1 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4 track if we pass or return IEEE 128-bit floating point.
5 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
6 C++ mangling that is compatible with GCC 8.1.
7 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
8 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10 (rs6000_function_arg_advance_1): Likewise.
11 (rs6000_mangle_type): Optionally generate mangled names that match
12 what GCC 8.1 generated for IEEE 128-bit floating point types.
13 (rs6000_globalize_decl_name): If we have an external function that
14 passes or returns IEEE 128-bit types, generate a weak reference
15 from the mangled name used in GCC 8.1 to the current mangled
17 (rs6000_init_builtins): Make __ibm128 use the long double type if
18 long double is IBM extended double. Make __float128 use the long
19 double type if long double is IEEE 128-bit.
21 2018-06-06 Jim Wilson <jimw@sifive.com>
23 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
24 (struct machine_function): New field interrupt_mode.
25 (riscv_handle_type_attribute): New function. Add forward declaration.
26 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
27 (riscv_expand_epilogue): Check interrupt_mode field.
28 (riscv_set_current_function): Check interrupt attribute args and
29 set interrupt_mode field.
30 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
31 (riscv_sret, riscv_uret): New.
32 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
33 new arguments to interrupt attribute.
35 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
38 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
39 Don't handle -mcpu=power8 if -mpower9-vector is also used.
41 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
43 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
44 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
45 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
46 several redundant entries.
48 2018-06-06 David Malcolm <dmalcolm@redhat.com>
50 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
51 type from "rtx" to "rtx_insn *".
52 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
53 for local "call_insn", removing cast.
54 (ix86_expand_call): Likewise, introducing a "call_insn" local.
56 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
58 PR tree-optimization/86066
59 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
60 for BIT_INSERT_EXPR stores.
62 2018-06-06 Richard Biener <rguenther@suse.de>
64 PR tree-optimization/86062
65 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
67 of to be offsetted base.
69 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
71 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
72 to be static and remove check on interrupt attribute name.
74 gcc/testsuite/gcc.target/msp430/
75 * function-attributes-4.c: New test.
76 * static-interrupts.c: New test.
78 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
80 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
81 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
83 2018-06-05 Steve Ellcey <sellcey@cavium.com>
86 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
88 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
89 Remove second argument, change how error is called.
90 (aarch64_layout_arg): Remove second argument from
91 aarch64_err_no_fpadvsimd call.
92 (aarch64_init_cumulative_args): Ditto.
93 (aarch64_gimplify_va_arg_expr): Ditto.
94 * config/aarch64/aarch64.md (mov<mode>): Ditto.
96 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
98 * config/i386/i386.md (simple_return_indirect_internal): New expander.
99 (*simple_return_indirect_internal<mode>): Rename from
100 simple_return_indirect_internal. Use W mode iterator.
101 (rstorssp): New expander.
102 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
103 (clrssbsy): New expander.
104 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
106 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
108 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
110 (cmse_check_pointed_object): Likewise.
112 2018-06-05 Martin Liska <mliska@suse.cz>
114 PR gcov-profile/47618
115 * doc/invoke.texi: Document how -fprofile-dir format
118 2018-06-05 Richard Biener <rguenther@suse.de>
120 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
121 removal pretend DOM info isn't available so we do not update
122 it and only remove edges, not dominated blocks. Actually free
123 DOM info in case we removed something. Remove unreachable blocks.
124 (mfb_keep_latches): Work with either DOM info or marked backedges.
125 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
126 first. Mark backedges if DOM info isn't available.
127 (Re-)compute DOM info after cleanup_control_flow_pre.
129 2018-06-05 Richard Biener <rguenther@suse.de>
131 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
132 (locus_discrim_hasher::hash): Adjust.
133 (locus_discrim_hasher::equal): Likewise.
134 (next_discriminator_for_locus): Work on line directly.
135 (same_line_p): Pass in expanded locus1 as well.
136 (assign_discriminators): Avoid redundant location expansions.
138 2018-06-05 Richard Biener <rguenther@suse.de>
140 PR tree-optimization/86046
141 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
142 if required after clearing TREE_ADDRESSABLE.
144 2018-06-05 Richard Biener <rguenther@suse.de>
146 PR tree-optimization/86047
147 * tree-ssa-loop.c (for_each_index): Glob handling of all
148 decls and constants and really handle all of them.
150 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
153 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
154 qualifier_void_pointer and qualifier_const_void_pointer.
155 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
156 (arm_init_builtins): Handle the above.
157 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
158 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
159 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
162 2018-06-05 Martin Liska <mliska@suse.cz>
164 * auto-profile.c (read_autofdo_file): Do not use
165 gcov_ctr_summary struct.
166 (afdo_callsite_hot_enough_for_early_inline): Likewise.
167 * coverage.c (struct counts_entry): Likewise.
168 (read_counts_file): Read just single summary entry.
169 (get_coverage_counts): Use gcov_summary struct.
170 * coverage.h (get_coverage_counts): Likewise.
171 * gcov-dump.c (dump_working_sets): Likewise.
172 (tag_summary): Dump just single summary.
173 * gcov-io.c (gcov_write_summary): Write just histogram
175 (gcov_read_summary): Read just single summary.
176 (compute_working_sets): Use gcov_summary struct.
177 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
178 of GCOV_COUNTERS_SUMMABLE.
179 (GCOV_COUNTERS_SUMMABLE): Remove.
180 (GCOV_FIRST_VALUE_COUNTER): Replace with
181 GCOV_COUNTER_V_INTERVAL.
182 (struct gcov_ctr_summary): Remove.
183 (struct gcov_summary): Directly use fields of former
185 (compute_working_sets): Use gcov_summary struct.
186 * gcov.c (read_count_file): Do not use ctrs fields.
187 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
189 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
191 * profile.c: Likewise.
192 * profile.h: Likewise.
194 2018-06-05 Martin Liska <mliska@suse.cz>
196 PR gcov-profile/84846
197 * gcov.c (output_lines): Print working directory only
198 in intermediate format.
200 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
202 * config/s390/s390-builtin-types.def: Add void function type.
203 * config/s390/s390-builtins.def: Use the function type for the
206 2018-06-04 Jim Wilson <jimw@sifive.com>
208 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
210 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
211 and maybe_eh_return. Change regno to unsigned int. Use new args to
212 handle EH_RETURN_DATA_REGNO registers properly.
213 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
214 (riscv_expand_epilogue): Update comment. Change argument name and
215 type. Update code to use new name and type. Pass new args to
216 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
218 * config/riscv/riscv.md (NORMAL_RETURN): New.
219 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
220 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
221 (eh_return): Call gen_eh_return_internal and emit barrier.
222 (eh_return_internal): Call riscv_expand_epilogue.
224 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
226 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
227 bit_insertion field and declare can_be_merged_into method.
228 (merged_store_group::can_be_merged_into): New method.
229 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
230 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
231 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
233 2018-06-04 Richard Biener <rguenther@suse.de>
235 PR tree-optimization/85955
236 * builtins.c (fold_builtin_sincos): Convert pointers to
237 destination to appropriate type before dereferencing.
239 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
241 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
243 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
245 * expr.c (expand_expr_real_1): Force the operand into memory if
246 its TYPE_MODE is BLKmode and if there is no integer mode for
247 the number of bits being extracted.
249 2018-06-04 Jakub Jelinek <jakub@redhat.com>
253 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
254 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
256 2018-06-04 Richard Biener <rguenther@suse.de>
258 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
259 (cleanup_tree_cfg_noloop): ... single caller. Do
260 start_recording_case_labels later.
262 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
264 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
265 to _IMMINTRIN_H_INCLUDED.
266 * config/i386/pconfigintrin.h: Ditto.
267 * config/i386/waitpkgintrin.h: Ditto.
268 * config/i386/immintrin.h: Add includes for sgxintrin.h,
269 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
270 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
271 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
272 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
273 waitpkgintrin.h and cldemoteintrin.h.
275 2018-06-04 Richard Biener <rguenther@suse.de>
277 PR tree-optimization/86038
278 * tracer.c (find_best_successor): Check probability for
279 being initialized, bail out if not.
281 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
284 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
285 of bits to ignore when comparing architectures.
287 2018-06-04 Jakub Jelinek <jakub@redhat.com>
289 PR tree-optimization/69615
290 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
291 maximum or minimum of the type, try to merge it also as if
292 range1 is + [-, x - 1] or + [x + 1, -].
295 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
297 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
299 PR tree-optimization/86034
300 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
301 the unsigned bitfield type in a bit insertion sequence if it does not
302 have a larger precision than the bitfield size.
303 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
305 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
307 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
309 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
311 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
312 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
313 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
314 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
316 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
318 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
319 Disable -fdelete-null-pointer-checks for ELF toolchain.
321 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
322 Kito Cheng <kito.cheng@gmail.com>
324 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
325 (nds32le-*-*, nds32be-*-*): Integrate checking process.
326 (nds32*-*-*): Add glibc and uclibc conditions.
327 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
328 (TARGET_EXCEPT_UNWIND_INFO): Define.
329 * config/nds32/elf.h: New file.
330 * config/nds32/linux.h: New file.
331 * config/nds32/nds32-elf.opt: New file.
332 * config/nds32/nds32-linux.opt: New file.
333 * config/nds32/nds32-fp-as-gp.c
334 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
335 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
337 (nds32_asm_file_end): Ditto.
338 (nds32_print_operand): Ditto.
339 (nds32_insert_attributes): Ditto.
340 (nds32_init_libfuncs): New function.
341 (TARGET_HAVE_TLS): Define.
342 (TARGET_INIT_LIBFUNCS): Define.
343 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
345 (TARGET_ELF): Apply different mcmodel setting.
346 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
347 been migrated into elf.h and linux.h files.
348 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
349 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
350 (mcmodel): The content has been migrated into nds32-elf.opt and
351 nds32-linux.opt files.
352 * config/nds32/t-elf: New file.
353 * config/nds32/t-linux: New file.
355 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
356 Shiva Chen <shiva0217@gmail.com>
358 * config/nds32/constants.md (unspec_volatile_element): Add
359 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
360 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
362 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
363 (make_pass_nds32_fp_as_gp): Declare.
364 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
366 (nds32_asm_function_end_prologue): Remove unused asm output.
367 (nds32_asm_function_begin_epilogue): Remove unused asm output.
368 (nds32_asm_file_start): Output necessary fp_as_gp information.
369 (nds32_option_override): Adjust register usage.
370 (nds32_expand_prologue): Consider fp_as_gp situation.
371 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
372 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
375 (simple_return): Ditto.
376 (omit_fp_begin): Output special directive for fp_as_gp.
377 (omit_fp_end): Output special directive for fp_as_gp.
378 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
379 mforbid-fp-as-gp): New options.
381 2018-06-01 Mark Wielaard <mark@klomp.org>
383 * dwarf2out.c (dwarf2out_finish): Remove generation of
386 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
388 * gimple-ssa-store-merging.c: Include gimple-fold.h.
389 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
390 (struct merged_store_group): Add bit_insertion field.
391 (dump_char_array): Use standard hexadecimal format.
392 (merged_store_group::merged_store_group): Set bit_insertion to false.
393 (merged_store_group::apply_stores): Use optimal buffer size. Deal
394 with BIT_INSERT_EXPR stores. Move up code updating the mask and
395 also print the mask in the dump file.
396 (pass_store_merging::gate): Minor tweak.
397 (imm_store_chain_info::coalesce_immediate): Fix wrong association
398 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
399 stores with INTEGER_CST stores.
400 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
401 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
402 and use it throughout. Generate bit insertion sequences if need be.
403 (pass_store_merging::process_store): Remove redundant condition.
404 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
406 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
408 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
409 the 128-bit floating point types. Fix function comment.
411 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
413 * config/aarch64/aarch64-simd.md
414 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
416 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
419 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
421 PR tree-optimization/85989
422 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
424 (backprop::intersect_uses): Check it when deciding whether this
425 is a backedge reference.
426 (backprop::process_block): Add each phi to m_visited_phis
427 after visiting it, then clear it at the end.
429 2018-06-01 Richard Biener <rguenther@suse.de>
431 * tree-vectorizer.h (vect_dr_stmt): New function.
432 (vect_get_load_cost): Adjust.
433 (vect_get_store_cost): Likewise.
434 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
435 Use vect_dr_stmt instead of DR_SMTT.
436 (vect_record_base_alignments): Likewise.
437 (vect_calculate_target_alignment): Likewise.
438 (vect_compute_data_ref_alignment): Likewise and make static.
439 (vect_update_misalignment_for_peel): Likewise.
440 (vect_verify_datarefs_alignment): Likewise.
441 (vector_alignment_reachable_p): Likewise.
442 (vect_get_data_access_cost): Likewise. Pass down
443 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
444 (vect_get_peeling_costs_all_drs): Likewise.
445 (vect_peeling_hash_get_lowest_cost): Likewise.
446 (vect_enhance_data_refs_alignment): Likewise.
447 (vect_find_same_alignment_drs): Likewise.
448 (vect_analyze_data_refs_alignment): Likewise.
449 (vect_analyze_group_access_1): Likewise.
450 (vect_analyze_group_access): Likewise.
451 (vect_analyze_data_ref_access): Likewise.
452 (vect_analyze_data_ref_accesses): Likewise.
453 (vect_vfa_segment_size): Likewise.
454 (vect_small_gap_p): Likewise.
455 (vectorizable_with_step_bound_p): Likewise.
456 (vect_prune_runtime_alias_test_list): Likewise.
457 (vect_analyze_data_refs): Likewise.
458 (vect_supportable_dr_alignment): Likewise.
459 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
460 (vect_gen_prolog_loop_niters): Likewise.
461 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
462 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
464 (vect_recog_mask_conversion_pattern): Likewise.
465 (vect_try_gather_scatter_pattern): Likewise.
466 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
467 to vect_get_store_cost.
468 (vect_get_store_cost): Get stmt_info instead of DR.
469 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
470 (vect_get_load_cost): Get stmt_info instead of DR.
472 2018-06-01 Richard Biener <rguenther@suse.de>
475 * gimple-fold.c (var_decl_component_p): Also allow offsetted
476 vars wrapped in MEM_REFs.
478 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
480 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
481 Fix subreg tests so that we only return a choice between
482 GENERAL_REGS and FP_REGS if the original classes included both.
484 2018-06-01 Richard Biener <rguenther@suse.de>
487 * tree-ssa-structalias.c (get_function_part_constraint):
488 Handle NULL fi->decl.
489 (find_func_aliases_for_call): Properly handle indirect
491 (find_func_clobbers): Likewise.
492 (ipa_pta_execute): Likewise.
493 (create_variable_info_for): For functions that are ifunc_resolver
494 resolve to a varinfo that contains the result of the resolver call.
495 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
498 2018-05-31 Michael Collison <michael.collison@arm.com>
500 * config/aarch64/aarch64.md:
501 (*fix_to_zero_extenddfdi2): New pattern.
502 * gcc.target/aarch64/fix_extend1.c: New testcase.
504 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
508 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
509 and BUILT_IN_STRNCMP_EQ.
510 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
512 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
513 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
514 (gimple_fold_builtin): Likewise.
515 * tree-ssa-strlen.c (compute_string_length): New function.
516 (determine_min_obsize): New function.
517 (handle_builtin_string_cmp): New function to handle calls to
518 string compare functions.
519 (strlen_optimize_stmt): Add handling to builtin string compare
521 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
522 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
523 * tree.c (build_common_builtin_nodes): Add new defines of
524 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
526 2018-05-31 Jakub Jelinek <jakub@redhat.com>
529 * bb-reorder.c (pass_partition_blocks::gate): Return false for
530 functions with naked attribute.
532 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
534 * config/i386/sse.md (avx_vec_concat<mode>):
535 Substitute concat_tg_mode mode attribute with xtg_mode.
536 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
537 (concat_tg_mode): Remove mode attribute.
539 2018-05-31 Martin Sebor <msebor@redhat.com>
542 * calls.c (alloc_max_size): Correct a logic error/typo.
543 Treat excessive arguments as infinite. Warn for invalid arguments.
544 * doc/invoke.texi (-Walloc-size-larger-than): Update.
546 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
549 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
550 and movx for Haswell.
552 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
553 Cesar Philippidis <cesar@codesourcery.com>
556 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
557 when emitting error on private/firstprivate reductions.
558 * omp-low.c (lower_omp_target): Avoid reference-type processing
559 on pointers for firstprivate clause.
561 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
563 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
566 * config/aarch64/aarch64-simd.md
567 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
568 (aarch64_ld1_x3_<mode>): Likewise
569 (aarch64_st1x2<VALLDIF:mode>): Likewise
570 (aarch64_st1_x2_<mode>): Likewise
571 (aarch64_st1x3<VALLDIF:mode>): Likewise
572 (aarch64_st1_x3_<mode>): Likewise
573 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
574 (vld1_s8_x3): Likewise.
