Factor out parts into
7991 set_base_and_offset and call it.
7993 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7996 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7997 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7998 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7999 search at the associated MODE_INT.
8001 2018-02-20 Jeff Law <law@redhat.com>
8004 PR tree-optimization/81592
8006 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8007 for range data rather than using global data.
8008 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8009 range data rather than using global data.
8010 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8011 pass it to children as needed.
8012 (struct directive::fmtresult): Similarly.
8013 (struct directive::set_width): Similarly.
8014 (struct directive::set_precision): Similarly.
8015 (format_integer, format_directive, parse_directive): Similarly.
8016 (format_none): Accept unnamed vr_values parameter.
8017 (format_percent, format_floating, format_character): Similarly.
8018 (format_string, format_plain): Similarly.
8019 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8020 the EVRP range analyzer for range data rather than using global data.
8021 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8022 gimple-ssa-evrp-analyze.h
8023 (class sprintf_dom_walker): Add after_dom_children member function.
8024 Add evrp_range_analyzer member.
8025 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8026 range analyzer as needed.
8027 (sprintf_dom_walker::after_dom_children): New member function.
8028 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8030 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8031 (evrp_range_analyzer::pop_to_marker): Likewise.
8033 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8035 PR tree-optimization/84419
8036 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8037 with the required type if its current type is compatible but
8040 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8043 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8044 after vectorization.
8046 2018-02-20 Martin Liska <mliska@suse.cz>
8049 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8050 possible values if we don't have a hint.
8052 2018-02-20 Martin Liska <mliska@suse.cz>
8056 * final.c (shorten_branches): Build align_tab array with one
8058 * opts.c (finish_options): Add alignment option limit check.
8059 (MAX_CODE_ALIGN): Likewise.
8060 (MAX_CODE_ALIGN_VALUE): Likewise.
8061 * doc/invoke.texi: Document maximum allowed option value for
8062 all -falign-* options.
8064 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8067 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8068 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8069 * var-tracking.c (emit_note_insn_var_location): Remove all references
8070 to NOTE_INSN_CALL_ARG_LOCATION.
8071 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8072 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8073 Use copy_rtx_if_shared.
8074 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8075 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8076 (dwarf2out_var_location): Remove handling of
8077 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8079 * final.c (final_scan_insn): Remove all references to
8080 NOTE_INSN_CALL_ARG_LOCATION.
8081 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8082 before dumping final insns.
8083 * except.c (emit_note_eh_region_end): Remove all references to
8084 NOTE_INSN_CALL_ARG_LOCATION.
8085 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8086 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8087 * config/arc/arc.c (hwloop_optimize): Likewise.
8088 * config/arm/arm.c (create_fix_barrier): Likewise.
8089 * config/s390/s390.c (s390_chunkify_start): Likewise.
8090 * config/sh/sh.c (find_barrier): Likewise.
8091 * config/i386/i386.c (rest_of_insert_endbranch,
8092 ix86_seh_fixup_eh_fallthru): Likewise.
8093 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8094 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8095 * config/frv/frv.c (frv_function_prologue): Likewise.
8096 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8098 (note_outside_basic_block_p): Remove all references to
8099 NOTE_INSN_CALL_ARG_LOCATION.
8100 * gengtype.c (adjust_field_rtx_def): Likewise.
8101 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8103 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8104 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8107 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8110 PR tree-optimization/84452
8111 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8112 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8115 2018-02-19 Martin Liska <mliska@suse.cz>
8118 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8120 2018-02-19 Martin Liska <mliska@suse.cz>
8121 Richard Sandiford <richard.sandiford@linaro.org>
8123 PR tree-optimization/82491
8124 * gimple-fold.c (get_base_constructor): Make earlier bail out
8127 2018-02-19 Carl Love <cel@us.ibm.com>
8129 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8130 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8132 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8133 P8V_BUILTIN_VEC_NEG.
8135 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8137 * config/rl78/rl78.md (movdf): New define expand.
8139 2018-02-19 Martin Liska <mliska@suse.cz>
8142 * doc/invoke.texi: Fix typo.
8143 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8145 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8147 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8148 handle rs6000_single_float and rs6000_double_float specially for
8151 2018-02-16 Jeff Law <law@redhat.com>
8153 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8156 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8159 * ipa-inline.c (inline_small_functions): Fix a typo.
8161 2018-02-16 Nathan Sidwell <nathan@acm.org>
8163 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8165 2018-02-16 Carl Love <cel@us.ibm.com>
8167 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8168 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8169 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8170 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8171 expansion to P8V_BUILTIN_VEC_FLOAT2.
8173 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8175 PR rtl-optimization/70023
8176 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8177 src_regno into account.
8179 2018-02-16 Carl Love <cel@us.ibm.com>
8181 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8182 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8183 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8184 * config/rs6000/rs6000.c: Remove case statements for
8185 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8186 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8187 and P9V_BUILTIN_VEC_VINSERT4B.
8188 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8189 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8190 * config/rs6000/vsx.md:
8191 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8194 2018-02-16 Carl Love <cel@us.ibm.com>
8196 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8198 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8200 * config/rs6000/rs6000-c.c: Add the definitions for
8201 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8202 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8203 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8204 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8205 definition for insert4b and define insn *insert3b_internal.
8206 * doc/extend.texi: Add documentation for vec_extract4b.
8208 2018-02-16 Nathan Sidwell <nathan@acm.org>
8210 * doc/extend.texi (Backwards Compatibility): Mention friend
8211 injection. Note for-scope is deprecated.
8212 * doc/invoke.texi (-ffriend-injection): Deprecate.
8214 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8216 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8217 that moved to I2, also allow destinations that are a paradoxical
8218 subreg (instead of a normal reg).
8220 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8223 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8226 2018-02-16 Richard Biener <rguenther@suse.de>
8228 PR tree-optimization/84037
8229 PR tree-optimization/84016
8231 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8232 Adjust vec_construct for the fact we need additional higher latency
8233 128bit inserts for AVX256 and AVX512 vector builds.
8234 (ix86_add_stmt_cost): Scale vector construction cost for
8237 2018-02-16 Richard Biener <rguenther@suse.de>
8239 PR tree-optimization/84417
8240 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8241 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8242 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8244 2018-02-16 Martin Liska <mliska@suse.cz>
8247 * internal-fn.def (ASAN_CHECK): Set proper flags.
8248 (ASAN_MARK): Likewise.
8250 2018-02-16 Julia Koval <julia.koval@intel.com>
8252 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8253 from PTA_CANNONLAKE.
8255 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8258 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8259 Use ++iter rather than iter++ for std::list iterators.
8260 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8261 defer deleting them until all nodes in the forest are processed. Do
8262 free even leaf nodes. Change to_process into auto_vec.
8265 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8266 * vec.h (vec_default_construct): Use memset instead of placement new
8267 if BROKEN_VALUE_INITIALIZATION is defined.
8268 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8269 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8272 PR rtl-optimization/83723
8273 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8274 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8275 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8277 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8279 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8281 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8283 PR rtl-optimization/81443
8284 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8285 from inner REGs to paradoxical SUBREGs.
8287 2018-02-16 Richard Biener <rguenther@suse.de>
8289 PR tree-optimization/84399
8290 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8291 For operands we can analyze at their definition make sure we can
8292 analyze them at each use as well.
8294 2018-02-16 Richard Biener <rguenther@suse.de>
8296 PR tree-optimization/84190
8297 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8298 volatile accesses if the decl isn't volatile.
8300 2018-02-15 Jason Merrill <jason@redhat.com>
8302 PR c++/84314 - ICE with templates and fastcall attribute.
8303 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8304 TYPE_CANONICAL on an existing type.
8306 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8308 PR tree-optimization/84383
8309 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8310 dstoff nor call operand_equal_p if dstbase is NULL.
8312 PR tree-optimization/84334
8313 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8314 also a CONSTANT_CLASS_P, punt.
8316 2018-02-14 Jim Wilson <jimw@sifive.com>
8318 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8319 first SMALL_OPERAND check. New local min_second_step. Move assert
8320 to where locals are set. Add TARGET_RVC support.
8321 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8323 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8325 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8327 2018-02-14 Martin Sebor <msebor@redhat.com>
8329 PR tree-optimization/83698
8330 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8331 arrays constrain the offset range to their bounds.
8332 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8333 (builtin_access::overlap): Avoid setting the size of overlap if it's
8335 (maybe_diag_overlap): Also consider arrays when deciding what values
8336 of offsets to include in diagnostics.
8338 2018-02-14 Martin Sebor <msebor@redhat.com>
8341 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8342 that correspond to the kind of a declaration.
8344 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8347 * config/pa/pa.md: Load address of PIC label using the linkage table
8348 if the label is nonlocal.
8350 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8352 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8353 warning message if user requests -maltivec=be.
8354 * doc/invoke.texi: Document deprecation of -maltivec=be.
8356 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8359 * config/rs6000/rs6000-c.c: Update definitions for
8360 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8361 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8363 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8366 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8367 add _get_ssp intrinsics. Remove argument from
8368 __builtin_ia32_rdssp[d|q].
8369 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8370 * config/i386/i386-builtin.def: Remove argument from
8371 __builtin_ia32_rdssp[d|q].
8372 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8373 ix86_expand_special_args_builtin for _rdssp[d|q].
8374 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8375 Clear register before usage.
8376 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8377 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8379 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8381 PR tree-optimization/84357
8382 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8383 operand 1 of an ARRAY_REF too.
8385 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8388 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8389 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8391 (set_of_reg): New struct.
8392 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8393 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8394 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8396 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8397 Split into bitclr, bitset, bitinvert patterns if appropriate.
8398 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8399 use rx_fuse_in_memory_bitop.
8400 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8401 to named insn, correct maximum insn length.
8403 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8406 * machmode.def: Define a complex mode for PARTIAL_INT.
8407 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8409 * doc/rtl.texi: Document CSPImode.
8410 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8412 (msp430_hard_regno_nregs_with_padding): Likewise.
8414 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8417 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8419 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8421 PR rtl-optimization/84169
8422 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8423 we generated a parallel as new i3 and we split that to new i2 and i3
8424 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8425 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8426 those to i2, not i1. Partially rewrite this scan code.
8428 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8431 * stor-layout.c (place_field): For variable length fields, adjust
8432 offset_align afterwards not just based on the field's alignment,
8433 but also on the size.
8436 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8437 of exps and logs in the use_exp2 case.
8439 2018-02-13 Jeff Law <law@redhat.com>
8441 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8444 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8447 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8451 * common.opt (gas-loc-support, gas-locview-support): New.
8452 (ginline-points, ginternal-reset-location-views): New.
8453 * doc/invoke.texi: Document them. Use @itemx where intended.
8454 (gvariable-location-views): Adjust.
8455 * target.def (reset_location_view): New.
8456 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8457 (TARGET_RESET_LOCATION_VIEW): New.
8458 * doc/tm.texi: Rebuilt.
8459 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8460 (dwarf2out_default_as_locview_support): New.
8461 (output_asm_line_debug_info): Use option variables.
8462 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8463 (output_loc_list): Likewise.
8464 (add_high_low_attributes): Check option variables.
8465 Don't output entry view attribute in strict mode.
8466 (gen_inlined_subroutine_die): Check option variables.
8467 (dwarf2out_inline_entry): Likewise.
8468 (init_sections_and_labels): Likewise.
8469 (dwarf2out_early_finish): Likewise.
8470 (maybe_reset_location_view): New, from...
8471 (dwarf2out_var_location): ... here. Call it.
8472 * debug.h (dwarf2out_default_as_loc_support): Declare.
8473 (dwarf2out_default_as_locview_support): Declare.
8474 * hooks.c (hook_int_rtx_insn_0): New.
8475 * hooks.h (hook_int_rtx_insn_0): Declare.
8476 * toplev.c (process_options): Take -gas-loc-support and
8477 -gas-locview-support from dwarf2out. Enable
8478 -gvariable-location-views by default only with locview
8479 assembler support. Enable -ginternal-reset-location-views by
8480 default only if the target defines the corresponding hook.
8481 Enable -ginline-points by default if location views are
8482 enabled; force it disabled if statement frontiers are
8484 * tree-inline.c (expand_call_inline): Check option variables.
8485 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8487 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8489 PR tree-optimization/84321
8490 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8491 handling. Also check whether the anti-range contains any values
8492 that satisfy the mask; switch to a VR_RANGE if not.
8494 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8497 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8499 2018-02-13 Martin Jambor <mjambor@suse.cz>
8502 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8503 of call statements, also set location of a load to a temporary.
8505 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8507 * config/rl78/rl78.c (add_vector_labels): New function.
8508 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8509 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8510 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8511 which checks that no arguments are passed.
8512 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8513 * doc/extend.texi: Documentation for the new attribute.
8515 2018-02-13 Andreas Schwab <schwab@suse.de>
8517 * config/riscv/linux.h (CPP_SPEC): Define.
8519 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8522 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8523 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8524 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8525 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8526 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8527 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8528 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8529 temporarily for AES and PCLMUL builtins.
8531 PR tree-optimization/84339
8532 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8533 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8537 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8538 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8539 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8541 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8543 * omp-simd-clone.h: New file.
8544 * omp-simd-clone.c: Include omp-simd-clone.h.
8545 (expand_simd_clones): No longer static.
8546 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8547 cgraph.h and omp-simd-clone.h.
8548 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8549 (vect_recog_widen_shift_pattern): Formatting fix.
8550 (vect_pattern_recog_1): Don't check optab for calls.
8553 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8554 operands[2] into a REG before using gen_lowpart on it.
8556 2018-02-12 Jeff Law <law@redhat.com>
8559 * config/sh/sh.c (find_barrier): Consider a sibling call
8562 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8563 successfully back substituting a reg.
8565 2018-02-12 Richard Biener <rguenther@suse.de>
8567 PR tree-optimization/84037
8568 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8569 parameter, move visited init to caller.
8570 (vect_slp_analyze_operations): Separate cost from validity
8571 check, initialize visited once for all instances.
8572 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8574 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8576 (vect_model_promotion_demotion_cost): Likewise.
8577 (vectorizable_bswap): Guard cost modeling with !slp_node
8578 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8580 (vectorizable_call): Likewise.
8581 (vectorizable_conversion): Likewise.
8582 (vectorizable_assignment): Likewise.
8583 (vectorizable_shift): Likewise.
8584 (vectorizable_operation): Likewise.
8585 (vectorizable_store): Likewise.
8586 (vectorizable_load): Likewise.
8587 (vectorizable_condition): Likewise.
8588 (vectorizable_comparison): Likewise.
8590 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8593 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8594 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8595 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8597 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8600 * params.def (inline-min-speedup): Increase from 8 to 15.
8601 (max-inline-insns-auto): Decrease from 40 to 30.
8602 * ipa-split.c (consider_split): Add some buffer for function to
8603 be considered inlining candidate.
8604 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8607 2018-02-12 Richard Biener <rguenther@suse.de>
8609 PR tree-optimization/84037
8610 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8611 matched stmts if we cannot swap the non-matched ones.
8613 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8615 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8616 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8617 _mm_maskz_scalef_round_ss): New intrinsics.
8618 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8619 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8620 __builtin_ia32_scalefss_round): Remove.
8621 (__builtin_ia32_scalefsd_mask_round,
8622 __builtin_ia32_scalefss_mask_round): New intrinsics.
8623 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8624 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8625 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8626 "<round_constraint>")): Changed to ...
8627 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8628 "<round_scalar_constraint>")): ... this.
8629 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8630 %0, %1, %2<round_op3>}"): Changed to ...
8631 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8632 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8633 %2<round_scalar_mask_op3>}"): ... this.
8634 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8636 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8638 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8639 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8640 (_mm_maskz_sqrt_round_ss): New intrinsics.
8641 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8642 (__builtin_ia32_sqrtsd_mask_round)
8643 (__builtin_ia32_sqrtss_mask_round): New builtins.
8644 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8645 (__builtin_ia32_sqrtss_round): Remove.
8646 (__builtin_ia32_sqrtsd_mask_round)
8647 (__builtin_ia32_sqrtss_mask_round): New builtins.
8648 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8649 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8650 ((match_operand:VF_128 1 "vector_operand"
8651 "xBm,<round_constraint>")): Changed to ...
8652 ((match_operand:VF_128 1 "vector_operand"
8653 "xBm,<round_scalar_constraint>")): ... this.
8654 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8655 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8656 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8657 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8658 %<iptr>1<round_scalar_mask_op3>}): ... this.
8659 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8660 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8662 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8665 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8666 Cast vec_cmpeq result to correct type.
8667 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8668 Cast vec_cmpgt result to correct type.
8670 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8672 * final.c (final_scan_insn_1): Renamed from...
8673 (final_scan_insn): ... this. New wrapper, to recover
8674 seen from the outermost call in recursive ones.
8675 * config/sparc/sparc.c (output_return): Drop seen from call.
8676 (output_sibcall): Likewise.
8677 * config/visium/visium.c (output_branch): Likewise.
8679 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8681 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8684 2018-02-10 Alan Modra <amodra@gmail.com>
8687 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8688 Specify LR as an input.
8690 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8693 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8694 remove_member_access_dummy_vars): New functions.
8695 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8696 lower_omp_1, execute_lower_omp): Use them.
8698 PR rtl-optimization/84308
8699 * shrink-wrap.c (spread_components): Release todo vector.
8701 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8703 PR rtl-optimization/57193
8704 * ira-color.c (struct allocno_color_data): Add member
8705 conflict_allocno_hard_prefs.
8706 (update_conflict_allocno_hard_prefs): New.
8707 (bucket_allocno_compare_func): Add a preference based on
8708 conflict_allocno_hard_prefs.
8709 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8710 (color_allocnos): Remove a dead code. Initiate
8711 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8713 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8716 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8717 constraint from =wa to wa. Avoid a subreg on the output operand,
8718 instead use a pseudo and subreg it in a move.
8719 (p9_xxbrd_<mode>): Changed to ...
8720 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8721 (p9_xxbrd_v2df): New expander.
8722 (p9_xxbrw_<mode>): Changed to ...
8723 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8724 (p9_xxbrw_v4sf): New expander.
8726 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8728 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8730 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8733 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8734 multiply in 32-bit mode.
8735 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8736 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8739 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8741 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8742 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8743 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8744 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8746 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8749 * dwarf2out.c (is_trivial_indirect_ref): New function.
8750 (dwarf2out_late_global_decl): Do not generate a location
8751 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8752 and that are not defined in the current unit.
8754 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8756 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8757 instead of a libcall for UNORDERED.
8759 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8762 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8763 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8765 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8768 * config/s390/s390.c (s390_set_current_function): Invoke
8769 s390_indirect_branch_settings also if fndecl didn't change.
8771 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8773 * config/rs6000/rs6000.md (blockage): Set length to zero.
8775 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8777 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8779 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8782 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8783 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8787 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8788 PARALLEL incoming that failed vt_get_decl_and_offset check.
8791 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8792 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8794 (get_variable_section): For decls in named .bss* sections pass true as
8795 second argument to bss_initializer_p.
8797 2018-02-09 Marek Polacek <polacek@redhat.com>
8798 Jakub Jelinek <jakub@redhat.com>
8801 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8802 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8803 Sync some changes from cxx_fold_indirect_ref.
8805 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8807 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8809 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8810 (BLOCK_INLINE_ENTRY_LABEL): New.
8811 (dwarf2out_var_location): Disregard inline entry markers.
8812 (inline_entry_data): New struct.
8813 (inline_entry_data_hasher): New hashtable type.
8814 (inline_entry_data_hasher::hash): New.
8815 (inline_entry_data_hasher::equal): New.
8816 (inline_entry_data_table): New variable.
8817 (add_high_low_attributes): Add DW_AT_entry_pc and
8818 DW_AT_GNU_entry_view attributes if a pending entry is found
8819 in inline_entry_data_table. Add old entry_pc attribute only
8820 if debug nonbinding markers are disabled.
8821 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8822 markers are enabled.
8823 (block_within_block_p, dwarf2out_inline_entry): New.
8824 (dwarf2out_finish): Check that no entries remained in
8825 inline_entry_data_table.
8826 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8827 (final_scan_insn, notice_source_line): Likewise.
8828 (rest_of_clean_state): Skip inline entry markers.
8829 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8831 * gimple.c (gimple_build_debug_inline_entry): New.
8832 * gimple.h (enum gimple_debug_subcode): Add
8833 GIMPLE_DEBUG_INLINE_ENTRY.
8834 (gimple_build_debug_inline_entry): Declare.
8835 (gimple_debug_inline_entry_p): New.
8836 (gimple_debug_nonbind_marker_p): Adjust.
8837 * insn-notes.def (INLINE_ENTRY): New.
8838 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8839 inline entry marker notes.
8840 (print_insn): Likewise.
8841 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8842 (INSN_DEBUG_MARKER_KIND): Likewise.
8843 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8844 * tree-inline.c (expand_call_inline): Build and insert
8845 debug_inline_entry stmt.
8846 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8847 inline entry blocks early, if nonbind markers are enabled.
8848 (dump_scope_block): Dump fragment info.
8849 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8850 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8851 (gimple_build_debug_inline_entry): New.
8852 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8853 Enable/disable inline entry points too.
8854 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8855 (DEBUG_INSN): Describe inline entry markers.
8857 * common.opt (gvariable-location-views): New.
8858 (gvariable-location-views=incompat5): New.
8859 * config.in: Rebuilt.
8860 * configure: Rebuilt.
8861 * configure.ac: Test assembler for view support.
8862 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8863 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8864 * dwarf2out.c (var_loc_view): New typedef.
8865 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8866 (dwarf2out_locviews_in_attribute): New.
8867 (dwarf2out_locviews_in_loclist): New.
8868 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8869 (enum dw_line_info_opcode): Add LI_adv_address.
8870 (struct dw_line_info_table): Add view.
8871 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8872 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8873 (zero_view_p): New variable.
8874 (ZERO_VIEW_P): New macro.
8875 (output_asm_line_debug_info): New.
8876 (struct var_loc_node): Add view.
8877 (add_AT_view_list, AT_loc_list): New.
8878 (add_var_loc_to_decl): Add view param. Test it against last.
8879 (new_loc_list): Add view params. Record them.
8880 (AT_loc_list_ptr): Handle loc and view lists.
8881 (view_list_to_loc_list_val_node): New.
8882 (print_dw_val): Handle dw_val_class_view_list.
8883 (size_of_die): Likewise.
8884 (value_format): Likewise.
8885 (loc_list_has_views): New.
8886 (gen_llsym): Set vl_symbol too.
8887 (maybe_gen_llsym, skip_loc_list_entry): New.
8888 (dwarf2out_maybe_output_loclist_view_pair): New.
8889 (output_loc_list): Output view list or entries too.
8890 (output_view_list_offset): New.
8891 (output_die): Handle dw_val_class_view_list.
8892 (output_dwarf_version): New.
8893 (output_compilation_unit_header): Use it.
8894 (output_skeleton_debug_sections): Likewise.
8895 (output_rnglists, output_line_info): Likewise.
8896 (output_pubnames, output_aranges): Update version comments.
8897 (output_one_line_info_table): Output view numbers in asm comments.
8898 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8899 Call maybe_gen_llsym.
8900 (loc_list_from_tree_1): Adjust.
8901 (add_AT_location_description): Create view list attribute if
8902 needed, check it's absent otherwise.
8903 (convert_cfa_to_fb_loc_list): Adjust.
8904 (maybe_emit_file): Call output_asm_line_debug_info for test.
8905 (dwarf2out_var_location): Reset views as needed. Precompute
8906 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8907 attribute. Set view.
8908 (new_line_info_table): Reset next view.
8909 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8910 (dwarf2out_source_line): Likewise. Output view resets and labels to
8911 the assembler, or select appropriate line info opcodes.
8912 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8913 (optimize_string_length): Catch it. Adjust.
8914 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8915 dw_val_class_view_list, and remove it if no longer needed.
8916 (hash_loc_list): Hash view numbers.
8917 (loc_list_hasher::equal): Compare them.
8918 (optimize_location_lists): Check whether a view list symbol is
8919 needed, and whether the locview attribute is present, and
8920 whether they match. Remove the locview attribute if no longer
8922 (index_location_lists): Call skip_loc_list_entry for test.
8923 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8924 Use output_dwarf_version.
8925 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8926 (struct dw_val_node): Add val_view_list.
8927 * final.c (SEEN_NEXT_VIEW): New.
8928 (set_next_view_needed): New.
8929 (clear_next_view_needed): New.
8930 (maybe_output_next_view): New.
8931 (final_start_function): Rename to...
8932 (final_start_function_1): ... this. Take pointer to FIRST,
8933 add SEEN parameter. Emit param bindings in the initial view.
8934 (final_start_function): Reintroduce SEEN-less interface.
8935 (final): Rename to...
8936 (final_1): ... this. Take SEEN parameter. Output final pending
8937 next view at the end.
8938 (final): Reintroduce seen-less interface.
8939 (final_scan_insn): Output pending next view before switching
8940 sections or ending a block. Mark the next view as needed when
8941 outputting variable locations. Notify debug backend of section
8942 changes, and of location view changes.
8943 (rest_of_handle_final): Adjust.
8944 * toplev.c (process_options): Autodetect value for debug variable
8945 location views option. Warn on incompat5 without -gdwarf-5.
8946 * doc/invoke.texi (gvariable-location-views): New.
8947 (gvariable-location-views=incompat5): New.
8948 (gno-variable-location-views): New.
8950 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8952 PR tree-optimization/84136
8953 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8954 that the result of find_edge is non-NULL.
8956 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8959 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8960 storing integer register in SImode. Fix cost of 256 and 512
8961 byte aligned SSE register store.
8963 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8965 * config/i386/i386.c (ix86_multiplication_cost): Fix
8966 multiplication cost for TARGET_AVX512DQ.
8968 2018-02-08 Marek Polacek <polacek@redhat.com>
8970 PR tree-optimization/84238
8971 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8974 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8976 PR tree-optimization/84265
8977 * tree-vect-stmts.c (vectorizable_store): Don't treat
8978 VMAT_CONTIGUOUS accesses as grouped.
8979 (vectorizable_load): Likewise.
8981 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8983 PR tree-optimization/81635
8984 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8985 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8986 (test_round_for_mask): New functions.
8987 (wide_int_cc_tests): Call test_round_for_mask.
8988 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8989 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8990 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8991 range returned by get_range_info.
8993 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8996 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8997 * symtab.c: Include builtins.h
8998 (symtab_node::output_to_lto_symbol_table_p): Move here
8999 from lto-streamer-out.c:output_symbol_p.
9000 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9001 (output_symbol_p): Move all logic to symtab.c
9002 (produce_symtab): Update.
9004 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9006 * config/s390/s390-opts.h (enum indirect_branch): Define.
9007 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9008 (s390_indirect_branch_via_thunk)
9009 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9010 (enum s390_indirect_branch_type): Define.
9011 * config/s390/s390.c (struct s390_frame_layout, struct
9012 machine_function): Remove.
9013 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9014 (indirect_branch_table_label_no, indirect_branch_table_name):
9016 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9017 (enum s390_indirect_branch_option): Define.
9018 (s390_return_addr_from_memory): New function.
9019 (s390_handle_string_attribute): New function.
9020 (s390_attribute_table): Add new attribute handler.
9021 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9022 (s390_indirect_branch_via_thunk): New function.
9023 (s390_indirect_branch_via_inline_thunk): New function.
9024 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9025 sibling call optimization for non z10 compiles.
9026 (s390_emit_call): Force indirect branch target to be a single
9027 register. Add r1 clobber for non-z10 compiles.
9028 (s390_emit_epilogue): Emit return jump via return_use expander.
9029 (s390_reorg): Handle JUMP_INSNs as execute targets.
9030 (s390_option_override_internal): Perform validity checks for the
9031 new command line options.
9032 (s390_indirect_branch_attrvalue): New function.
9033 (s390_indirect_branch_settings): New function.
9034 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9035 (s390_output_indirect_thunk_function): New function.
9036 (s390_code_end): Implement target hook.
9037 (s390_case_values_threshold): Implement target hook.
9038 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9040 * config/s390/s390.h (struct s390_frame_layout)
9041 (struct machine_function): Move here from s390.c.
9042 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9043 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9044 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9045 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9046 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9047 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9048 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9049 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9050 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9051 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9052 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9053 (mnemonic attribute): Add values which aren't recognized
9055 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9056 pattern for branch conversion. Fix mnemonic attribute.
9057 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9058 indirect branch via thunk if requested.
9059 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9060 ("*indirect_jump"): Disable for branch conversion using out of
9062 ("indirect_jump_via_thunk<mode>_z10")
9063 ("indirect_jump_via_thunk<mode>")
9064 ("indirect_jump_via_inlinethunk<mode>_z10")
9065 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9066 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9067 ("casesi_jump_via_inlinethunk<mode>_z10")
9068 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9069 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9070 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9071 ("*indirect2_jump"): Disable for branch conversion.
9072 ("casesi_jump"): Turn into expander and expand patterns for branch
9074 ("return_use"): New expander.
9075 ("*return"): Emit return via thunk and rename it to ...
9076 ("*return<mode>"): ... this one.
9077 * config/s390/s390.opt: Add new options and and enum for the
9080 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9082 * lra-constraints.c (match_reload): Unconditionally use
9083 gen_lowpart_SUBREG, rather than selecting between that
9084 and equivalent gen_rtx_SUBREG code.
9086 2018-02-08 Richard Biener <rguenther@suse.de>
9088 PR tree-optimization/84233
9089 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9090 changed flag instead of boguously re-using phi_inserted.
9092 2018-02-08 Martin Jambor <mjambor@suse.cz>
9094 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9095 static local variables.
9097 2018-02-08 Richard Biener <rguenther@suse.de>
9099 PR tree-optimization/84278
9100 * tree-vect-stmts.c (vectorizable_store): When looking for
9101 smaller vector types to perform grouped strided loads/stores
9102 make sure the mode is supported by the target.
9103 (vectorizable_load): Likewise.
9105 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9107 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9108 Increase LDP/STP opportunities by adding adjacent callee-saves.
9110 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9112 PR rtl-optimization/84068
9113 PR rtl-optimization/83459
9114 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9116 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9118 PR tree-optimization/84224
9119 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9120 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9123 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9126 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9127 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9128 count, remove one USE.
9130 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9132 * doc/install.texi (Configuration): Document the
9133 --with-long-double-format={ibm,ieee} PowerPC configuration
9137 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9138 Convert from define_expand to be define_insn_and_split. Rework
9139 float/double/_Float128 conversions to QI/HI/SImode to work with
9140 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9141 conversions to QI/HImode types did a store and then a load to
9142 truncate the value. For conversions to VSX registers, don't split
9143 the insn, instead emit the code directly. Use the code iterator
9144 any_fix to combine signed and unsigned conversions.
9145 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9146 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9147 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9148 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9149 (fix_<mode>di2_hw): Likewise.
9150 (fixuns_<mode>di2_hw): Likewise.
9151 (fix_<mode>si2_hw): Likewise.
9152 (fixuns_<mode>si2_hw): Likewise.
9153 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9154 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9155 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9156 fix<uns>_trunc<SFDF:mode>si2_p8.
9157 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9158 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9159 (fix<uns>_<mode>_mem): Likewise.
9160 (fctiw<u>z_<mode>_mem): Likewise.
9161 (fix<uns>_<mode>_mem): Likewise.
9162 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9163 the register allocator from doing a direct move to the GPRs to do
9164 a store, and instead use the ISA 3.0 store byte/half-word from
9165 vector register instruction. For IEEE 128-bit floating point,
9166 also optimize stores of 32-bit ints.
9167 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9169 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9171 * genextract.c (push_pathstr_operand): New function to support
9173 (walk_rtx): Call push_pathstr_operand.
9174 (print_path): Support [a-zA-Z].
9176 2018-02-07 Richard Biener <rguenther@suse.de>
9178 PR tree-optimization/84037
9179 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9180 (cse_and_gimplify_to_preheader): Declare.
9181 (vect_get_place_in_interleaving_chain): Likewise.
9182 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9184 (_loop_vec_info::~_loop_vec_info): Delete it.
9185 (cse_and_gimplify_to_preheader): New function.
9186 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9187 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9188 (vectorizable_load): Likewise. For grouped stores always base
9189 the IV on the first element.
9190 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9191 condition before gimplifying.
9193 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9195 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9196 *DIV_EXPR and *MOD_EXPR.
9198 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9201 * config/i386/i386.c (ix86_option_override_internal): Mask out
9202 the CF_SET bit when checking -fcf-protection.
9204 2018-02-07 Tom de Vries <tom@codesourcery.com>
9207 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9210 2018-02-07 Richard Biener <rguenther@suse.de>
9212 PR tree-optimization/84204
9213 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9216 PR tree-optimization/84205
9217 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9218 special-case isl_ast_op_zdiv_r.
9220 PR tree-optimization/84223
9221 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9222 Only add conditions from within the region.
9223 (gather_bbs::after_dom_children): Adjust.
9225 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9228 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9229 * config/avr/avr.md: Only post-reload split REG-REG moves if
9230 either register is GENERAL_REG_P.
9232 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9234 PR tree-optimization/84235
9235 * tree-ssa-scopedtables.c
9236 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9237 if the subtraction is performed in floating point type where NaNs are
9238 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9239 build 1. Formatting fix.
9241 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9244 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9245 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9246 and skip it regardless of bb boundaries. Use CALL_P macro,
9247 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9248 unnecessarily, formatting fix.
9250 2018-02-06 Michael Collison <michael.collison@arm.com>
9252 * config/arm/thumb2.md:
9253 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9254 (*thumb_mov_notscc): Ditto.
9256 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9259 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9260 unsigned_fix, not "s".
9262 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9264 * configure.ac (gcc_fn_eh_frame_ro): New function.
9265 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9266 correct .eh_frame permissions.
9267 * configure: Regenerate.
9269 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9271 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9274 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9276 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9277 Display warning message for -mno-speculate-indirect-jumps.
9279 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9281 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9283 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9285 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9287 PR tree-optimization/84225
9288 * tree-eh.c (find_trapping_overflow): Only call
9289 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9291 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9294 * config/i386/i386.c: Reimplement the check of possible options
9295 -mibt/-mshstk conbination. Change error messages.
9296 * doc/invoke.texi: Fix a typo: remove extra '='.
9298 2018-02-06 Marek Polacek <polacek@redhat.com>
9300 PR tree-optimization/84228
9301 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9303 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9306 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9307 emitted arch directives.
9308 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9309 __ARM_FEATURE_COPROC before changing architectures.
9311 2018-02-06 Richard Biener <rguenther@suse.de>
9313 * config/i386/i386.c (print_reg): Fix typo.
9314 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9316 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9318 * configure: Regenerate.
9320 2018-02-05 Martin Sebor <msebor@redhat.com>
9322 PR tree-optimization/83369
9323 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9326 2018-02-05 Martin Liska <mliska@suse.cz>
9328 * doc/invoke.texi: Cherry-pick upstream r323995.
9330 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9332 * ira.c (ira_init_register_move_cost): Adjust comment.
9334 2018-02-05 Martin Liska <mliska@suse.cz>
9336 PR gcov-profile/84137
9337 * doc/gcov.texi: Fix typo in documentation.
9339 2018-02-05 Martin Liska <mliska@suse.cz>
9341 PR gcov-profile/83879
9342 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9343 using dlopen functionality.
9345 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9347 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9348 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9349 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9350 _mm_maskz_range_round_ss): New intrinsics.
9351 (__builtin_ia32_rangesd128_round)
9352 (__builtin_ia32_rangess128_round): Remove.
9353 (__builtin_ia32_rangesd128_mask_round,
9354 __builtin_ia32_rangess128_mask_round): New builtins.
9355 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9356 __builtin_ia32_rangess128_round): Remove.
9357 (__builtin_ia32_rangesd128_mask_round,
9358 __builtin_ia32_rangess128_mask_round): New builtins.
9359 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9360 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9361 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9362 "<round_saeonly_constraint>")): Changed to ...
9363 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9364 "<round_saeonly_scalar_constraint>")): ... this.
9365 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9366 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9367 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9368 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9369 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9371 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9373 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9375 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9376 Remove all values except native, 8540 and 8548.
9378 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9380 * config/i386/i386.c (ix86_output_function_return): Pass
9381 INVALID_REGNUM, instead of -1, as invalid register number to
9382 indirect_thunk_name and output_indirect_thunk.
9384 2018-02-02 Julia Koval <julia.koval@intel.com>
9386 * config.gcc: Add -march=icelake.
9387 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9388 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9389 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9390 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9391 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9392 (processor_target_table): Add icelake.
9393 (ix86_option_override_internal): Handle new PTAs.
9394 (get_builtin_code_for_version): Handle icelake.
9395 (M_INTEL_COREI7_ICELAKE): New.
9396 (fold_builtin_cpu): Handle icelake.
9397 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9398 * doc/invoke.texi: Add -march=icelake.
9400 2018-02-02 Julia Koval <julia.koval@intel.com>
9402 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9403 to wide_int_bitmask.
9404 * wide-int-bitmask.h: New.
9406 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9409 * config/i386/i386.md: Replace Pmode with word_mode in
9410 builtin_setjmp_setup and builtin_longjmp to support x32.
9412 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9416 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9418 (rs6000_supported_cpu_names): New static variable.
9419 (linux_cpu_translation_table): Likewise.
9420 (elf_platform) <cpu>: Define new static variable and use it.
9421 Translate kernel AT_PLATFORM name to canonical name if needed.
9422 Error if platform name is unknown.
9424 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9427 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9429 2018-02-01 Jeff Law <law@redhat.com>
9432 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9433 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9434 the scratch if RELEASE_VIA_POP is false.
9435 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9436 If we have to save a temporary register, decrement SIZE appropriately.
9437 Pass new arguments to release_scratch_register_on_entry.
9438 (ix86_adjust_stack_and_probe): Likewise.
9439 (ix86_emit_probe_stack_range): Pass new arguments to
9440 release_scratch_register_on_entry.
9442 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9444 PR rtl-optimization/84157
9445 * combine.c (change_zero_ext): Use REG_P predicate in
9446 front of HARD_REGISTER_P predicate.
9448 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9450 * config/avr/avr.c (avr_option_override): Move disabling of
9451 -fdelete-null-pointer-checks to...
9452 * common/config/avr/avr-common.c (avr_option_optimization_table):
9455 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9457 PR tree-optimization/81635
9458 * tree-data-ref.c (split_constant_offset_1): For types that
9459 wrap on overflow, try to use range info to prove that wrapping
9462 2018-02-01 Renlin Li <renlin.li@arm.com>
9465 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9467 (aarch64_register_move_cost): Likewise.
9468 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9470 (REG_CLASS_NAMES): Likewise.
9471 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9472 TAILCALL_ADDR_REGS. Remove IP registers.
9473 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9475 2018-02-01 Richard Biener <rguenther@suse.de>
9477 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9478 for specifying RPO order and allow NULL for that.
9479 * domwalk.c (dom_walker::dom_walker): Likewise.
9480 (dom_walker::walk): Handle NULL RPO order.
9481 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9483 (rewrite_update_dom_walker): Likewise.
9484 (mark_def_dom_walker): Likewise.
9486 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9488 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9489 (aarch64_maybe_expand_sve_subreg_move): Declare.
9490 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9491 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9493 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9494 that are semantically a reverse operation.
9495 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9496 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9497 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9499 (aarch64_can_change_mode_class): For big-endian, forbid changes
9500 between two SVE modes if they have different element sizes.
9502 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9504 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9505 the TImode handling for big-endian targets.
9507 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9509 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9510 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9512 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9513 Remove BSWAP handing for big-endian targets and use the form of
9514 LD1RQ appropariate for the mode.
9516 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9518 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9519 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9522 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9525 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9526 check for operands that need to go through aarch64_sve_reload_be.
9528 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9530 PR tree-optimization/81661
9531 PR tree-optimization/84117
9532 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9533 * tree-eh.c: Include gimplify.h.
9534 (find_trapping_overflow, replace_trapping_overflow,
9535 rewrite_to_non_trapping_overflow): New functions.
9536 * tree-vect-loop.c: Include tree-eh.h.
9537 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9538 * tree-data-ref.c: Include tree-eh.h.
9539 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9541 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9543 PR rtl-optimization/84123
9544 * combine.c (change_zero_ext): Check if hard register satisfies
9545 can_change_dest_mode before calling gen_lowpart_SUBREG.
9547 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9550 * ira.c (ira_init_register_move_cost): Remove assert.
9552 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9554 PR rtl-optimization/84071
9555 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9556 * doc/tm.texi: Regenerate.
9558 2018-01-31 Richard Biener <rguenther@suse.de>
9560 PR tree-optimization/84132
9561 * tree-data-ref.c (analyze_miv_subscript): Properly
9562 check whether evolution_function_is_affine_multivariate_p
9563 before calling gcd_of_steps_may_divide_p.
9565 2018-01-31 Julia Koval <julia.koval@intel.com>
9568 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9569 * config/i386/i386.md (rdpid_rex64) New.
9570 (rdpid): Make 32bit only.
9572 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9575 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9576 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9578 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9581 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9583 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9585 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9587 PR rtl-optimization/84071
9588 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9589 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9591 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9593 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9594 (arc_attribute_table): Add 'aux' attribute.
9595 (arc_in_small_data_p): Consider aux like variables.
9596 (arc_is_aux_reg_p): New function.
9597 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9598 (arc_get_aux_arg): New function.
9599 (prepare_move_operands): Handle aux-register access.
9600 (arc_handle_aux_attribute): New function.
9601 * doc/extend.texi (ARC Variable attributes): Add subsection.
9603 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9605 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9606 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9607 (arc_attribute_table): Add 'uncached' attribute.
9608 (arc_print_operand): Print '.di' flag for uncached memory
9610 (arc_in_small_data_p): Do not consider for small data the uncached
9612 (arc_is_uncached_mem_p): New function.
9613 * config/arc/predicates.md (compact_store_memory_operand): Check
9614 for uncached memory accesses.
9615 (nonvol_nonimm_operand): Likewise.
9616 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9618 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9621 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9622 falign-loops=): Add Optimization flag.
9624 2018-01-30 Jeff Law <law@redhat.com>
9627 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9628 INT_REGISTERS_SAVED. Check it prior to calling
9629 get_scratch_register_on_entry.
9630 (ix86_adjust_stack_and_probe): Similarly.
9631 (ix86_emit_probe_stack_range): Similarly.
9632 (ix86_expand_prologue): Corresponding changes.
9634 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9637 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9638 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9640 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9643 * lra-constraints.c (curr_insn_transform): Process AND in the
9646 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9648 PR rtl-optimization/83986
9649 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9650 dependence against last_pending_memory_flush in addition to
9653 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9655 PR tree-optimization/81611
9656 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9659 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9662 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9665 2018-01-30 Richard Biener <rguenther@suse.de>
9666 Jakub Jelinek <jakub@redhat.com>
9668 PR tree-optimization/84111
9669 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9670 inner loops added during recursion, as they don't have up-to-date
9673 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9676 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9677 (can_inline_edge_by_limits_p): ... here.
9678 (can_early_inline_edge_p, check_callers,
9679 update_caller_keys, update_callee_keys, recursive_inlining,
9680 add_new_edges_to_heap, speculation_useful_p,
9681 inline_small_functions,
9682 inline_small_functions, flatten_function,
9683 inline_to_all_callers_1): Update.
9685 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9687 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9690 2018-01-30 Richard Biener <rguenther@suse.de>
9692 PR tree-optimization/83008
9693 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9694 invariant and constant vector uses in stmts when they need
9697 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9700 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9701 * configure: Regenerate.
9703 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9705 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9707 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9708 Use gen_rtx_REG rather than gen_lowpart.
9710 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9712 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9713 rather than 0 when creating partial subregs.
9715 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9717 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9720 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9723 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9724 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9725 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9726 flags. This restores the settings used before the 2017-07-24.
9727 Turning off pre increment/decrement/modify allows IVOPTS to
9728 optimize DF/SF loops where the index is an int.
9730 2018-01-29 Richard Biener <rguenther@suse.de>
9731 Kelvin Nilsen <kelvin@gcc.gnu.org>
9734 * tree-vect-stmts.c (vectorizable_call): Don't call
9735 targetm.vectorize_builtin_md_vectorized_function if callee is
9738 2018-01-22 Carl Love <cel@us.ibm.com>
9740 * doc/extend.tex: Fix typo in second arg in
9741 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9743 2018-01-29 Richard Biener <rguenther@suse.de>
9745 PR tree-optimization/84086
9746 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9747 (flush_ssaname_freelist): When SSA names were released reset
9748 the SCEV hash table.
9750 2018-01-29 Richard Biener <rguenther@suse.de>
9752 PR tree-optimization/84057
9753 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9754 removed paths when removing edges.
9756 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9758 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9759 -mfunction-return=@var{choice}.
9761 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9764 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9766 (layout::print_source_line): Likewise.
9767 (test_get_line_width_without_trailing_whitespace): Add test cases.
9769 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9772 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9775 2018-01-26 Jim Wilson <jimw@sifive.com>
9777 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9779 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9782 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9784 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9785 and CMP + SUB-immediate -> SUBS.
9787 2018-01-26 Martin Sebor <msebor@redhat.com>
9789 PR tree-optimization/83896
9790 * tree-ssa-strlen.c (get_string_len): Rename...
9791 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9792 Avoid assuming length is constant.
9793 (handle_char_store): Use HOST_WIDE_INT for string length.
9795 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9798 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9799 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9801 2018-01-26 Richard Biener <rguenther@suse.de>
9803 PR rtl-optimization/84003
9804 * dse.c (record_store): Only record redundant stores when
9805 the earlier store aliases at least all accesses the later one does.
9807 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9809 PR rtl-optimization/83985
9810 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9811 REG_CFA_RESTORE insns.
9812 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9813 REG_CFA_RESTORE insns here.
9816 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9817 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9819 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9821 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9822 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9823 (arc_init): Likewise.
9824 (arc_override_options): Likewise.
9825 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9827 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9829 * config/arc/arc.h (TARGET_DBNZ): Define.
9830 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9831 properly set the tune attribute.
9832 (dbnz): Use TARGET_DBNZ guard.
9833 * config/arc/arc.opt (mtune): Add core3 option.
9835 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9837 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9838 recognize new pic like addresses.
9839 (arc_delegitimize_address): Clean up.
9841 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9843 * config/arc/arc-arches.def: Option mrf16 valid for all
9845 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9846 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9847 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9848 * config/arc/arc-tables.opt: Regenerate.
9849 * config/arc/arc.c (arc_conditional_register_usage): Handle
9850 reduced register file case.
9851 (arc_file_start): Set must have build attributes.
9852 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9854 * config/arc/arc.opt (mrf16): Add new option.
9855 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9856 * config/arc/genmultilib.awk: Handle new mrf16 option.
9857 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9858 * config/arc/t-multilib: Regenerate.
9859 * doc/invoke.texi (ARC Options): Document mrf16 option.
9861 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9863 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9864 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9865 (arc_attribute_table): Add 'secure_call' attribute.
9866 (arc_print_operand): Print secure call operand.
9867 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9869 (arc_is_secure_call_p): New function. * config/arc/arc.md
9870 (call_i): Add support for sjli instruction.
9871 (call_value_i): Likewise.
9872 * config/arc/constraints.md (Csc): New constraint.
9874 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9875 John Eric Martin <John.Martin@emmicro-us.com>
9877 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9878 * config/arc/arc.c (_arc_jli_section): New struct.
9879 (arc_jli_section): New type.
9880 (rc_jli_sections): New static variable.
9881 (arc_handle_jli_attribute): New function.
9882 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9883 (arc_file_end): New function.
9884 (TARGET_ASM_FILE_END): Define.
9885 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9886 (arc_add_jli_section): New function.
9887 (jli_call_scan): Likewise.
9888 (arc_reorg): Call jli_call_scan.
9889 (arc_output_addsi): Remove 'S' from printing asm operand.
9890 (arc_is_jli_call_p): New function.
9891 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9893 (movhi_insn): Likewise.
9894 (movsi_insn): Likewise.
9895 (movsi_set_cc_insn): Likewise.
9896 (loadqi_update): Likewise.
9897 (load_zeroextendqisi_update): Likewise.
9898 (load_signextendqisi_update): Likewise.
9899 (loadhi_update): Likewise.
9900 (load_zeroextendhisi_update): Likewise.
9901 (load_signextendhisi_update): Likewise.
9902 (loadsi_update): Likewise.
9903 (loadsf_update): Likewise.
9904 (movsicc_insn): Likewise.
9905 (bset_insn): Likewise.
9906 (bxor_insn): Likewise.
9907 (bclr_insn): Likewise.
9908 (bmsk_insn): Likewise.
9909 (bicsi3_insn): Likewise.
9910 (cmpsi_cc_c_insn): Likewise.
9911 (movsi_ne): Likewise.
9912 (movsi_cond_exec): Likewise.
9913 (clrsbsi2): Likewise.
9924 (call_i): Remove 'S' asm letter, add jli instruction.
9925 (call_value_i): Likewise.
9926 * config/arc/arc.op (mjli-always): New option.
9927 * config/arc/constraints.md (Cji): New constraint.
9928 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9930 (subsf3_fpx): Likewise.
9931 (mulsf3_fpx): Likewise.
9932 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9934 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9936 * doc/invoke.texi (ARC): Document mjli-always option.
9938 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9940 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9941 avoid addition with 0 and use incw and decw where possible.
9943 2018-01-26 Richard Biener <rguenther@suse.de>
9945 PR tree-optimization/81082
9946 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9947 association if it requires casting to unsigned.
9948 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9949 from fold_plusminus_mult_expr to catch important cases late when
9950 range info is available.
9952 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9954 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9955 * configure.ac (hidden_linkonce): New test.
9956 * configure: Regenerate.
9957 * config.in: Regenerate.
9959 2018-01-26 Julia Koval <julia.koval@intel.com>
9961 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9962 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9963 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9964 _mm_mask_bitshuffle_epi64_mask): Fix type.
9965 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9966 USI_FTYPE_V4DI_V4DI_USI): Remove.
9967 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9968 __builtin_ia32_vpshufbitqmb256_mask,
9969 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9970 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9971 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9973 2018-01-26 Alan Modra <amodra@gmail.com>
9976 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9977 UNSPEC_VBPERMQ. Sort other unspecs.
9979 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9981 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9983 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9986 * predict.c (drop_profile): Do not push/pop cfun; update also
9988 (handle_missing_profiles): Fix logic looking for zero profiles.
9990 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9993 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9994 on functions with #pragma omp declare simd or functions with simd
9996 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9997 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9998 Remove trailing \n from warning_at calls.
10000 2018-01-25 Tom de Vries <tom@codesourcery.com>
10003 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10004 for neutered workers.
10006 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10009 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10010 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10012 2018-01-24 Jeff Law <law@redhat.com>
10015 * i386.c (get_probe_interval): Move to earlier point.
10016 (ix86_compute_frame_layout): If -fstack-clash-protection and
10017 the frame is larger than the probe interval, then use pushes
10018 to save registers rather than reg->mem moves.
10019 (ix86_expand_prologue): Remove conditional for int_registers_saved
10022 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10025 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10026 min/max for never referenced object.
10028 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10030 PR middle-end/83977
10031 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10033 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10034 attributes from DECL_ATTRIBUTES (decl) without affecting
10035 DECL_ATTRIBUTES (current_function_decl).
10036 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10037 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10039 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10041 PR tree-optimization/83979
10042 * fold-const.c (fold_comparison): Use constant_boolean_node
10043 instead of boolean_{true,false}_node.
10045 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10047 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10050 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10052 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10053 Simplify the clause that sets the length attribute.