575 (vld1_u16_x3): Likewise.
576 (vld1_s16_x3): Likewise.
577 (vld1_u32_x3): Likewise.
578 (vld1_s32_x3): Likewise.
579 (vld1_u64_x3): Likewise.
580 (vld1_s64_x3): Likewise.
581 (vld1_f16_x3): Likewise.
582 (vld1_f32_x3): Likewise.
583 (vld1_f64_x3): Likewise.
584 (vld1_p8_x3): Likewise.
585 (vld1_p16_x3): Likewise.
586 (vld1_p64_x3): Likewise.
587 (vld1q_u8_x3): Likewise.
588 (vld1q_s8_x3): Likewise.
589 (vld1q_u16_x3): Likewise.
590 (vld1q_s16_x3): Likewise.
591 (vld1q_u32_x3): Likewise.
592 (vld1q_s32_x3): Likewise.
593 (vld1q_u64_x3): Likewise.
594 (vld1q_s64_x3): Likewise.
595 (vld1q_f16_x3): Likewise.
596 (vld1q_f32_x3): Likewise.
597 (vld1q_f64_x3): Likewise.
598 (vld1q_p8_x3): Likewise.
599 (vld1q_p16_x3): Likewise.
600 (vld1q_p64_x3): Likewise.
601 (vst1_s64_x2): Likewise.
602 (vst1_u64_x2): Likewise.
603 (vst1_f64_x2): Likewise.
604 (vst1_s8_x2): Likewise.
605 (vst1_p8_x2): Likewise.
606 (vst1_s16_x2): Likewise.
607 (vst1_p16_x2): Likewise.
608 (vst1_s32_x2): Likewise.
609 (vst1_u8_x2): Likewise.
610 (vst1_u16_x2): Likewise.
611 (vst1_u32_x2): Likewise.
612 (vst1_f16_x2): Likewise.
613 (vst1_f32_x2): Likewise.
614 (vst1_p64_x2): Likewise.
615 (vst1q_s8_x2): Likewise.
616 (vst1q_p8_x2): Likewise.
617 (vst1q_s16_x2): Likewise.
618 (vst1q_p16_x2): Likewise.
619 (vst1q_s32_x2): Likewise.
620 (vst1q_s64_x2): Likewise.
621 (vst1q_u8_x2): Likewise.
622 (vst1q_u16_x2): Likewise.
623 (vst1q_u32_x2): Likewise.
624 (vst1q_u64_x2): Likewise.
625 (vst1q_f16_x2): Likewise.
626 (vst1q_f32_x2): Likewise.
627 (vst1q_f64_x2): Likewise.
628 (vst1q_p64_x2): Likewise.
629 (vst1_s64_x3): Likewise.
630 (vst1_u64_x3): Likewise.
631 (vst1_f64_x3): Likewise.
632 (vst1_s8_x3): Likewise.
633 (vst1_p8_x3): Likewise.
634 (vst1_s16_x3): Likewise.
635 (vst1_p16_x3): Likewise.
636 (vst1_s32_x3): Likewise.
637 (vst1_u8_x3): Likewise.
638 (vst1_u16_x3): Likewise.
639 (vst1_u32_x3): Likewise.
640 (vst1_f16_x3): Likewise.
641 (vst1_f32_x3): Likewise.
642 (vst1_p64_x3): Likewise.
643 (vst1q_s8_x3): Likewise.
644 (vst1q_p8_x3): Likewise.
645 (vst1q_s16_x3): Likewise.
646 (vst1q_p16_x3): Likewise.
647 (vst1q_s32_x3): Likewise.
648 (vst1q_s64_x3): Likewise.
649 (vst1q_u8_x3): Likewise.
650 (vst1q_u16_x3): Likewise.
651 (vst1q_u32_x3): Likewise.
652 (vst1q_u64_x3): Likewise.
653 (vst1q_f16_x3): Likewise.
654 (vst1q_f32_x3): Likewise.
655 (vst1q_f64_x3): Likewise.
656 (vst1q_p64_x3): Likewise.
658 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
660 * config/msp430/msp430.c (msp430_output_labelref): Prepend
661 user_label_prefix to name.
663 * tree-core.h: Update comment about the format of NAME string
664 passed to handler in attribute_spec.
666 * config/msp430/msp430.md: Remove erroneous subreg expression from
667 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
668 zero_extend{q,h}isi2.
670 2018-05-30 Borislav Petkov <bp@suse.de>
672 * doc/extend.texi: Document some architecture specific
673 constraints and sort entries.
675 2018-05-30 Martin Sebor <msebor@redhat.com>
678 * builtins.c (expand_builtin_stpcpy_1): New function.
679 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
680 only if the former succeeds.
682 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
684 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
687 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
689 * doc/invoke.texi (-flinker-output): Document
691 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
693 * passes.c (ipa_write_summaries): Only modify statements if body
695 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
696 incrementally linking.
697 (ipa_passes): Likewise.
698 * lto-cgraph.c (lto_output_node): When incrementally linking do not
699 pass down resolution info.
700 * common.opt (flag_incremental_link): Update info.
701 * gcc.c (plugin specs): Turn flinker-output=* to
702 -plugin-opt=-linker-output-known
703 * toplev.c (compile_file): Also cut compilation when doing incremental
705 * flag-types. (enum lto_partition_model): Add
706 LTO_LINKER_OUTPUT_NOLTOREL.
707 (invoke.texi): Add -flinker-output docs.
708 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
709 link same way as WPA; do not stream in dead initializers.
711 * dwarf2out.c (dwarf2out_die_ref_for_decl,
712 darf2out_register_external_decl): Support incremental link.
714 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
716 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
718 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
720 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
721 it down to simple_object_copy_lto_debug_sections.
722 (run_gcc): Determine incremental LTO link time and configure
723 lto1 into non-wpa mode, disable renaming of debug sections.
725 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
727 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
728 descriptions of various incorrectly documented functions.
730 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
733 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
734 address check not strict.
736 2018-05-30 Richard Biener <rguenther@suse.de>
738 PR tree-optimization/85964
739 * tracer.c (better_p): Drop initialized count check, we only
740 call the function with initialized counts now.
741 (find_best_successor): Do find a best edge if one
742 has uninitialized count.
743 (find_best_predecessor): Likewise. Do BB frequency check only
744 if count is initialized.
746 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
748 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
749 (aarch64_ldrstr_offset_compare): New.
750 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
751 load/store orderings.
752 (aarch64_gen_adjusted_ldpstp): Likewise.
754 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
756 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
757 Check for subset of GENERAL_REGS and FP_REGS.
758 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
761 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
763 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
764 and wi::to_poly_offset. Add the current offset and then check
765 whether the sum fits, rather than using an unchecked addition of
766 a checked term. Check for a shwi rather than a uhwi.
767 * expr.c (get_bit_range): Use tree_to_poly_uint64.
768 (store_constructor): Use poly_int_tree_p.
769 (expand_expr_real_1): Likewise.
770 * function.c (assign_temp): Likewise.
771 * fold-const.c (const_binop): Use poly_int_tree_p and
773 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
775 * ipa-icf-gimple.c (func_checker::compare_operand): Use
776 to_poly_offset for MEM offsets.
777 * ipa-icf.c (sem_variable::equals): Likewise.
778 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
779 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
780 wi::to_poly_offset for BIT_FIELD_REF offsets.
781 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
783 * var-tracking.c (emit_note_insn_var_location): Use
786 2018-05-29 Jim Wilson <jimw@sifive.com>
788 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
790 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
793 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
794 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
796 (sse4_1_round<mode>2): Use nonimmediate_operand
797 for operand 1 predicate.
799 2018-05-29 Martin Sebor <msebor@redhat.com>
800 Richard Biener <rguenther@suse.de>
803 * calls.c (get_size_range): Call determine_value_range instead
805 * tree-vrp.h (determine_value_range): Declared new function.
806 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
808 2018-05-29 Richard Biener <rguenther@suse.de>
810 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
811 sure to use non-pattern stmts for get_earlier_stmt arguments.
812 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
813 called on pattern stmts.
814 (get_later_stmt): Likewise.
816 2018-05-29 Martin Liska <mliska@suse.cz>
818 PR gcov-profile/85759
819 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
822 2018-05-29 Jakub Jelinek <jakub@redhat.com>
824 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
826 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
830 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
831 VEC_PACK_FLOAT_EXPR): New tree codes.
832 * tree-pretty-print.c (op_code_prio): Handle
833 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
834 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
835 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
836 * tree-inline.c (estimate_operator_cost): Likewise.
837 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
838 * fold-const.c (const_binop): Likewise.
839 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
840 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
841 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
842 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
843 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
844 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
845 * expr.c (expand_expr_real_2): Likewise.
846 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
847 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
848 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
850 * optabs.c (expand_widen_pattern_expr): For
851 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
852 sign from result type rather than operand's type.
853 (expand_binop_directly): For vec_packu_float_optab and
854 vec_packs_float_optab allow result type to be different from operand's
856 * optabs-tree.c (optab_for_tree_code): Handle
857 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
858 VEC_PACK_FLOAT_EXPR. Formatting fixes.
859 * tree-vect-generic.c (expand_vector_operations_1): Handle
860 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
862 * tree-vect-stmts.c (supportable_widening_operation): Handle
864 (supportable_narrowing_operation): Handle FLOAT_EXPR.
865 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
866 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
867 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
868 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
870 (vec_pack<floatprefix>_float_<mode>): New expander.
871 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
873 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
874 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
875 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
876 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
877 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
879 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
880 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
881 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
882 VEC_PACK_FLOAT_EXPR): Document.
884 2018-05-29 Richard Biener <rguenther@suse.de>
886 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
888 (stmt_vec_info_vec): Make pointer.
889 (init_stmt_vec_info_vec): Remove.
890 (free_stmt_vec_info_vec): Likewise.
891 (set_stmt_vec_info_vec): New function.
892 (free_stmt_vec_infos): Likewise.
893 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
894 (set_vinfo_for_stmt): Likewise.
895 (get_earlier_stmt): Likewise.
896 (get_later_stmt): Likewise.
897 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
898 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
899 (vec_info::~vec_info): Free stmt_vec_infos.
900 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
901 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
902 (pass_slp_vectorize::execute): Likewise.
903 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
904 (free_stmt_vec_info_vec): Likewise.
905 (set_stmt_vec_info_vec): New function.
906 (free_stmt_vec_infos): Likewise.
907 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
908 the global stmt_vec_info_vec.
909 * tree-parloops.c (gather_scalar_reductions): Use
910 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
913 2018-05-29 Richard Biener <rguenther@suse.de>
915 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
917 2018-05-29 Martin Liska <mliska@suse.cz>
918 David Malcolm <dmalcolm@redhat.com>
920 * vec.c (test_reverse): New.
921 (vec_c_tests): Add new test.
922 * vec.h (vl_ptr>::reverse): New function.
924 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
926 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
928 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
931 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
933 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
935 2018-05-28 Richard Biener <rguenther@suse.de>
937 PR tree-optimization/85933
938 * tree-vect-data-refs.c (vect_record_base_alignments): Only
939 look at stmts marked as vectorizable.
941 2018-05-28 Richard Biener <rguenther@suse.de>
943 PR tree-optimization/85934
944 * tree-vect-generic.c (expand_vector_operations_1): Hoist
945 vector boolean check before scalar optimization.
947 2018-05-28 Jakub Jelinek <jakub@redhat.com>
949 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
952 2018-05-28 Mark Wielaard <mark@klomp.org>
954 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
955 if it is an expression containing a minus sign.
957 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
959 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
961 2018-05-27 Paul Koning <ni1d@arrl.net>
963 * config/pdp11/pdp11.md (truncsihi2): Remove.
965 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
966 Chung-Ju Wu <jasonwucj@gmail.com>
968 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
970 (unaligned_store_dw): Ditto.
971 * config/nds32/nds32-memory-manipulation.c
972 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
973 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
974 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
975 (emit_setmem_word_loop): Rename to ...
976 (emit_setmem_doubleword_loop): ... this.
977 (nds32_gen_dup_4_byte_to_word_value): New function.
978 (nds32_gen_dup_8_byte_to_double_word_value): New function.
979 (nds32_expand_setmem_loop): Refine implementation.
980 (nds32_expand_setmem_loop_v3m): Ditto.
981 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
984 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
986 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
988 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
990 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
991 (nds32_init_machine_status): Initialize machine->attr_naked_p and
992 machine->attr_no_prologue_p.
993 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
994 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
995 (nds32_expand_epilogue): Consider attr_naked_p.
996 (nds32_expand_epilogue_v3pop): Likewise.
997 (nds32_can_use_return_insn): Likewise.
998 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
999 attr_no_prologue_p fields.
1000 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1002 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1005 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1007 * config/i386/sse.md
1008 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1010 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1012 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1014 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1016 (*<floatsuffix>floatv2div2sf2): Rename to ...
1017 (*float<floatunssuffix>v2div2sf2): ... this.
1018 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1019 (float<floatunssuffix>v2div2sf2_mask): ... this.
1020 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1021 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1022 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1024 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1026 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1028 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1030 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1032 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1034 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1035 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1036 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1037 gen_ufix_truncv8dfv8si2.
1038 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1039 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1040 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1041 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1042 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1043 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1044 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1045 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1047 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1051 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1053 2018-05-25 Jim Wilson <jimw@sifive.com>
1055 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1056 * config/riscv/riscv.c (struct machine_function): Add
1057 interrupt_handler_p and attribute_checked_p fields.
1058 (riscv_attribute_table): Add interrupt.
1059 (riscv_interrupt_type_p): New.
1060 (riscv_save_reg_p): Save extra regs for interrupt handler.
1061 (riscv_use_save_libcall): Return false for interrupt handler.
1062 (riscv_first_stack_step): Add forward declaration.
1063 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1064 for interrupt handler with large frame. Use it for saved reg list.
1065 (riscv_expand_prologue): Move flag_stack_usage_info support to
1066 eliminate duplication.
1067 (riscv_expand_epilogue): Generate mret for interrupt handler.
1068 (riscv_epilogue_uses): New.
1069 (riscv_can_use_return_insn): Return false for interrupt handler.
1070 (riscv_function_ok_for_sibcall): Likewise.
1071 (riscv_set_current_function): Add interrupt handler support.
1072 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1073 * config/riscv/riscv.md (UNSPECV_MRET): New.
1075 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1077 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1079 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1081 PR tree-optimization/85712
1082 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1083 this candidate has already been replaced in-situ by a copy.
1085 2018-05-25 Jason Merrill <jason@redhat.com>
1087 PR c++/80485 - inline function non-zero address.
1088 * symtab.c (nonzero_address): Check DECL_COMDAT.
1090 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1093 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1094 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1095 extension of SImode operation. Use const123_operand predicate.
1096 (*saddsi_1): Remove.
1097 (*saddl_se_1): Ditto.
1099 (*ssubl_se_1): Ditto.
1100 * config/alpha/predicates.md (const123_operand): New predicate.
1101 * config/alpha/constraints.md (P): Use IN_RANGE.
1103 2018-05-25 Richard Biener <rguenther@suse.de>
1105 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1107 (ref_maybe_used_by_stmt_p): Likewise.
1108 (stmt_may_clobber_ref_p): Likewise.
1109 (stmt_may_clobber_ref_p_1): Likewise.
1110 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1112 (ref_maybe_used_by_stmt_p): Likewise.
1113 (stmt_may_clobber_ref_p): Likewise.
1114 (stmt_may_clobber_ref_p_1): Likewise.
1115 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1116 the alias oracle to disambiguate DRs with stmts DR analysis
1118 (vect_analyze_data_refs): Do not give up on not analyzable
1119 DRs for BB vectorization. Remove code truncating the dataref
1122 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1125 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1126 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1127 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1129 2018-05-25 Richard Biener <rguenther@suse.de>
1131 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1132 function, combining stmt data ref gathering and fatal analysis
1134 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1135 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1136 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1137 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1138 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1140 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1142 PR tree-optimization/85720
1143 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1144 SCC if all partitions are builtins.
1145 (version_loop_by_alias_check): New parameter. Generate cancelable
1146 runtime alias check if all partitions are builtins.
1147 (distribute_loop): Update call to above function.
1149 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1151 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1152 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1153 (parm_default_def_partition_arg): Ditto.
1154 (set_parm_default_def_partition): Ditto.
1155 (get_parm_default_def_partitions): Ditto and make it static.
1156 (get_undefined_value_partitions): Ditto and make it static.
1157 (remove_ssa_form): Refactor call to init_var_map here.
1158 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1159 computation for loop region.
1160 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1161 (register_default_def): Delete.
1162 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1163 (parm_default_def_partition_arg): Ditto.
1164 (set_parm_default_def_partition): Ditto.
1165 (get_parm_default_def_partitions): Ditto and make it static.
1166 (get_undefined_value_partitions): Ditto and make it static.
1167 (coalesce_with_default, coalesce_with_default): Update comment.
1168 (create_coalesce_list_for_region): New func factored out from
1169 create_outofssa_var_map.
1170 (populate_coalesce_list_for_outofssa): New func factored out from
1171 create_outofssa_var_map and coalesce_ssa_name.
1172 (create_outofssa_var_map): Delete.
1173 (coalesce_ssa_name): Refactor to support live range computation.
1174 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1175 (get_parm_default_def_partitions): Delete.
1176 (get_undefined_value_partitions): Ditto.
1177 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1178 computation for loop region.
1179 (new_tree_live_info, loe_visit_block): Ditto.
1180 (live_worklist, set_var_live_on_entry): Ditto.
1181 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1182 * tree-ssa-live.h (struct _var_map): New fields.
1183 (init_var_map): Change decl.
1184 (region_contains_p): New.
1186 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1188 * tree-ssa-live.h (live_merge_and_clear): Delete.
1190 2018-05-25 Richard Biener <rguenther@suse.de>
1193 * tree-dump.c (dequeue_and_dump): Remove access to removed
1194 operand 2 of a SWITCH_EXPR.
1196 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1198 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1199 multiplication and division.
1200 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1201 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1202 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1203 (cond_udiv_optab, cond_umod_optab): New optabs.
1204 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1205 (IFN_COND_RDIV): New internal functions.
1206 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1207 TRUNC_MOD_EXPR and RDIV_EXPR.
1208 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1209 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1211 (SVE_INT_BINARY): Include mult.
1212 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1213 (optab, sve_int_op): Handle mult.
1214 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1216 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1217 for SVE_INT_BINARY_SD.
1219 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1221 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1222 (optab, sve_int_op): Handle div and udiv.
1223 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1224 for SVE_INT_BINARY_SD.
1225 (*<optab><mode>3): New insn for the same.
1227 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1229 * tree-vect-patterns.c: Include predict.h.
1230 (vect_recog_divmod_pattern): Restrict check for division support
1231 to when optimizing for size.
1233 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1235 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1236 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1237 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1238 (gimple_match_op::set_op): Likewise.
1239 (gimple_resimplify4): Declare.
1240 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1241 (expr::gen_transform): Likewise.
1242 (decision_tree::gen): Generate a simplification routine for 4 operands.
1243 * gimple-match-head.c (gimple_simplify): Add an overload for
1244 4 operands. In the top-level function, handle up to 4 call
1245 arguments and call gimple_resimplify4.
1246 (gimple_resimplify4): New function.
1247 (build_call_internal): Pass a fourth operand.
1248 (maybe_push_to_seq): Likewise.
1249 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1250 Fold VEC_COND_EXPRs of an operation and a default value into
1251 an IFN_COND_* function if possible.
1252 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1254 (SVE_COND_FP_BINARY): Include them.
1255 (optab, sve_fp_op): Handle them.
1256 (SVE_INT_BINARY_REV): New code iterator.
1257 (SVE_COND_FP_BINARY_REV): New int iterator.
1258 (commutative): New int attribute.
1259 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1261 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1263 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1264 (*cond_<optab><mode>): New patterns for reversed operands.
1266 2018-05-25 Richard Biener <rguenther@suse.de>
1268 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1269 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1270 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1271 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1272 * tree-vect-data-refs.c (everywhere): Adjust users.
1273 * tree-vect-loop.c (everywhere): Likewise.
1274 * tree-vect-slp.c (everywhere): Likewise.
1275 * tree-vect-stmts.c (everywhere): Likewise.
1276 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1278 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1280 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1282 (gcc_cv_as_section_exclude): ... this.
1283 Try Solaris as #exclude syntax.
1284 * configure: Regenerate.
1285 * config.in: Regenerate.
1286 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1288 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1289 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1291 * varasm.c (default_elf_asm_named_section): Don't check if
1292 HAVE_GAS_SECTION_EXCLUDE is defined.
1294 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1296 * doc/md.texi: Update the documentation of the cond_* optabs
1297 to mention the new final operand. Fix GET_MODE_NUNITS call.
1298 Describe the scalar case too.
1299 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1300 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1302 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1303 (get_conditional_internal_fn): Update comment.
1304 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1305 accumulator value as a final argument to conditional functions.
1306 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1307 a define_expand and add an "else" operand. Assert for now that
1308 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1309 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1310 (*cond_<optab><mode>): New patterns.
1311 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1312 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1313 (UNSPEC_COND_EOR): Delete.
1314 (optab): Remove associated mappings.
1315 (SVE_INT_BINARY): New code iterator.
1316 (sve_int_op): Remove int attribute and add "minus" to the code
1318 (SVE_COND_INT_OP): Delete.
1319 (SVE_COND_FP_OP): Rename to...
1320 (SVE_COND_FP_BINARY): ...this.
1322 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1324 * optabs.c (can_reuse_operands_p): New function.
1325 (maybe_legitimize_operands): Try to reuse the results for
1328 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1330 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1331 Add {q} suffix to insn mnemonic.
1333 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1335 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1336 (msp430_warn_func_return): New.
1338 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1340 * fold-const.c (tree_nonzero_bits): New function.
1341 * fold-const.h (tree_nonzero_bits): Likewise.
1342 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1343 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1345 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1349 * varasm.c (assemble_alias): Check ifunc_resolver only on
1352 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1355 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1356 when memory input operand is handled.
1358 2018-05-24 Luis Machado <luis.machado@linaro.org>
1360 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1362 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1364 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1366 * match.pd: Delay FMA folds until after vectorization.
1368 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1371 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1372 address check not strict.
1374 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1376 * gimple-match.h (gimple_match_op): New class.
1377 (mprts_hook): Replace parameters with a gimple_match_op *.
1378 (maybe_build_generic_op): Likewise.
1379 (gimple_simplified_result_is_gimple_val): Replace parameters with
1380 a const gimple_match_op *.
1381 (gimple_simplify): Replace code_helper * and tree * parameters with
1382 a gimple_match_op * parameter.
1383 (gimple_resimplify1): Replace code_helper *, tree and tree *
1384 parameters with a gimple_match_op * parameter.
1385 (gimple_resimplify2): Likewise.
1386 (gimple_resimplify3): Likewise.
1387 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1388 parameters with a gimple_match_op * parameter.
1389 * gimple-match-head.c (gimple_simplify): Change prototypes of
1390 auto-generated functions to take a gimple_match_op * instead of
1391 separate code_helper * and tree * parameters. Make the same
1392 change in the top-level overload and update calls to the
1393 gimple_resimplify routines. Update calls to the auto-generated
1394 functions and to maybe_push_res_to_seq in the publicly-facing
1395 operation-specific gimple_simplify overloads.
1396 (gimple_match_op::MAX_NUM_OPS): Define.
1397 (gimple_resimplify1): Replace rcode and ops with a single res_op
1398 parameter. Update call to gimple_simplify.
1399 (gimple_resimplify2): Likewise.
1400 (gimple_resimplify3): Likewise.
1401 (mprts_hook): Replace parameters with a gimple_match_op *.
1402 (maybe_build_generic_op): Likewise.
1403 (build_call_internal): Replace type, nargs and ops with
1404 a gimple_match_op *.
1405 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1406 with a single gimple_match_op *. Update calls to mprts_hook,
1407 build_call_internal and gimple_simplified_result_is_gimple_val.
1408 Factor out code that is common to the tree_code and combined_fn cases.
1409 * genmatch.c (expr::gen_transform): Replace tem_code and
1410 tem_ops with a gimple_match_op called tem_op. Update calls
1411 to the gimple_resimplify functions and maybe_push_res_to_seq.
1412 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1413 res_ops. Update call to the gimple_resimplify functions.
1414 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1415 (decision_tree::gen): Make the functions take a gimple_match_op *
1416 called res_op instead of separate res_code and res_ops parameters.
1417 Update call accordingly.
1418 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1419 and ops with a single res_op parameter. Update calls to
1420 maybe_build_generic_op and maybe_push_res_to_seq.
1421 (fold_stmt_1): Update calls to gimple_simplify and
1422 replace_stmt_with_simplification.
1423 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1424 and gimple_simplified_result_is_gimple_val.
1425 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1427 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1428 with a gimple_match_op *.
1429 (vn_nary_build_or_lookup): Likewise. Update call to
1430 vn_nary_build_or_lookup_1.
1431 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1432 gimple_match_op *. Update calls to the gimple_resimplify routines
1433 and to gimple_simplified_result_is_gimple_val.
1434 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1435 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1436 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1437 (visit_nary_op): Likewise.
1438 (visit_reference_op_load): Likewise.
1440 2018-05-23 Luis Machado <luis.machado@linaro.org>
1442 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1443 modifier for printing the step amount.
1445 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1448 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1451 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1453 * doc/sourcebuild.texi (Endianness): New subsubsection.
1455 2018-05-23 Luis Machado <luis.machado@linaro.org>
1457 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1458 <prefetch_dynamic_strides>: New const bool field.
1459 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1460 prefetch_dynamic_strides.
1461 (exynosm1_prefetch_tune): Likewise.
1462 (thunderxt88_prefetch_tune): Likewise.
1463 (thunderx_prefetch_tune): Likewise.
1464 (thunderx2t99_prefetch_tune): Likewise.
1465 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1467 (aarch64_override_options_internal): Update to set
1468 PARAM_PREFETCH_DYNAMIC_STRIDES.
1469 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1470 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1471 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1472 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1473 prefetch-dynamic-strides setting.
1475 2018-05-23 Luis Machado <luis.machado@linaro.org>
1477 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1478 <minimum_stride>: New const int field.
1479 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1480 minimum_stride field defaulting to -1.
1481 (exynosm1_prefetch_tune): Likewise.
1482 (thunderxt88_prefetch_tune): Likewise.
1483 (thunderx_prefetch_tune): Likewise.
1484 (thunderx2t99_prefetch_tune): Likewise.
1485 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1486 <default_opt_level>: Set to 3.
1487 (aarch64_override_options_internal): Update to set
1488 PARAM_PREFETCH_MINIMUM_STRIDE.
1489 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1490 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1491 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1492 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1493 stride is constant and is below the minimum stride threshold.
1495 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1497 * config/arm/arm-cpus.in (mode26): Delete.
1498 (armv4): Delete mode26 reference.
1499 * config/arm/arm.c (arm_configure_build_target): Delete use of
1502 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1504 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1506 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1507 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1509 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1510 (floatunsdidf2): Ditto.
1512 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1514 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1515 (fixuns_trunc<mode>si2_avx512f): Ditto.
1516 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1517 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1518 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1520 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1522 PR rtl-optimization/79985
1523 * df-scan.c (df_insn_refs_collect): Remove special case for
1524 global registers and asm statements.
1526 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1528 * extend.texi (Global Register Variables): Rewrite the bullet list.
1529 Note that the register is available for allocation. Note that access
1530 via inline asm must use constraints. Add note about async-signal
1531 handlers. Remove paragraph about automagic register selection.
1533 2018-05-23 Richard Biener <rguenther@suse.de>
1535 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1536 of fixed offset from memset VN.
1538 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1540 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1542 (alloc_cand_and_find_basis): Initialize first_interp field.
1543 (slsr_process_mul): Modify first_interp field.
1544 (slsr_process_add): Likewise.
1545 (slsr_process_cast): Modify first_interp field for each new
1547 (slsr_process_copy): Likewise.
1548 (dump_candidate): Dump first_interp field.
1549 (replace_mult_candidate): Process all interpretations, not just
1551 (replace_rhs_if_not_dup): Likewise.
1552 (replace_one_candidate): Likewise.
1554 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1556 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1558 (aarch64_needs_frame_chain): New function.
1559 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1561 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1564 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1565 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1566 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1567 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1568 as true for strict-align.
1569 (aarch64_can_inline_p): Perform checks even when callee has no
1570 attributes to check for strict alignment.
1571 * doc/extend.texi (AArch64 Function Attributes): Document
1573 * doc/invoke.texi: (AArch64 Options): Likewise.
1575 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1577 PR tree-optimization/85853
1578 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1579 the handling of the root of the node to...
1580 (vect_slp_analyze_node_operations_1): ...this new function,
1581 and run the whole thing with the child nodes' def types
1582 set according to their SLP node's def type.
1584 2018-05-23 Richard Biener <rguenther@suse.de>
1587 * tree-data-ref.c (create_runtime_alias_checks): Defer
1588 and ignore overflow warnings.
1590 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1592 PR tree-optimization/85822
1593 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1596 2018-05-23 Richard Biener <rguenther@suse.de>
1598 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1599 memset constants via native_interpret_expr.
1601 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1604 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1606 (cgraph_node::create_alias): Likewise.
1607 (cgraph_node::get_availability): Check ifunc_resolver instead
1608 of looking up ifunc attribute.
1609 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1610 * varasm.c (do_assemble_alias): Likewise.
1611 (assemble_alias): Likewise.
1612 (default_binds_local_p_3): Likewise.
1613 * cgraph.h (cgraph_node): Add ifunc_resolver.
1614 (cgraph_node::only_called_directly_or_aliased_p): Return false
1616 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1618 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1619 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1620 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1621 instead of looking up ifunc attribute.
1623 2018-05-22 Luis Machado <luis.machado@linaro.org>
1625 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1627 2018-05-22 Martin Sebor <msebor@redhat.com>
1630 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1631 only when expasion succeeds.
1632 (expand_builtin_strcmp): Same.
1633 (expand_builtin_strncmp): Same.
1635 2018-05-22 Martin Sebor <msebor@redhat.com>
1637 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1639 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1640 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1642 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1643 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1645 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1646 Add check for legitimate_address.
1647 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1648 (aarch64_swap_ldrstr_operands): New.
1649 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1652 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1653 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1655 * config/aarch64/aarch64.md: New patterns to generate stp
1657 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1658 single words and double words.
1659 (load_pair_sw, load_pair_dw): Likewise.
1660 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1662 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1664 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1665 for different mode ldpstp and add peephole for merged zero stores.
1667 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1669 (aarch64_gen_store_pair): Rename calls to match new patterns.
1670 (aarch64_gen_load_pair): Rename calls to match new patterns.
1671 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1672 (load_pair<DREG:mode><DREG2:mode>): ... This.
1673 (store_pair<mode>): Rename to...
1674 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1675 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1677 (V_INT_EQUIV): Handle SImode.
1678 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1681 2018-05-22 Martin Sebor <msebor@redhat.com>
1684 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1685 or ajust the presumed bound on an operation to avoid unnecessary
1688 2018-05-22 Martin Sebor <msebor@redhat.com>
1690 PR tree-optimization/85826
1691 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1692 assuming that a DECL necesarily has a constant size.
1694 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1697 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1699 2018-05-22 Richard Biener <rguenther@suse.de>
1701 PR tree-optimization/85834
1702 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1703 non-constant and non-zero memset arguments.
1705 2018-05-22 Martin Liska <mliska@suse.cz>
1708 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1710 2018-05-22 Richard Biener <rguenther@suse.de>
1712 PR tree-optimization/85863
1713 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1714 comparisons when vectype is specified.
1715 (vectorizable_condition): Do not specify vectype for
1716 vect_is_simple_cond when SLP vectorizing.
1718 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1721 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1722 define __ibm128 as long double.
1723 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1724 as a distinct type when IEEE 128-bit support is enabled.
1725 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1726 128-bit types to use the correct functions.
1727 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1728 convert between 128-bit floating point types that have different
1729 modes but the same representation, instead of using gen_lowpart to
1731 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1733 (IFKF_reg): New attributes to give the register constraints for
1735 (extend<mode>tf2_internal): New insns to mark an explicit
1736 conversion between 128-bit floating point types that have a
1737 different mode but share the same representation.
1739 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1741 PR tree-optimization/85814
1742 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1743 a null return from get_strinfo when unsharing the next
1744 strinfo in the chain.
1746 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1749 * varasm.c (weak_finish): Clean up weak_decls.
1751 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1753 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1754 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1755 UNSPEC_UADALP values.
1756 * config/aarch64/iterators.md (ABAL): New int iterator.
1759 (sur): Add mappings for the above.
1760 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1762 (aarch64_<sur>abal<mode>_4): Likewise.
1763 (aarch64_<sur>adalp<mode>_3): Likewise.
1764 (<sur>sadv16qi): New define_expand.
1766 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1768 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1769 (*movdf_internal): Ditto.
1770 (*rcpsf2_sse): Ditto.
1771 (*rsqrtsf2_sse): Ditto.
1772 (*sqrt<mode>2_sse): Ditto.
1774 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1776 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1778 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1779 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1780 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1781 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1783 * config/aarch64/arm_neon.h: Likewise.
1784 * config/aarch64/iterators.md (VQ_I): New.