10054 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10055 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10056 clause that sets the length attribute.
10057 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10059 2018-01-24 Tom de Vries <tom@codesourcery.com>
10062 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10063 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10064 Add strict parameter.
10065 (prevent_branch_around_nothing): Insert dummy insn between branch to
10066 label and label with no ptx insn inbetween.
10067 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10069 2018-01-24 Tom de Vries <tom@codesourcery.com>
10072 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10073 for neutered threads in warp.
10074 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10076 2018-01-24 Richard Biener <rguenther@suse.de>
10078 PR tree-optimization/83176
10079 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10082 2018-01-24 Richard Biener <rguenther@suse.de>
10084 PR tree-optimization/82819
10085 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10086 code generating pluses that are no-ops in the target precision.
10088 2018-01-24 Richard Biener <rguenther@suse.de>
10090 PR middle-end/84000
10091 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10093 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10095 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10096 to merge probabilities.
10097 * predict.c (probably_never_executed): Also mark as cold functions
10098 with global 0 profile and guessed local profile.
10099 * profile-count.c (profile_probability::combine_with_count): New
10101 * profile-count.h (profile_probability::operator*,
10102 profile_probability::operator*=, profile_probability::operator/,
10103 profile_probability::operator/=): Reduce precision to adjusted
10104 and set value to guessed on contradictory divisions.
10105 (profile_probability::combine_with_freq): Remove.
10106 (profile_probability::combine_wiht_count): Declare.
10107 (profile_count::force_nonzero):: Set to adjusted.
10108 (profile_count::probability_in):: Set quality to adjusted.
10109 * tree-ssa-tail-merge.c (replace_block_by): Use
10110 combine_with_count.
10112 2018-01-23 Andrew Waterman <andrew@sifive.com>
10113 Jim Wilson <jimw@sifive.com>
10115 * config/riscv/riscv.c (riscv_stack_boundary): New.
10116 (riscv_option_override): Set riscv_stack_boundary. Handle
10117 riscv_preferred_stack_boundary_arg.
10118 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10119 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10120 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10121 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10122 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10123 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10125 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10128 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10129 of struct ix86_frame.
10130 (ix86_expand_epilogue): Likewise. Add a local variable for
10131 the reg_save_offset field in struct ix86_frame.
10133 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10135 PR tree-optimization/82604
10136 * tree-loop-distribution.c (enum partition_kind): New enum item
10137 PKIND_PARTIAL_MEMSET.
10138 (partition_builtin_p): Support above new enum item.
10139 (generate_code_for_partition): Ditto.
10140 (compute_access_range): Differentiate cases that equality can be
10141 proven at all loops, the innermost loops or no loops.
10142 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10143 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10144 (finalize_partitions, distribute_loop): Don't fuse partition of
10145 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10146 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10147 parloop is enabled.
10149 2018-01-23 Martin Liska <mliska@suse.cz>
10151 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10152 order to ignore the predictor.
10153 (PRED_POLYMORPHIC_CALL): Likewise.
10154 (PRED_RECURSIVE_CALL): Likewise.
10156 2018-01-23 Martin Liska <mliska@suse.cz>
10158 * tree-profile.c (tree_profiling): Print function header to
10159 aware reader which function we are working on.
10160 * value-prof.c (gimple_find_values_to_profile): Do not print
10161 not interesting value histograms.
10163 2018-01-23 Martin Liska <mliska@suse.cz>
10165 * profile-count.h (enum profile_quality): Add
10166 profile_uninitialized as the first value. Do not number values
10167 as they are zero based.
10168 (profile_count::verify): Update sanity check.
10169 (profile_probability::verify): Likewise.
10171 2018-01-23 Nathan Sidwell <nathan@acm.org>
10173 * doc/invoke.texi (ffor-scope): Deprecate.
10175 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10177 PR tree-optimization/83510
10178 * domwalk.c (set_all_edges_as_executable): New function.
10179 (dom_walker::dom_walker): Convert bool param
10180 "skip_unreachable_blocks" to enum reachability. Move setup of
10181 edge flags to set_all_edges_as_executable and only do it when
10182 reachability is REACHABLE_BLOCKS.
10183 * domwalk.h (enum dom_walker::reachability): New enum.
10184 (dom_walker::dom_walker): Convert bool param
10185 "skip_unreachable_blocks" to enum reachability.
10186 (set_all_edges_as_executable): New decl.
10187 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10188 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10190 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10191 but converting true to REACHABLE_BLOCKS.
10192 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10194 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10195 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10196 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10198 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10199 if check_all_array_refs will be called.
10201 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10203 * tree.c (selftest::test_location_wrappers): Add more test
10206 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10208 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10209 (selftest::test_bit_in_range): Likewise.
10211 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10214 * doc/sourcebuild.texi (vect_float): Say that the selector
10215 only describes the situation when -funsafe-math-optimizations is on.
10216 (vect_float_strict): Document.
10218 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10220 PR tree-optimization/83965
10221 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10222 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10223 instead of checking only for a reduction.
10224 (vect_recog_widen_sum_pattern): Likewise.
10226 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10228 * predict.c (probably_never_executed): Only use precise profile info.
10229 (compute_function_frequency): Skip after inlining hack since we now
10230 have quality checking.
10232 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10234 * profile-count.h (profile_probability::very_unlikely,
10235 profile_probability::unlikely, profile_probability::even): Set
10236 precision to guessed.
10238 2018-01-23 Richard Biener <rguenther@suse.de>
10240 PR tree-optimization/83963
10241 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10242 Properly terminate dominator walk when crossing the exit edge not
10243 when visiting its source block.
10245 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10248 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10249 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10251 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10253 PR tree-optimization/83957
10254 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10255 semicolon after for body surrounded by braces.
10257 PR tree-optimization/83081
10258 * profile-count.h (profile_probability::split): New method.
10259 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10260 Use profile_probability::split.
10261 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10262 when splitting a single conditional jump into 2.
10264 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10266 PR tree-optimization/69452
10267 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10270 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10272 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10273 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10274 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10276 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10278 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10280 * config/rl78/rl78.md (movdi): New define_expand.
10281 * config/rl78/rl78.c (rl78_split_movdi): New function.
10283 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10286 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10288 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10289 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10290 128-bit to produce an UNSPEC move to get the double word with the
10291 signbit and then a shift directly to do signbit.
10292 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10293 implementation with a new version that just does either a direct
10294 move or a regular move. Move memory interface to separate insns.
10295 Move insns so they are next to the expander.
10296 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10297 with signbit move. Split big and little endian case.
10298 (signbit<mode>2_dm_mem_le): Likewise.
10299 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10300 (signbit<mode>2_dm2): Likewise.
10302 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10304 * config/rl78/rl78.md (anddi3): New define_expand.
10306 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10308 * config/rl78/rl78.md (umindi3): New define_expand.
10310 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10312 * config/rl78/rl78.md (smindi3): New define_expand.
10314 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10316 * config/rl78/rl78.md (smaxdi3): New define_expand.
10318 2018-01-22 Carl Love <cel@us.ibm.com>
10320 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10321 LVX_V1TI): Add macro expansion.
10322 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10323 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10324 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10325 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10326 Change check to determine if the instruction is a byte reversing
10327 entry. Fix typo in comment.
10328 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10329 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10330 Add def_builtin calls for new builtins.
10331 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10332 Add define_insn expansion.
10334 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10336 * config/rl78/rl78.md (umaxdi3): New define_expand.
10338 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10340 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10341 for non-QImode registers.
10343 2018-01-22 Richard Biener <rguenther@suse.de>
10345 PR tree-optimization/83963
10346 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10347 including the loop exit block.
10348 (scop_detection::merge_sese): Likewise.
10349 (scop_detection::add_scop): Do it here instead.
10351 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10353 * doc/sourcebuild.texi (arm_softfloat): Document.
10355 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10358 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10359 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10360 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10362 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10363 David Edelsohn <dje.gcc@gmail.com>
10366 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10367 Change "crset eq" to "crset 2".
10368 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10369 (*call_indirect_aix<mode>_nospec): Likewise.
10370 (*call_value_indirect_aix<mode>_nospec): Likewise.
10371 (*call_indirect_elfv2<mode>_nospec): Likewise.
10372 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10373 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10374 change assembly output from . to $.
10375 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10376 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10377 (*tablejump<mode>_internal1_nospec): Likewise.
10379 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10382 * config/sh/sh_optimize_sett_clrt.cc:
10383 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10385 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10387 PR tree-optimization/83940
10388 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10389 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10390 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10391 use it to pass back the definition type.
10392 (vect_check_store_rhs): Likewise rhs_dt_out.
10393 (vect_build_gather_load_calls): Add a mask_dt argument and use
10394 it instead of a call to vect_is_simple_use.
10395 (vectorizable_store): Update calls to vect_check_load_store_mask
10396 and vect_check_store_rhs. Use the dt returned by the latter instead
10397 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10398 instead of calls to vect_is_simple_use. Pass the scalar rather
10399 than the vector operand to vect_is_simple_use when handling
10400 second and subsequent copies of an rhs value.
10401 (vectorizable_load): Update calls to vect_check_load_store_mask
10402 and vect_build_gather_load_calls. Use the cached mask_dt and
10403 gs_info.offset_dt instead of calls to vect_is_simple_use.
10405 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10407 PR middle-end/83945
10408 * tree-emutls.c: Include gimplify.h.
10409 (lower_emutls_2): New function.
10410 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10411 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10412 it before further processing.
10415 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10416 UINTVAL (trueop1) instead of INTVAL (op1).
10418 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10422 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10423 INCOMING_FRAME_SP_OFFSET if not defined.
10424 (scan_trace): Add ENTRY argument. If true and
10425 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10426 emit a note to adjust the CFA offset.
10427 (create_cfi_notes): Adjust scan_trace callers.
10428 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10429 INCOMING_FRAME_SP_OFFSET in the CIE.
10430 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10431 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10433 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10434 * doc/tm.texi: Regenerated.
10436 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10438 PR rtl-optimization/83147
10439 * lra-constraints.c (remove_inheritance_pseudos): Use
10440 lra_substitute_pseudo_within_insn.
10442 2018-01-19 Tom de Vries <tom@codesourcery.com>
10443 Cesar Philippidis <cesar@codesourcery.com>
10446 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10448 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10451 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10452 spaces for function labels.
10454 2018-01-19 Martin Liska <mliska@suse.cz>
10456 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10457 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10458 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10459 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10460 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10461 (PRED_CONST_RETURN): Change from 69 to 65.
10462 (PRED_NULL_RETURN): Change from 91 to 71.
10463 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10464 (PRED_LOOP_GUARD): Change from 66 to 73.
10466 2018-01-19 Martin Liska <mliska@suse.cz>
10468 * predict.c (predict_insn_def): Add new assert.
10469 (struct branch_predictor): Change type to signed integer.
10470 (test_prediction_value_range): Amend test to cover
10471 PROB_UNINITIALIZED.
10472 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10473 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10474 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10475 (PRED_LOOP_IV_COMPARE): Likewise.
10476 * predict.h (PROB_UNINITIALIZED): Define new constant.
10478 2018-01-19 Martin Liska <mliska@suse.cz>
10480 * predict.c (dump_prediction): Add new format for
10481 analyze_brprob.py script which is enabled with -details
10483 * profile-count.h (precise_p): New function.
10485 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10487 PR tree-optimization/83922
10488 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10489 there are no statements that need masking.
10490 (vect_active_double_reduction_p): New function.
10491 (vect_analyze_loop_operations): Use it when handling phis that
10492 are not in the loop header.
10494 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10496 PR tree-optimization/83914
10497 * tree-vect-loop.c (vectorizable_induction): Don't convert
10498 init_expr or apply the peeling adjustment for inductions
10499 that are nested within the vectorized loop.
10501 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10503 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10506 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10510 * function.h (gimplify_parameters): Add gimple_seq * argument.
10511 * function.c: Include gimple.h and options.h.
10512 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10513 for the added local temporaries if needed.
10514 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10515 if there are any parameter cleanups, wrap whole body into a
10516 try/finally with the cleanups.
10518 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10521 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10522 Use GET_MODE_CLASS for scalar floating point.
10524 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10528 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10529 Fix call of call_cgraph_insertion_hooks.
10531 2018-01-18 Martin Sebor <msebor@redhat.com>
10533 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10535 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10538 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10541 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10544 * common.opt: (-ffile-prefix-map): New option.
10545 * opts.c (common_handle_option): Defer it.
10546 * opts-global.c (handle_common_deferred_options): Handle it.
10547 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10548 * file-prefix-map.h: New file.
10549 (remap_debug_filename, add_debug_prefix_map): ...here.
10550 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10551 * final.c (debug_prefix_map, add_debug_prefix_map
10552 remap_debug_filename): Move to...
10553 * file-prefix-map.c: New file.
10554 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10555 generalize, get rid of alloca(), use strrchr() instead of strchr().
10556 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10557 Implement in terms of add_prefix_map().
10558 (remap_macro_filename, remap_debug_filename): Implement in term of
10560 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10561 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10562 * dbxout.c: Include file-prefix-map.h.
10563 * varasm.c: Likewise.
10564 * vmsdbgout.c: Likewise.
10565 * xcoffout.c: Likewise.
10566 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10567 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10568 * doc/invoke.texi (-ffile-prefix-map): Document.
10569 (-fdebug-prefix-map): Update description.
10571 2018-01-18 Martin Liska <mliska@suse.cz>
10573 * config/i386/i386.c (indirect_thunk_name): Document that also
10575 (output_indirect_thunk): Document why both instructions
10576 (pause and lfence) are generated.
10578 2018-01-18 Richard Biener <rguenther@suse.de>
10580 PR tree-optimization/83887
10581 * graphite-scop-detection.c
10582 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10583 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10584 (scop_detection::merge_sese): Re-implement with a flood-fill
10585 algorithm that properly finds a SESE region if it exists.
10587 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10590 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10591 pointer_diff optimizations use view_convert instead of convert.
10593 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10595 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10596 Generate different code for -mno-speculate-indirect-jumps.
10597 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10598 (*call_indirect_aix<mode>): Disable for
10599 -mno-speculate-indirect-jumps.
10600 (*call_indirect_aix<mode>_nospec): New define_insn.
10601 (*call_value_indirect_aix<mode>): Disable for
10602 -mno-speculate-indirect-jumps.
10603 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10604 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10605 -mno-speculate-indirect-jumps.
10606 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10608 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10610 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10611 long double type, set the flags for noting the default long double
10612 type, even if we don't pass or return a long double type.
10614 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10617 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10620 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10622 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10623 support for merge[hl].
10624 (fold_mergehl_helper): New helper function.
10625 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10626 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10627 (altivec_vmrglw_direct): Add xxmrglw insn.
10629 2018-01-17 Andrew Waterman <andrew@sifive.com>
10631 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10632 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10634 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10637 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10638 call the lto_location_cache before reading the
10639 DECL_SOURCE_LOCATION of the types.
10641 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10642 Richard Sandiford <richard.sandiford@linaro.org>
10644 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10645 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10646 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10647 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10648 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10650 * config/aarch64/constraints.md (aarch64_movti_operand):
10652 * config/aarch64/predicates.md (Uti): Add new constraint.
10654 2018-01-17 Carl Love <cel@us.ibm.com>
10656 * config/rs6000/vsx.md (define_expand xl_len_r,
10657 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10658 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10660 (define_expand, define_insn): Move the shift left from the
10661 define_insn to the define_expand for lxvl and stxvl instructions.
10662 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10663 and XL_LEN_R definitions to PURE.
10665 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10667 * config/i386/i386.c (indirect_thunk_name): Declare regno
10668 as unsigned int. Compare regno with INVALID_REGNUM.
10669 (output_indirect_thunk): Ditto.
10670 (output_indirect_thunk_function): Ditto.
10671 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10672 in the call to output_indirect_thunk_function.
10674 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10676 PR middle-end/83884
10677 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10678 rather than the size of inner_type to determine the stack slot size
10679 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10681 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10684 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10687 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10689 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10690 endian Linux systems to optionally enable multilibs for selecting
10691 the long double type if the user configured an explicit type.
10692 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10693 have no long double multilibs if not defined.
10694 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10695 warn if the user used -mabi={ieee,ibm}longdouble and we built
10696 multilibs for long double.
10697 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10698 appropriate multilib option.
10699 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10701 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10702 for building long double multilibs.
10703 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10705 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10707 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10710 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10712 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10715 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10718 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10721 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10723 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10724 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10726 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10728 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10729 different rtl trees depending on TARGET_64BIT.
10730 (rs6000_gen_lvx): Likewise.
10732 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10734 * config/visium/visium.md (nop): Tweak comment.
10735 (hazard_nop): Likewise.
10737 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10739 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10740 -mspeculate-indirect-jumps.
10741 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10742 for -mno-speculate-indirect-jumps.
10743 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10744 (*call_value_indirect_elfv2<mode>): Disable for
10745 -mno-speculate-indirect-jumps.
10746 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10747 (indirect_jump): Emit different RTL for
10748 -mno-speculate-indirect-jumps.
10749 (*indirect_jump<mode>): Disable for
10750 -mno-speculate-indirect-jumps.
10751 (*indirect_jump<mode>_nospec): New define_insn.
10752 (tablejump): Emit different RTL for
10753 -mno-speculate-indirect-jumps.
10754 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10755 (tablejumpsi_nospec): New define_expand.
10756 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10757 (tablejumpdi_nospec): New define_expand.
10758 (*tablejump<mode>_internal1): Disable for
10759 -mno-speculate-indirect-jumps.
10760 (*tablejump<mode>_internal1_nospec): New define_insn.
10761 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10764 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10766 * caller-save.c (insert_save): Drop unnecessary parameter. All
10769 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10770 Richard Biener <rguenth@suse.de>
10773 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10774 return early, inline manually is_gimple_sizepos. Make sure if we
10775 call gimplify_expr we don't end up with a gimple constant.
10776 * tree.c (variably_modified_type_p): Don't return true for
10777 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10778 * gimplify.h (is_gimple_sizepos): Remove.
10780 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10782 PR tree-optimization/83857
10783 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10784 vectorizable_live_operation for pure SLP statements.
10785 (vectorizable_live_operation): Handle PHIs.
10787 2018-01-16 Richard Biener <rguenther@suse.de>
10789 PR tree-optimization/83867
10790 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10791 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10793 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10796 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10797 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10798 If off is not INTEGER_CST, issue a may not be aligned warning
10799 rather than isn't aligned. Use isn%'t rather than isn't.
10800 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10802 <case MULT_EXPR>: Improve the case when bottom and one of the
10803 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10804 operand, in that case check if the other operand is multiple of
10805 bottom divided by the INTEGER_CST operand.
10807 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10810 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10811 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10812 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10813 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10814 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10815 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10816 (pa_function_arg_size): New function.
10818 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10820 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10821 in a separate statement.
10823 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10825 PR tree-optimization/83847
10826 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10827 group gathers and scatters.
10829 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10831 PR rtl-optimization/86620
10832 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10834 PR rtl-optimization/83213
10835 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10836 to last if both are JUMP_INSNs.
10838 PR tree-optimization/83843
10839 * gimple-ssa-store-merging.c
10840 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10841 store_immediate_info for bswap/nop orig_stores.
10843 2018-01-15 Andrew Waterman <andrew@sifive.com>
10845 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10847 <UDIV>: Increase cost if !TARGET_DIV.
10849 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10851 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10852 (define_attr "cr_logical_3op"): New.
10853 (cceq_ior_compare): Adjust.
10854 (cceq_ior_compare_complement): Adjust.
10855 (*cceq_rev_compare): Adjust.
10856 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10857 (is_cracked_insn): Adjust.
10858 (insn_must_be_first_in_group): Adjust.
10859 * config/rs6000/40x.md: Adjust.
10860 * config/rs6000/440.md: Adjust.
10861 * config/rs6000/476.md: Adjust.
10862 * config/rs6000/601.md: Adjust.
10863 * config/rs6000/603.md: Adjust.
10864 * config/rs6000/6xx.md: Adjust.
10865 * config/rs6000/7450.md: Adjust.
10866 * config/rs6000/7xx.md: Adjust.
10867 * config/rs6000/8540.md: Adjust.
10868 * config/rs6000/cell.md: Adjust.
10869 * config/rs6000/e300c2c3.md: Adjust.
10870 * config/rs6000/e500mc.md: Adjust.
10871 * config/rs6000/e500mc64.md: Adjust.
10872 * config/rs6000/e5500.md: Adjust.
10873 * config/rs6000/e6500.md: Adjust.
10874 * config/rs6000/mpc.md: Adjust.
10875 * config/rs6000/power4.md: Adjust.
10876 * config/rs6000/power5.md: Adjust.
10877 * config/rs6000/power6.md: Adjust.
10878 * config/rs6000/power7.md: Adjust.
10879 * config/rs6000/power8.md: Adjust.
10880 * config/rs6000/power9.md: Adjust.
10881 * config/rs6000/rs64.md: Adjust.
10882 * config/rs6000/titan.md: Adjust.
10884 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10886 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10887 ix86_indirect_branch_register logic.
10889 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10891 * config/i386/constraints.md (Bs): Update
10892 ix86_indirect_branch_register check. Don't check
10893 ix86_indirect_branch_register with GOT_memory_operand.
10895 * config/i386/predicates.md (GOT_memory_operand): Don't check
10896 ix86_indirect_branch_register here.
10897 (GOT32_symbol_operand): Likewise.
10899 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10901 * config/i386/predicates.md (constant_call_address_operand):
10902 Rewrite ix86_indirect_branch_register logic.
10903 (sibcall_insn_operand): Likewise.
10905 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10907 * config/i386/constraints.md (Bs): Replace
10908 ix86_indirect_branch_thunk_register with
10909 ix86_indirect_branch_register.
10911 * config/i386/i386.md (indirect_jump): Likewise.
10912 (tablejump): Likewise.
10913 (*sibcall_memory): Likewise.
10914 (*sibcall_value_memory): Likewise.
10915 Peepholes of indirect call and jump via memory: Likewise.
10916 * config/i386/i386.opt: Likewise.
10917 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10918 (GOT_memory_operand): Likewise.
10919 (call_insn_operand): Likewise.
10920 (sibcall_insn_operand): Likewise.
10921 (GOT32_symbol_operand): Likewise.
10923 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10925 PR middle-end/83837
10926 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10927 type rather than type addr's type points to.
10928 (expand_omp_atomic_mutex): Likewise.
10929 (expand_omp_atomic): Likewise.
10931 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10934 * config/i386/i386.c (output_indirect_thunk_function): Use
10935 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10936 for __x86_return_thunk.
10938 2018-01-15 Richard Biener <rguenther@suse.de>
10940 PR middle-end/83850
10941 * expmed.c (extract_bit_field_1): Fix typo.
10943 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10946 * config/arm/iterators.md (VF): New mode iterator.
10947 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10948 Remove integer-related logic from pattern.
10949 (neon_vabd<mode>_3): Likewise.
10951 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10953 PR middle-end/82694
10954 * common.opt (fstrict-overflow): No longer an alias.
10955 (fwrapv-pointer): New option.
10956 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10957 also for pointer types based on flag_wrapv_pointer.
10958 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10959 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10960 opts->x_flag_wrapv got set.
10961 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10962 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10963 POINTER_TYPE_OVERFLOW_UNDEFINED.
10964 * match.pd: Likewise in address comparison pattern.
10965 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10967 2018-01-15 Richard Biener <rguenther@suse.de>
10970 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10971 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10972 Reset type names to their identifier if their TYPE_DECL doesn't
10973 have linkage (and thus is used for ODR and devirt).
10974 (save_debug_info_for_decl): Remove.
10975 (save_debug_info_for_type): Likewise.
10976 (add_tree_to_fld_list): Adjust.
10977 * tree-pretty-print.c (dump_generic_node): Make dumping of
10978 type names more robust.
10980 2018-01-15 Richard Biener <rguenther@suse.de>
10982 * BASE-VER: Bump to 8.0.1.
10984 2018-01-14 Martin Sebor <msebor@redhat.com>
10987 * builtins.c (check_access): Avoid warning when the no-warning bit
10990 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10992 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10993 * ira-color (allocno_hard_regs_compare): Likewise.
10995 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10998 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10999 Use .pushsection/.popsection.
11001 2018-01-14 Martin Sebor <msebor@redhat.com>
11004 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11006 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11008 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11009 entry from extra_headers.
11010 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11011 extra_headers, make the list bitwise identical to the i?86-*-* one.
11013 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11015 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11016 -mcmodel=large with -mindirect-branch=thunk,
11017 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11018 -mfunction-return=thunk-extern.
11019 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11020 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11021 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11023 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11025 * config/i386/i386.c (print_reg): Print the name of the full
11026 integer register without '%'.
11027 (ix86_print_operand): Handle 'V'.
11028 * doc/extend.texi: Document 'V' modifier.
11030 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11032 * config/i386/constraints.md (Bs): Disallow memory operand for
11033 -mindirect-branch-register.
11035 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11036 (GOT_memory_operand): Likewise.
11037 (call_insn_operand): Likewise.
11038 (sibcall_insn_operand): Likewise.
11039 (GOT32_symbol_operand): Likewise.
11040 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11041 for -mindirect-branch-register.
11042 (tablejump): Likewise.
11043 (*sibcall_memory): Likewise.
11044 (*sibcall_value_memory): Likewise.
11045 Disallow peepholes of indirect call and jump via memory for
11046 -mindirect-branch-register.
11047 (*call_pop): Replace m with Bw.
11048 (*call_value_pop): Likewise.
11049 (*sibcall_pop_memory): Replace m with Bs.
11050 * config/i386/i386.opt (mindirect-branch-register): New option.
11051 * doc/invoke.texi: Document -mindirect-branch-register option.
11053 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11055 * config/i386/i386-protos.h (ix86_output_function_return): New.
11056 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11057 set function_return_type.
11058 (indirect_thunk_name): Add ret_p to indicate thunk for function
11060 (output_indirect_thunk_function): Pass false to
11061 indirect_thunk_name.
11062 (ix86_output_indirect_branch_via_reg): Likewise.
11063 (ix86_output_indirect_branch_via_push): Likewise.
11064 (output_indirect_thunk_function): Create alias for function
11065 return thunk if regno < 0.
11066 (ix86_output_function_return): New function.
11067 (ix86_handle_fndecl_attribute): Handle function_return.
11068 (ix86_attribute_table): Add function_return.
11069 * config/i386/i386.h (machine_function): Add
11070 function_return_type.
11071 * config/i386/i386.md (simple_return_internal): Use
11072 ix86_output_function_return.
11073 (simple_return_internal_long): Likewise.
11074 * config/i386/i386.opt (mfunction-return=): New option.
11075 (indirect_branch): Mention -mfunction-return=.
11076 * doc/extend.texi: Document function_return function attribute.
11077 * doc/invoke.texi: Document -mfunction-return= option.
11079 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11081 * config/i386/i386-opts.h (indirect_branch): New.
11082 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11083 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11084 with local indirect jump when converting indirect call and jump.
11085 (ix86_set_indirect_branch_type): New.
11086 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11087 (indirectlabelno): New.
11088 (indirect_thunk_needed): Likewise.
11089 (indirect_thunk_bnd_needed): Likewise.
11090 (indirect_thunks_used): Likewise.
11091 (indirect_thunks_bnd_used): Likewise.
11092 (INDIRECT_LABEL): Likewise.
11093 (indirect_thunk_name): Likewise.
11094 (output_indirect_thunk): Likewise.
11095 (output_indirect_thunk_function): Likewise.
11096 (ix86_output_indirect_branch_via_reg): Likewise.
11097 (ix86_output_indirect_branch_via_push): Likewise.
11098 (ix86_output_indirect_branch): Likewise.
11099 (ix86_output_indirect_jmp): Likewise.
11100 (ix86_code_end): Call output_indirect_thunk_function if needed.
11101 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11103 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11104 (ix86_attribute_table): Add indirect_branch.
11105 * config/i386/i386.h (machine_function): Add indirect_branch_type
11106 and has_local_indirect_jump.
11107 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11109 (tablejump): Likewise.
11110 (*indirect_jump): Use ix86_output_indirect_jmp.
11111 (*tablejump_1): Likewise.
11112 (simple_return_indirect_internal): Likewise.
11113 * config/i386/i386.opt (mindirect-branch=): New option.
11114 (indirect_branch): New.
11117 (thunk-inline): Likewise.
11118 (thunk-extern): Likewise.
11119 * doc/extend.texi: Document indirect_branch function attribute.
11120 * doc/invoke.texi: Document -mindirect-branch= option.
11122 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11125 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11127 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11129 * ipa-inline.c (want_inline_small_function_p): Return false if
11130 inlining has already failed with CIF_FINAL_ERROR.
11131 (update_caller_keys): Call want_inline_small_function_p before
11133 (update_callee_keys): Likewise.
11135 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11137 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11139 (rs6000_quadword_masked_address_p): Likewise.
11140 (quad_aligned_load_p): Likewise.
11141 (quad_aligned_store_p): Likewise.
11142 (const_load_sequence_p): Add comment to describe the outer-most loop.
11143 (mimic_memory_attributes_and_flags): New function.
11144 (rs6000_gen_stvx): Likewise.
11145 (replace_swapped_aligned_store): Likewise.
11146 (rs6000_gen_lvx): Likewise.
11147 (replace_swapped_aligned_load): Likewise.
11148 (replace_swapped_load_constant): Capitalize argument name in
11149 comment describing this function.
11150 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11151 and stores that access quad-word aligned addresses and replace
11152 with stvx or lvx instructions when appropriate.
11153 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11154 New function prototype.
11155 (rs6000_quadword_masked_address_p): Likewise.
11156 (rs6000_gen_lvx): Likewise.
11157 (rs6000_gen_stvx): Likewise.
11158 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11159 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11160 when memory address is aligned.
11161 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11162 this split to select lvx instruction when memory address is aligned.
11163 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11164 instruction when memory address is aligned.
11165 (*vsx_le_perm_load_v16qi): Likewise.
11166 (four unnamed splitters): Modify to select the stvx instruction
11167 when memory is aligned.
11169 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11171 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11172 which appears in the queue multiple times.
11174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11175 Alan Hayward <alan.hayward@arm.com>
11176 David Sherwood <david.sherwood@arm.com>
11178 * tree-vectorizer.h (vec_lower_bound): New structure.
11179 (_loop_vec_info): Add check_nonzero and lower_bounds.
11180 (LOOP_VINFO_CHECK_NONZERO): New macro.
11181 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11182 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11183 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11184 fields. Make seg_len the distance travelled, not including the
11186 (dr_direction_indicator): Declare.
11187 (dr_zero_step_indicator): Likewise.
11188 (dr_known_forward_stride_p): Likewise.
11189 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11191 (runtime_alias_check_p): Allow runtime alias checks with
11193 (operator ==): Compare access_size and align.
11194 (prune_runtime_alias_test_list): Rework for new distinction between
11195 the access_size and seg_len.
11196 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11198 (get_segment_min_max): New function.
11199 (create_intersect_range_checks): Use it.
11200 (dr_step_indicator): New function.
11201 (dr_direction_indicator): Likewise.
11202 (dr_zero_step_indicator): Likewise.
11203 (dr_known_forward_stride_p): Likewise.
11204 * tree-loop-distribution.c (data_ref_segment_size): Return
11205 DR_STEP * (niters - 1).
11206 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11208 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11209 (vect_preserves_scalar_order_p): New function, split out from...
11210 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11211 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11212 (vect_vfa_access_size): New function.
11213 (vect_vfa_align): Likewise.
11214 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11215 (dump_lower_bound): New function.
11216 (vect_check_lower_bound): Likewise.
11217 (vect_small_gap_p): Likewise.
11218 (vectorizable_with_step_bound_p): Likewise.
11219 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11220 depencies if the vectorization factor is 1. Convert the checks
11221 for nonzero steps into checks on the bounds of DR_STEP. Try using
11222 a bunds check for variable steps if the minimum required step is
11223 relatively small. Update calls to the dr_with_seg_len
11224 constructor and to vect_compile_time_alias.
11225 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11227 (vect_loop_versioning): Call it.
11228 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11230 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11233 Alan Hayward <alan.hayward@arm.com>
11234 David Sherwood <david.sherwood@arm.com>
11236 * doc/sourcebuild.texi (vect_scatter_store): Document.
11237 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11239 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11241 * genopinit.c (main): Add supports_vec_scatter_store and
11242 supports_vec_scatter_store_cached to target_optabs.
11243 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11244 IFN_MASK_SCATTER_STORE.
11245 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11247 * internal-fn.h (internal_store_fn_p): Declare.
11248 (internal_fn_stored_value_index): Likewise.
11249 * internal-fn.c (scatter_store_direct): New macro.
11250 (expand_scatter_store_optab_fn): New function.
11251 (direct_scatter_store_optab_supported_p): New macro.
11252 (internal_store_fn_p): New function.
11253 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11254 IFN_MASK_SCATTER_STORE.
11255 (internal_fn_mask_index): Likewise.
11256 (internal_fn_stored_value_index): New function.
11257 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11258 for scatter stores.
11259 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11260 * optabs-query.c (supports_vec_scatter_store_p): New function.
11261 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11262 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11263 true for scatter stores.
11264 (vect_gather_scatter_fn_p): Handle scatter stores too.
11265 (vect_check_gather_scatter): Consider using scatter stores if
11266 supports_vec_scatter_store_p.
11267 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11268 scatter stores too.
11269 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11270 internal_fn_stored_value_index.
11271 (check_load_store_masking): Handle scatter stores too.
11272 (vect_get_store_rhs): Make public.
11273 (vectorizable_call): Use internal_store_fn_p.
11274 (vectorizable_store): Handle scatter store internal functions.
11275 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11276 when deciding whether the end of the group has been reached.
11277 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11278 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11279 (mask_scatter_store<mode>): New insns.
11281 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11282 Alan Hayward <alan.hayward@arm.com>
11283 David Sherwood <david.sherwood@arm.com>
11285 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11286 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11287 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11289 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11290 Use vect_truncate_gather_scatter_offset if we can't treat the
11291 operation as a normal gather load or scatter store.
11292 (get_group_load_store_type): Take the gather_scatter_info
11293 as argument. Try using a gather load or scatter store for
11294 single-element groups.
11295 (get_load_store_type): Update calls to get_group_load_store_type
11296 and vect_use_strided_gather_scatters_p.
11298 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11299 Alan Hayward <alan.hayward@arm.com>
11300 David Sherwood <david.sherwood@arm.com>
11302 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11303 optional tree argument.
11304 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11306 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11307 but continue to use the current value as a fallback.
11308 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11309 to compare the updates.
11310 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11311 (get_load_store_type): Use it when handling a strided access.
11312 (vect_get_strided_load_store_ops): New function.
11313 (vect_get_data_ptr_increment): Likewise.
11314 (vectorizable_load): Handle strided gather loads. Always pass
11315 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11318 Alan Hayward <alan.hayward@arm.com>
11319 David Sherwood <david.sherwood@arm.com>
11321 * doc/md.texi (gather_load@var{m}): Document.
11322 (mask_gather_load@var{m}): Likewise.
11323 * genopinit.c (main): Add supports_vec_gather_load and
11324 supports_vec_gather_load_cached to target_optabs.
11325 * optabs-tree.c (init_tree_optimization_optabs): Use
11326 ggc_cleared_alloc to allocate target_optabs.
11327 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11328 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11330 * internal-fn.h (internal_load_fn_p): Declare.
11331 (internal_gather_scatter_fn_p): Likewise.
11332 (internal_fn_mask_index): Likewise.
11333 (internal_gather_scatter_fn_supported_p): Likewise.
11334 * internal-fn.c (gather_load_direct): New macro.
11335 (expand_gather_load_optab_fn): New function.
11336 (direct_gather_load_optab_supported_p): New macro.
11337 (direct_internal_fn_optab): New function.
11338 (internal_load_fn_p): Likewise.
11339 (internal_gather_scatter_fn_p): Likewise.
11340 (internal_fn_mask_index): Likewise.
11341 (internal_gather_scatter_fn_supported_p): Likewise.
11342 * optabs-query.c (supports_at_least_one_mode_p): New function.
11343 (supports_vec_gather_load_p): Likewise.
11344 * optabs-query.h (supports_vec_gather_load_p): Declare.
11345 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11346 and memory_type field.
11347 (NUM_PATTERNS): Bump to 15.
11348 * tree-vect-data-refs.c: Include internal-fn.h.
11349 (vect_gather_scatter_fn_p): New function.
11350 (vect_describe_gather_scatter_call): Likewise.
11351 (vect_check_gather_scatter): Try using internal functions for
11352 gather loads. Recognize existing calls to a gather load function.
11353 (vect_analyze_data_refs): Consider using gather loads if
11354 supports_vec_gather_load_p.
11355 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11356 (vect_get_gather_scatter_offset_type): Likewise.
11357 (vect_convert_mask_for_vectype): Likewise.
11358 (vect_add_conversion_to_patterm): Likewise.
11359 (vect_try_gather_scatter_pattern): Likewise.
11360 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11361 (vect_vect_recog_func_ptrs): Add it.
11362 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11363 internal_fn_mask_index and internal_gather_scatter_fn_p.
11364 (check_load_store_masking): Take the gather_scatter_info as an
11365 argument and handle gather loads.
11366 (vect_get_gather_scatter_ops): New function.
11367 (vectorizable_call): Check internal_load_fn_p.
11368 (vectorizable_load): Likewise. Handle gather load internal
11370 (vectorizable_store): Update call to check_load_store_masking.
11371 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11372 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11373 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11374 (aarch64_gather_scale_operand_d): New predicates.
11375 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11376 (mask_gather_load<mode>): New insns.
11378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11379 Alan Hayward <alan.hayward@arm.com>
11380 David Sherwood <david.sherwood@arm.com>
11382 * optabs.def (fold_left_plus_optab): New optab.
11383 * doc/md.texi (fold_left_plus_@var{m}): Document.
11384 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11385 * internal-fn.c (fold_left_direct): Define.
11386 (expand_fold_left_optab_fn): Likewise.
11387 (direct_fold_left_optab_supported_p): Likewise.
11388 * fold-const-call.c (fold_const_fold_left): New function.
11389 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11390 * tree-parloops.c (valid_reduction_p): New function.
11391 (gather_scalar_reductions): Use it.
11392 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11393 (vect_finish_replace_stmt): Declare.
11394 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11395 (needs_fold_left_reduction_p): New function, split out from...
11396 (vect_is_simple_reduction): ...here. Accept reductions that
11397 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11398 (vect_force_simple_reduction): Also store the reduction type in
11399 the assignment's STMT_VINFO_REDUC_TYPE.
11400 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11401 (merge_with_identity): New function.
11402 (vect_expand_fold_left): Likewise.
11403 (vectorize_fold_left_reduction): Likewise.
11404 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11405 scalar phi in place for it. Check for target support and reject
11406 cases that would reassociate the operation. Defer the transform
11407 phase to vectorize_fold_left_reduction.
11408 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11409 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11410 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11412 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11414 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11415 call to ifc_temp_var.
11417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11418 Alan Hayward <alan.hayward@arm.com>
11419 David Sherwood <david.sherwood@arm.com>
11421 * target.def (legitimize_address_displacement): Take the original
11422 offset as a poly_int.
11423 * targhooks.h (default_legitimize_address_displacement): Update
11425 * targhooks.c (default_legitimize_address_displacement): Likewise.
11426 * doc/tm.texi: Regenerate.
11427 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11428 as an argument, moving assert of ad->disp == ad->disp_term to...
11429 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11430 Try calling targetm.legitimize_address_displacement before expanding
11431 the address rather than afterwards, and adjust for the new interface.
11432 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11433 Match the new hook interface. Handle SVE addresses.
11434 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11435 new hook interface.
11437 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11439 * Makefile.in (OBJS): Add early-remat.o.
11440 * target.def (select_early_remat_modes): New hook.
11441 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11442 * doc/tm.texi: Regenerate.
11443 * targhooks.h (default_select_early_remat_modes): Declare.
11444 * targhooks.c (default_select_early_remat_modes): New function.
11445 * timevar.def (TV_EARLY_REMAT): New timevar.
11446 * passes.def (pass_early_remat): New pass.
11447 * tree-pass.h (make_pass_early_remat): Declare.
11448 * early-remat.c: New file.
11449 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11451 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11454 Alan Hayward <alan.hayward@arm.com>
11455 David Sherwood <david.sherwood@arm.com>
11457 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11458 vfm1 with a bound_epilog parameter.
11459 (vect_do_peeling): Update calls accordingly, and move the prologue
11460 call earlier in the function. Treat the base bound_epilog as 0 for
11461 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11462 this base when peeling for gaps.
11463 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11464 with fully-masked loops.
11465 (vect_estimate_min_profitable_iters): Handle the single peeled
11466 iteration in that case.
11468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11469 Alan Hayward <alan.hayward@arm.com>
11470 David Sherwood <david.sherwood@arm.com>
11472 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11473 single-element interleaving even if the size is not a power of 2.
11474 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11475 accesses for single-element interleaving if the group size is
11478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11479 Alan Hayward <alan.hayward@arm.com>
11480 David Sherwood <david.sherwood@arm.com>
11482 * doc/md.texi (fold_extract_last_@var{m}): Document.
11483 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11484 * optabs.def (fold_extract_last_optab): New optab.
11485 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11486 * internal-fn.c (fold_extract_direct): New macro.
11487 (expand_fold_extract_optab_fn): Likewise.
11488 (direct_fold_extract_optab_supported_p): Likewise.
11489 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11490 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11491 EXTRACT_LAST_REDUCTION.
11492 (get_initial_def_for_reduction): Do not create an initial vector
11493 for EXTRACT_LAST_REDUCTION reductions.
11494 (vectorizable_reduction): Leave the scalar phi in place for
11495 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11496 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11497 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11498 transform phase to vectorizable_condition.
11499 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11501 (vect_finish_stmt_generation): ...here.
11502 (vect_finish_replace_stmt): New function.
11503 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11504 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11506 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11509 Alan Hayward <alan.hayward@arm.com>
11510 David Sherwood <david.sherwood@arm.com>
11512 * doc/md.texi (extract_last_@var{m}): Document.
11513 * optabs.def (extract_last_optab): New optab.
11514 * internal-fn.def (EXTRACT_LAST): New internal function.
11515 * internal-fn.c (cond_unary_direct): New macro.
11516 (expand_cond_unary_optab_fn): Likewise.
11517 (direct_cond_unary_optab_supported_p): Likewise.
11518 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11519 loops using EXTRACT_LAST.
11520 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11521 (extract_last_<mode>): ...this optab.
11522 (vec_extract<mode><Vel>): Update accordingly.
11524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11525 Alan Hayward <alan.hayward@arm.com>
11526 David Sherwood <david.sherwood@arm.com>
11528 * target.def (empty_mask_is_expensive): New hook.
11529 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11530 * doc/tm.texi: Regenerate.
11531 * targhooks.h (default_empty_mask_is_expensive): Declare.
11532 * targhooks.c (default_empty_mask_is_expensive): New function.
11533 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11534 if the target says that empty masks are expensive.
11535 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11537 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11540 Alan Hayward <alan.hayward@arm.com>
11541 David Sherwood <david.sherwood@arm.com>
11543 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11544 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11545 (vect_use_loop_mask_for_alignment_p): New function.
11546 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11547 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11548 niters_skip argument. Make sure that the first niters_skip elements
11549 of the first iteration are inactive.
11550 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11551 Update call to vect_set_loop_masks_directly.
11552 (get_misalign_in_elems): New function, split out from...
11553 (vect_gen_prolog_loop_niters): ...here.
11554 (vect_update_init_of_dr): Take a code argument that specifies whether
11555 the adjustment should be added or subtracted.
11556 (vect_update_init_of_drs): Likewise.
11557 (vect_prepare_for_masked_peels): New function.
11558 (vect_do_peeling): Skip prologue peeling if we're using a mask
11559 instead. Update call to vect_update_inits_of_drs.
11560 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11562 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11563 alignment. Do not include the number of peeled iterations in
11564 the minimum threshold in that case.
11565 (vectorizable_induction): Adjust the start value down by
11566 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11567 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11568 Take the number of skipped iterations into account when calculating
11570 * tree-vect-stmts.c (vect_gen_while_not): New function.
11572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11573 Alan Hayward <alan.hayward@arm.com>
11574 David Sherwood <david.sherwood@arm.com>
11576 * doc/sourcebuild.texi (vect_fully_masked): Document.
11577 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11578 default value to 0.
11579 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11581 (vect_analyze_loop_2): ...here. Don't check the vectorization
11582 factor against the number of loop iterations if the loop is
11585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11586 Alan Hayward <alan.hayward@arm.com>
11587 David Sherwood <david.sherwood@arm.com>
11589 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11590 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11591 (dump_groups): Update accordingly.
11592 (iv_use::mem_type): New member variable.
11593 (address_p): New function.
11594 (record_use): Add a mem_type argument and initialize the new
11596 (record_group_use): Add a mem_type argument. Use address_p.
11597 Remove obsolete null checks of base_object. Update call to record_use.
11598 (find_interesting_uses_op): Update call to record_group_use.
11599 (find_interesting_uses_cond): Likewise.
11600 (find_interesting_uses_address): Likewise.
11601 (get_mem_type_for_internal_fn): New function.
11602 (find_address_like_use): Likewise.
11603 (find_interesting_uses_stmt): Try find_address_like_use before
11604 calling find_interesting_uses_op.
11605 (addr_offset_valid_p): Use the iv mem_type field as the type
11606 of the addressed memory.
11607 (add_autoinc_candidates): Likewise.
11608 (get_address_cost): Likewise.
11609 (split_small_address_groups_p): Use address_p.
11610 (split_address_groups): Likewise.
11611 (add_iv_candidate_for_use): Likewise.
11612 (autoinc_possible_for_pair): Likewise.
11613 (rewrite_groups): Likewise.
11614 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11615 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11616 (get_alias_ptr_type_for_ptr_address): New function.
11617 (rewrite_use_address): Rewrite address uses in calls that were
11618 identified by find_address_like_use.
11620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11621 Alan Hayward <alan.hayward@arm.com>
11622 David Sherwood <david.sherwood@arm.com>
11624 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11626 * gimple-expr.h (is_gimple_addressable: Likewise.
11627 * gimple-expr.c (is_gimple_address): Likewise.
11628 * internal-fn.c (expand_call_mem_ref): New function.
11629 (expand_mask_load_optab_fn): Use it.
11630 (expand_mask_store_optab_fn): Likewise.
11632 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11633 Alan Hayward <alan.hayward@arm.com>
11634 David Sherwood <david.sherwood@arm.com>
11636 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11637 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11638 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11639 (cond_umax@var{mode}): Document.
11640 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11641 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11642 (cond_umin_optab, cond_umax_optab): New optabs.
11643 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11644 (COND_IOR, COND_XOR): New internal functions.
11645 * internal-fn.h (get_conditional_internal_fn): Declare.
11646 * internal-fn.c (cond_binary_direct): New macro.
11647 (expand_cond_binary_optab_fn): Likewise.
11648 (direct_cond_binary_optab_supported_p): Likewise.
11649 (get_conditional_internal_fn): New function.
11650 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11651 Cope with reduction statements that are vectorized as calls rather
11653 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11654 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11655 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11656 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11657 (UNSPEC_COND_EOR): New unspecs.
11658 (optab): Add mappings for them.
11659 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11660 (sve_int_op, sve_fp_op): New int attributes.
11662 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11663 Alan Hayward <alan.hayward@arm.com>
11664 David Sherwood <david.sherwood@arm.com>
11666 * optabs.def (while_ult_optab): New optab.
11667 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11668 * internal-fn.def (WHILE_ULT): New internal function.
11669 * internal-fn.h (direct_internal_fn_supported_p): New override
11670 that takes two types as argument.
11671 * internal-fn.c (while_direct): New macro.
11672 (expand_while_optab_fn): New function.
11673 (convert_optab_supported_p): Likewise.
11674 (direct_while_optab_supported_p): New macro.
11675 * wide-int.h (wi::udiv_ceil): New function.
11676 * tree-vectorizer.h (rgroup_masks): New structure.
11677 (vec_loop_masks): New typedef.
11678 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11679 and fully_masked_p.
11680 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11681 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11682 (vect_max_vf): New function.
11683 (slpeel_make_loop_iterate_ntimes): Delete.
11684 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11685 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11686 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11687 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11688 internal-fn.h, stor-layout.h and optabs-query.h.
11689 (vect_set_loop_mask): New function.
11690 (add_preheader_seq): Likewise.
11691 (add_header_seq): Likewise.
11692 (interleave_supported_p): Likewise.
11693 (vect_maybe_permute_loop_masks): Likewise.
11694 (vect_set_loop_masks_directly): Likewise.
11695 (vect_set_loop_condition_masked): Likewise.
11696 (vect_set_loop_condition_unmasked): New function, split out from
11697 slpeel_make_loop_iterate_ntimes.
11698 (slpeel_make_loop_iterate_ntimes): Rename to..
11699 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11700 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11701 (vect_do_peeling): Update call accordingly.
11702 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11704 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11705 mask_compare_type, can_fully_mask_p and fully_masked_p.
11706 (release_vec_loop_masks): New function.
11707 (_loop_vec_info): Use it to free the loop masks.
11708 (can_produce_all_loop_masks_p): New function.
11709 (vect_get_max_nscalars_per_iter): Likewise.
11710 (vect_verify_full_masking): Likewise.
11711 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11712 retries, and free the mask rgroups before retrying. Check loop-wide
11713 reasons for disallowing fully-masked loops. Make the final decision
11714 about whether use a fully-masked loop or not.
11715 (vect_estimate_min_profitable_iters): Do not assume that peeling
11716 for the number of iterations will be needed for fully-masked loops.
11717 (vectorizable_reduction): Disable fully-masked loops.
11718 (vectorizable_live_operation): Likewise.
11719 (vect_halve_mask_nunits): New function.
11720 (vect_double_mask_nunits): Likewise.
11721 (vect_record_loop_mask): Likewise.
11722 (vect_get_loop_mask): Likewise.
11723 (vect_transform_loop): Handle the case in which the final loop
11724 iteration might handle a partial vector. Call vect_set_loop_condition
11725 instead of slpeel_make_loop_iterate_ntimes.
11726 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11727 (check_load_store_masking): New function.
11728 (prepare_load_store_mask): Likewise.
11729 (vectorizable_store): Handle fully-masked loops.
11730 (vectorizable_load): Likewise.
11731 (supportable_widening_operation): Use vect_halve_mask_nunits for
11733 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11734 (vect_gen_while): New function.
11735 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11736 (aarch64_uqdec<mode>): New insn.
11738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11739 Alan Hayward <alan.hayward@arm.com>
11740 David Sherwood <david.sherwood@arm.com>
11742 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11743 (reduc_xor_scal_optab): New optabs.
11744 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11745 (reduc_xor_scal_@var{m}): Document.
11746 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11747 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11748 internal functions.
11749 * fold-const-call.c (fold_const_call): Handle them.
11750 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11751 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11752 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11753 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11754 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11755 (UNSPEC_XORV): New unspecs.
11756 (optab): Add entries for them.
11757 (BITWISEV): New int iterator.
11758 (bit_reduc_op): New int attributes.
11760 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11761 Alan Hayward <alan.hayward@arm.com>
11762 David Sherwood <david.sherwood@arm.com>
11764 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11765 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11766 * optabs.def (vec_shl_insert_optab): New optab.
11767 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11768 (duplicate_and_interleave): Likewise.
11769 * tree-vect-loop.c: Include internal-fn.h.
11770 (neutral_op_for_slp_reduction): New function, split out from
11771 get_initial_defs_for_reduction.