1786 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1788 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1790 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1793 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1795 * config/nds32/constraints.md (S): New constraint.
1796 * config/nds32/nds32.md (call_internal): Use constraint S.
1797 (call_value_internal): Likewise.
1798 (sibcall_internal): Likewise.
1799 (sibcall_value_internal): Likewise.
1801 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1802 Chung-Ju Wu <jasonwucj@gmail.com>
1804 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1807 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1808 Chung-Ju Wu <jasonwucj@gmail.com>
1810 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1811 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1812 (nds32_rtx_costs_impl): Simplify.
1813 (nds32_address_cost_impl): Simplify.
1814 (nds32_init_rtx_costs): New function.
1815 (nds32_rtx_costs_speed_prefer): Likewise.
1816 (nds32_rtx_costs_size_prefer): Likewise.
1817 (nds32_address_cost_speed_prefer): Likewise.
1818 (nds32_address_cost_speed_fwprop): Likewise.
1819 (nds32_address_cost_size_prefer): Likewise.
1820 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1821 * config/nds32/nds32.c (nds32_option_override): Use
1822 nds32_init_rtx_costs function.
1824 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1826 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1827 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1828 (TARGET_PIPELINE_N8): Likewise.
1829 (TARGET_PIPELINE_N10): Likewise.
1830 (TARGET_PIPELINE_N13): Likewise.
1831 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1833 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1835 * config/nds32/nds32-fpu.md: Update copyright year.
1837 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1839 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1841 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/nds32.c
1844 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1845 * config/nds32/nds32.opt (minline-asm-r15): New option.
1847 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1849 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1851 * config/nds32/nds32.md (abssi2): New pattern.
1853 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1855 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1856 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1857 Merge insn pattern from sse_cvtsi2ss<round_name> and
1858 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1859 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1860 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1861 using SWI48 mode iterator.
1862 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1863 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1864 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1865 pattern from sse_cvttss2si<round_saeonly_name>
1866 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1867 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1868 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1869 using SWI48 mode iterator.
1870 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1871 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1872 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1873 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1874 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1875 using SWI48 mode iterator.
1876 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1877 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1878 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1879 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1880 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1881 SWI48 mode iterator.
1882 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1883 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1884 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1885 pattern from sse_cvttsd2si<round_saeonly_name>
1886 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1888 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1890 * config/nds32/nds32-md-auxiliary.c
1891 (nds32_valid_smw_lwm_base_p): Refine.
1892 (nds32_output_smw_single_word): Refine.
1893 (nds32_output_smw_double_word): New.
1894 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1896 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1898 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1899 (nds32_output_stack_pop): Refine.
1900 (nds32_expand_unaligned_load): Refine.
1901 (nds32_expand_unaligned_store): Refine.
1903 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1904 Chung-Ju Wu <jasonwucj@gmail.com>
1906 * config/nds32/constants.md: Add TP_REGNUM constant.
1907 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1908 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1910 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1911 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1912 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1913 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1914 and PIC code generation.
1915 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1917 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1919 * config/nds32/nds32.md: Support TLS and PIC.
1920 * config/nds32/nds32.c: Support TLS and PIC.
1921 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1922 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1925 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1927 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1928 mode with E_ prefix.
1930 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1931 Chung-Ju Wu <jasonwucj@gmail.com>
1933 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1934 * config/nds32/nds32-md-auxiliary.c
1935 (symbolic_reference_mentioned_p): New.
1936 (nds32_legitimize_ict_address): New.
1937 (nds32_expand_ict_move): New.
1938 (nds32_indirect_call_referenced_p): New.
1939 (nds32_symbol_binds_local_p): Delete.
1940 (nds32_long_call_p): Modify.
1941 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1942 * config/nds32/nds32-protos.h
1943 (symbolic_reference_mentioned_p): Declare.
1944 (nds32_legitimize_ict_address): Declare.
1945 (nds32_expand_ict_move): Declare.
1946 (nds32_indirect_call_referenced_p): Declare.
1947 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1948 (nds32_relax_group): Use nds32_ict_const_p as condition.
1949 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1950 (nds32_asm_file_start): Output ict_model directive in asm code.
1951 (nds32_legitimate_address_p): Consider indirect call.
1952 (nds32_print_operand): Consider indirect call.
1953 (nds32_print_operand_address): Consider indirect call.
1954 (nds32_insert_attributes): Handle "indirect_call" attribute.
1955 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1956 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1957 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1958 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1959 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1960 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1961 (TARGET_ICT_MODEL_SMALL): Define.
1962 (TARGET_ICT_MODEL_LARGE): Define.
1963 * config/nds32/nds32.md (movsi): Consider ict model.
1964 (call, call_value): Consider ict model.
1965 (sibcall, sibcall_value): Consider ict model.
1966 * config/nds32/nds32.opt (mict-model): New option.
1967 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1970 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1971 Monk Chiang <sh.chiang04@gmail.com>
1972 Jim Wilson <jimw@sifive.com>
1974 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1975 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1976 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1977 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1978 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1979 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1980 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1981 compute save_libcall_adjustment properly.
1982 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1983 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1984 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1985 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1986 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1987 (ABI_SPEC): Handle mabi=ilp32e.
1988 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1989 (RVE): Add RVE mask.
1990 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1991 <-march>: Add rv32e as an example.
1993 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1996 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1997 (intra_create_variable_infos): Handle C++ constructors.
1999 2018-05-18 Martin Liska <mliska@suse.cz>
2001 * passes.def: Remove a redundant pass.
2003 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2006 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2008 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2010 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2012 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2014 (armv2, armv2a, armv3, armv3m): Delete architectures.
2015 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2016 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2017 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2019 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2020 (*mulsidi3adddi): Likewise.
2021 (mulsidi3): Likewise.
2022 (*mulsidi3_nov6): Likewise.
2023 (umulsidi3): Likewise.
2024 (umulsidi3_nov6): Likewise.
2025 (umaddsidi4): Likewise.
2026 (*umulsidi3adddi): Likewise.
2027 (smulsi3_highpart): Likewise.
2028 (*smulsi3_highpart_nov6): Likewise.
2029 (umulsi3_highpart): Likewise.
2030 (*umulsi3_highpart_nov6): Likewise.
2031 * config/arm/arm.h (arm_arch3m): Delete.
2032 * config/arm/arm.c (arm_arch3m): Delete.
2033 (arm_option_override_internal): Update armv3-related comment.
2034 (arm_configure_build_target): Delete use of isa_bit_mode32.
2035 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2036 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2037 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2039 (mulusa3): Likewise.
2040 * config/arm/arm-protos.h (arm_arch3m): Delete.
2041 * config/arm/arm-tables.opt: Regenerate.
2042 * config/arm/arm-tune.md: Likewise.
2043 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2044 deleted architectures.
2046 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2048 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2049 (armv5t, armv5te): New features.
2050 (ARMv5, ARMv5e): Delete fgroups.
2051 (ARMv5t, ARMv5te): Adjust for above changes.
2053 (armv5, armv5e): Delete arches.
2054 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2056 (*call_reg_arm): Likewise.
2057 (*call_value_reg_armv5): Likewise.
2058 (*call_value_reg_arm): Likewise.
2059 (*call_symbol): Likewise.
2060 (*call_value_symbol): Likewise.
2061 (*sibcall_insn): Likewise.
2062 (*sibcall_value_insn): Likewise.
2064 (prefetch): Likewise.
2065 (define_split and define_peephole2 dependent on arm_arch5):
2067 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2069 (TARGET_ARM_QBIT): Likewise.
2070 (TARGET_DSP_MULTIPLY): Likewise.
2071 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2072 (arm_arch5, arm_arch5e): Delete.
2073 (arm_arch5t, arm_arch5te): Declare.
2074 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2075 (arm_arch5t): Declare.
2076 (arm_option_reconfigure_globals): Update for the above.
2077 (arm_options_perform_arch_sanity_checks): Update comment, replace
2078 use of arm_arch5 with arm_arch5t.
2079 (use_return_insn): Likewise.
2080 (arm_emit_call_insn): Likewise.
2081 (output_return_instruction): Likewise.
2082 (arm_final_prescan_insn): Likewise.
2083 (arm_coproc_builtin_available): Likewise.
2084 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2085 arm_arch5e with arm_arch5t and arm_arch5te.
2086 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2087 (arm_arch5t, arm_arch5te): Declare.
2088 * config/arm/arm-tables.opt: Regenerate.
2089 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2090 * config/arm/t-multilib: Likewise.
2091 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2092 instead of arm_arch5.
2093 (*call_reg_thumb1): Likewise.
2094 (*call_value_reg_thumb1_v5): Likewise.
2095 (*call_value_reg_thumb1): Likewise.
2096 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2098 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2100 2018-05-18 Martin Liska <mliska@suse.cz>
2102 PR gcov-profile/84846
2103 * doc/gcov.texi: Document -t option of gcov tool.
2105 2018-05-18 Martin Liska <mliska@suse.cz>
2107 PR gcov-profile/84846
2108 * gcov.c (print_usage): Add new -t option.
2109 (process_args): Handle the option.
2110 (generate_results): Use stdout as output when requested by
2113 2018-05-18 Martin Liska <mliska@suse.cz>
2115 PR gcov-profile/84846
2116 * coverage.c (coverage_init): Write PWD to .gcno file.
2117 * doc/gcov.texi: Document how working directory is printed.
2118 * gcov-dump.c (dump_gcov_file): Print PWD.
2119 * gcov.c (output_intermediate_file): Likewise.
2120 (read_graph_file): Read PWD string.
2121 (output_lines): Print PWD.
2123 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2126 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2127 for retval and return false if all args to phi are zero.
2129 2018-05-18 Richard Biener <rguenther@suse.de>
2131 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2133 (evrp_dom_walker::before_dom_children): Call it.
2135 2018-05-18 Richard Biener <rguenther@suse.de>
2137 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2138 results when processing array refs with variable index.
2140 2018-05-18 Toon Moene <toon@moene.org>
2142 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2143 directly after that of -floop-interchange. Indicate that both
2144 options are enabled by default when specifying -O3.
2146 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2148 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2149 iterator. Delete separate integer-mode vec_set<mode> expander.
2150 (aarch64_simd_vec_setv2di): Delete.
2151 (vec_setv2di): Delete.
2152 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2153 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2154 the "w, r" alternative.
2156 2018-05-18 Martin Liska <mliska@suse.cz>
2158 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2159 * tree-pass.h (make_pass_lower_switch_O0): New function.
2160 * tree-switch-conversion.c (node_has_low_bound): Remove.
2161 (node_has_high_bound): Likewise.
2162 (node_is_bounded): Likewise.
2163 (class pass_lower_switch): Make it a template type and create
2165 (pass_lower_switch::execute): Add template argument.
2166 (make_pass_lower_switch): New function.
2167 (make_pass_lower_switch_O0): New function.
2168 (do_jump_if_equal): Remove.
2169 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2170 all the hard work to tree optimization passes.
2172 2018-05-18 Martin Liska <mliska@suse.cz>
2174 * dbgcnt.c (limit_low): Renamed from limit.
2175 (limit_high): New variable.
2176 (dbg_cnt_is_enabled): Check for upper limit.
2177 (dbg_cnt): Adjust dumping.
2178 (dbg_cnt_set_limit_by_index): Add new argument for high
2180 (dbg_cnt_set_limit_by_name): Likewise.
2181 (dbg_cnt_process_single_pair): Parse new format.
2182 (dbg_cnt_process_opt): Use strtok.
2183 (dbg_cnt_list_all_counters): Remove 'value' and add
2185 * doc/invoke.texi: Document changes.
2187 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2189 * doc/sourcebuild.texi (scalar_all_fma): Document.
2190 * tree.def (FMA_EXPR): Delete.
2191 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2192 * internal-fn.c (ternary_direct): New macro.
2193 (expand_ternary_optab_fn): Likewise.
2194 (direct_ternary_optab_supported_p): Likewise.
2195 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2196 * builtins.c (fold_builtin_fma): Delete.
2197 (fold_builtin_3): Don't call it.
2198 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2199 * expr.c (expand_expr_real_2): Likewise.
2200 * fold-const.c (operand_equal_p): Likewise.
2201 (fold_ternary_loc): Likewise.
2202 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2203 * gimple.c (DEFTREECODE): Likewise.
2204 * gimplify.c (gimplify_expr): Likewise.
2205 * optabs-tree.c (optab_for_tree_code): Likewise.
2206 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2207 * tree-eh.c (operation_could_trap_p): Likewise.
2208 (stmt_could_throw_1_p): Likewise.
2209 * tree-inline.c (estimate_operator_cost): Likewise.
2210 * tree-pretty-print.c (dump_generic_node): Likewise.
2211 (op_code_prio): Likewise.
2212 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2213 * tree-ssa-operands.c (get_expr_operands): Likewise.
2214 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2215 * fold-const-call.h (fold_fma): Delete.
2216 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2217 CFN_FNMA and CFN_FNMS.
2219 * genmatch.c (combined_fn): New enum.
2220 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2221 (commutative_op): New function.
2222 (commutate): Use it. Handle more than 2 operands.
2223 (dt_operand::gen_gimple_expr): Use commutative_op.
2224 (parser::parse_expr): Allow :c to be used with non-binary
2225 operators if the commutative operand is known.
2226 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2227 CFN_FMS, CFN_FNMA and CFN_FNMS.
2228 (backprop::process_assign_use): Remove FMA_EXPR handling.
2229 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2230 (gen_hsa_fma): New function.
2231 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2232 IFN_FNMA and IFN_FNMS.
2233 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2234 * gimple-fold.h (follow_all_ssa_edges): Declare.
2235 * gimple-fold.c (follow_all_ssa_edges): New function.
2236 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2237 gimple_build interface and use follow_all_ssa_edges to fold the result.
2238 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2239 instead of checking for optabs directly.
2240 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2241 rather than FMA_EXPRs.
2242 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2243 call to IFN_FMA instead of an FMA_EXPR.
2245 2018-05-17 Jim Wilson <jimw@sifive.com>
2247 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2248 sign extended promoted subreg, and the range does not have the sign bit
2249 set, then do a sign extend.
2251 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2252 test, check for sign extended subreg and/or constant operands, and
2253 do a sign extend in that case.
2255 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2257 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2258 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2260 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2261 Change logics_shift_reg to logics_shift_imm.
2262 (thunderx2t99_fp_loadpair_basic): Delete.
2263 (thunderx2t99_fp_storepair_basic): Delete.
2264 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2265 (thunderx2t99_asimd_polynomial): Delete.
2266 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2267 and neon_fp_mul_d_scalar_q.
2268 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2269 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2270 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2271 (thunderx2t99_asimd_lut): Add missing tbl types.
2272 (thunderx2t99_asimd_ext): Delete.
2273 (thunderx2t99_asimd_load1_1_mult): Delete.
2274 (thunderx2t99_asimd_load1_2_mult): Delete.
2275 (thunderx2t99_asimd_load1_ldp): New.
2276 (thunderx2t99_asimd_load1): New.
2277 (thunderx2t99_asimd_load2): Add missing *load2* types.
2278 (thunderx2t99_asimd_load3): New.
2279 (thunderx2t99_asimd_load4): New.
2280 (thunderx2t99_asimd_store1_1_mult): Delete.
2281 (thunderx2t99_asimd_store1_2_mult): Delete.
2282 (thunderx2t99_asimd_store2_mult): Delete.
2283 (thunderx2t99_asimd_store2_onelane): Delete.
2284 (thunderx2t99_asimd_store_stp): New.
2285 (thunderx2t99_asimd_store1): New.
2286 (thunderx2t99_asimd_store2): New.
2287 (thunderx2t99_asimd_store3): New.
2288 (thunderx2t99_asimd_store4): New.
2290 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2292 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2293 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2295 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2296 Segher Boessenkool <segher@kernel.crashing.org>
2299 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2302 2018-05-17 Richard Biener <rguenther@suse.de>
2304 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2305 for pruning loop and prune defs feeding only already visited PHIs.
2307 2018-05-17 Richard Biener <rguenther@suse.de>
2309 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2311 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2312 Richard Biener <rguenther@suse.de>
2314 PR tree-optimization/85793
2315 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2316 for VMAT_ELEMENTWISE.
2318 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2320 * internal-fn.h (lookup_internal_fn): Declare
2321 * internal-fn.c (lookup_internal_fn): New function.
2322 * gimple.c (gimple_build_call_from_tree): Handle calls to
2324 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2325 internal function names.
2326 * tree-pretty-print.c (dump_generic_node): Likewise.
2327 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2329 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2331 * gimple-fold.h (gimple_build): Make the function forms take
2332 combined_fn rather than built_in_function.
2333 (gimple_simplify): Likewise.
2334 * gimple-match-head.c (gimple_simplify): Likewise.