11772 (get_initial_def_for_reduction): Handle option 2 for variable-length
11773 vectors by loading the neutral value into a vector and then shifting
11774 the initial value into element 0.
11775 (get_initial_defs_for_reduction): Replace the code argument with
11776 the neutral value calculated by neutral_op_for_slp_reduction.
11777 Use gimple_build_vector for constant-length vectors.
11778 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11779 but the first group_size elements have a neutral value.
11780 Use duplicate_and_interleave otherwise.
11781 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11782 Update call to get_initial_defs_for_reduction. Handle SLP
11783 reductions for variable-length vectors by creating one vector
11784 result for each scalar result, with the elements associated
11785 with other scalar results stubbed out with the neutral value.
11786 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11787 Require IFN_VEC_SHL_INSERT for double reductions on
11788 variable-length vectors, or SLP reductions that have
11789 a neutral value. Require can_duplicate_and_interleave_p
11790 support for variable-length unchained SLP reductions if there
11791 is no neutral value, such as for MIN/MAX reductions. Also require
11792 the number of vector elements to be a multiple of the number of
11793 SLP statements when doing variable-length unchained SLP reductions.
11794 Update call to vect_create_epilog_for_reduction.
11795 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11796 and remove initial values.
11797 (duplicate_and_interleave): Make public.
11798 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11799 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11802 Alan Hayward <alan.hayward@arm.com>
11803 David Sherwood <david.sherwood@arm.com>
11805 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11806 (can_duplicate_and_interleave_p): New function.
11807 (vect_get_and_check_slp_defs): Take the vector of statements
11808 rather than just the current one. Remove excess parentheses.
11809 Restriction rejectinon of vect_constant_def and vect_external_def
11810 for variable-length vectors to boolean types, or types for which
11811 can_duplicate_and_interleave_p is false.
11812 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11813 (duplicate_and_interleave): New function.
11814 (vect_get_constant_vectors): Use gimple_build_vector for
11815 constant-length vectors and suitable variable-length constant
11816 vectors. Use duplicate_and_interleave for other variable-length
11817 vectors. Don't defer the update when inserting new statements.
11819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11820 Alan Hayward <alan.hayward@arm.com>
11821 David Sherwood <david.sherwood@arm.com>
11823 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11824 min_profitable_iters doesn't go negative.
11826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11827 Alan Hayward <alan.hayward@arm.com>
11828 David Sherwood <david.sherwood@arm.com>
11830 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11831 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11832 * optabs.def (vec_mask_load_lanes_optab): New optab.
11833 (vec_mask_store_lanes_optab): Likewise.
11834 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11835 (MASK_STORE_LANES): Likewise.
11836 * internal-fn.c (mask_load_lanes_direct): New macro.
11837 (mask_store_lanes_direct): Likewise.
11838 (expand_mask_load_optab_fn): Handle masked operations.
11839 (expand_mask_load_lanes_optab_fn): New macro.
11840 (expand_mask_store_optab_fn): Handle masked operations.
11841 (expand_mask_store_lanes_optab_fn): New macro.
11842 (direct_mask_load_lanes_optab_supported_p): Likewise.
11843 (direct_mask_store_lanes_optab_supported_p): Likewise.
11844 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11846 (vect_load_lanes_supported): Likewise.
11847 * tree-vect-data-refs.c (strip_conversion): New function.
11848 (can_group_stmts_p): Likewise.
11849 (vect_analyze_data_ref_accesses): Use it instead of checking
11850 for a pair of assignments.
11851 (vect_store_lanes_supported): Take a masked_p parameter.
11852 (vect_load_lanes_supported): Likewise.
11853 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11854 vect_store_lanes_supported and vect_load_lanes_supported.
11855 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11856 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11857 parameter. Don't allow gaps for masked accesses.
11858 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11859 and vect_load_lanes_supported.
11860 (get_load_store_type): Take a masked_p parameter and update
11861 call to get_group_load_store_type.
11862 (vectorizable_store): Update call to get_load_store_type.
11863 Handle IFN_MASK_STORE_LANES.
11864 (vectorizable_load): Update call to get_load_store_type.
11865 Handle IFN_MASK_LOAD_LANES.
11867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11868 Alan Hayward <alan.hayward@arm.com>
11869 David Sherwood <david.sherwood@arm.com>
11871 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11873 * config/aarch64/aarch64-protos.h
11874 (aarch64_sve_struct_memory_operand_p): Declare.
11875 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11876 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11877 (VPRED, vpred): Handle SVE structure modes.
11878 * config/aarch64/constraints.md (Utx): New constraint.
11879 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11880 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11881 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11882 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11883 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11884 structure modes. Split into pieces after RA.
11885 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11886 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11888 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11889 SVE structure modes.
11890 (aarch64_classify_address): Likewise.
11891 (sizetochar): Move earlier in file.
11892 (aarch64_print_operand): Handle SVE register lists.
11893 (aarch64_array_mode): New function.
11894 (aarch64_sve_struct_memory_operand_p): Likewise.
11895 (TARGET_ARRAY_MODE): Redefine.
11897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11898 Alan Hayward <alan.hayward@arm.com>
11899 David Sherwood <david.sherwood@arm.com>
11901 * target.def (array_mode): New target hook.
11902 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11903 * doc/tm.texi: Regenerate.
11904 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11905 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11906 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11907 targetm.array_mode.
11908 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11912 Alan Hayward <alan.hayward@arm.com>
11913 David Sherwood <david.sherwood@arm.com>
11915 * fold-const.c (fold_binary_loc): Check the argument types
11916 rather than the result type when testing for a vector operation.
11918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11920 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11921 * doc/tm.texi: Regenerate.
11923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11924 Alan Hayward <alan.hayward@arm.com>
11925 David Sherwood <david.sherwood@arm.com>
11927 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11928 (sve): Document new AArch64 extension.
11929 * doc/md.texi (w): Extend the description of the AArch64
11930 constraint to include SVE vectors.
11931 (Upl, Upa): Document new AArch64 predicate constraints.
11932 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11934 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11935 (msve-vector-bits=): New option.
11936 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11937 SVE when these are disabled.
11938 (sve): New extension.
11939 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11940 modes. Adjust their number of units based on aarch64_sve_vg.
11941 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11942 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11943 aarch64_addr_query_type.
11944 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11945 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11946 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11947 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11948 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11949 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11950 (aarch64_simd_imm_zero_p): Delete.
11951 (aarch64_check_zero_based_sve_index_immediate): Declare.
11952 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11953 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11954 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11955 (aarch64_sve_float_mul_immediate_p): Likewise.
11956 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11957 rather than an rtx.
11958 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11959 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11960 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11961 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11962 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11963 (aarch64_regmode_natural_size): Likewise.
11964 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11965 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11967 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11968 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11969 for VG and the SVE predicate registers.
11970 (V_ALIASES): Add a "z"-prefixed alias.
11971 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11972 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11973 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11974 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11975 (REG_CLASS_NAMES): Add entries for them.
11976 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11977 and the predicate registers.
11978 (aarch64_sve_vg): Declare.
11979 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11980 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11981 (REGMODE_NATURAL_SIZE): Define.
11982 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11984 * config/aarch64/aarch64.c: Include cfgrtl.h.
11985 (simd_immediate_info): Add a constructor for series vectors,
11986 and an associated step field.
11987 (aarch64_sve_vg): New variable.
11988 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11989 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11990 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11991 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11992 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11993 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11994 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11995 (aarch64_get_mask_mode): New functions.
11996 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11997 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11998 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11999 predicate modes and predicate registers. Explicitly restrict
12000 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12001 to store a vector mode if it is recognized by
12002 aarch64_classify_vector_mode.
12003 (aarch64_regmode_natural_size): New function.
12004 (aarch64_hard_regno_caller_save_mode): Return the original mode
12006 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12007 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12008 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12009 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12011 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12012 does not overlap dest if the function is frame-related. Handle
12014 (aarch64_split_add_offset): New function.
12015 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12016 them aarch64_add_offset.
12017 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12018 and update call to aarch64_sub_sp.
12019 (aarch64_add_cfa_expression): New function.
12020 (aarch64_expand_prologue): Pass extra temporary registers to the
12021 functions above. Handle the case in which we need to emit new
12022 DW_CFA_expressions for registers that were originally saved
12023 relative to the stack pointer, but now have to be expressed
12024 relative to the frame pointer.
12025 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12027 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12028 IP0 and IP1 values for SVE frames.
12029 (aarch64_expand_vec_series): New function.
12030 (aarch64_expand_sve_widened_duplicate): Likewise.
12031 (aarch64_expand_sve_const_vector): Likewise.
12032 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12033 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12034 into the register, rather than emitting a SET directly.
12035 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12036 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12037 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12038 (offset_9bit_signed_scaled_p): New functions.
12039 (aarch64_replicate_bitmask_imm): New function.
12040 (aarch64_bitmask_imm): Use it.
12041 (aarch64_cannot_force_const_mem): Reject expressions involving
12042 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12043 (aarch64_classify_index): Handle SVE indices, by requiring
12044 a plain register index with a scale that matches the element size.
12045 (aarch64_classify_address): Handle SVE addresses. Assert that
12046 the mode of the address is VOIDmode or an integer mode.
12047 Update call to aarch64_classify_symbol.
12048 (aarch64_classify_symbolic_expression): Update call to
12049 aarch64_classify_symbol.
12050 (aarch64_const_vec_all_in_range_p): New function.
12051 (aarch64_print_vector_float_operand): Likewise.
12052 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12053 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12054 and the FP immediates 1.0 and 0.5.
12055 (aarch64_print_address_internal): Handle SVE addresses.
12056 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12057 (aarch64_regno_regclass): Handle predicate registers.
12058 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12060 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12061 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12062 (aarch64_convert_sve_vector_bits): New function.
12063 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12064 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12065 rather than an rtx.
12066 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12067 Handle SVE vector and predicate modes. Accept VL-based constants
12068 that need only one temporary register, and VL offsets that require
12069 no temporary registers.
12070 (aarch64_conditional_register_usage): Mark the predicate registers
12071 as fixed if SVE isn't available.
12072 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12073 Return true for SVE vector and predicate modes.
12074 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12075 rather than an unsigned int. Handle SVE modes.
12076 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12078 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12080 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12081 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12082 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12083 (aarch64_sve_float_mul_immediate_p): New functions.
12084 (aarch64_sve_valid_immediate): New function.
12085 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12086 Explicitly reject structure modes. Check for INDEX constants.
12087 Handle PTRUE and PFALSE constants.
12088 (aarch64_check_zero_based_sve_index_immediate): New function.
12089 (aarch64_simd_imm_zero_p): Delete.
12090 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12091 vector modes. Accept constants in the range of CNT[BHWD].
12092 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12093 ask for an Advanced SIMD mode.
12094 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12095 (aarch64_simd_vector_alignment): Handle SVE predicates.
12096 (aarch64_vectorize_preferred_vector_alignment): New function.
12097 (aarch64_simd_vector_alignment_reachable): Use it instead of
12099 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12100 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12102 (MAX_VECT_LEN): Delete.
12103 (expand_vec_perm_d): Add a vec_flags field.
12104 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12105 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12106 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12108 (aarch64_evpc_rev): Rename to...
12109 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12110 (aarch64_evpc_rev_global): New function.
12111 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12112 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12114 (aarch64_evpc_sve_tbl): New function.
12115 (aarch64_expand_vec_perm_const_1): Update after rename of
12116 aarch64_evpc_rev. Handle SVE permutes too, trying
12117 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12118 than aarch64_evpc_tbl.
12119 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12120 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12121 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12122 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12123 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12124 (aarch64_expand_sve_vcond): New functions.
12125 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12126 of aarch64_vector_mode_p.
12127 (aarch64_dwarf_poly_indeterminate_value): New function.
12128 (aarch64_compute_pressure_classes): Likewise.
12129 (aarch64_can_change_mode_class): Likewise.
12130 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12131 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12132 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12133 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12134 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12135 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12136 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12137 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12139 (Dn, Dl, Dr): Accept const as well as const_vector.
12140 (Dz): Likewise. Compare against CONST0_RTX.
12141 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12142 of "vector" where appropriate.
12143 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12144 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12145 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12146 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12147 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12148 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12149 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12150 (v_int_equiv): Extend to SVE modes.
12151 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12153 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12154 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12155 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12156 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12157 (SVE_COND_FP_CMP): New int iterators.
12158 (perm_hilo): Handle the new unpack unspecs.
12159 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12161 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12162 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12163 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12164 (aarch64_equality_operator, aarch64_constant_vector_operand)
12165 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12166 (aarch64_sve_nonimmediate_operand): Likewise.
12167 (aarch64_sve_general_operand): Likewise.
12168 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12169 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12170 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12171 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12172 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12173 (aarch64_sve_float_arith_immediate): Likewise.
12174 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12175 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12176 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12177 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12178 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12179 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12180 (aarch64_sve_float_arith_operand): Likewise.
12181 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12182 (aarch64_sve_float_mul_operand): Likewise.
12183 (aarch64_sve_vec_perm_operand): Likewise.
12184 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12185 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12186 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12187 as well as const_vector.
12188 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12189 in file. Use CONST0_RTX and CONSTM1_RTX.
12190 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12191 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12192 Use aarch64_simd_imm_zero.
12193 * config/aarch64/aarch64-sve.md: New file.
12194 * config/aarch64/aarch64.md: Include it.
12195 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12196 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12197 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12198 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12199 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12200 (sve): New attribute.
12201 (enabled): Disable instructions with the sve attribute unless
12203 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12204 aarch64_expand_mov_immediate.
12205 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12206 CNT[BHSD] immediates.
12207 (movti): Split CONST_POLY_INT moves into two halves.
12208 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12209 Split additions that need a temporary here if the destination
12210 is the stack pointer.
12211 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12212 (*add<mode>3_poly_1): New instruction.
12213 (set_clobber_cc): New expander.
12215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12217 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12218 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12219 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12220 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12221 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12222 Change innermode from fixed_mode_size to machine_mode.
12223 (simplify_subreg): Update call accordingly. Handle a constant-sized
12224 subreg of a variable-length CONST_VECTOR.
12226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12227 Alan Hayward <alan.hayward@arm.com>
12228 David Sherwood <david.sherwood@arm.com>
12230 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12231 (add_offset_to_base): New function, split out from...
12232 (create_mem_ref): ...here. When handling a scale other than 1,
12233 check first whether the address is valid without the offset.
12234 Add it into the base if so, leaving the index and scale as-is.
12236 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12239 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12240 fold_for_warn before checking if arg2 is INTEGER_CST.
12242 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12244 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12245 (store_multiple_operation): Delete.
12246 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12247 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12248 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12249 guarded by TARGET_STRING.
12250 (rs6000_output_load_multiple): Delete.
12251 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12252 OPTION_MASK_STRING / TARGET_STRING handling.
12253 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12254 (const rs6000_opt_masks) <"string">: Change mask to 0.
12255 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12256 (MASK_STRING): Delete.
12257 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12259 (load_multiple): Delete.
12266 (store_multiple): Delete.
12273 (movmemsi_8reg): Delete.
12274 (corresponding unnamed define_insn): Delete.
12275 (movmemsi_6reg): Delete.
12276 (corresponding unnamed define_insn): Delete.
12277 (movmemsi_4reg): Delete.
12278 (corresponding unnamed define_insn): Delete.
12279 (movmemsi_2reg): Delete.
12280 (corresponding unnamed define_insn): Delete.
12281 (movmemsi_1reg): Delete.
12282 (corresponding unnamed define_insn): Delete.
12283 * config/rs6000/rs6000.opt (mno-string): New.
12284 (mstring): Replace by deprecation warning stub.
12285 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12287 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12289 * regrename.c (regrename_do_replace): If replacing the same
12290 reg multiple times, try to reuse last created gen_raw_REG.
12293 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12294 main to workaround a bug in GDB.
12296 2018-01-12 Tom de Vries <tom@codesourcery.com>
12299 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12301 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12303 PR rtl-optimization/80481
12304 * ira-color.c (get_cap_member): New function.
12305 (allocnos_conflict_by_live_ranges_p): Use it.
12306 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12307 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12309 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12312 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12313 (*saddl_se_1): Ditto.
12314 (*ssubsi_1): Ditto.
12315 (*ssubl_se_1): Ditto.
12317 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12319 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12320 rather than wi::to_widest for DR_INITs.
12321 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12322 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12323 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12325 (vect_analyze_group_access_1): Note that here.
12327 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12329 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12330 polynomial type sizes.
12332 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12334 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12335 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12336 (gimple_add_tmp_var): Likewise.
12338 2018-01-12 Martin Liska <mliska@suse.cz>
12340 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12341 (gimple_alloc_sizes): Likewise.
12342 (dump_gimple_statistics): Use PRIu64 in printf format.
12343 * gimple.h: Change uint64_t to int.
12345 2018-01-12 Martin Liska <mliska@suse.cz>
12347 * tree-core.h: Use uint64_t instead of int.
12348 * tree.c (tree_node_counts): Likewise.
12349 (tree_node_sizes): Likewise.
12350 (dump_tree_statistics): Use PRIu64 in printf format.
12352 2018-01-12 Martin Liska <mliska@suse.cz>
12354 * Makefile.in: As qsort_chk is implemented in vec.c, add
12355 vec.o to linkage of gencfn-macros.
12356 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12357 passing the info to record_node_allocation_statistics.
12358 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12360 * ggc-common.c (struct ggc_usage): Add operator== and use
12361 it in operator< and compare function.
12362 * mem-stats.h (struct mem_usage): Likewise.
12363 * vec.c (struct vec_usage): Remove operator< and compare
12364 function. Can be simply inherited.
12366 2018-01-12 Martin Jambor <mjambor@suse.cz>
12369 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12370 * tree-ssa-math-opts.c: Include domwalk.h.
12371 (convert_mult_to_fma_1): New function.
12372 (fma_transformation_info): New type.
12373 (fma_deferring_state): Likewise.
12374 (cancel_fma_deferring): New function.
12375 (result_of_phi): Likewise.
12376 (last_fma_candidate_feeds_initial_phi): Likewise.
12377 (convert_mult_to_fma): Added deferring logic, split actual
12378 transformation to convert_mult_to_fma_1.
12379 (math_opts_dom_walker): New type.
12380 (math_opts_dom_walker::after_dom_children): New method, body moved
12381 here from pass_optimize_widening_mul::execute, added deferring logic
12383 (pass_optimize_widening_mul::execute): Moved most of code to
12384 math_opts_dom_walker::after_dom_children.
12385 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12386 * config/i386/i386.c (ix86_option_override_internal): Added
12387 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12389 2018-01-12 Richard Biener <rguenther@suse.de>
12392 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12393 inline instance vars.
12395 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12398 * config/rx/rx.c (rx_is_restricted_memory_address):
12399 Handle SUBREG case.
12401 2018-01-12 Richard Biener <rguenther@suse.de>
12403 PR tree-optimization/80846
12404 * target.def (split_reduction): New target hook.
12405 * targhooks.c (default_split_reduction): New function.
12406 * targhooks.h (default_split_reduction): Declare.
12407 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12408 target requests first reduce vectors by combining low and high
12410 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12411 (get_vectype_for_scalar_type_and_size): Export.
12412 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12413 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12414 * doc/tm.texi: Regenerate.
12415 * config/i386/i386.c (ix86_split_reduction): Implement
12416 TARGET_VECTORIZE_SPLIT_REDUCTION.
12418 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12421 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12422 in PIC mode except for TARGET_VXWORKS_RTP.
12423 * config/sparc/sparc.c: Include cfgrtl.h.
12424 (TARGET_INIT_PIC_REG): Define.
12425 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12426 (sparc_pic_register_p): New predicate.
12427 (sparc_legitimate_address_p): Use it.
12428 (sparc_legitimize_pic_address): Likewise.
12429 (sparc_delegitimize_address): Likewise.
12430 (sparc_mode_dependent_address_p): Likewise.
12431 (gen_load_pcrel_sym): Remove 4th parameter.
12432 (load_got_register): Adjust call to above. Remove obsolete stuff.
12433 (sparc_expand_prologue): Do not call load_got_register here.
12434 (sparc_flat_expand_prologue): Likewise.
12435 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12436 (sparc_use_pseudo_pic_reg): New function.
12437 (sparc_init_pic_reg): Likewise.
12438 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12439 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12441 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12443 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12444 Add item for branch_cost.
12446 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12448 PR rtl-optimization/83565
12449 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12450 not extend the result to a larger mode for rotate operations.
12451 (num_sign_bit_copies1): Likewise.
12453 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12456 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12458 Use values-Xc.o for -pedantic.
12459 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12461 2018-01-12 Martin Liska <mliska@suse.cz>
12464 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12466 (possible_polymorphic_call_targets): Use it.
12467 (ipa_devirt): Likewise.
12469 2018-01-12 Martin Liska <mliska@suse.cz>
12471 * profile-count.h (enum profile_quality): Use 0 as invalid
12472 enum value of profile_quality.
12474 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12476 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12477 -mext-string options.
12479 2018-01-12 Richard Biener <rguenther@suse.de>
12481 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12482 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12483 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12485 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12487 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12489 * configure.ac (--with-long-double-format): Add support for the
12490 configuration option to change the default long double format on
12492 * config.gcc (powerpc*-linux*-*): Likewise.
12493 * configure: Regenerate.
12494 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12495 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12496 used without modification.
12498 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12500 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12501 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12502 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12503 MISC_BUILTIN_SPEC_BARRIER.
12504 (rs6000_init_builtins): Likewise.
12505 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12507 (speculation_barrier): New define_insn.
12508 * doc/extend.texi: Document __builtin_speculation_barrier.
12510 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12513 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12514 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12515 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12517 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12518 integral modes instead of "ss" and "sd".
12519 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12520 vectors with 32-bit and 64-bit elements.
12521 (vecdupssescalarmodesuffix): New mode attribute.
12522 (vec_dup<mode>): Use it.
12524 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12527 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12528 frame if argument is passed on stack.
12530 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12533 * ree.c (combine_reaching_defs): Optimize also
12534 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12535 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12537 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12539 PR middle-end/83189
12540 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12542 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12544 PR middle-end/83718
12545 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12546 after they are computed.
12548 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12550 PR tree-optimization/83695
12551 * gimple-loop-linterchange.cc
12552 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12553 reset cached scev information after interchange.
12554 (pass_linterchange::execute): Remove call to scev_reset_htab.
12556 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12558 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12559 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12560 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12561 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12562 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12563 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12564 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12565 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12566 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12567 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12568 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12569 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12570 (V_lane_reg): Likewise.
12571 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12573 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12574 (vfmal_lane_low<mode>_intrinsic,
12575 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12576 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12577 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12578 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12579 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12580 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12582 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12584 * config/arm/arm-cpus.in (fp16fml): New feature.
12585 (ALL_SIMD): Add fp16fml.
12586 (armv8.2-a): Add fp16fml as an option.
12587 (armv8.3-a): Likewise.
12588 (armv8.4-a): Add fp16fml as part of fp16.
12589 * config/arm/arm.h (TARGET_FP16FML): Define.
12590 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12592 * config/arm/arm-modes.def (V2HF): Define.
12593 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12594 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12595 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12596 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12597 vfmsl_low, vfmsl_high): New set of builtins.
12598 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12599 (vfml_op): New code attribute.
12600 (VFMLHALVES): New int iterator.
12601 (VFML, VFMLSEL): New mode attributes.
12602 (V_reg): Define mapping for V2HF.
12603 (V_hi, V_lo): New mode attributes.
12604 (VF_constraint): Likewise.
12605 (vfml_half, vfml_half_selector): New int attributes.
12606 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12608 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12609 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12611 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12612 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12613 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12614 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12616 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12617 Document new effective target and option set.
12619 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12621 * config/arm/arm-cpus.in (armv8_4): New feature.
12622 (ARMv8_4a): New fgroup.
12623 (armv8.4-a): New arch.
12624 * config/arm/arm-tables.opt: Regenerate.
12625 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12626 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12627 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12628 Add matching rules for -march=armv8.4-a and extensions.
12629 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12631 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12634 * config/rx/rx.md (BW): New mode attribute.
12635 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12637 2018-01-11 Richard Biener <rguenther@suse.de>
12639 PR tree-optimization/83435
12640 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12641 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12642 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12644 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12645 Alan Hayward <alan.hayward@arm.com>
12646 David Sherwood <david.sherwood@arm.com>
12648 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12650 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12651 (aarch64_print_address_internal): Use it to check for a zero offset.
12653 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12654 Alan Hayward <alan.hayward@arm.com>
12655 David Sherwood <david.sherwood@arm.com>
12657 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12658 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12659 Return a poly_int64 rather than a HOST_WIDE_INT.
12660 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12661 rather than a HOST_WIDE_INT.
12662 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12663 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12664 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12665 final_offset from HOST_WIDE_INT to poly_int64.
12666 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12667 to_constant when getting the number of units in an Advanced SIMD
12669 (aarch64_builtin_vectorized_function): Check for a constant number
12671 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12673 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12674 attribute instead of GET_MODE_NUNITS.
12675 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12676 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12677 GET_MODE_SIZE for fixed-size registers.
12678 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12679 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12680 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12681 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12682 (aarch64_print_operand, aarch64_print_address_internal)
12683 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12684 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12685 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12686 Handle polynomial GET_MODE_SIZE.
12687 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12688 wider than SImode without modification.
12689 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12690 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12691 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12692 passing and returning SVE modes.
12693 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12694 rather than GEN_INT.
12695 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12696 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12697 (aarch64_allocate_and_probe_stack_space): Likewise.
12698 (aarch64_layout_frame): Cope with polynomial offsets.
12699 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12700 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12701 polynomial offsets.
12702 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12703 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12704 poly_int64 rather than a HOST_WIDE_INT.
12705 (aarch64_get_separate_components, aarch64_process_components)
12706 (aarch64_expand_prologue, aarch64_expand_epilogue)
12707 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12708 (aarch64_anchor_offset): New function, split out from...