2335 * gimple-fold.c (gimple_build): Likewise.
2336 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2337 rather than gimple_build_call_internal.
2338 (get_initial_defs_for_reduction): Likewise.
2339 (vect_create_epilog_for_reduction): Likewise.
2340 (vectorizable_live_operation): Likewise.
2342 2018-05-17 Martin Liska <mliska@suse.cz>
2344 * gimple-ssa-sprintf.c (format_directive): Do not use
2345 space in between 'G_' and '('.
2347 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2350 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2351 even if the mask is not all ones.
2354 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2356 (ix86_gimple_fold_builtin): Likewise.
2359 * config/i386/i386.c: Include tree-vector-builder.h.
2360 (ix86_vector_shift_count): New function.
2361 (ix86_fold_builtin): Fold shift builtins by scalar count.
2362 (ix86_gimple_fold_builtin): Likewise.
2364 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2365 _mm512_setzero): New intrinsics.
2367 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2368 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2370 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2371 code generation for cases where splatting a value is not useful.
2372 * simplify-rtx.c (simplify_ternary_operation): Simplify
2373 vec_merge across a vec_duplicate and a paradoxical subreg forming
2374 a vector mode to a vec_concat.
2376 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2378 * config.gcc: Support "goldmont-plus".
2379 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2381 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2382 PROCESSOR_GOLDMONT_PLUS.
2383 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2384 (processor_target_table): Add "goldmont-plus".
2385 (PTA_GOLDMONT_PLUS): Define.
2386 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2387 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2388 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2389 (fold_builtin_cpu): Add "goldmont-plus".
2390 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2391 (ix86_option_override_internal): Add "goldmont-plus".
2392 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2393 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2394 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2395 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2397 2018-05-17 Richard Biener <rguenther@suse.de>
2399 PR tree-optimization/85757
2400 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2401 remove defs that only feed that PHI from further processing.
2403 2018-05-16 Jim Wilson <jimw@sifive.com>
2405 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2407 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2409 2018-05-16 Mark Wielaard <mark@klomp.org>
2411 * dwarf2out.c (count_index_strings): New function.
2412 (output_indirect_strings): Call count_index_strings and generate
2413 header for dwarf_version >= 5.
2415 2018-05-16 Mark Wielaard <mark@klomp.org>
2417 * dwarf2out.c (dwarf_FORM): New function.
2418 (set_indirect_string): Use dwarf_FORM.
2419 (reset_indirect_string): Likewise.
2420 (size_of_die): Likewise.
2421 (value_format): Likewise.
2422 (output_die): Likewise.
2423 (add_skeleton_AT_string): Likewise.
2424 (output_macinfo_op): Likewise.
2425 (index_string): Likewise.
2426 (output_index_string_offset): Likewise.
2427 (output_index_string): Likewise.
2428 (count_index_strings): Likewise.
2430 2018-05-16 Carl Love <cel@us.ibm.com>
2432 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2433 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2435 2018-05-16 Martin Jambor <mjambor@suse.cz>
2437 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2438 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2440 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2442 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2443 (fnma<mode>4): Likewise.
2444 (fms<mode>4): Likewise.
2445 (fnms<mode>4): Likewise.
2446 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2447 (aarch64_fnma<mode>4): Likewise.
2448 (aarch64_fms<mode>4): Likewise.
2449 (aarch64_fnms<mode>4): Likewise.
2450 (aarch64_fnmadd<mode>4): Likewise.
2452 2018-05-16 Jason Merrill <jason@redhat.com>
2454 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2456 2018-05-16 Richard Biener <rguenther@suse.de>
2458 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2459 (dump_stmt_cost): Declare.
2460 (add_stmt_cost): Dump cost we add.
2461 (add_stmt_costs): New function.
2462 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2464 (vect_analyze_stmt): Adjust prototype.
2465 (vectorizable_condition): Likewise.
2466 (vectorizable_live_operation): Likewise.
2467 (vectorizable_reduction): Likewise.
2468 (vectorizable_induction): Likewise.
2469 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2470 cost vector to pass to vectorizable_ and record afterwards.
2471 (vect_model_reduction_cost): Take cost vector argument and adjust.
2472 (vect_model_induction_cost): Likewise.
2473 (vectorizable_reduction): Likewise.
2474 (vectorizable_induction): Likewise.
2475 (vectorizable_live_operation): Likewise.
2476 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2477 SLP_TREE_NUMBER_OF_VEC_STMTS.
2478 (vect_analyze_slp_cost_1): Remove.
2479 (vect_analyze_slp_cost): Likewise.
2480 (vect_slp_analyze_node_operations): Take visited args and
2481 a target cost vector. Avoid processing already visited stmt sets.
2482 (vect_slp_analyze_operations): Use a local cost vector to gather
2483 costs and register those of non-discarded instances.
2484 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2485 (vect_schedule_slp_instance): Remove copying of
2486 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2488 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2489 adding cost. Record cost entry location.
2490 (vect_prologue_cost_for_slp_op): Function to compute cost of
2491 a constant or invariant generated for SLP vect in the prologue,
2492 split out from vect_analyze_slp_cost_1.
2493 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2494 (vect_model_promotion_demotion_cost): Likewise.
2495 (vect_model_store_cost): Likewise, make static.
2496 (vect_model_load_cost): Likewise.
2497 (vectorizable_bswap): Add cost vector arg and adjust.
2498 (vectorizable_call): Likewise.
2499 (vectorizable_simd_clone_call): Likewise.
2500 (vectorizable_conversion): Likewise.
2501 (vectorizable_assignment): Likewise.
2502 (vectorizable_shift): Likewise.
2503 (vectorizable_operation): Likewise.
2504 (vectorizable_store): Likewise.
2505 (vectorizable_load): Likewise.
2506 (vectorizable_condition): Likewise.
2507 (vectorizable_comparison): Likewise.
2508 (can_vectorize_live_stmts): Likewise.
2509 (vect_analyze_stmt): Likewise.
2510 (vect_transform_stmt): Adjust calls to vectorizable_*.
2511 * tree-vectorizer.c: Include gimple-pretty-print.h.
2512 (dump_stmt_cost): New function.
2514 2018-05-16 Richard Biener <rguenther@suse.de>
2516 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2517 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2518 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2519 (check_name): New callback.
2520 (dse_classify_store): Track cycles via a visited bitmap of PHI
2521 defs and simplify handling of in-loop and across loop dead stores
2522 and properly fail for loop-variant refs. Handle byte-tracking with
2523 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2526 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2528 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2529 (vect_get_mask_type_for_stmt): Likewise.
2530 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2532 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2533 to determine the statement's vector type and the vector type that
2534 should be used for calculating nunits. Deal with cases in which
2535 the type has to be deferred.
2536 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2537 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2538 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2539 (vect_determine_vf_for_stmt): New functions, split out from...
2540 (vect_determine_vectorization_factor): ...here.
2541 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2542 (vect_get_mask_type_for_stmt): New functions, split out from
2543 vect_determine_vectorization_factor.
2545 2018-05-16 Richard Biener <rguenther@suse.de>
2547 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2548 verify the [VEC_]COND_EXPR embedded comparison.
2550 2018-05-15 Martin Sebor <msebor@redhat.com>
2552 PR tree-optimization/85753
2553 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2554 RECORD_TYPE in addition to ARRAY_TYPE.
2556 2018-05-15 Martin Sebor <msebor@redhat.com>
2559 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2561 2018-05-15 Richard Biener <rguenther@suse.de>
2563 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2564 add by_clobber_p one. Change algorithm to collect all defs
2565 representing uses we need to walk and try reducing them to
2566 a single one before failing.
2567 (dse_dom_walker::dse_optimize_stmt): Adjust.
2569 2018-05-13 Mark Wielaard <mark@klomp.org>
2571 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2572 (size_of_loc_descr): Likewise.
2573 (output_loc_operands): Likewise.
2574 (output_loc_operands_raw): Likewise.
2575 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2576 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2577 (hash_loc_operands): Likewise.
2578 (compare_loc_operands): Likewise.
2580 2018-05-14 Mark Wielaard <mark@klomp.org>
2582 * dwarf2out.c (count_index_addrs): New function.
2583 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2585 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2587 PR tree-optimization/83648
2588 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2589 return value as malloc candidate.
2591 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2594 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2595 param as true in call to suggest_attribute.
2597 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2599 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2600 -mreadonly-in-sdata.
2602 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2604 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2606 (aarch64_crypto_aesd_fused): Likewise.
2608 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2610 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2611 (movsi_aarch64): Likewise.
2612 (load_pairsi): Likewise.
2613 (load_pairdi): Likewise.
2614 (store_pairsi): Likewise.
2615 (store_pairdi): Likewise.
2616 (load_pairsf): Likewise.
2617 (load_pairdf): Likewise.
2618 (store_pairsf): Likewise.
2619 (store_pairdf): Likewise.
2620 (zero_extend): Likewise.
2621 (trunc): Swap alternatives.
2622 (fcvt_target): Add '?' to prefer w over r.
2624 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2627 * config/i386/i386.md: Disallow non-commutative arithmetics in
2628 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2629 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2630 in the peephole2 before it.
2632 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2634 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2635 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2636 (ix86_handle_option): Handle -mcldemote.
2637 * config.gcc: New header.
2638 * config/i386/cldemoteintrin.h: New file.
2639 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2640 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2642 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2643 OPTION_MASK_ISA_CLDEMOTE.
2644 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2645 (ix86_valid_target_attribute_inner_p): Ditto.
2646 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2647 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2648 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2649 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2650 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2652 * config/i386/i386.opt: Add -mcldemote.
2653 * config/i386/x86intrin.h: New header.
2654 * doc/invoke.texi: Add -mcldemote.
2656 2018-05-14 Richard Biener <rguenther@suse.de>
2658 * doc/match-and-simplify.texi: Adjust :s documentation.
2660 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2662 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2663 intended memcpy size.
2664 (REORDER_45): Likewise.
2666 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2668 * sort.cc: New file.
2669 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2670 * vec.c (qsort_chk): Use gcc_qsort.
2671 * Makefile.in (OBJS-libcommon): Add sort.o.
2672 (build/sort.o): New target. Use it...
2673 (BUILD_RTL): ... here, and...
2674 (build/gencfn-macros): ... here, and...
2675 (build/genmatch): ... here.
2677 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2678 Chung-Ju Wu <jasonwucj@gmail.com>
2680 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2681 * config/nds32/nds32-graywolf.md: New file.
2682 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2683 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2685 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2686 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2687 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2688 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2689 * config/nds32/pipelines.md: Include n15 settings.
2691 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2692 Chung-Ju Wu <jasonwucj@gmail.com>
2694 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2695 * config/nds32/nds32-n13.md: New file.
2696 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2697 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2699 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2700 * config/nds32/nds32.md (pipeline_model): Add n13.
2701 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2702 * config/nds32/pipelines.md: Include n13 settings.
2704 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2705 Chung-Ju Wu <jasonwucj@gmail.com>
2707 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2708 * config/nds32/nds32-n10.md: New file.
2709 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2710 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2712 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2713 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2714 * config/nds32/nds32.md (pipeline_model): Add n10.
2715 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2716 * config/nds32/pipelines.md: Include n10 settings.
2718 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2719 Kito Cheng <kito.cheng@gmail.com>
2720 Chung-Ju Wu <jasonwucj@gmail.com>
2722 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2723 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2724 Add enum values for DSP extension instructions.
2725 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2727 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2728 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2730 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2731 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2732 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2733 * config/nds32/nds32-intrinsic.md: Likewise.
2734 * config/nds32/nds32_intrinsic.h: Likewise.
2735 * config/nds32/nds32-md-auxiliary.c: Likewise.
2736 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2737 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2738 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2739 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2740 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2741 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2742 TYPE_DMAC in switch statement.
2743 * config/nds32/nds32.c: New checking and implementation for DSP
2744 extension instructions.
2745 * config/nds32/nds32.h: Likewise.
2746 * config/nds32/nds32.md: Likewise.
2747 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2748 * config/nds32/predicates.md: Implement new predicates for DSP
2751 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2753 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2754 Reformat alternatives and attributes so it is easier to identify
2755 which constraints/attributes go with which instruction.
2756 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2757 (mov<mode>_softfloat32, FMOVE64): Likewise.
2758 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2759 (mov<mode>_softfloat64, FMOVE64): Likewise.
2761 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2763 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2765 (Basic PowerPC Built-in Functions): The new name of the
2766 subsection previously known as "PowerPC Built-in Functions".
2767 (Basic PowerPC Built-in Functions Available on all Configurations):
2769 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2770 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2771 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2772 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2774 2018-05-11 Martin Jambor <mjambor@suse.cz>
2777 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2780 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2783 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2785 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2787 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2788 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2789 (ix86_handle_option): Handle -mwaitpkg.
2790 * config.gcc: New header.
2791 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2792 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2793 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2795 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2796 OPTION_MASK_ISA_WAITPKG.
2797 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2798 (ix86_option_override_internal): Add PTA_WAITPKG.
2799 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2800 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2801 IX86_BUILTIN_TPAUSE.
2802 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2803 __builtin_ia32_umwait and __builtin_ia32_tpause.
2804 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2805 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2806 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2807 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2808 UNSPECV_TPAUSE): New.
2809 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2810 * config/i386/i386.opt: Add -mwaitpkg.
2811 * config/i386/waitpkgintrin.h: New file.
2812 * config/i386/x86intrin.h: New header.
2813 * doc/invoke.texi: Add -mwaitpkg.
2815 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2818 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2820 (cortex-m0): Use armv6s-m isa.
2821 (cortex-m0plus): Likewise.
2822 (cortex-m1): Likewise.
2823 (cortex-m0.small-multiply): Likewise.
2824 (cortex-m0plus.small-multiply): Likewise.
2825 (cortex-m1.small-multiply): Likewise.
2827 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2828 Jakub Jelinek <jakub@redhat.com>
2830 PR tree-optimization/85692
2831 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2832 source permute as well.
2834 2018-05-11 Martin Liska <mliska@suse.cz>
2837 * doc/extend.texi: Document LLVM style format for no_sanitize
2840 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2842 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2843 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2844 (mode_supports_vsx_dform_quad): Likewise.
2845 (mode_supports_vmx_dform): Move these functions to be next to the
2846 other mode_supports functions.
2847 (mode_supports_dq_form): Likewise.
2848 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2849 mode_supports_dq_form.
2850 (reg_offset_addressing_ok_p): Likewise.
2851 (offsettable_ok_by_alignment): Likewise.
2852 (rs6000_legitimate_offset_address_p): Likewise.
2853 (legitimate_lo_sum_address_p): Likewise.
2854 (rs6000_legitimize_address): Likewise.
2855 (rs6000_legitimize_reload_address): Likewise.
2856 (rs6000_secondary_reload_inner): Likewise.
2857 (rs6000_preferred_reload_class): Likewise.
2858 (rs6000_output_move_128bit): Likewise.
2860 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2862 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2863 Generate SImode target register for null target.
2864 <case IX86_BUILTIN_XGETBV>: Ditto.
2865 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2866 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2868 2018-05-10 Carl Love <cel@us.ibm.com>
2870 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2871 dcbtt and dcbtstt if operands[2] is 0.
2873 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2876 * config/i386/sse.md (usadv64qi): New expander.
2878 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2880 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2881 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2882 -maltivec=be support.
2883 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2884 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2885 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2886 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2887 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2888 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2889 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2890 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2891 altivec_vsumsws): Adjust.
2892 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2893 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2894 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2896 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2897 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2898 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2899 (altivec_lve<VI_char>x): Delete expand.
2900 (*altivec_lve<VI_char>x_internal): Rename to...
2901 (altivec_lve<VI_char>x): ... this.
2902 (altivec_lvxl_<mode>): Delete expand.
2903 (*altivec_lvxl_<mode>_internal): Rename to ...
2904 (altivec_lvxl_<mode>): ... this.
2905 (altivec_stvxl_<mode>): Delete expand.
2906 (*altivec_stvxl_<mode>_internal): Rename to ...
2907 (altivec_stvxl_<mode>): ... this.
2908 (altivec_stve<VI_char>x): Delete expand.
2909 (*altivec_stve<VI_char>x_internal): Rename to ...
2910 (altivec_stve<VI_char>x): ... this.
2911 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2912 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2913 reduc_plus_scal_<mode>): Adjust.
2914 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2916 (rs6000_cpu_cpp_builtins): Adjust.
2917 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2918 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2919 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2920 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2921 -maltivec=be support.
2922 (rs6000_split_vec_extract_var): Adjust.
2923 (rs6000_split_v4si_init): Adjust.
2924 (swap_selector_for_mode): Delete.
2925 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2926 altivec_expand_stvex_be): Delete.
2927 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2928 -maltivec=be support.
2929 (rs6000_gimple_fold_builtin): Ditto.