12709 (aarch64_legitimize_address): ...here.
12710 (aarch64_builtin_vectorization_cost): Handle polynomial
12711 TYPE_VECTOR_SUBPARTS.
12712 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12714 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12715 number of elements from the PARALLEL rather than the mode.
12716 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12717 rather than GET_MODE_BITSIZE.
12718 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12719 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12720 (aarch64_expand_vec_perm_const_1): Handle polynomial
12721 d->perm.length () and d->perm elements.
12722 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12723 Apply to_constant to d->perm elements.
12724 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12725 polynomial CONST_VECTOR_NUNITS.
12726 (aarch64_move_pointer): Take amount as a poly_int64 rather
12728 (aarch64_progress_pointer): Avoid temporary variable.
12729 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12730 the mode attribute instead of GET_MODE.
12732 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12733 Alan Hayward <alan.hayward@arm.com>
12734 David Sherwood <david.sherwood@arm.com>
12736 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12737 x exists before using it.
12738 (aarch64_add_constant_internal): Rename to...
12739 (aarch64_add_offset_1): ...this. Replace regnum with separate
12740 src and dest rtxes. Handle the case in which they're different,
12741 including when the offset is zero. Replace scratchreg with an rtx.
12742 Use 2 additions if there is no spare register into which we can
12743 move a 16-bit constant.
12744 (aarch64_add_constant): Delete.
12745 (aarch64_add_offset): Replace reg with separate src and dest
12746 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12747 Use aarch64_add_offset_1.
12748 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12749 an rtx rather than an int. Take the delta as a poly_int64
12750 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12751 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12752 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12753 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12754 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12755 and aarch64_add_sp.
12756 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12757 aarch64_add_constant.
12759 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12761 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12762 Use scalar_float_mode.
12764 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12766 * config/aarch64/aarch64-simd.md
12767 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12768 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12769 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12770 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12771 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12772 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12773 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12774 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12775 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12776 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12778 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12781 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12782 targ_options->x_arm_arch_string is non NULL.
12784 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12786 * config/aarch64/aarch64.h
12787 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12789 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12792 * expmed.c (emit_store_flag_force): Swap if const op0
12793 and change VOIDmode to mode of op0.
12795 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12797 PR rtl-optimization/83761
12798 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12799 than bytes to mode_for_size.
12801 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12803 PR middle-end/83189
12804 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12805 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12808 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12810 PR middle-end/83575
12811 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12812 when in layout mode.
12813 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12814 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12817 2018-01-10 Michael Collison <michael.collison@arm.com>
12819 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12820 * config/aarch64/aarch64-option-extension.def: Add
12821 AARCH64_OPT_EXTENSION of 'fp16fml'.
12822 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12823 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12824 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12825 * config/aarch64/constraints.md (Ui7): New constraint.
12826 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12827 (VFMLA_SEL_W): Ditto.
12830 (VFMLA16_LOW): New int iterator.
12831 (VFMLA16_HIGH): Ditto.
12832 (UNSPEC_FMLAL): New unspec.
12833 (UNSPEC_FMLSL): Ditto.
12834 (UNSPEC_FMLAL2): Ditto.
12835 (UNSPEC_FMLSL2): Ditto.
12836 (f16mac): New code attribute.
12837 * config/aarch64/aarch64-simd-builtins.def
12838 (aarch64_fmlal_lowv2sf): Ditto.
12839 (aarch64_fmlsl_lowv2sf): Ditto.
12840 (aarch64_fmlalq_lowv4sf): Ditto.
12841 (aarch64_fmlslq_lowv4sf): Ditto.
12842 (aarch64_fmlal_highv2sf): Ditto.
12843 (aarch64_fmlsl_highv2sf): Ditto.
12844 (aarch64_fmlalq_highv4sf): Ditto.
12845 (aarch64_fmlslq_highv4sf): Ditto.
12846 (aarch64_fmlal_lane_lowv2sf): Ditto.
12847 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12848 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12849 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12850 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12851 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12852 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12853 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12854 (aarch64_fmlal_lane_highv2sf): Ditto.
12855 (aarch64_fmlsl_lane_highv2sf): Ditto.
12856 (aarch64_fmlal_laneq_highv2sf): Ditto.
12857 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12858 (aarch64_fmlalq_lane_highv4sf): Ditto.
12859 (aarch64_fmlsl_lane_highv4sf): Ditto.
12860 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12861 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12862 * config/aarch64/aarch64-simd.md:
12863 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12864 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12865 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12866 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12867 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12868 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12869 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12870 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12871 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12872 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12873 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12874 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12875 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12876 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12877 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12878 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12879 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12880 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12881 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12882 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12883 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12884 (vfmlsl_low_u32): Ditto.
12885 (vfmlalq_low_u32): Ditto.
12886 (vfmlslq_low_u32): Ditto.
12887 (vfmlal_high_u32): Ditto.
12888 (vfmlsl_high_u32): Ditto.
12889 (vfmlalq_high_u32): Ditto.
12890 (vfmlslq_high_u32): Ditto.
12891 (vfmlal_lane_low_u32): Ditto.
12892 (vfmlsl_lane_low_u32): Ditto.
12893 (vfmlal_laneq_low_u32): Ditto.
12894 (vfmlsl_laneq_low_u32): Ditto.
12895 (vfmlalq_lane_low_u32): Ditto.
12896 (vfmlslq_lane_low_u32): Ditto.
12897 (vfmlalq_laneq_low_u32): Ditto.
12898 (vfmlslq_laneq_low_u32): Ditto.
12899 (vfmlal_lane_high_u32): Ditto.
12900 (vfmlsl_lane_high_u32): Ditto.
12901 (vfmlal_laneq_high_u32): Ditto.
12902 (vfmlsl_laneq_high_u32): Ditto.
12903 (vfmlalq_lane_high_u32): Ditto.
12904 (vfmlslq_lane_high_u32): Ditto.
12905 (vfmlalq_laneq_high_u32): Ditto.
12906 (vfmlslq_laneq_high_u32): Ditto.
12907 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12908 (AARCH64_FL_FOR_ARCH8_4): New.
12909 (AARCH64_ISA_F16FML): New ISA flag.
12910 (TARGET_F16FML): New feature flag for fp16fml.
12911 (doc/invoke.texi): Document new fp16fml option.
12913 2018-01-10 Michael Collison <michael.collison@arm.com>
12915 * config/aarch64/aarch64-builtins.c:
12916 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12917 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12918 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12919 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12920 (AARCH64_ISA_SHA3): New ISA flag.
12921 (TARGET_SHA3): New feature flag for sha3.
12922 * config/aarch64/iterators.md (sha512_op): New int attribute.
12923 (CRYPTO_SHA512): New int iterator.
12924 (UNSPEC_SHA512H): New unspec.
12925 (UNSPEC_SHA512H2): Ditto.
12926 (UNSPEC_SHA512SU0): Ditto.
12927 (UNSPEC_SHA512SU1): Ditto.
12928 * config/aarch64/aarch64-simd-builtins.def
12929 (aarch64_crypto_sha512hqv2di): New builtin.
12930 (aarch64_crypto_sha512h2qv2di): Ditto.
12931 (aarch64_crypto_sha512su0qv2di): Ditto.
12932 (aarch64_crypto_sha512su1qv2di): Ditto.
12933 (aarch64_eor3qv8hi): Ditto.
12934 (aarch64_rax1qv2di): Ditto.
12935 (aarch64_xarqv2di): Ditto.
12936 (aarch64_bcaxqv8hi): Ditto.
12937 * config/aarch64/aarch64-simd.md:
12938 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12939 (aarch64_crypto_sha512su0qv2di): Ditto.
12940 (aarch64_crypto_sha512su1qv2di): Ditto.
12941 (aarch64_eor3qv8hi): Ditto.
12942 (aarch64_rax1qv2di): Ditto.
12943 (aarch64_xarqv2di): Ditto.
12944 (aarch64_bcaxqv8hi): Ditto.
12945 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12946 (vsha512h2q_u64): Ditto.
12947 (vsha512su0q_u64): Ditto.
12948 (vsha512su1q_u64): Ditto.
12949 (veor3q_u16): Ditto.
12950 (vrax1q_u64): Ditto.
12951 (vxarq_u64): Ditto.
12952 (vbcaxq_u16): Ditto.
12953 * config/arm/types.md (crypto_sha512): New type attribute.
12954 (crypto_sha3): Ditto.
12955 (doc/invoke.texi): Document new sha3 option.
12957 2018-01-10 Michael Collison <michael.collison@arm.com>
12959 * config/aarch64/aarch64-builtins.c:
12960 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12961 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12962 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12963 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12964 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12965 (AARCH64_ISA_SM4): New ISA flag.
12966 (TARGET_SM4): New feature flag for sm4.
12967 * config/aarch64/aarch64-simd-builtins.def
12968 (aarch64_sm3ss1qv4si): Ditto.
12969 (aarch64_sm3tt1aq4si): Ditto.
12970 (aarch64_sm3tt1bq4si): Ditto.
12971 (aarch64_sm3tt2aq4si): Ditto.
12972 (aarch64_sm3tt2bq4si): Ditto.
12973 (aarch64_sm3partw1qv4si): Ditto.
12974 (aarch64_sm3partw2qv4si): Ditto.
12975 (aarch64_sm4eqv4si): Ditto.
12976 (aarch64_sm4ekeyqv4si): Ditto.
12977 * config/aarch64/aarch64-simd.md:
12978 (aarch64_sm3ss1qv4si): Ditto.
12979 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12980 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12981 (aarch64_sm4eqv4si): Ditto.
12982 (aarch64_sm4ekeyqv4si): Ditto.
12983 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12984 (sm3part_op): Ditto.
12985 (CRYPTO_SM3TT): Ditto.
12986 (CRYPTO_SM3PART): Ditto.
12987 (UNSPEC_SM3SS1): New unspec.
12988 (UNSPEC_SM3TT1A): Ditto.
12989 (UNSPEC_SM3TT1B): Ditto.
12990 (UNSPEC_SM3TT2A): Ditto.
12991 (UNSPEC_SM3TT2B): Ditto.
12992 (UNSPEC_SM3PARTW1): Ditto.
12993 (UNSPEC_SM3PARTW2): Ditto.
12994 (UNSPEC_SM4E): Ditto.
12995 (UNSPEC_SM4EKEY): Ditto.
12996 * config/aarch64/constraints.md (Ui2): New constraint.
12997 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12998 * config/arm/types.md (crypto_sm3): New type attribute.
12999 (crypto_sm4): Ditto.
13000 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13001 (vsm3tt1aq_u32): Ditto.
13002 (vsm3tt1bq_u32): Ditto.
13003 (vsm3tt2aq_u32): Ditto.
13004 (vsm3tt2bq_u32): Ditto.
13005 (vsm3partw1q_u32): Ditto.
13006 (vsm3partw2q_u32): Ditto.
13007 (vsm4eq_u32): Ditto.
13008 (vsm4ekeyq_u32): Ditto.
13009 (doc/invoke.texi): Document new sm4 option.
13011 2018-01-10 Michael Collison <michael.collison@arm.com>
13013 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13014 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13015 (AARCH64_FL_FOR_ARCH8_4): New.
13016 (AARCH64_FL_V8_4): New flag.
13017 (doc/invoke.texi): Document new armv8.4-a option.
13019 2018-01-10 Michael Collison <michael.collison@arm.com>
13021 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13022 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13023 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13024 * config/aarch64/aarch64-option-extension.def: Add
13025 AARCH64_OPT_EXTENSION of 'sha2'.
13026 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13027 (crypto): Disable sha2 and aes if crypto disabled.
13028 (crypto): Enable aes and sha2 if enabled.
13029 (simd): Disable sha2 and aes if simd disabled.
13030 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13032 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13033 (TARGET_SHA2): New feature flag for sha2.
13034 (TARGET_AES): New feature flag for aes.
13035 * config/aarch64/aarch64-simd.md:
13036 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13037 conditional on TARGET_AES.
13038 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13039 (aarch64_crypto_sha1hsi): Make pattern conditional
13041 (aarch64_crypto_sha1hv4si): Ditto.
13042 (aarch64_be_crypto_sha1hv4si): Ditto.
13043 (aarch64_crypto_sha1su1v4si): Ditto.
13044 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13045 (aarch64_crypto_sha1su0v4si): Ditto.
13046 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13047 (aarch64_crypto_sha256su0v4si): Ditto.
13048 (aarch64_crypto_sha256su1v4si): Ditto.
13049 (doc/invoke.texi): Document new aes and sha2 options.
13051 2018-01-10 Martin Sebor <msebor@redhat.com>
13053 PR tree-optimization/83781
13054 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13057 2018-01-11 Martin Sebor <msebor@gmail.com>
13058 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13060 PR tree-optimization/83501
13061 PR tree-optimization/81703
13063 * tree-ssa-strlen.c (get_string_cst): Rename...
13064 (get_string_len): ...to this. Handle global constants.
13065 (handle_char_store): Adjust.
13067 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13068 Jim Wilson <jimw@sifive.com>
13070 * config/riscv/riscv-protos.h (riscv_output_return): New.
13071 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13072 (riscv_attribute_table, riscv_output_return),
13073 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13074 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13075 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13077 (riscv_expand_prologue): Add early return for naked function.
13078 (riscv_expand_epilogue): Likewise.
13079 (riscv_function_ok_for_sibcall): Return false for naked function.
13080 (riscv_set_current_function): New.
13081 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13082 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13083 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13084 * doc/extend.texi (RISC-V Function Attributes): New.
13086 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13088 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13089 check for 128-bit long double before checking TCmode.
13090 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13091 128-bit long doubles before checking TFmode or TCmode.
13092 (FLOAT128_IBM_P): Likewise.
13094 2018-01-10 Martin Sebor <msebor@redhat.com>
13096 PR tree-optimization/83671
13097 * builtins.c (c_strlen): Unconditionally return zero for the empty
13099 Use -Warray-bounds for warnings.
13100 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13101 for non-constant array indices with COMPONENT_REF, arrays of
13102 arrays, and pointers to arrays.
13103 (gimple_fold_builtin_strlen): Determine and set length range for
13104 non-constant character arrays.
13106 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13108 PR middle-end/81897
13109 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13112 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13114 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13116 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13119 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13120 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13121 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13122 indexed_or_indirect_operand predicate.
13123 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13124 (*vsx_le_perm_load_v8hi): Likewise.
13125 (*vsx_le_perm_load_v16qi): Likewise.
13126 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13127 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13128 (*vsx_le_perm_store_v8hi): Likewise.
13129 (*vsx_le_perm_store_v16qi): Likewise.
13130 (eight unnamed splitters): Likewise.
13132 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13134 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13135 * config/rs6000/emmintrin.h: Likewise.
13136 * config/rs6000/mmintrin.h: Likewise.
13137 * config/rs6000/xmmintrin.h: Likewise.
13139 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13142 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13144 * tree.c (tree_nop_conversion): Return true for location wrapper
13146 (maybe_wrap_with_location): New function.
13147 (selftest::check_strip_nops): New function.
13148 (selftest::test_location_wrappers): New function.
13149 (selftest::tree_c_tests): Call it.
13150 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13151 (maybe_wrap_with_location): New decl.
13152 (EXPR_LOCATION_WRAPPER_P): New macro.
13153 (location_wrapper_p): New inline function.
13154 (tree_strip_any_location_wrapper): New inline function.
13156 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13159 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13160 stack_realign_offset for the largest alignment of stack slot
13162 (ix86_find_max_used_stack_alignment): New function.
13163 (ix86_finalize_stack_frame_flags): Use it. Set
13164 max_used_stack_alignment if we don't realign stack.
13165 * config/i386/i386.h (machine_function): Add
13166 max_used_stack_alignment.
13168 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13170 * config/arm/arm.opt (-mbranch-cost): New option.
13171 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13174 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13177 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13178 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13180 2018-01-10 Richard Biener <rguenther@suse.de>
13183 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13184 early out so it also covers the case where we have a non-NULL
13187 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13189 PR tree-optimization/83753
13190 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13191 for non-strided grouped accesses if the number of elements is 1.
13193 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13196 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13197 * i386.h (TARGET_USE_GATHER): Define.
13198 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13200 2018-01-10 Martin Liska <mliska@suse.cz>
13203 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13204 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13206 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13207 CLEANUP_NO_PARTITIONING is not set.
13209 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13211 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13212 for vectors, as a partial revert of r254296.
13213 * rtl.h (const_vec_p): Delete.
13214 (const_vec_duplicate_p): Don't test for vector CONSTs.
13215 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13216 * expmed.c (make_tree): Likewise.
13219 * common.md (E, F): Use CONSTANT_P instead of checking for
13221 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13222 checking for CONST_VECTOR.
13224 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13226 PR middle-end/83575
13227 * predict.c (force_edge_cold): Handle in more sane way edges
13228 with no prediction.
13230 2018-01-09 Carl Love <cel@us.ibm.com>
13232 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13234 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13235 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13236 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13237 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13238 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13239 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13240 * config/rs6000/rs6000-protos.h: Add extern defition for
13241 rs6000_generate_float2_double_code.
13242 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13244 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13245 (float2_v2df): Add define_expand.
13247 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13250 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13251 op_mode in the force_to_mode call.
13253 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13255 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13256 instead of checking each element individually.
13257 (aarch64_evpc_uzp): Likewise.
13258 (aarch64_evpc_zip): Likewise.
13259 (aarch64_evpc_ext): Likewise.
13260 (aarch64_evpc_rev): Likewise.
13261 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13262 instead of checking each element individually. Return true without
13264 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13265 whether all selected elements come from the same input, instead of
13266 checking each element individually. Remove calls to gen_rtx_REG,
13267 start_sequence and end_sequence and instead assert that no rtl is
13270 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13272 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13273 order of HIGH and CONST checks.
13275 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13277 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13278 if the destination isn't an SSA_NAME.
13280 2018-01-09 Richard Biener <rguenther@suse.de>
13282 PR tree-optimization/83668
13283 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13285 (canonicalize_loop_form): ... here, renamed from ...
13286 (canonicalize_loop_closed_ssa_form): ... this and amended to
13287 swap successor edges for loop exit blocks to make us use
13288 the RPO order we need for initial schedule generation.
13290 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13292 PR tree-optimization/64811
13293 * match.pd: When optimizing comparisons with Inf, avoid
13294 introducing or losing exceptions from comparisons with NaN.
13296 2018-01-09 Martin Liska <mliska@suse.cz>
13299 * asan.c (shadow_mem_size): Add gcc_assert.
13301 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13303 Don't save registers in main().
13306 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13307 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13308 * config/avr/avr.c (avr_set_current_function): Don't error if
13309 naked, OS_task or OS_main are specified at the same time.
13310 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13312 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13314 * common/config/avr/avr-common.c (avr_option_optimization_table):
13315 Switch on -mmain-is-OS_task for optimizing compilations.
13317 2018-01-09 Richard Biener <rguenther@suse.de>
13319 PR tree-optimization/83572
13320 * graphite.c: Include cfganal.h.
13321 (graphite_transform_loops): Connect infinite loops to exit
13322 and remove fake edges at the end.
13324 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13326 * ipa-inline.c (edge_badness): Revert accidental checkin.
13328 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13331 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13332 symbols; not inline clones.
13334 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13337 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13338 hard registers. Formatting fixes.
13340 PR preprocessor/83722
13341 * gcc.c (try_generate_repro): Pass
13342 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13343 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13346 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13347 Kito Cheng <kito.cheng@gmail.com>
13349 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13350 (riscv_leaf_function_p): Delete.
13351 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13353 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13355 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13357 (do_ifelse): New function.
13358 (do_isel): New function.
13359 (do_sub3): New function.
13360 (do_add3): New function.
13361 (do_load_mask_compare): New function.
13362 (do_overlap_load_compare): New function.
13363 (expand_compare_loop): New function.
13364 (expand_block_compare): Call expand_compare_loop() when appropriate.
13365 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13366 option description.
13367 (-mblock-compare-inline-loop-limit): New option.
13369 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13372 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13373 Reverse order of second and third operands in first alternative.
13374 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13375 of first and second elements in UNSPEC_VPERMR vector.
13376 (altivec_expand_vec_perm_le): Likewise.
13378 2018-01-08 Jeff Law <law@redhat.com>
13380 PR rtl-optimizatin/81308
13381 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13382 (process_switch): If group_case_labels makes a change, then set
13384 (pass_convert_switch::execute): If a switch is converted, then
13385 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13387 PR rtl-optimization/81308
13388 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13391 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13393 PR target/83663 - Revert r255946
13394 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13395 generation for cases where splatting a value is not useful.
13396 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13397 across a vec_duplicate and a paradoxical subreg forming a vector
13398 mode to a vec_concat.
13400 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13402 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13403 -march=armv8.3-a variants.
13404 * config/arm/t-multilib: Likewise.
13405 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13407 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13409 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13411 (cceq_ior_compare_complement): Give it a name so I can use it, and
13412 change boolean_or_operator predicate to boolean_operator so it can
13413 be used to generate a crand.
13414 (eqne): New code iterator.
13415 (bd/bd_neg): New code_attrs.
13416 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13417 a single define_insn.
13418 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13419 decrement (bdnzt/bdnzf/bdzt/bdzf).
13420 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13421 with the new names of the branch decrement patterns, and added the
13422 names of the branch decrement conditional patterns.
13424 2018-01-08 Richard Biener <rguenther@suse.de>
13426 PR tree-optimization/83563
13427 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13430 2018-01-08 Richard Biener <rguenther@suse.de>
13432 PR middle-end/83713
13433 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13435 2018-01-08 Richard Biener <rguenther@suse.de>
13437 PR tree-optimization/83685
13438 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13439 references to abnormals.
13441 2018-01-08 Richard Biener <rguenther@suse.de>
13444 * dwarf2out.c (output_indirect_strings): Handle empty
13445 skeleton_debug_str_hash.
13446 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13448 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13450 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13451 (emit_store_direct): Likewise.
13452 (arc_trampoline_adjust_address): Likewise.
13453 (arc_asm_trampoline_template): New function.
13454 (arc_initialize_trampoline): Use asm_trampoline_template.
13455 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13456 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13457 * config/arc/arc.md (flush_icache): Delete pattern.
13459 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13461 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13462 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13465 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13468 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13469 by not USED_FOR_TARGET.
13470 (make_pass_resolve_sw_modes): Likewise.
13472 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13474 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13477 2018-01-08 Richard Biener <rguenther@suse.de>
13479 PR middle-end/83580
13480 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13482 2018-01-08 Richard Biener <rguenther@suse.de>
13484 PR middle-end/83517
13485 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13487 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13489 PR middle-end/81897
13490 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13491 basic blocks with a small number of successors.
13492 (convert_control_dep_chain_into_preds): Improve handling of
13494 (dump_predicates): Split apart into...
13495 (dump_pred_chain): ...here...
13496 (dump_pred_info): ...and here.
13497 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13498 (can_chain_union_be_invalidated_p): Improve check for invalidation
13500 (uninit_uses_cannot_happen): Avoid unnecessary if
13501 convert_control_dep_chain_into_preds yielded nothing.
13503 2018-01-06 Martin Sebor <msebor@redhat.com>
13505 PR tree-optimization/83640
13506 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13507 subtracting negative offset from size.
13508 (builtin_access::overlap): Adjust offset bounds of the access to fall
13509 within the size of the object if possible.
13511 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13513 PR rtl-optimization/83699
13514 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13515 extract_bit_field_as_subreg to cases in which the extracted
13516 value is also a vector.
13518 * lra-constraints.c (process_alt_operands): Test for the equivalence
13519 substitutions when detecting a possible reload cycle.
13521 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13524 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13525 by default if flag_selective_schedling{,2}. Formatting fixes.
13527 PR rtl-optimization/83682
13528 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13529 if it has non-VECTOR_MODE element mode.
13530 (vec_duplicate_p): Likewise.
13532 PR middle-end/83694
13533 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13534 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13536 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13539 * config/i386/i386-builtin.def
13540 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13541 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13542 Require also OPTION_MASK_ISA_AVX512F in addition to
13543 OPTION_MASK_ISA_GFNI.
13544 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13545 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13546 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13547 to OPTION_MASK_ISA_GFNI.
13548 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13549 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13550 OPTION_MASK_ISA_AVX512BW.
13551 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13552 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13553 addition to OPTION_MASK_ISA_GFNI.
13554 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13555 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13556 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13557 to OPTION_MASK_ISA_GFNI.
13558 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13559 a requirement for all ISAs rather than any of them with a few
13561 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13563 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13564 bitmasks to be enabled with 3 exceptions, instead of requiring any
13565 enabled ISA with lots of exceptions.
13566 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13567 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13568 Change avx512bw in isa attribute to avx512f.
13569 * config/i386/sgxintrin.h: Add license boilerplate.
13570 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13571 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13572 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13573 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13575 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13576 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13577 temporarily sse2 rather than sse if not enabled already.
13580 * config/i386/sse.md (VI248_VLBW): Rename to ...
13581 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13582 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13583 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13584 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13585 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13586 mode iterator instead of VI248_VLBW.
13588 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13590 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13591 (record_modified): Skip clobbers; add debug output.
13592 (param_change_prob): Use sreal frequencies.
13594 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13596 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13597 punt for user-aligned variables.
13599 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13601 * tree-chrec.c (chrec_contains_symbols): Return true for
13604 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13607 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13608 of (x|y) == x for BICS pattern.
13610 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13612 PR tree-optimization/83605
13613 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13614 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13617 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13619 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13620 * config/epiphany/rtems.h: New file.
13622 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13623 Uros Bizjak <ubizjak@gmail.com>
13626 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13627 QIreg_operand instead of register_operand predicate.
13628 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13629 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13630 comments instead of -fmitigate[-_]rop.
13632 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13635 * cgraphunit.c (symbol_table::compile): Switch to text_section
13636 before calling assembly_start debug hook.
13637 * run-rtl-passes.c (run_rtl_passes): Likewise.