2930 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2932 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2933 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2934 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2935 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2936 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2937 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2938 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2939 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2940 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2941 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2942 anonymous split): Adjust.
2943 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2944 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2946 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2948 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2949 when --with-gxx-include-dir is also specified.
2950 * configure: Regenerate.
2952 2018-05-09 Jim Wilson <jimw@sifive.com>
2955 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2956 * config/riscv/t-withmultilib: New.
2957 * config/riscv/withmultilib.h: New.
2958 * doc/install.texi: Document RISC-V --with-multilib-list support.
2960 2018-05-09 Richard Biener <rguenther@suse.de>
2962 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2964 (vect_bb_vectorization_profitable_p): Adjust. Compute
2965 actual scalar cost using the cost vector and the add_stmt_cost
2968 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2970 PR rtl-optimization/85645
2971 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2972 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2974 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2976 PR rtl-optimization/85645
2977 * shrink-wrap.c (spread_components): Return a boolean saying if
2978 anything was changed.
2979 (try_shrink_wrapping_separate): Iterate spread_components until
2980 nothing changes anymore.
2982 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2984 PR rtl-optimization/85645
2985 * regrename.c (build_def_use): Also kill the chains that include the
2986 destination of a REG_CFA_REGISTER note.
2988 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2990 PR rtl-optimization/85645
2991 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2992 insn that has a REG_CFA_REGISTER note.
2994 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2996 * cfgexpand.c (expand_clobber): New function.
2997 (expand_gimple_stmt_1): Use it.
2998 * tree-vect-stmts.c (vect_clobber_variable): New function,
3000 (vectorizable_simd_clone_call): ...here.
3001 (vectorizable_store): Emit a clobber either side of an
3002 IFN_STORE_LANES sequence.
3003 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3005 2018-05-09 Tom de Vries <tom@codesourcery.com>
3008 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3009 (define_insn "trap_if_false"): Add exit after trap.
3011 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3013 PR rtl-optimization/85638
3014 * bb-reorder.c: Include common/common-target.h.
3015 (create_forwarder_block): New function extracted from...
3016 (fix_up_crossing_landing_pad): ...here. Rename into...
3017 (dw2_fix_up_crossing_landing_pad): ...this.
3018 (sjlj_fix_up_crossing_landing_pad): New function.
3019 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3020 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3021 from both partitions and exit the loop after one iteration.
3023 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3026 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3028 (Basic PowerPC Built-in Functions): The new name of the
3029 subsection previously known as "PowerPC Built-in Functions".
3030 (Basic PowerPC Built-in Functions Available on all Configurations):
3032 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3034 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3036 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3038 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3041 2018-05-08 Jim Wilson <jimw@sifive.com>
3043 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3044 (LD_EMUL_SUFFIX): New.
3045 (LINK_SPEC): Use it.
3047 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3049 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3051 (Basic PowerPC Built-in Functions): The new name of the
3052 subsection previously known as "PowerPC Built-in Functions".
3053 (Basic PowerPC Built-in Functions Available on all Configurations):
3055 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3057 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3059 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3061 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3064 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3067 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3068 after cmpelim optimization.
3070 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3072 * config.gcc: Support "goldmont".
3073 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3074 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3076 * config/i386/i386.c (m_GOLDMONT): Define.
3077 (processor_target_table): Add "goldmont".
3078 (PTA_GOLDMONT): Define.
3079 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3080 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3081 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3082 (fold_builtin_cpu): Add "goldmont".
3083 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3084 (ix86_option_override_internal): Add "goldmont".
3085 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3086 (processor_type): Add PROCESSOR_GOLDMONT.
3087 * config/i386/i386.md: Add CPU "glm".
3088 * config/i386/glm.md: New file.
3089 * config/i386/x86-tune.def: Add m_GOLDMONT.
3090 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3092 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3095 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3097 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3098 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3099 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3102 * config/i386/i386.c (ix86_fold_builtin): Handle
3103 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3106 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3107 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3109 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3112 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3113 (check_arch): Likewise.
3114 (check_fpu): Return the result rather than printing it.
3115 (end arch): Fix operator precedence.
3116 (end cpu): Likewise.
3117 (END): Print the result from check_fpu.
3119 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3120 Alan Hayward <alan.hayward@arm.com>
3121 David Sherwood <david.sherwood@arm.com>
3123 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3124 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3125 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3126 (*fcmuo<mode>_and): New patterns.
3128 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3130 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3131 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3132 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3133 (cmp_op, sve_imm_con): New code attributes.
3134 (SVE_COND_INT_CMP, imm_con): Delete.
3135 (cmp_op): Remove above unspecs from int attribute.
3136 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3138 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3139 comparison-specific unspecs.
3140 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3141 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3142 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3143 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3144 (*vec_fcm<cmp_op><mode>): Rename to...
3145 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3146 (*vec_fcmuo<mode>): Rename to...
3147 (*fcmuo<mode>): ...this and adjust likewise.
3148 (*pred_fcm<cmp_op><mode>): New pattern.
3149 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3150 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3152 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3154 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3155 (aarch64_emit_sve_predicated_cond): New function.
3156 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3157 (aarch64_emit_unspec_cond_or): Replace with...
3158 (aarch64_emit_sve_or_conds): ...this new function. Use
3159 aarch64_emit_sve_ptrue_op for the individual comparisons and
3160 aarch64_emit_binop to OR them together.
3161 (aarch64_emit_inverted_unspec_cond): Replace with...
3162 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3163 aarch64_emit_sve_ptrue_op for the comparison and
3164 aarch64_emit_unop to invert the result.
3165 (aarch64_expand_sve_vec_cmp_float): Update after the above
3166 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3168 2018-05-07 Nathan Sidwell <nathan@acm.org>
3170 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3171 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3172 (Backwards Compatibility): Likewise.
3174 2018-05-07 Luis Machado <luis.machado@linaro.org>
3178 2018-05-07 Luis Machado <luis.machado@linaro.org>
3180 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3181 <prefetch_dynamic_strides>: New const bool field.
3182 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3183 prefetch_dynamic_strides.
3184 (exynosm1_prefetch_tune): Likewise.
3185 (thunderxt88_prefetch_tune): Likewise.
3186 (thunderx_prefetch_tune): Likewise.
3187 (thunderx2t99_prefetch_tune): Likewise.
3188 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3190 (aarch64_override_options_internal): Update to set
3191 PARAM_PREFETCH_DYNAMIC_STRIDES.
3192 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3193 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3194 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3195 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3196 prefetch-dynamic-strides setting.
3198 2018-05-07 Luis Machado <luis.machado@linaro.org>
3200 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3201 <minimum_stride>: New const int field.
3202 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3203 minimum_stride field.
3204 (exynosm1_prefetch_tune): Likewise.
3205 (thunderxt88_prefetch_tune): Likewise.
3206 (thunderx_prefetch_tune): Likewise.
3207 (thunderx2t99_prefetch_tune): Likewise.
3208 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3209 (aarch64_override_options_internal): Update to set
3210 PARAM_PREFETCH_MINIMUM_STRIDE.
3211 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3212 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3213 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3214 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3215 stride is constant and is below the minimum stride threshold.
3217 2018-05-07 Luis Machado <luis.machado@linaro.org>
3219 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3222 2018-05-07 Luis Machado <luis.machado@linaro.org>
3224 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3225 <prefetch_dynamic_strides>: New const bool field.
3226 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3227 prefetch_dynamic_strides.
3228 (exynosm1_prefetch_tune): Likewise.
3229 (thunderxt88_prefetch_tune): Likewise.
3230 (thunderx_prefetch_tune): Likewise.
3231 (thunderx2t99_prefetch_tune): Likewise.
3232 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3234 (aarch64_override_options_internal): Update to set
3235 PARAM_PREFETCH_DYNAMIC_STRIDES.
3236 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3237 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3238 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3239 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3240 prefetch-dynamic-strides setting.
3242 2018-05-07 Luis Machado <luis.machado@linaro.org>
3244 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3245 <minimum_stride>: New const int field.
3246 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3247 minimum_stride field.
3248 (exynosm1_prefetch_tune): Likewise.
3249 (thunderxt88_prefetch_tune): Likewise.
3250 (thunderx_prefetch_tune): Likewise.
3251 (thunderx2t99_prefetch_tune): Likewise.
3252 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3253 (aarch64_override_options_internal): Update to set
3254 PARAM_PREFETCH_MINIMUM_STRIDE.
3255 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3256 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3257 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3258 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3259 stride is constant and is below the minimum stride threshold.
3261 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3264 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3265 the type is addressable. Don't force op into register if it has
3268 2018-05-05 Roland McGrath <mcgrathr@google.com>
3271 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3272 any section for which we don't know a specific type it should have,
3273 regardless of name. Previously this was done only for the exact
3274 names ".init_array", ".fini_array", and ".preinit_array".
3275 (default_elf_asm_named_section): Add comment about
3276 relationship with default_section_type_flags and SECTION_NOTYPE.
3277 (get_section): Don't consider it a type conflict if one side has
3278 SECTION_NOTYPE and the other doesn't, as long as neither has the
3279 SECTION_BSS et al used in the default_section_type_flags logic.
3281 2018-05-05 Tom de Vries <tom@codesourcery.com>
3284 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3285 (workaround_barsyncs): New function.
3286 (nvptx_reorg): Use workaround_barsyncs.
3287 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3288 (define_expand "nvptx_membar_cta"): New define_expand.
3289 (define_insn "*nvptx_membar_cta"): New insn.
3291 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3293 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3294 To improve optimization opportunities.
3295 * builtin-types.def: The new needed builtin types for the above.
3297 2018-05-04 Richard Biener <rguenther@suse.de>
3299 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3300 * gimple-ssa-store-merging.c
3301 (imm_store_chain_info::output_merged_store): Remove redundant create,
3302 release split_store vector contents on failure.
3303 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3304 scalar stmt vector on cache hit.
3306 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3308 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3310 * config.gcc (powerpc-xilinx-eabi*): Remove.
3311 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3313 (fusion_addis_mem_combo_load): Ditto.
3314 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3316 (rs6000_cpu_cpp_builtins): Ditto.
3317 * config/rs6000/rs6000-linux.c
3318 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3319 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3320 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3322 (rs6000_setup_reg_addr_masks): Ditto.
3323 (rs6000_init_hard_regno_mode_ok): Ditto.
3324 (rs6000_option_override_internal): Ditto.
3325 (legitimate_lo_sum_address_p): Ditto.
3326 (rs6000_legitimize_address): Ditto.
3327 (rs6000_legitimize_reload_address): Ditto.
3328 (rs6000_legitimate_address_p): Ditto.
3329 (abi_v4_pass_in_fpr): Ditto.
3330 (setup_incoming_varargs): Ditto.
3331 (rs6000_gimplify_va_arg): Ditto.
3332 (rs6000_split_multireg_move): Ditto.
3333 (rs6000_savres_strategy): Ditto.
3334 (rs6000_emit_prologue_components): Ditto.
3335 (rs6000_emit_epilogue_components): Ditto.
3336 (rs6000_emit_prologue): Ditto.
3337 (rs6000_emit_epilogue): Ditto.
3338 (rs6000_elf_file_end): Ditto.
3339 (rs6000_function_value): Ditto.
3340 (rs6000_libcall_value): Ditto.
3341 * config/rs6000/rs6000.h: Ditto.
3342 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3343 (TARGET_MINMAX): ... this. New.
3344 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3345 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3346 (*movsi_internal1_single): Delete.
3347 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3348 mfpu=, mxilinx-fpu): Delete.
3349 * config/rs6000/singlefp.h: Delete.
3350 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3351 * config/rs6000/t-rs6000: Ditto.
3352 * config/rs6000/t-xilinx: Delete.
3353 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3354 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3355 (VStype_simple): Delete.
3356 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3357 * config/rs6000/xfpu.h: Delete.
3358 * config/rs6000/xfpu.md: Delete.
3359 * config/rs6000/xilinx.h: Delete.
3360 * config/rs6000/xilinx.opt: Delete.
3361 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3362 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3364 2018-05-04 Tom de Vries <tom@codesourcery.com>
3367 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3370 2018-05-04 Richard Biener <rguenther@suse.de>
3373 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3374 (expand_complex_div_wide): Likewise.
3375 (expand_complex_operations_1): Likewise.
3376 (expand_complex_libcall): Preserve EH info of the original stmt.
3377 (tree_lower_complex): Handle removed blocks.
3378 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3379 on complex multiplication and division libcall builtins.
3381 2018-05-04 Richard Biener <rguenther@suse.de>
3384 * fold-const.c (negate_expr_p): Restrict negation of operand
3385 zero of a division to when we know that can happen without
3387 (fold_negate_expr_1): Likewise.
3389 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3392 * real.h (real_nextafter): Declare.
3393 * real.c (real_nextafter): New function.
3394 * fold-const-call.c (fold_const_nextafter): New function.
3395 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3396 CASE_CFN_NEXTTOWARD.
3397 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3398 even when arg1_mode is different from arg0_mode.
3400 2018-05-03 Nathan Sidwell <nathan@acm.org>
3402 * doc/extend.texi (Deprecated Features): Remove
3404 (Backwards Compatibility): Likewise.
3405 * doc/invoke.texi (C++ Language Options): Likewise.
3406 (C++ Dialect Options): Likewise.
3408 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3411 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3412 _mm512_mask_mullox_epi64): New intrinsics.
3414 2018-05-03 Tom de Vries <tom@codesourcery.com>
3417 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3418 dump files): Add offload-tree.
3420 2018-05-03 Richard Biener <rguenther@suse.de>
3422 PR tree-optimization/85615
3423 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3424 to loops not nested in BBs loop father to avoid creating multi-entry
3427 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3429 PR tree-optimization/70291
3430 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3431 arguments. Change return type to tree. Emit libcall as a new
3432 statement rather than replacing existing one when inplace_p is true.
3433 (expand_complex_multiplication_components): New function.
3434 (expand_complex_multiplication): Expand floating-point complex
3435 multiplication using the above.
3436 (expand_complex_division): Rename inner_type parameter to type.
3437 Update expand_complex_libcall call-site.
3438 (expand_complex_operations_1): Update expand_complex_multiplication
3439 and expand_complex_division call-sites.
3441 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3444 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3445 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3446 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3447 the highest significant bit of the shift count mask is clear. In
3448 check whether and[sq]i3 is needed verify that all significant bits
3449 of the shift count other than the highest are set.
3451 2018-05-02 Tom de Vries <tom@codesourcery.com>
3454 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3455 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3456 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3457 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3458 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3459 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3460 __builtin_goacc_parlevel_size.
3462 2018-05-02 Richard Biener <rguenther@suse.de>
3464 PR tree-optimization/85597
3465 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3466 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3469 2018-05-02 Tom de Vries <tom@codesourcery.com>
3472 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3473 dump files): Add ltrans-tree.
3475 2018-05-02 Tom de Vries <tom@codesourcery.com>
3478 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3479 dump files): Add wpa-ipa.
3481 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3483 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3484 powerpc*-*-linux*paired* target.
3485 * config/rs6000/750cl.h: Delete.
3486 * config/rs6000/paired.h: Delete.
3487 * config/rs6000/paired.md: Delete.
3488 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3490 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3491 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3492 comment. Remove paired float support.
3493 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3494 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3496 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3497 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3499 * config/rs6000/rs6000.c: Remove paired float support.
3500 (paired_expand_vector_init, paired_expand_vector_move,
3501 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3502 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3503 paired_expand_builtin, paired_expand_predicate_builtin,
3504 paired_init_builtins): Delete.
3505 * config/rs6000/rs6000.h: Remove paired float support.
3506 * config/rs6000/rs6000.md: Remove paired float support.
3507 (move_from_CR_ov_bit): Delete.
3508 * config/rs6000/rs6000.opt (mpaired): Delete.
3509 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3510 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3512 2018-05-02 Richard Biener <rguenther@suse.de>
3515 * gimplify.c (gimplify_save_expr): When in SSA form allow
3516 SAVE_EXPRs to compute to SSA vars.
3518 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3521 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3522 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3523 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3524 clobber operands[2], instead use a new pseudo. Formatting fixes.
3526 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3528 PR tree-optimization/85586
3529 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3530 exit early for statements in the same group if the accesses are
3533 2018-05-02 Tom de Vries <tom@codesourcery.com>
3536 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3539 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3541 PR tree-optimization/85143
3542 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3544 2018-05-01 Tom de Vries <tom@codesourcery.com>
3547 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3548 not found" error message.
3550 2018-05-01 Tom de Vries <tom@codesourcery.com>
3553 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3554 * vec.c (test_ordered_remove_if): New function.
3555 (vec_c_tests): Call test_ordered_remove_if.
3556 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3557 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3558 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3559 VEC_ORDERED_REMOVE_IF.
3561 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3563 PR tree-optimization/82665
3564 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3565 pointer subtraction where arguments come from a memchr call.
3567 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3569 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3570 --push-state --as-needed and --pop-state instead of --as-needed and
3571 --no-as-needed if ld supports it.
3572 * configure: Regenerated.
3575 * doc/install.texi2html: Replace _002d with - and _002a with * in
3576 generated html files using sed.
3578 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3581 * gcc-rich-location.c (blank_line_before_p): New function.
3582 (use_new_line): New function.
3583 (gcc_rich_location::add_fixit_insert_formatted): New function.
3584 * gcc-rich-location.h
3585 (gcc_rich_location::add_fixit_insert_formatted): New function.
3587 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3589 * selftest.c (assert_streq): Rename "expected" and "actual" to
3590 "val1" and "val2". Extend NULL-handling to cover both inputs
3591 symmetrically, while still requiring both to be non-NULL for a pass.
3592 * selftest.h (assert_streq): Rename "expected" and "actual" to
3594 (ASSERT_EQ): Likewise.
3595 (ASSERT_EQ_AT): Likewise.
3596 (ASSERT_KNOWN_EQ): Likewise.
3597 (ASSERT_KNOWN_EQ_AT): Likewise.
3598 (ASSERT_NE): Likewise.
3599 (ASSERT_MAYBE_NE): Likewise.
3600 (ASSERT_MAYBE_NE_AT): Likewise.
3601 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3602 the assertion to pass.
3603 (ASSERT_STREQ_AT): Likewise.
3605 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3607 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3608 interaction with -pie.
3610 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3612 * selftest.h: Fix alphabetization of per-source-file selftest
3615 2018-04-30 Jason Merrill <jason@redhat.com>
3617 PR c++/61982 - dead stores to destroyed objects.
3618 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3621 2018-04-30 Jason Merrill <jason@redhat.com>
3623 * tree.c (build_clobber): New.
3624 * tree.h: Declare it.
3625 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3627 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3629 * diagnostic-show-locus.c (layout::layout): Update for
3630 location_get_source_line returning a char_span.
3631 (struct char_span): Move to input.h.
3632 (struct correction): Update for fields in char_span becoming
3634 (struct source_line): Update for location_get_source_line
3635 returning a char_span.
3636 (layout::print_line): Likewise.
3637 * edit-context.c (edited_file::print_content): Likewise.
3638 (edited_file::print_diff_hunk): Likewise.
3639 (edited_file::print_run_of_changed_lines): Likewise.
3640 (edited_file::get_num_lines): Likewise.
3641 (edited_line::edited_line): Likewise.
3642 * final.c (asm_show_source): Likewise.
3643 * input.c (location_get_source_line): Convert return type
3644 from const char * to char_span, losing the final "line_len"
3646 (dump_location_info): Update for the above.
3647 (get_substring_ranges_for_loc): Likewise. Use a char_span
3648 when handling the literal within the line.
3649 (test_reading_source_line): Update for location_get_source_line
3650 returning a char_span.
3651 * input.h (class char_span): Move here from
3652 diagnostic-show-locus.c, converting from a struct to a class.
3653 Make data members private.
3654 (char_span::operator bool): New.
3655 (char_span::length): New.
3656 (char_span::get_buffer): New.
3657 (char_span::operator[]): New.
3658 (char_span::subspan): Make const.
3659 (char_span::xstrdup): New.
3660 (location_get_source_line): Convert return type from const char *
3661 to char_span, losing the final "line_size" param.
3663 2018-04-30 Jan Hubicka <jh@suse.cz>
3665 * lto-wrapper.c (ltrans_priorities): New static var.
3666 (cmp_priority): New.
3667 (run_gcc): Read priorities and if doing parallel build order
3668 the Makefile by them.
3670 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3672 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3674 2018-04-30 Richard Biener <rguenther@suse.de>
3676 * tree-cfg.c (verify_address): Remove base argument, add
3677 flag whether to check TREE_ADDRESSABLE and do that.
3678 (verify_expr): Remove.
3679 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3680 (verify_gimple_assign_single): Likewise.
3681 (verify_gimple_switch): Likewise.
3682 (verify_expr_location_1): Dereference tp once. Add (disabled)
3683 piece from verify_expr.
3684 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3686 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3688 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3690 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3692 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3693 (small_data_pattern): Likewise.
3694 (arc_rewrite_small_data): Likewise.
3695 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3696 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3697 (get_symbol_alignment): New function.
3698 (legitimate_small_data_address_p): Likewise.
3699 (legitimate_scaled_address): Update, call
3700 legitimate_small_data_address_p.
3701 (output_sdata): New static variable.
3702 (arc_print_operand): Update how we handle small data operands.
3703 (arc_print_operand_address): Likewise.
3704 (arc_legitimate_address_p): Update, use
3705 legitimate_small_data_address_p.
3706 (arc_rewrite_small_data_p): Remove.
3707 (arc_rewrite_small_data_1): Likewise.
3708 (arc_rewrite_small_data): Likewise.
3709 (small_data_pattern): Likewise.
3710 (compact_sda_memory_operand): Update to use
3711 legitimate_small_data_address_p and get_symbol_alignment.
3712 (prepare_move_operands): Don't rewite sdata pattern.
3713 (prepare_extend_operands): Remove.
3714 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3716 (zero_extendqisi2): Likewise.
3717 (zero_extendhisi2): Likewise.
3718 (extendqihi2): Likewise.
3719 (extendqisi2): Likewise.
3720 (extendhisi2): Likewise.
3724 * config/arc/constraints.md (Usd): Change it to memory constraint.
3726 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3728 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3729 as source of std instructions.
3730 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3731 allow 6-bit constants as source for store instructions.
3732 (movdi_insn): Update instruction pattern to allow 6-bit constants
3733 as source for store instructions.
3735 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3737 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3739 2018-04-30 Nathan Sidwell <nathan@acm.org>
3740 Sandra Loosemore <sandra@codesourcery.com>
3742 * dumpfile.c (dump_open): Allow '-' for stdout.
3743 * doc/invoke.texi (Developer Options): Document dump filename
3744 determination early. Document stdin/stdout selection.
3746 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3748 Microblaze Target: PIC data text relative
3750 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3751 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3753 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3754 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3755 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3756 New addressing mode for data-text relative position indepenedent code.
3757 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3758 'ADDRESS_SYMBOLIC_TXT_REL'.
3759 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3760 (microblaze_legitimate_pic_operand): Exclude function calls from
3761 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3762 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3764 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3765 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3766 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3767 for 'address + offset'.
3768 (microblaze_expand_prologue): Add new function prologue call for
3770 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3771 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3772 table in case of TARGET_PIC_DATA_TEXT_REL.
3773 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3774 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3775 Add new macros 'UNSPEC_TEXT',
3776 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3777 + exclude function calls from 'UNSPEC_PLT' in case of data text
3779 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3780 new target hook for generating address diff vector tables in case of
3782 * doc/tm.texi : Regenerate.
3783 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3784 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3785 of addr diff vector generation.
3786 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3787 target hook definition.
3788 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3789 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3790 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3791 option for data text relative.
3793 2018-04-30 Richard Biener <rguenther@suse.de>
3795 * tree-chrec.h (evolution_function_is_constant_p): Remove
3797 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3799 2018-04-30 Richard Biener <rguenther@suse.de>
3802 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3804 2018-04-30 Richard Biener <rguenther@suse.de>
3806 PR tree-optimization/28364
3807 PR tree-optimization/85275
3808 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3809 copying first exit test.
3811 2018-04-28 Mark Wielaard <mark@klomp.org>
3813 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3815 (dwarf_AT): Handle DW_AT_addr_base.
3816 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3818 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3821 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3822 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3823 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3824 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3826 2018-04-28 Richard Biener <rguenther@suse.de>
3828 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3829 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3830 to reflect use. Only add interesting stmts.
3832 2018-04-27 Martin Jambor <mjambor@suse.cz>
3835 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3836 the jump function allows for passing through aggregate values.
3838 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3840 * input.h (in_system_header_at): Convert from macro to inline
3842 (from_macro_expansion_at): Likewise.
3843 (from_macro_definition_at): Likewise.
3845 2018-04-27 Jeff Law <law@redhat.com>
3847 * config.gcc: Mark tile* targets as deprecated/obsolete.
3849 2018-04-27 Richard Biener <rguenther@suse.de>
3851 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3854 2018-04-27 Richard Biener <rguenther@suse.de>
3856 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3858 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3860 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3861 with Yd constraint. Set "preferred_for_speed" attribute from
3862 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3864 (*movdi_internal): Ditto.
3865 (movti_interunit splitters): Remove
3866 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3867 (movdi_interunit splitters): Ditto.
3868 * config/i386/constraints.md (Ye): Remove.
3869 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3871 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3874 * config/aarch64/constraints.md (Usg): Limit to 31.
3877 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3879 PR tree-optimization/85529
3880 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3881 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3882 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3883 zero extension or masking of the MSB bit.
3884 (optimize_range_tests): Add FIRST_BB argument, pass it through
3885 to optimize_range_tests_var_bound.
3886 (maybe_optimize_range_tests, reassociate_bb): Adjust
3887 optimize_range_tests callers.
3889 2018-04-26 Richard Biener <rguenther@suse.de>
3890 Jakub Jelinek <jakub@redhat.com>
3892 * cgraph.h (symbol_table): Just declare debug method here.
3893 * symtab.c (symbol_table::debug): Define.
3895 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3897 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3899 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3901 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3902 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3903 (*movdi_internal): Substitute Yi and Yj constraint with x
3904 and Ym and Yn constraint with y constraint. Update "isa"
3905 attribute and set "preferred_for_speed" attribute from
3906 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3907 (*movsi_internal): Ditto.
3908 (*movdf_internal): Ditto.
3909 (*movsf_internal): Ditto.
3910 (*zero_extendsidi2): Ditto.
3911 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3912 (sse2_loadld): Ditto.
3913 (*vec_extract<ssevecmodelower>_0): Ditto.
3914 (*vec_extractv4si_0_zext_sse4): Ditto.
3915 (vec_concatv2di): Ditto.
3916 (*vec_dup<mode>): Ditto.
3917 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3918 * config/i386/constraints.md (Yi): Remove.
3923 2018-04-26 Nathan Sidwell <nathan@acm.org>
3925 * dumpfile.c (dump_open): New.
3926 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3927 (dump_finish): Detect stdio/stderr by value not name.
3929 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3931 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3933 2018-04-26 Tom de Vries <tom@codesourcery.com>
3936 * config/nvptx/nvptx.c (verify_neutering_jumps)
3937 (verify_neutering_labels): New function
3938 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3940 2018-04-26 Tom de Vries <tom@codesourcery.com>
3943 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3944 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3947 2018-04-26 Richard Biener <rguenther@suse.de>
3948 Tom de Vries <tom@codesourcery.com>
3951 * lto-streamer-out.c (output_function): Fixup loops if required to match
3952 discovery done in the reader.
3954 2018-04-26 Richard Biener <rguenther@suse.de>
3956 PR tree-optimization/85116
3957 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3958 have a loop exit from the single latch predecessor. Remove
3959 case of header with just condition.
3960 (ch_base::copy_headers): Exclude infinite loops from any
3962 (pass_ch::execute): Record exits.
3964 2018-04-26 Richard Biener <rguenther@suse.de>
3966 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3967 prologue cost vector and pass it to vect_get_load_cost.
3968 (vect_get_peeling_costs_all_drs): Likewise.
3969 (vect_peeling_hash_get_lowest_cost): Likewise.
3970 (vect_enhance_data_refs_alignment): Likewise.
3972 2018-04-26 Richard Biener <rguenther@suse.de>
3975 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3976 checking of integer<->pointer conversions.
3977 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3978 sign-/zero-extending pointer types.
3979 (expand_omp_for_static_chunk): Likewise.
3981 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3982 Jean Lee <xiaoyur347@gmail.com>
3984 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3985 (TARGET_ASAN_SHADOW_OFFSET): Define.
3986 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3987 true for -fsanitize=address.
3989 2018-04-25 Mark Wielaard <mark@klomp.org>
3991 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3994 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3996 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3997 than "alu", remove explicit "memory" and "imm_disp" attributes.
3998 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4001 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4002 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4003 gen_lowpart_no_emit.
4005 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4008 * config/i386/i386.c (ix86_expand_builtin): Change memory
4009 operand to XI, extend p0 to Pmode.
4010 * config/i386/i386.md: Change unspec volatile and operand
4011 1 mode to XI, change operand 0 mode to P.
4013 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4015 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4016 GET_MODE_MASK before any checking.
4017 (nds32_can_use_bset_p): Likewise.
4018 (nds32_can_use_btgl_p): Likewise.
4020 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4022 * config/nds32/nds32-doubleword.md: New define_split pattern for
4023 illegal register number.
4025 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4027 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4029 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4031 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4033 2018-04-25 Richard Biener <rguenther@suse.de>
4035 * lto-streamer.h (LTO_major_version): Bump to 8.
4037 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4039 * BASE-VER: Set to 9.0.0.
4041 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4043 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4044 in __abskf2 and __powikf2.
4046 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4049 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4050 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4051 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4052 Use the above on operand 2. Reindent.
4053 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4055 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4058 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4060 * config/i386/i386.opt (mcet): Removed.
4061 * doc/install.texi: Remove -mcet documentation.
4062 * doc/invoke.texi: Likewise.
4064 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4067 * doc/install.texi: Remove -mcet from bootstrap-cet.
4069 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4072 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4073 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4077 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4078 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4079 containing a CONST_VECTOR.
4081 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4083 * doc/install.texi: Update newlib dependency for nvptx.
4085 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4088 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4089 instead of INTVAL when shifting x left.
4091 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4093 PR tree-optimization/85478
4094 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4095 vect_grouped_store_supported for single element vectors.
4097 2018-04-24 Richard Biener <rguenther@suse.de>
4100 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4101 load cost increase to the case of non-constant step.
4103 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4106 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4107 destination if any_malformed_asm.
4109 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4112 * expr.c (store_field): In the bitfield case, if the value comes from
4113 a function call and is returned in registers by means of a PARALLEL,
4114 do not change the mode of the temporary unless BLKmode and VOIDmode.
4116 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4118 PR rtl-optimization/85423
4119 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4120 dependencies to debug insns when the previous insn is non-debug.
4122 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4124 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4125 enums into a single definition.
4126 (fls): Fix predicates and printing.
4129 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4131 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4132 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4133 and short u6 immediate.
4134 (check_if_valid_sleep_operand): Remove.
4135 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4137 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4139 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4140 flag_always_save_lp condition.
4141 * config/nds32/nds32.opt (malways-save-lp): New option.
4143 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4145 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4146 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4147 * config/nds32/nds32.h
4148 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4149 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4151 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4153 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4154 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4156 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4157 Chung-Ju Wu <jasonwucj@gmail.com>
4159 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4161 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4162 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4164 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4166 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4168 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4170 * config/nds32/nds32-protos.h (nds32_data_alignment,
4171 nds32_local_alignment): Declare.
4172 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4173 nds32_local_alignment): New functions.
4174 (TARGET_CONSTANT_ALIGNMENT): Define.
4175 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4177 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4179 * config/nds32/nds32.c
4180 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4181 (TARGET_MODES_TIEABLE_P): Likewise.
4183 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4185 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4188 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4189 Chung-Ju Wu <jasonwucj@gmail.com>
4191 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4192 for unaligned access.
4193 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4195 * config/nds32/nds32-intrinsic.md: Likewise.
4196 * config/nds32/nds32_intrinsic.h: Likewise.
4197 * config/nds32/nds32.h (nds32_builtins): Likewise.
4198 * config/nds32/nds32.opt (munaligned-access): New option.
4199 * config/nds32/nds32.c (nds32_asm_file_start): Display
4200 flag_unaligned_access status.
4202 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4204 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4205 -mno-relax is present.
4206 * config/riscv/linux.h (LINK_SPEC): Ditto.
4208 2018-04-20 Martin Sebor <msebor@redhat.com>
4211 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4213 (gimple_fold_builtin_stxcpy_chk): Same.
4214 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4216 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4219 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4220 __powikf2 when long double is IEEE 128-bit.
4222 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4224 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4225 step to make sure stack always aligned.
4227 2018-04-20 Carl Love <cel@us.ibm.com>
4230 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4231 size check for arg0.
4233 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4234 Tom de Vries <tom@codesourcery.com>
4237 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4238 Emit insns for calls too.
4239 (nvptx_find_par): Always look for worker-level predecessor insn.
4240 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4242 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4243 (nvptx_process_pars): Propagate frames for calls.
4245 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4248 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4250 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4251 (ix86_handle_option): Don't handle OPT_mibt.
4252 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4254 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4256 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4257 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4258 (ix86_target_macros): Define __CET__ with flag_cf_protection
4259 for -fcf-protection.