13640 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13642 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13643 range_int_cst_p rather than !symbolic_range_p before calling
13644 extract_range_from_multiplicative_op_1.
13646 2018-01-04 Jeff Law <law@redhat.com>
13648 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13649 redundant test in assertion.
13651 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13653 * doc/rtl.texi: Document machine_mode wrapper classes.
13655 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13657 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13658 using tree_to_uhwi.
13660 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13662 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13663 the VEC_PERM_EXPR fold to fail.
13665 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13668 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13669 to switched_sections.
13671 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13674 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13675 test for d.testing.
13677 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13680 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13681 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13683 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13686 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13687 is BLKmode and bitpos not zero or mode change is needed.
13689 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13692 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13695 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13698 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13699 instead of MULT rtx. Update all corresponding splitters.
13700 (*saddl_se): Ditto.
13701 (*ssub<modesuffix>): Ditto.
13702 (*ssubl_se): Ditto.
13703 (*cmp_sadd_di): Update split patterns.
13704 (*cmp_sadd_si): Ditto.
13705 (*cmp_sadd_sidi): Ditto.
13706 (*cmp_ssub_di): Ditto.
13707 (*cmp_ssub_si): Ditto.
13708 (*cmp_ssub_sidi): Ditto.
13709 * config/alpha/predicates.md (const23_operand): New predicate.
13710 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13711 Look for ASHIFT, not MULT inner operand.
13712 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13714 2018-01-04 Martin Liska <mliska@suse.cz>
13716 PR gcov-profile/83669
13717 * gcov.c (output_intermediate_file): Add version to intermediate
13719 * doc/gcov.texi: Document new field 'version' in intermediate
13720 file format. Fix location of '-k' option of gcov command.
13722 2018-01-04 Martin Liska <mliska@suse.cz>
13725 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13727 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13729 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13731 2018-01-03 Martin Sebor <msebor@redhat.com>
13733 PR tree-optimization/83655
13734 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13735 checking calls with invalid arguments.
13737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13739 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13740 (vectorizable_mask_load_store): Delete.
13741 (vectorizable_call): Return false for masked loads and stores.
13742 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13743 instead of gimple_assign_rhs1.
13744 (vectorizable_load): Handle IFN_MASK_LOAD.
13745 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13749 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13751 (vectorizable_mask_load_store): ...here.
13752 (vectorizable_load): ...and here.
13754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13756 * tree-vect-stmts.c (vect_build_all_ones_mask)
13757 (vect_build_zero_merge_argument): New functions, split out from...
13758 (vectorizable_load): ...here.
13760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13762 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13764 (vectorizable_mask_load_store): ...here.
13765 (vectorizable_store): ...and here.
13767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13769 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13771 (vectorizable_mask_load_store): ...here.
13773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13775 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13776 (vect_model_store_cost): Take a vec_load_store_type instead of a
13778 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13779 (vect_model_store_cost): Take a vec_load_store_type instead of a
13781 (vectorizable_mask_load_store): Update accordingly.
13782 (vectorizable_store): Likewise.
13783 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13787 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13788 IFN_MASK_LOAD calls here rather than...
13789 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13792 Alan Hayward <alan.hayward@arm.com>
13793 David Sherwood <david.sherwood@arm.com>
13795 * expmed.c (extract_bit_field_1): For vector extracts,
13796 fall back to extract_bit_field_as_subreg if vec_extract
13799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13800 Alan Hayward <alan.hayward@arm.com>
13801 David Sherwood <david.sherwood@arm.com>
13803 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13804 they are variable or constant sized.
13805 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13806 slots for constant-sized data.
13808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13809 Alan Hayward <alan.hayward@arm.com>
13810 David Sherwood <david.sherwood@arm.com>
13812 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13813 handling COND_EXPRs with boolean comparisons, try to find a better
13814 basis for the mask type than the boolean itself.
13816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13818 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13819 is calculated and how it can be overridden.
13820 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13821 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13823 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13827 Alan Hayward <alan.hayward@arm.com>
13828 David Sherwood <david.sherwood@arm.com>
13830 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13831 Remove the mode argument.
13832 (aarch64_simd_valid_immediate): Remove the mode and inverse
13834 * config/aarch64/iterators.md (bitsize): New iterator.
13835 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13836 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13837 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13838 aarch64_simd_valid_immediate.
13839 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13840 (aarch64_reg_or_bic_imm): Likewise.
13841 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13842 with an insn_type enum and msl with a modifier_type enum.
13843 Replace element_width with a scalar_mode. Change the shift
13844 to unsigned int. Add constructors for scalar_float_mode and
13845 scalar_int_mode elements.
13846 (aarch64_vect_float_const_representable_p): Delete.
13847 (aarch64_can_const_movi_rtx_p)
13848 (aarch64_simd_scalar_immediate_valid_for_move)
13849 (aarch64_simd_make_constant): Update call to
13850 aarch64_simd_valid_immediate.
13851 (aarch64_advsimd_valid_immediate_hs): New function.
13852 (aarch64_advsimd_valid_immediate): Likewise.
13853 (aarch64_simd_valid_immediate): Remove mode and inverse
13854 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13855 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13856 and aarch64_float_const_representable_p on the result.
13857 (aarch64_output_simd_mov_immediate): Remove mode argument.
13858 Update call to aarch64_simd_valid_immediate and use of
13859 simd_immediate_info.
13860 (aarch64_output_scalar_simd_mov_immediate): Update call
13863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13864 Alan Hayward <alan.hayward@arm.com>
13865 David Sherwood <david.sherwood@arm.com>
13867 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13868 (mode_nunits): Likewise CONST_MODE_NUNITS.
13869 * machmode.def (ADJUST_NUNITS): Document.
13870 * genmodes.c (mode_data::need_nunits_adj): New field.
13871 (blank_mode): Update accordingly.
13872 (adj_nunits): New variable.
13873 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13875 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13876 listed in adj_nunits.
13877 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13878 listed in adj_nunits. Don't emit case statements for such modes.
13879 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13880 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13881 nothing if adj_nunits is nonnull.
13882 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13883 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13884 (emit_mode_fbit): Update use of print_maybe_const_decl.
13885 (emit_move_size): Likewise. Treat the array as non-const
13887 (emit_mode_adjustments): Handle adj_nunits.
13889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13891 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13892 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13893 (VECTOR_MODES): Use it.
13894 (make_vector_modes): Take the prefix as an argument.
13896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13897 Alan Hayward <alan.hayward@arm.com>
13898 David Sherwood <david.sherwood@arm.com>
13900 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13901 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13902 for MODE_VECTOR_BOOL.
13903 * machmode.def (VECTOR_BOOL_MODE): Document.
13904 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13905 (make_vector_bool_mode): New function.
13906 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13908 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13909 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13911 * stor-layout.c (int_mode_for_mode): Likewise.
13912 * tree.c (build_vector_type_for_mode): Likewise.
13913 * varasm.c (output_constant_pool_2): Likewise.
13914 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13915 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13916 for MODE_VECTOR_BOOL.
13917 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13918 of mode class checks.
13919 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13920 instead of a list of mode class checks.
13921 (expand_vector_scalar_condition): Likewise.
13922 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13925 Alan Hayward <alan.hayward@arm.com>
13926 David Sherwood <david.sherwood@arm.com>
13928 * machmode.h (mode_size): Change from unsigned short to
13930 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13931 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13932 or if measurement_type is not polynomial.
13933 (fixed_size_mode::includes_p): Check for constant-sized modes.
13934 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13935 return a poly_uint16 rather than an unsigned short.
13936 (emit_mode_size): Change the type of mode_size from unsigned short
13937 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13938 (emit_mode_adjustments): Cope with polynomial vector sizes.
13939 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13941 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13943 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13944 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13945 * caller-save.c (setup_save_areas): Likewise.
13946 (replace_reg_with_saved_mem): Likewise.
13947 * calls.c (emit_library_call_value_1): Likewise.
13948 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13949 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13950 (gen_lowpart_for_combine): Likewise.
13951 * convert.c (convert_to_integer_1): Likewise.
13952 * cse.c (equiv_constant, cse_insn): Likewise.
13953 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13954 (cselib_subst_to_values): Likewise.
13955 * dce.c (word_dce_process_block): Likewise.
13956 * df-problems.c (df_word_lr_mark_ref): Likewise.
13957 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13958 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13959 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13960 (rtl_for_decl_location): Likewise.
13961 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13962 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13963 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13964 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13965 (expand_expr_real_1): Likewise.
13966 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13967 (pad_below): Likewise.
13968 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13969 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13970 * ira.c (get_subreg_tracking_sizes): Likewise.
13971 * ira-build.c (ira_create_allocno_objects): Likewise.
13972 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13973 (ira_sort_regnos_for_alter_reg): Likewise.
13974 * ira-costs.c (record_operand_costs): Likewise.
13975 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13976 (resolve_simple_move): Likewise.
13977 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13978 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13979 (lra_constraints): Likewise.
13980 (CONST_POOL_OK_P): Reject variable-sized modes.
13981 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13982 (add_pseudo_to_slot, lra_spill): Likewise.
13983 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13984 * optabs-query.c (get_best_extraction_insn): Likewise.
13985 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13986 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13987 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13988 * recog.c (offsettable_address_addr_space_p): Likewise.
13989 * regcprop.c (maybe_mode_change): Likewise.
13990 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13991 * regrename.c (build_def_use): Likewise.
13992 * regstat.c (dump_reg_info): Likewise.
13993 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13994 (find_reloads, find_reloads_subreg_address): Likewise.
13995 * reload1.c (eliminate_regs_1): Likewise.
13996 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13997 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13998 (simplify_binary_operation_1, simplify_subreg): Likewise.
13999 * targhooks.c (default_function_arg_padding): Likewise.
14000 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14001 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14002 (verify_gimple_assign_ternary): Likewise.
14003 * tree-inline.c (estimate_move_cost): Likewise.
14004 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14005 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14006 (get_address_cost_ainc): Likewise.
14007 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14008 (vect_supportable_dr_alignment): Likewise.
14009 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14010 (vectorizable_reduction): Likewise.
14011 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14012 (vectorizable_operation, vectorizable_load): Likewise.
14013 * tree.c (build_same_sized_truth_vector_type): Likewise.
14014 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14015 * var-tracking.c (emit_note_insn_var_location): Likewise.
14016 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14017 (ADDR_VEC_ALIGN): Likewise.
14019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14020 Alan Hayward <alan.hayward@arm.com>
14021 David Sherwood <david.sherwood@arm.com>
14023 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14025 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14026 or if measurement_type is polynomial.
14027 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14028 * combine.c (make_extraction): Likewise.
14029 * dse.c (find_shift_sequence): Likewise.
14030 * dwarf2out.c (mem_loc_descriptor): Likewise.
14031 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14032 (extract_bit_field, extract_low_bits): Likewise.
14033 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14034 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14035 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14036 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14037 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14038 * reload.c (find_reloads): Likewise.
14039 * reload1.c (alter_reg): Likewise.
14040 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14041 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14042 * tree-if-conv.c (predicate_mem_writes): Likewise.
14043 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14044 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14045 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14046 * valtrack.c (dead_debug_insert_temp): Likewise.
14047 * varasm.c (mergeable_constant_section): Likewise.
14048 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14051 Alan Hayward <alan.hayward@arm.com>
14052 David Sherwood <david.sherwood@arm.com>
14054 * expr.c (expand_assignment): Cope with polynomial mode sizes
14055 when assigning to a CONCAT.
14057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14058 Alan Hayward <alan.hayward@arm.com>
14059 David Sherwood <david.sherwood@arm.com>
14061 * machmode.h (mode_precision): Change from unsigned short to
14063 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14065 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14066 or if measurement_type is not polynomial.
14067 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14068 in which the mode is already known to be a scalar_int_mode.
14069 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14070 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14072 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14073 for GET_MODE_PRECISION.
14074 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14075 for GET_MODE_PRECISION.
14076 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14078 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14079 (expand_field_assignment, make_extraction): Likewise.
14080 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14081 (get_last_value): Likewise.
14082 * convert.c (convert_to_integer_1): Likewise.
14083 * cse.c (cse_insn): Likewise.
14084 * expr.c (expand_expr_real_1): Likewise.
14085 * lra-constraints.c (simplify_operand_subreg): Likewise.
14086 * optabs-query.c (can_atomic_load_p): Likewise.
14087 * optabs.c (expand_atomic_load): Likewise.
14088 (expand_atomic_store): Likewise.
14089 * ree.c (combine_reaching_defs): Likewise.
14090 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14091 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14092 * tree.h (type_has_mode_precision_p): Likewise.
14093 * ubsan.c (instrument_si_overflow): Likewise.
14095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14096 Alan Hayward <alan.hayward@arm.com>
14097 David Sherwood <david.sherwood@arm.com>
14099 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14100 polynomial numbers of units.
14101 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14102 (valid_vector_subparts_p): New function.
14103 (build_vector_type): Remove temporary shim and take the number
14104 of units as a poly_uint64 rather than an int.
14105 (build_opaque_vector_type): Take the number of units as a
14106 poly_uint64 rather than an int.
14107 * tree.c (build_vector_from_ctor): Handle polynomial
14108 TYPE_VECTOR_SUBPARTS.
14109 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14110 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14111 (build_vector_from_val): If the number of units is variable,
14112 use build_vec_duplicate_cst for constant operands and
14113 VEC_DUPLICATE_EXPR otherwise.
14114 (make_vector_type): Remove temporary is_constant ().
14115 (build_vector_type, build_opaque_vector_type): Take the number of
14116 units as a poly_uint64 rather than an int.
14117 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14119 * cfgexpand.c (expand_debug_expr): Likewise.
14120 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14121 (store_constructor, expand_expr_real_1): Likewise.
14122 (const_scalar_mask_from_tree): Likewise.
14123 * fold-const-call.c (fold_const_reduction): Likewise.
14124 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14125 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14126 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14127 (fold_relational_const): Likewise.
14128 (native_interpret_vector): Likewise. Change the size from an
14129 int to an unsigned int.
14130 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14131 TYPE_VECTOR_SUBPARTS.
14132 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14133 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14134 duplicating a non-constant operand into a variable-length vector.
14135 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14136 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14137 * ipa-icf.c (sem_variable::equals): Likewise.
14138 * match.pd: Likewise.
14139 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14140 * print-tree.c (print_node): Likewise.
14141 * stor-layout.c (layout_type): Likewise.
14142 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14143 * tree-cfg.c (verify_gimple_comparison): Likewise.
14144 (verify_gimple_assign_binary): Likewise.
14145 (verify_gimple_assign_ternary): Likewise.
14146 (verify_gimple_assign_single): Likewise.
14147 * tree-pretty-print.c (dump_generic_node): Likewise.
14148 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14149 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14150 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14151 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14152 (vect_shift_permute_load_chain): Likewise.
14153 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14154 (expand_vector_condition, optimize_vector_constructor): Likewise.
14155 (lower_vec_perm, get_compute_type): Likewise.
14156 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14157 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14158 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14159 (vect_recog_mask_conversion_pattern): Likewise.
14160 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14161 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14162 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14163 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14164 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14165 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14166 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14167 (supportable_widening_operation): Likewise.
14168 (supportable_narrowing_operation): Likewise.
14169 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14171 * varasm.c (output_constant): Likewise.
14173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14174 Alan Hayward <alan.hayward@arm.com>
14175 David Sherwood <david.sherwood@arm.com>
14177 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14178 so that both the length == 3 and length != 3 cases set up their
14179 own permute vectors. Add comments explaining why we know the
14180 number of elements is constant.
14181 (vect_permute_load_chain): Likewise.
14183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14184 Alan Hayward <alan.hayward@arm.com>
14185 David Sherwood <david.sherwood@arm.com>
14187 * machmode.h (mode_nunits): Change from unsigned char to
14189 (ONLY_FIXED_SIZE_MODES): New macro.
14190 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14191 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14192 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14194 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14195 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14196 or if measurement_type is not polynomial.
14197 * genmodes.c (ZERO_COEFFS): New macro.
14198 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14200 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14201 Use ZERO_COEFFS when emitting initializers.
14202 * data-streamer.h (bp_pack_poly_value): New function.
14203 (bp_unpack_poly_value): Likewise.
14204 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14205 for GET_MODE_NUNITS.
14206 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14207 for GET_MODE_NUNITS.
14208 * tree.c (make_vector_type): Remove temporary shim and make
14209 the real function take the number of units as a poly_uint64
14210 rather than an int.
14211 (build_vector_type_for_mode): Handle polynomial nunits.
14212 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14213 * emit-rtl.c (const_vec_series_p_1): Likewise.
14214 (gen_rtx_CONST_VECTOR): Likewise.
14215 * fold-const.c (test_vec_duplicate_folding): Likewise.
14216 * genrecog.c (validate_pattern): Likewise.
14217 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14218 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14219 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14220 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14221 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14222 * rtlanal.c (subreg_get_info): Likewise.
14223 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14224 (vect_grouped_load_supported): Likewise.
14225 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14226 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14227 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14228 (simplify_const_unary_operation, simplify_binary_operation_1)
14229 (simplify_const_binary_operation, simplify_ternary_operation)
14230 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14231 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14232 instead of CONST_VECTOR_NUNITS.
14233 * varasm.c (output_constant_pool_2): Likewise.
14234 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14235 explicit-encoded elements in the XVEC for variable-length vectors.
14237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14239 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14242 Alan Hayward <alan.hayward@arm.com>
14243 David Sherwood <david.sherwood@arm.com>
14245 * coretypes.h (fixed_size_mode): Declare.
14246 (fixed_size_mode_pod): New typedef.
14247 * builtins.h (target_builtins::x_apply_args_mode)
14248 (target_builtins::x_apply_result_mode): Change type to
14249 fixed_size_mode_pod.
14250 * builtins.c (apply_args_size, apply_result_size, result_vector)
14251 (expand_builtin_apply_args_1, expand_builtin_apply)
14252 (expand_builtin_return): Update accordingly.
14254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14256 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14257 * cselib.c (cselib_hash_rtx): Likewise.
14258 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14259 CONST_VECTOR encoding.
14261 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14262 Jeff Law <law@redhat.com>
14265 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14266 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14267 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14268 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14271 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14272 explicitly probe *sp in a noreturn function if there were any callee
14273 register saves or frame pointer is needed.
14275 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14278 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14279 BLKmode for ternary, binary or unary expressions.
14282 * var-tracking.c (delete_vta_debug_insn): New inline function.
14283 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14284 insns from get_insns () to NULL instead of each bb separately.
14285 Use delete_vta_debug_insn. No longer static.
14286 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14287 delete_vta_debug_insns callers.
14288 * rtl.h (delete_vta_debug_insns): Declare.
14289 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14290 instead of variable_tracking_main.
14292 2018-01-03 Martin Sebor <msebor@redhat.com>
14294 PR tree-optimization/83603
14295 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14296 arguments past the endof the argument list in functions declared
14297 without a prototype.
14298 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14299 Avoid checking when arguments are null.
14301 2018-01-03 Martin Sebor <msebor@redhat.com>
14304 * doc/extend.texi (attribute const): Fix a typo.
14305 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14306 issuing -Wsuggest-attribute for void functions.
14308 2018-01-03 Martin Sebor <msebor@redhat.com>
14310 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14311 offset_int::from instead of wide_int::to_shwi.
14312 (maybe_diag_overlap): Remove assertion.
14313 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14314 * gimple-ssa-sprintf.c (format_directive): Same.
14315 (parse_directive): Same.
14316 (sprintf_dom_walker::compute_format_length): Same.
14317 (try_substitute_return_value): Same.
14319 2018-01-03 Jeff Law <law@redhat.com>
14321 PR middle-end/83654
14322 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14323 non-constant residual for zero at runtime and avoid probing in
14324 that case. Reorganize code for trailing problem to mirror handling
14327 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14329 PR tree-optimization/83501
14330 * tree-ssa-strlen.c (get_string_cst): New.
14331 (handle_char_store): Call get_string_cst.
14333 2018-01-03 Martin Liska <mliska@suse.cz>
14335 PR tree-optimization/83593
14336 * tree-ssa-strlen.c: Include tree-cfg.h.
14337 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14338 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14339 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14341 (strlen_dom_walker::before_dom_children): Call
14342 gimple_purge_dead_eh_edges. Dump tranformation with details
14344 (strlen_dom_walker::before_dom_children): Update call by adding
14345 new argument cleanup_eh.
14346 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14348 2018-01-03 Martin Liska <mliska@suse.cz>
14351 * cif-code.def (VARIADIC_THUNK): New enum value.
14352 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14355 2018-01-03 Jan Beulich <jbeulich@suse.com>
14357 * sse.md (mov<mode>_internal): Tighten condition for when to use
14358 vmovdqu<ssescalarsize> for TI and OI modes.
14360 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14362 Update copyright years.
14364 2018-01-03 Martin Liska <mliska@suse.cz>
14367 * ipa-visibility.c (function_and_variable_visibility): Skip
14368 functions with noipa attribure.
14370 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14372 * gcc.c (process_command): Update copyright notice dates.
14373 * gcov-dump.c (print_version): Ditto.
14374 * gcov.c (print_version): Ditto.
14375 * gcov-tool.c (print_version): Ditto.
14376 * gengtype.c (create_file): Ditto.
14377 * doc/cpp.texi: Bump @copying's copyright year.
14378 * doc/cppinternals.texi: Ditto.
14379 * doc/gcc.texi: Ditto.
14380 * doc/gccint.texi: Ditto.
14381 * doc/gcov.texi: Ditto.
14382 * doc/install.texi: Ditto.
14383 * doc/invoke.texi: Ditto.
14385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14387 * vector-builder.h (vector_builder::m_full_nelts): Change from
14388 unsigned int to poly_uint64.
14389 (vector_builder::full_nelts): Update prototype accordingly.
14390 (vector_builder::new_vector): Likewise.
14391 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14392 (vector_builder::operator ==): Likewise.
14393 (vector_builder::finalize): Likewise.
14394 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14395 Take the number of elements as a poly_uint64 rather than an
14397 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14398 from unsigned int to poly_uint64.
14399 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14400 (vec_perm_indices::new_vector): Likewise.
14401 (vec_perm_indices::length): Likewise.
14402 (vec_perm_indices::nelts_per_input): Likewise.
14403 (vec_perm_indices::input_nelts): Likewise.
14404 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14405 number of elements per input as a poly_uint64 rather than an
14406 unsigned int. Use the original encoding for variable-length
14407 vectors, rather than clamping each individual element.
14408 For the second and subsequent elements in each pattern,
14409 clamp the step and base before clamping their sum.
14410 (vec_perm_indices::series_p): Handle polynomial element counts.
14411 (vec_perm_indices::all_in_range_p): Likewise.
14412 (vec_perm_indices_to_tree): Likewise.
14413 (vec_perm_indices_to_rtx): Likewise.
14414 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14415 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14416 (tree_vector_builder::new_binary_operation): Handle polynomial
14417 element counts. Return false if we need to know the number
14418 of elements at compile time.
14419 * fold-const.c (fold_vec_perm): Punt if the number of elements
14420 isn't known at compile time.
14422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14424 * vec-perm-indices.h (vec_perm_builder): Change element type
14425 from HOST_WIDE_INT to poly_int64.
14426 (vec_perm_indices::element_type): Update accordingly.
14427 (vec_perm_indices::clamp): Handle polynomial element_types.
14428 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14429 (vec_perm_indices::all_in_range_p): Likewise.
14430 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14432 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14433 polynomial vec_perm_indices element types.
14434 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14435 * fold-const.c (fold_vec_perm): Likewise.
14436 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14437 * tree-vect-generic.c (lower_vec_perm): Likewise.
14438 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14439 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14440 element type to HOST_WIDE_INT.
14442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14443 Alan Hayward <alan.hayward@arm.com>
14444 David Sherwood <david.sherwood@arm.com>
14446 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14447 rather than an int. Use plus_constant.
14448 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14449 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14452 Alan Hayward <alan.hayward@arm.com>
14453 David Sherwood <david.sherwood@arm.com>
14455 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14456 a HOST_WIDE_INT to a poly_int64.
14458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14459 Alan Hayward <alan.hayward@arm.com>
14460 David Sherwood <david.sherwood@arm.com>
14462 * calls.c (load_register_parameters): Cope with polynomial
14463 mode sizes. Require a constant size for BLKmode parameters
14464 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14465 forces a parameter to be padded at the lsb end in order to
14466 fill a complete number of words, require the parameter size
14467 to be ordered wrt UNITS_PER_WORD.
14469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14470 Alan Hayward <alan.hayward@arm.com>
14471 David Sherwood <david.sherwood@arm.com>
14473 * reload1.c (spill_stack_slot_width): Change element type
14474 from unsigned int to poly_uint64_pod.
14475 (alter_reg): Treat mode sizes as polynomial.
14477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14478 Alan Hayward <alan.hayward@arm.com>
14479 David Sherwood <david.sherwood@arm.com>
14481 * reload.c (complex_word_subreg_p): New function.
14482 (reload_inner_reg_of_subreg, push_reload): Use it.
14484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14485 Alan Hayward <alan.hayward@arm.com>
14486 David Sherwood <david.sherwood@arm.com>
14488 * lra-constraints.c (process_alt_operands): Reject matched
14489 operands whose sizes aren't ordered.
14490 (match_reload): Refer to this check here.
14492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14493 Alan Hayward <alan.hayward@arm.com>
14494 David Sherwood <david.sherwood@arm.com>
14496 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14497 that the mode size is in the set {1, 2, 4, 8, 16}.
14499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14500 Alan Hayward <alan.hayward@arm.com>
14501 David Sherwood <david.sherwood@arm.com>
14503 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14504 Use plus_constant instead of gen_rtx_PLUS.
14506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14507 Alan Hayward <alan.hayward@arm.com>
14508 David Sherwood <david.sherwood@arm.com>
14510 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14511 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14512 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14513 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14514 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14515 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14516 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14517 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14518 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14519 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14521 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14522 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14523 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14524 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14525 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14526 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14527 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14528 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14529 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14530 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14532 * expr.c (emit_move_resolve_push): Treat the input and result
14533 of PUSH_ROUNDING as a poly_int64.
14534 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14535 (emit_push_insn): Likewise.
14536 * lra-eliminations.c (mark_not_eliminable): Likewise.
14537 * recog.c (push_operand): Likewise.
14538 * reload1.c (elimination_effects): Likewise.
14539 * rtlanal.c (nonzero_bits1): Likewise.
14540 * calls.c (store_one_arg): Likewise. Require the padding to be
14541 known at compile time.
14543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14544 Alan Hayward <alan.hayward@arm.com>
14545 David Sherwood <david.sherwood@arm.com>
14547 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14548 Use plus_constant instead of gen_rtx_PLUS.
14550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14551 Alan Hayward <alan.hayward@arm.com>
14552 David Sherwood <david.sherwood@arm.com>
14554 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14555 rather than an int.
14557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14558 Alan Hayward <alan.hayward@arm.com>
14559 David Sherwood <david.sherwood@arm.com>
14561 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14562 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14563 via stack temporaries. Treat the mode size as polynomial too.
14565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14566 Alan Hayward <alan.hayward@arm.com>
14567 David Sherwood <david.sherwood@arm.com>
14569 * expr.c (expand_expr_real_2): When handling conversions involving
14570 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14571 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14572 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14581 Alan Hayward <alan.hayward@arm.com>
14582 David Sherwood <david.sherwood@arm.com>
14584 * combine.c (can_change_dest_mode): Handle polynomial
14585 REGMODE_NATURAL_SIZE.
14586 * expmed.c (store_bit_field_1): Likewise.
14587 * expr.c (store_constructor): Likewise.
14588 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14589 and polynomial REGMODE_NATURAL_SIZE.
14590 (gen_lowpart_common): Likewise.
14591 * reginfo.c (record_subregs_of_mode): Likewise.
14592 * rtlanal.c (read_modify_subreg_p): Likewise.
14594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14595 Alan Hayward <alan.hayward@arm.com>
14596 David Sherwood <david.sherwood@arm.com>
14598 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14599 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
14607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14608 Alan Hayward <alan.hayward@arm.com>
14609 David Sherwood <david.sherwood@arm.com>
14611 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14612 in a POINTER_PLUS_EXPR.
14614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14615 Alan Hayward <alan.hayward@arm.com>
14616 David Sherwood <david.sherwood@arm.com>
14618 * omp-simd-clone.c (simd_clone_subparts): New function.
14619 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14620 (ipa_simd_modify_function_body): Likewise.
14622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14623 Alan Hayward <alan.hayward@arm.com>
14624 David Sherwood <david.sherwood@arm.com>
14626 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14627 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14628 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14629 (expand_vector_condition, vector_element): Likewise.
14630 (subparts_gt): New function.
14631 (get_compute_type): Use subparts_gt.
14632 (count_type_subparts): Delete.
14633 (expand_vector_operations_1): Use subparts_gt instead of
14634 count_type_subparts.
14636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14637 Alan Hayward <alan.hayward@arm.com>
14638 David Sherwood <david.sherwood@arm.com>
14640 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14641 (vect_compile_time_alias): ...this new function. Do the calculation
14642 on poly_ints rather than trees.
14643 (vect_prune_runtime_alias_test_list): Update call accordingly.
14645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14646 Alan Hayward <alan.hayward@arm.com>
14647 David Sherwood <david.sherwood@arm.com>
14649 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14651 (vect_schedule_slp_instance): Likewise.
14653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14654 Alan Hayward <alan.hayward@arm.com>
14655 David Sherwood <david.sherwood@arm.com>
14657 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14658 constant and extern definitions for variable-length vectors.
14659 (vect_get_constant_vectors): Note that the number of units
14660 is known to be constant.
14662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14663 Alan Hayward <alan.hayward@arm.com>
14664 David Sherwood <david.sherwood@arm.com>
14666 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14667 of units as polynomial. Choose between WIDE and NARROW based
14670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14671 Alan Hayward <alan.hayward@arm.com>
14672 David Sherwood <david.sherwood@arm.com>
14674 * tree-vect-stmts.c (simd_clone_subparts): New function.
14675 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14678 Alan Hayward <alan.hayward@arm.com>
14679 David Sherwood <david.sherwood@arm.com>
14681 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14682 vectors as polynomial. Use build_index_vector for
14683 IFN_GOMP_SIMD_LANE.
14685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14686 Alan Hayward <alan.hayward@arm.com>
14687 David Sherwood <david.sherwood@arm.com>
14689 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14690 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14691 for variable-length vectors.
14692 (vectorizable_mask_load_store): Treat the number of units as
14693 polynomial, asserting that it is constant if the condition has
14694 already been enforced.
14695 (vectorizable_store, vectorizable_load): Likewise.
14697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14698 Alan Hayward <alan.hayward@arm.com>
14699 David Sherwood <david.sherwood@arm.com>
14701 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14702 of units as polynomial. Punt if we can't tell at compile time
14703 which vector contains the final result.
14705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14706 Alan Hayward <alan.hayward@arm.com>
14707 David Sherwood <david.sherwood@arm.com>
14709 * tree-vect-loop.c (vectorizable_induction): Treat the number
14710 of units as polynomial. Punt on SLP inductions. Use an integer
14711 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14712 cast of such a series for variable-length floating-point
14715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14716 Alan Hayward <alan.hayward@arm.com>
14717 David Sherwood <david.sherwood@arm.com>
14719 * tree.h (build_index_vector): Declare.
14720 * tree.c (build_index_vector): New function.
14721 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14722 of units as polynomial, forcibly converting it to a constant if
14723 vectorizable_reduction has already enforced the condition.
14724 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14725 to create a {1,2,3,...} vector.
14726 (vectorizable_reduction): Treat the number of units as polynomial.
14727 Choose vectype_in based on the largest scalar element size rather
14728 than the smallest number of units. Enforce the restrictions
14731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14732 Alan Hayward <alan.hayward@arm.com>
14733 David Sherwood <david.sherwood@arm.com>
14735 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14736 number of units as polynomial.
14738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14739 Alan Hayward <alan.hayward@arm.com>
14740 David Sherwood <david.sherwood@arm.com>
14742 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14743 * target.def (autovectorize_vector_sizes): Return the vector sizes
14744 by pointer, using vector_sizes rather than a bitmask.
14745 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14746 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14747 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14749 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14750 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14751 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14752 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14753 * omp-general.c (omp_max_vf): Likewise.
14754 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14755 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14756 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14757 * tree-vect-slp.c (vect_slp_bb): Likewise.
14758 * doc/tm.texi: Regenerate.
14759 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14761 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14762 the vector size as a poly_uint64 rather than an unsigned int.
14763 (current_vector_size): Change from an unsigned int to a poly_uint64.
14764 (get_vectype_for_scalar_type): Update accordingly.
14765 * tree.h (build_truth_vector_type): Take the size and number of
14766 units as a poly_uint64 rather than an unsigned int.
14767 (build_vector_type): Add a temporary overload that takes
14768 the number of units as a poly_uint64 rather than an unsigned int.
14769 * tree.c (make_vector_type): Likewise.
14770 (build_truth_vector_type): Take the number of units as a poly_uint64
14771 rather than an unsigned int.
14773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14774 Alan Hayward <alan.hayward@arm.com>
14775 David Sherwood <david.sherwood@arm.com>
14777 * target.def (get_mask_mode): Take the number of units and length
14778 as poly_uint64s rather than unsigned ints.
14779 * targhooks.h (default_get_mask_mode): Update accordingly.
14780 * targhooks.c (default_get_mask_mode): Likewise.
14781 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14782 * doc/tm.texi: Regenerate.
14784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14785 Alan Hayward <alan.hayward@arm.com>
14786 David Sherwood <david.sherwood@arm.com>
14788 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14789 * omp-general.c (omp_max_vf): Likewise.
14790 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14791 (expand_omp_simd): Handle polynomial safelen.
14792 * omp-low.c (omplow_simd_context): Add a default constructor.
14793 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14794 (lower_rec_simd_input_clauses): Update accordingly.
14795 (lower_rec_input_clauses): Likewise.
14797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14798 Alan Hayward <alan.hayward@arm.com>
14799 David Sherwood <david.sherwood@arm.com>
14801 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14802 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14803 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14804 (vect_analyze_slp_cost): Likewise.
14805 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14806 (vect_model_load_cost): Likewise.
14808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14809 Alan Hayward <alan.hayward@arm.com>
14810 David Sherwood <david.sherwood@arm.com>
14812 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14813 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14814 from an unsigned int * to a poly_uint64_pod *.
14815 (calculate_unrolling_factor): New function.
14816 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14819 Alan Hayward <alan.hayward@arm.com>
14820 David Sherwood <david.sherwood@arm.com>
14822 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14823 from an unsigned int to a poly_uint64.
14824 (_loop_vec_info::slp_unrolling_factor): Likewise.
14825 (_loop_vec_info::vectorization_factor): Change from an int
14827 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14828 (vect_get_num_vectors): New function.
14829 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14830 (vect_get_num_copies): Use vect_get_num_vectors.
14831 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14832 to an unsigned int *.
14833 (vect_analyze_data_refs): Change min_vf from an int * to a
14835 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14836 than an unsigned HOST_WIDE_INT.
14837 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14838 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14839 to an unsigned int *.
14840 (vect_analyze_data_ref_dependences): Likewise.
14841 (vect_compute_data_ref_alignment): Handle polynomial vf.
14842 (vect_enhance_data_refs_alignment): Likewise.
14843 (vect_prune_runtime_alias_test_list): Likewise.
14844 (vect_shift_permute_load_chain): Likewise.
14845 (vect_supportable_dr_alignment): Likewise.
14846 (dependence_distance_ge_vf): Take the vectorization factor as a
14847 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14848 (vect_analyze_data_refs): Change min_vf from an int * to a
14850 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14851 vfm1 as a poly_uint64 rather than an int. Make the same change
14852 for the returned bound_scalar.
14853 (vect_gen_vector_loop_niters): Handle polynomial vf.
14854 (vect_do_peeling): Likewise. Update call to
14855 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14856 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14858 * tree-vect-loop.c (vect_determine_vectorization_factor)
14859 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14860 (vect_get_known_peeling_cost): Likewise.
14861 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14862 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14863 (vect_transform_loop): Likewise. Use the lowest possible VF when
14864 updating the upper bounds of the loop.
14865 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14866 rather than an int.
14867 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14868 polynomial unroll factors.
14869 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14870 (vect_make_slp_decision): Likewise.
14871 (vect_supported_load_permutation_p): Likewise, and polynomial
14873 (vect_analyze_slp_cost): Handle polynomial vf.
14874 (vect_slp_analyze_node_operations): Likewise.
14875 (vect_slp_analyze_bb_1): Likewise.
14876 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14877 than an unsigned HOST_WIDE_INT.
14878 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14879 (vectorizable_load): Handle polynomial vf.
14880 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14882 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14885 Alan Hayward <alan.hayward@arm.com>
14886 David Sherwood <david.sherwood@arm.com>
14888 * match.pd: Handle bit operations involving three constants
14889 and try to fold one pair.
14891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14893 * tree-vect-loop-manip.c: Include gimple-fold.h.
14894 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14895 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14896 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14897 Add a path that uses a step of VF instead of 1, but disable it
14899 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14900 and niters_no_overflow parameters. Update calls to
14901 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14902 Create a new SSA name if the latter choses to use a ste other
14903 than zero, and return it via niters_vector_mult_vf_var.
14904 * tree-vect-loop.c (vect_transform_loop): Update calls to
14905 vect_do_peeling, vect_gen_vector_loop_niters and
14906 slpeel_make_loop_iterate_ntimes.
14907 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14908 (vect_gen_vector_loop_niters): Update declarations after above changes.
14910 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14912 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14913 128-bit round to integer instructions.
14914 (ceil<mode>2): Likewise.
14915 (btrunc<mode>2): Likewise.
14916 (round<mode>2): Likewise.
14918 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14920 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14921 unaligned VSX load/store on P8/P9.
14922 (expand_block_clear): Allow the use of unaligned VSX
14923 load/store on P8/P9.
14925 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14927 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14929 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14930 swap associated with both a load and a store.
14932 2018-01-02 Andrew Waterman <andrew@sifive.com>
14934 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14935 * config/riscv/riscv.md (clear_cache): Use it.
14937 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14939 * web.c: Remove out-of-date comment.
14941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14943 * expr.c (fixup_args_size_notes): Check that any existing
14944 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14945 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14946 (emit_single_push_insn): ...here.
14948 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14950 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14951 (const_vector_encoded_nelts): New function.
14952 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14953 (const_vector_int_elt, const_vector_elt): Declare.
14954 * emit-rtl.c (const_vector_int_elt_1): New function.
14955 (const_vector_elt): Likewise.
14956 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14957 of CONST_VECTOR_ELT.
14959 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14961 * expr.c: Include rtx-vector-builder.h.
14962 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14963 directly on the tree encoding.
14964 (const_vector_from_tree): Likewise.
14965 * optabs.c: Include rtx-vector-builder.h.
14966 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14967 sequence of "u" values.
14968 * vec-perm-indices.c: Include rtx-vector-builder.h.
14969 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14970 directly on the vec_perm_indices encoding.
14972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14974 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14975 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14976 * rtx-vector-builder.h: New file.
14977 * rtx-vector-builder.c: Likewise.
14978 * rtl.h (rtx_def::u2): Add a const_vector field.
14979 (CONST_VECTOR_NPATTERNS): New macro.
14980 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14981 (CONST_VECTOR_DUPLICATE_P): Likewise.
14982 (CONST_VECTOR_STEPPED_P): Likewise.
14983 (CONST_VECTOR_ENCODED_ELT): Likewise.
14984 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14985 (unwrap_const_vec_duplicate): Likewise.
14986 (const_vec_series_p): Check for a non-duplicated vector encoding.
14987 Say that the function only returns true for integer vectors.
14988 * emit-rtl.c: Include rtx-vector-builder.h.
14989 (gen_const_vec_duplicate_1): Delete.
14990 (gen_const_vector): Call gen_const_vec_duplicate instead of
14991 gen_const_vec_duplicate_1.
14992 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14993 (gen_const_vec_duplicate): Use rtx_vector_builder.
14994 (gen_const_vec_series): Likewise.
14995 (gen_rtx_CONST_VECTOR): Likewise.
14996 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14997 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14998 Build a new vector rather than modifying a CONST_VECTOR in-place.
14999 (handle_special_swappables): Update call accordingly.
15000 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15001 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15002 Build a new vector rather than modifying a CONST_VECTOR in-place.
15003 (handle_special_swappables): Update call accordingly.
15005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15007 * simplify-rtx.c (simplify_const_binary_operation): Use
15008 CONST_VECTOR_ELT instead of XVECEXP.
15010 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15012 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15013 the selector elements to be different from the data elements
15014 if the selector is a VECTOR_CST.
15015 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15016 ssizetype for the selector.
15018 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15020 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15021 before testing each element individually.
15022 * tree-vect-generic.c (lower_vec_perm): Likewise.
15024 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15026 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15027 * selftest-run-tests.c (selftest::run_tests): Call it.
15028 * vector-builder.h (vector_builder::operator ==): New function.
15029 (vector_builder::operator !=): Likewise.
15030 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15031 (vec_perm_indices::all_from_input_p): New function.
15032 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15033 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15034 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15035 instead of reading the VECTOR_CST directly. Detect whether both
15036 vector inputs are the same before constructing the vec_perm_indices,
15037 and update the number of inputs argument accordingly. Use the
15038 utility functions added above. Only construct sel2 if we need to.
15040 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15042 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15043 the broadcast of the low byte.
15044 (expand_mult_highpart): Use an explicit encoding for the permutes.
15045 * optabs-query.c (can_mult_highpart_p): Likewise.
15046 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15047 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15048 (vectorizable_bswap): Likewise.
15049 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15050 explicit encoding for the power-of-2 permutes.
15051 (vect_permute_store_chain): Likewise.
15052 (vect_grouped_load_supported): Likewise.
15053 (vect_permute_load_chain): Likewise.
15055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15057 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15058 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15059 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15060 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15061 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15062 (vect_gen_perm_mask_any): Likewise.
15064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15066 * int-vector-builder.h: New file.
15067 * vec-perm-indices.h: Include int-vector-builder.h.
15068 (vec_perm_indices): Redefine as an int_vector_builder.
15069 (auto_vec_perm_indices): Delete.
15070 (vec_perm_builder): Redefine as a stand-alone class.
15071 (vec_perm_indices::vec_perm_indices): New function.
15072 (vec_perm_indices::clamp): Likewise.
15073 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15074 (vec_perm_indices::new_vector): New function.
15075 (vec_perm_indices::new_expanded_vector): Update for new
15076 vec_perm_indices class.
15077 (vec_perm_indices::rotate_inputs): New function.
15078 (vec_perm_indices::all_in_range_p): Operate directly on the
15079 encoded form, without computing elided elements.
15080 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15081 encoding. Update for new vec_perm_indices class.
15082 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15083 the given vec_perm_builder.
15084 (expand_vec_perm_var): Update vec_perm_builder constructor.
15085 (expand_mult_highpart): Use vec_perm_builder instead of
15086 auto_vec_perm_indices.
15087 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15088 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15089 or double series encoding as appropriate.
15090 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15091 vec_perm_indices instead of auto_vec_perm_indices.
15092 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15093 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15094 (vect_permute_store_chain): Likewise.
15095 (vect_grouped_load_supported): Likewise.
15096 (vect_permute_load_chain): Likewise.
15097 (vect_shift_permute_load_chain): Likewise.
15098 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15099 (vect_transform_slp_perm_load): Likewise.
15100 (vect_schedule_slp_instance): Likewise.
15101 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15102 (vectorizable_mask_load_store): Likewise.
15103 (vectorizable_bswap): Likewise.
15104 (vectorizable_store): Likewise.
15105 (vectorizable_load): Likewise.
15106 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15107 vec_perm_indices instead of auto_vec_perm_indices. Use
15108 tree_to_vec_perm_builder to read the vector from a tree.
15109 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15110 vec_perm_builder instead of a vec_perm_indices.
15111 (have_whole_vector_shift): Use vec_perm_builder and
15112 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15113 truncation to calc_vec_perm_mask_for_shift.
15114 (vect_create_epilog_for_reduction): Likewise.
15115 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15116 from auto_vec_perm_indices to vec_perm_indices.
15117 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15118 instead of changing individual elements.
15119 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15120 the vector in d.perm.
15121 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15122 from auto_vec_perm_indices to vec_perm_indices.
15123 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15124 instead of changing individual elements.
15125 (arm_vectorize_vec_perm_const): Use new_vector to install
15126 the vector in d.perm.
15127 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15128 Update vec_perm_builder constructor.
15129 (rs6000_expand_interleave): Likewise.
15130 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15131 (rs6000_expand_interleave): Likewise.
15133 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15135 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15136 to qimode could truncate the indices.
15137 * optabs.c (expand_vec_perm_var): Likewise.
15139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15141 * Makefile.in (OBJS): Add vec-perm-indices.o.
15142 * vec-perm-indices.h: New file.
15143 * vec-perm-indices.c: Likewise.
15144 * target.h (vec_perm_indices): Replace with a forward class
15146 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15147 * optabs.h: Include vec-perm-indices.h.
15148 (expand_vec_perm): Delete.
15149 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15150 (expand_vec_perm_const): Declare.
15151 * target.def (vec_perm_const_ok): Replace with...
15152 (vec_perm_const): ...this new hook.
15153 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15154 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15155 * doc/tm.texi: Regenerate.
15156 * optabs.def (vec_perm_const): Delete.
15157 * doc/md.texi (vec_perm_const): Likewise.
15158 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15159 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15160 expand_vec_perm for constant permutation vectors. Assert that
15161 the mode of variable permutation vectors is the integer equivalent
15162 of the mode that is being permuted.
15163 * optabs-query.h (selector_fits_mode_p): Declare.
15164 * optabs-query.c: Include vec-perm-indices.h.
15165 (selector_fits_mode_p): New function.
15166 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15167 is defined, instead of checking whether the vec_perm_const_optab
15168 exists. Use targetm.vectorize.vec_perm_const instead of
15169 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15170 fit in the vector mode before using a variable permute.
15171 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15172 vec_perm_indices instead of an rtx.
15173 (expand_vec_perm): Replace with...
15174 (expand_vec_perm_const): ...this new function. Take the selector
15175 as a vec_perm_indices rather than an rtx. Also take the mode of
15176 the selector. Update call to shift_amt_for_vec_perm_mask.
15177 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15178 Use vec_perm_indices::new_expanded_vector to expand the original
15179 selector into bytes. Check whether the indices fit in the vector
15180 mode before using a variable permute.
15181 (expand_vec_perm_var): Make global.
15182 (expand_mult_highpart): Use expand_vec_perm_const.
15183 * fold-const.c: Includes vec-perm-indices.h.
15184 * tree-ssa-forwprop.c: Likewise.
15185 * tree-vect-data-refs.c: Likewise.
15186 * tree-vect-generic.c: Likewise.
15187 * tree-vect-loop.c: Likewise.
15188 * tree-vect-slp.c: Likewise.
15189 * tree-vect-stmts.c: Likewise.
15190 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15192 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15193 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15194 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15195 (aarch64_vectorize_vec_perm_const): ...this new function.
15196 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15197 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15198 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15199 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15200 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15201 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15202 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15204 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15205 check for NEON modes.
15206 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15207 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15208 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15209 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15211 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15212 the old VEC_PERM_CONST conditions.
15213 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15214 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15215 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15216 (ia64_vectorize_vec_perm_const_ok): Merge into...
15217 (ia64_vectorize_vec_perm_const): ...this new function.
15218 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15219 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15220 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15221 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15222 * config/mips/mips.c (mips_expand_vec_perm_const)
15223 (mips_vectorize_vec_perm_const_ok): Merge into...
15224 (mips_vectorize_vec_perm_const): ...this new function.
15225 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15226 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15227 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15228 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15229 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15230 (rs6000_expand_vec_perm_const): Delete.
15231 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15233 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15234 (altivec_expand_vec_perm_const_le): Take each operand individually.
15235 Operate on constant selectors rather than rtxes.
15236 (altivec_expand_vec_perm_const): Likewise. Update call to
15237 altivec_expand_vec_perm_const_le.
15238 (rs6000_expand_vec_perm_const): Delete.
15239 (rs6000_vectorize_vec_perm_const_ok): Delete.
15240 (rs6000_vectorize_vec_perm_const): New function.
15241 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15242 an element count and rtx array.
15243 (rs6000_expand_extract_even): Update call accordingly.
15244 (rs6000_expand_interleave): Likewise.
15245 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15246 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15247 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15248 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15249 (rs6000_expand_vec_perm_const): Delete.
15250 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15251 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15252 (altivec_expand_vec_perm_const_le): Take each operand individually.
15253 Operate on constant selectors rather than rtxes.
15254 (altivec_expand_vec_perm_const): Likewise. Update call to
15255 altivec_expand_vec_perm_const_le.
15256 (rs6000_expand_vec_perm_const): Delete.
15257 (rs6000_vectorize_vec_perm_const_ok): Delete.
15258 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15259 reference to the SPE evmerge intructions.
15260 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15261 an element count and rtx array.
15262 (rs6000_expand_extract_even): Update call accordingly.
15263 (rs6000_expand_interleave): Likewise.
15264 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15265 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15267 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15269 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15271 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15272 vector mode and that that mode matches the mode of the data
15274 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15275 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15276 directly using expand_vec_perm_1 when forcing selectors into
15278 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15280 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15282 * optabs-query.h (can_vec_perm_p): Delete.
15283 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15284 * optabs-query.c (can_vec_perm_p): Split into...
15285 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15286 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15287 particular selector is valid.
15288 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15289 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15290 (vect_grouped_load_supported): Likewise.
15291 (vect_shift_permute_load_chain): Likewise.
15292 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15293 (vect_transform_slp_perm_load): Likewise.
15294 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15295 (vectorizable_bswap): Likewise.
15296 (vect_gen_perm_mask_checked): Likewise.
15297 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15298 implementations of variable permutation vectors into account
15299 when deciding which selector to use.
15300 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15301 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15302 with a false third argument.
15303 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15304 to test whether the constant selector is valid and can_vec_perm_var_p
15305 to test whether a variable selector is valid.
15307 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15309 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15310 * optabs-query.c (can_vec_perm_p): Likewise.
15311 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15312 instead of vec_perm_indices.
15313 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15314 (vect_gen_perm_mask_checked): Likewise,
15315 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15316 (vect_gen_perm_mask_checked): Likewise,
15318 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15320 * optabs-query.h (qimode_for_vec_perm): Declare.
15321 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15322 (qimode_for_vec_perm): ...this new function.
15323 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15325 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15327 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15328 does not have a conditional at the top.
15330 2018-01-02 Richard Biener <rguenther@suse.de>
15332 * ipa-inline.c (big_speedup_p): Fix expression.
15334 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15337 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15340 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15344 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15345 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15346 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15347 cond_taken_branch_cost 3->4.
15349 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15351 PR tree-optimization/83581
15352 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15353 TODO_cleanup_cfg if any changes have been made.
15355 PR middle-end/83608
15356 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15357 convert_modes if target mode has the right side, but different mode
15360 PR middle-end/83609
15361 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15362 last argument when extracting from CONCAT. If either from_real or
15363 from_imag is NULL, use expansion through memory. If result is not
15364 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15365 the parts directly to inner mode, if even that fails, use expansion
15368 PR middle-end/83623
15369 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15370 check for bswap in mode rather than HImode and use that in expand_unop
15373 Copyright (C) 2018 Free Software Foundation, Inc.
15375 Copying and distribution of this file, with or without modification,
15376 are permitted in any medium without royalty provided the copyright
15377 notice and this notice are preserved.