4260 * config/i386/i386.c (isa2_opts): Remove -mibt.
4261 * config/i386/i386.h (TARGET_IBT): Removed.
4262 (TARGET_IBT_P): Likewise.
4263 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4264 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4265 * config/i386/i386.opt (mcet): Update help message.
4268 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4269 -mcet as an alias for -mshstk.
4271 2018-04-20 Richard Biener <rguenther@suse.de>
4274 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4275 complexity by forcing a single use of the multiply operand.
4277 2018-04-20 Martin Jambor <mjambor@suse.cz>
4280 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4281 recursion dependency to only apply to non-clones.
4283 2018-04-20 Martin Jambor <mjambor@suse.cz>
4286 * ipa-cp.c (create_specialized_node): Check that clones of
4287 self-recursive edges exist during IPA-CP.
4289 2018-04-19 Toon Moene <toon@moene.org>
4291 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4294 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4296 PR tree-optimization/85467
4297 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4298 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4299 VECTOR_CST element to type.
4301 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4304 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4305 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4306 (builtin_longjmp): Likewise.
4307 (save_stack_nonlocal): New pattern.
4308 (restore_stack_nonlocal): Likewise.
4310 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4313 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4314 Replace ASM_OUTPUT_LABEL with fprintf.
4316 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4319 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4320 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4321 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4322 define __IBT__ and __SHSTK__ for -fcf-protection.
4323 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4325 (ix86_trampoline_init): Likewise.
4326 (x86_output_mi_thunk): Likewise.
4327 (ix86_notrack_prefixed_insn_p): Likewise.
4328 (ix86_option_override_internal): Don't disallow -fcf-protection.
4329 * config/i386/i386.md (rdssp<mode>): Also enable for
4331 (incssp<mode>): Likewise.
4332 (nop_endbr): Likewise.
4333 * config/i386/i386.opt (mcet): Change help message to built-in
4337 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4338 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4339 enable CET built-in functions.
4341 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4343 * common/config/i386/i386-common.c
4344 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4345 OPTION_MASK_ISA_MOVDIRI_UNSET,
4346 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4347 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4348 * config.gcc (movdirintrin.h): New header.
4349 * config/i386/cpuid.h (bit_MOVDIRI,
4350 bit_MOVDIR64B): New bits.
4351 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4353 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4354 (VOID, PVOID, PCVOID)): New function types.
4355 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4356 __builtin_ia32_directstoreu_u64,
4357 __builtin_ia32_movdir64b): New builtins.
4358 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4359 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4361 (ix86_valid_target_attribute_inner_p): Ditto.
4362 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4363 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4364 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4365 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4366 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4367 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4368 (movdiri<mode>, movdir64b_<mode>): New.
4369 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4370 * config/i386/immintrin.h: Include movdirintrin.h.
4371 * config/i386/movdirintrin.h: New file.
4372 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4374 2018-04-19 Richard Biener <rguenther@suse.de>
4377 * cfg.c (clear_bb_flags): When loop state says we have
4378 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4380 2018-04-19 Richard Biener <rguenther@suse.de>
4382 PR tree-optimization/84737
4383 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4384 copying restrict info.
4385 (vect_setup_realignment): Use it.
4386 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4387 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4388 the first DR to all generated stores.
4389 (vectorizable_load): Likewise for loads.
4391 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4393 PR tree-optimization/85446
4394 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4395 the integral and pointer types to have the same precision.
4397 * doc/install.texi: Document --disable-cet being the default and
4400 2018-04-18 Martin Liska <mliska@suse.cz>
4402 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4405 2018-04-18 Martin Liska <mliska@suse.cz>
4408 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4411 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4412 arguments if they are comparable.
4414 2018-04-18 Martin Liska <mliska@suse.cz>
4417 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4420 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4423 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4426 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4427 ENDBR after calling __morestack.
4429 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4432 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4433 by using gcc_base_ver to generate a gcc_driver_version, and use
4434 it when generating GCC_DRIVER_NAME.
4435 * configure: Regenerate.
4437 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4440 * config.gcc: Obsolete powerpc*-*-*spe*.
4442 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4445 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4446 (stabstr_D): Change type of unum from unsigned int to
4447 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4450 2018-04-17 Jim Wilson <jimw@sifive.com>
4453 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4454 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4455 Set arg_pointer_offset after using pretend_args_size.
4457 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4459 PR rtl-optimization/85431
4460 * dse.c (record_store): Ignore zero width stores.
4463 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4464 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4465 __builtin_stack_restore rather than after it.
4466 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4467 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4468 argument instead of virtual_dynamic_stack_rtx.
4470 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4472 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4474 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4475 Add note to error message to explain internal mapping of overloaded
4476 built-in function name to non-overloaded built-in function name.
4477 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4480 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4483 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4484 where the inputs overlap with the output.
4486 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4488 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4489 (=v, v) alternative and explicit "memory" attribute.
4490 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4491 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4493 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4494 "sselog1" type instead of "sselog".
4495 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4496 "sselog". Remove explicit "memory" attribute.
4497 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4498 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4500 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4501 "sselog1" type instead of "sselog". Remove explicit "memory"
4503 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4504 use "sselog1" type instead of "sselog". Remove explicit "memory"
4506 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4507 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4509 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4510 "sselog1" type instead of "sselog". Remove explicit "memory"
4512 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4513 use "sselog1" type instead of "sselog". Remove explicit "memory"
4517 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4520 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4523 2018-04-17 Martin Jambor <mjambor@suse.cz>
4526 * ipa-cp.c (create_specialized_node): Call
4527 expand_all_artificial_thunks if necessary.
4529 2018-04-17 Martin Liska <mliska@suse.cz>
4532 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4533 in message, remote space in between '_G' and '('.
4535 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4538 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4539 avx512f_vmcmp<mode>3<round_saeonly_name>,
4540 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4541 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4542 avx512f_rndscale<mode><round_saeonly_name>,
4543 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4544 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4545 Use %<iptr>2 instead of %2 for -masm=intel.
4546 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4547 avx512f_vcvttss2usi<round_saeonly_name>,
4548 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4550 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4551 avx512f_vcvttsd2usi<round_saeonly_name>,
4552 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4553 Use %q1 instead of %1 for -masm=intel.
4554 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4555 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4556 of %3 for -masm=intel.
4557 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4559 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4561 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4563 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4564 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4566 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4568 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4569 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4570 %g1 and one with %0 and %1.
4571 (avx512er_vmrcp28<mode><round_saeonly_name>,
4572 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4574 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4575 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4576 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4577 of %0 and %{%4%} for -masm=intel.
4578 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4579 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4580 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4581 order of %0 and %{%5%}%{z%} for -masm=intel.
4583 2018-04-17 Jan Hubicka <jh@suse.cz>
4586 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4588 2018-04-17 Martin Liska <mliska@suse.cz>
4591 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4592 for target_clone error message. Make default implementation
4593 clone to be a local declaration.
4594 (separate_attrs): Add new argument and check for an empty
4596 (expand_target_clones): Handle it.
4597 (ipa_target_clone): Make redirection just for target_clones
4600 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4601 Tom de Vries <tom@codesourcery.com>
4604 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4605 tiled basic blocks without omp continue statements.
4607 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4610 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4611 vec_extract expression as having side effects to make sure it gets
4614 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4617 * config/i386/i386.c (get_builtin_code_for_version): Check
4620 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4623 * gcc/config.gcc: Support "skylake".
4624 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4626 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4627 (processor_target_table): Add "skylake".
4628 (ix86_option_override_internal): Add "skylake".
4629 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4630 PROCESSOR_CANNONLAKE.
4631 (get_builtin_code_for_version): Fix priority for
4632 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4633 PROCESSOR_SKYLAKE-AVX512.
4634 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4635 (processor_type): Add PROCESSOR_SKYLAKE.
4637 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4638 Jason Merrill <jason@redhat.com>
4641 * convert.c (convert_to_integer_1): Use direct recursion for
4642 enumeral types and types with a precision less than the number
4643 of bits in their mode.
4645 2018-04-16 Julia Koval <julia.koval@intel.com>
4648 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4649 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4651 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4654 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4655 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4656 and -mno-direct-move.
4658 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4661 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4662 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4663 Ensure negative shifts result in {0}.
4665 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4667 PR rtl-optimization/79916
4668 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4669 regs (if any) to define how to gnerate SD moves when LRA is in
4672 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4674 PR rtl-optimization/85393
4675 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4676 * except.c (expand_dw2_landing_pad_for_region): Make static.
4677 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4678 a label and unconditional jump to old_bb, rather than
4679 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4682 PR rtl-optimization/85376
4683 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4684 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4685 instead of a specific value.
4687 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4688 Bin Cheng <bin.cheng@arm.com>
4690 PR tree-optimization/82965
4691 PR tree-optimization/83991
4692 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4693 by_profile_only parameter.
4694 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4695 information if the loop was predicted to iterate too many times.
4696 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4698 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4701 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4704 2018-04-13 Martin Liska <mliska@suse.cz>
4705 Jakub Jelinek <jakub@redhat.com>
4708 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4709 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4710 * builtins.c (expand_builtin_memory_copy_args): Use
4711 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4712 handle dest_addr == pc_rtx.
4714 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4717 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4718 asked to not generate direct moves.
4719 (fix_trunc<mode>si2_stfiwx): Similar.
4720 (fix_trunc<mode>si2_internal): Similar.
4722 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4725 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4726 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4727 lookup if dest in some wider mode is known to be const0_rtx and
4728 if so, record permanent equivalence for it to be ZERO_EXTEND of
4729 the narrower mode destination.
4731 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4733 * lto-streamer-out.c (output_function): Revert 259346.
4734 * omp-expand.c (expand_oacc_for): Likewise.
4736 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4738 PR rtl-optimization/85354
4739 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4740 * sel-sched.c (sel_global_init): ... here.
4742 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4745 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4746 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4747 mode for PE-COFF targets.
4748 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4749 (i386_pe_asm_lto_end): Likewise.
4750 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4751 (TARGET_ASM_LTO_END): Likewise.
4752 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4753 (i386_pe_asm_lto_start): New function.
4754 (i386_pe_asm_lto_end): Likewise.
4756 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4757 Richard Biener <rguenther@suse.de>
4760 * lto-streamer-out.c (output_function): Fix CFG loop state before
4762 * omp-expand.c (expand_oacc_for): Handle calls to internal
4763 functions like regular functions.
4765 2018-04-12 Richard Biener <rguenther@suse.de>
4768 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4769 for the early LTO debug to properly generate references to it
4770 during DIE emission. Do not re-use that for the skeleton for
4772 (dwarf2out_early_finish): Likewise.
4774 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4777 * config/i386/sse.md
4778 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4779 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4780 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4781 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4782 and output is a reg, avoid creating invalid lowpart subreg, but
4783 instead split into a 512-bit move. Don't split if not AVX512VL,
4784 input is xmm16+ reg and output is a mem.
4785 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4786 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4787 xmm16+ reg and output is a mem.
4789 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4791 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4792 also for flag_dwarf2_cfi_asm.
4794 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4796 PR rtl-optimization/85342
4797 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4798 a bool scalar var inside of the loop instead. Don't try to update
4799 recog_data.operand after failed apply_change_group.
4801 2018-04-12 Tom de Vries <tom@codesourcery.com>
4804 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4805 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4806 array with flexible array member as array without given dimension.
4807 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4808 argument for undefined param to true.
4810 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4813 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4814 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4815 from PowerPC section.
4816 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4817 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4818 help text that is too long.
4819 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4820 help text that is too long.
4821 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4822 help text that is too long.
4824 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4826 * config/alpha/alpha.md (stack_probe_internal): Rename
4827 from "probe_stack". Update all callers.
4829 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4831 PR rtl-optimization/84566
4832 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4833 sched_macro_fuse_insns.
4835 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4838 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4839 (compute_block_dependences): ... from here.
4841 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4843 PR tree-optimization/85331
4844 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4845 from int to HOST_WIDE_INT.
4847 2018-04-11 Martin Jambor <mjambor@suse.cz>
4850 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4851 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4852 not the same as the source val.
4853 (cgraph_edge_brings_value_p): New parameter.
4854 (gather_edges_for_value): Pass destination value to
4855 cgraph_edge_brings_value_p.
4856 (perhaps_add_new_callers): Likewise.
4857 (get_info_about_necessary_edges): Likewise and exclude values brought
4858 only by self-recursive edges.
4859 (create_specialized_node): Redirect only clones of self-calling edges.
4860 (+self_recursive_pass_through_p): New function.
4861 (find_more_scalar_values_for_callers_subset): Use it.
4862 (find_aggregate_values_for_callers_subset): Likewise.
4863 (known_aggs_to_agg_replacement_list): Removed.
4864 (decide_whether_version_node): Re-calculate known constants for all
4865 remaining context clones.
4867 2018-04-11 Richard Biener <rguenther@suse.de>
4870 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4871 from early DWARF output.
4872 (dwarf2out_early_finish): Output line info unconditionally into
4873 early DWARF and add reference to it.
4875 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4878 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4879 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4880 other than V2DFmode using iptr mode attribute.
4881 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4883 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4885 PR rtl-optimization/84659
4886 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4888 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4891 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4893 (output_loc_list): Pass address of a dummy size variable even in the
4894 locview handling loop.
4895 (index_location_lists): Add comment on why skip_loc_list_entry can't
4898 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4901 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4904 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4907 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4908 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4909 and -mstring-compare-inline-limit.
4911 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4914 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4915 for stack clash protection in a register whenever we need it to be in
4918 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4920 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4921 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4923 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4926 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4928 (mlong-double-): Ditto.
4929 * config/rs6000/sysv4.opt (msdata=): Ditto.
4930 (mtls-size=): Ditto.
4932 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4934 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4935 erroneous entries for
4936 "vector int vec_ldl (int, long int *)", and
4937 "vector unsigned int vec_ldl (int, unsigned long int *)".
4938 Add comments and entries for
4939 "vector bool char vec_ldl (int, bool char *)",
4940 "vector bool short vec_ldl (int, bool short *)",
4941 "vector bool int vec_ldl (int, bool int *)",
4942 "vector bool long long vec_ldl (int, bool long long *)",
4943 "vector pixel vec_ldl (int, pixel *)",
4944 "vector long long vec_ldl (int, long long *)",
4945 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4946 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4947 type tree bool_long_long_type_node and correct definition of
4948 bool_V2DI_type_node to make reference to this new type tree.
4949 (rs6000_mangle_type): Replace erroneous reference to
4950 bool_long_type_node with bool_long_long_type_node.
4951 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4952 comments to emphasize sign distinctions for char and int types and
4953 replace RS6000_BTI_bool_long constant with
4954 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4955 use of RS6000_BTI_pixel.
4956 (bool_long_type_node): Remove this macro definition.
4957 (bool_long_long_type_node): New macro definition
4959 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4961 PR rtl-optimization/85300
4962 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4963 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4964 simplify_unary_operation fails.
4966 2018-04-10 Martin Liska <mliska@suse.cz>
4968 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4969 cgraph_edge and ipa_ref.
4971 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4975 * config/i386/sse.md
4976 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4977 computation of the VEC_MERGE selector from mask.
4978 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4979 Fix decoding of the VEC_MERGE selector into mask.
4981 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4983 PR tree-optimization/85286
4984 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4986 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4988 * final.c (final_1): Set insn_last_address as well as
4989 insn_current_address.
4991 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4994 * explow.c (emit_stack_probe): Call validize_mem on memory location
4995 before passing it to gen_probe_stack. Create address operand and
4996 legitimize it for the probe_stack_address case.
4998 2018-04-09 Jan Hubicka <jh@suse.cz>
5001 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5002 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5003 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5005 (free_lang_data): Rebuild type inheritance graph.
5007 2018-04-09 Martin Sebor <msebor@redhat.com>
5009 * invoke.texi (-finline-small-functions): Mention other optimization
5011 (-findirect-inlining, -fpartial-inlining): Same.
5012 (-finline-functions-called-once): Same.
5013 (-freorder-blocks-and-partition): Same.
5015 2018-04-09 Jan Hubicka <jh@suse.cz>
5018 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5019 jumps; choose last target that matches the criteria (i.e.
5020 no partition changes for non-crossing jumps).
5021 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5022 support for redirecting crossing jumps to non-crossing.
5024 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5026 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5027 also for naked functions.
5029 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5031 * config/arc/arc.md (add_shift): New pattern.
5032 (add_shift2): Likewise.
5033 (sub_shift): Likewise.
5034 (sub_shift_cmp0_noout): Likewise.
5035 (compare_si_ashiftsi): Likewise.
5036 (xbfu_cmp0_noout): New combine pattern.
5037 (xbfu_cmp0"): Likewise.
5038 (movsi_set_cc_insn): Place the predicable variant first.
5039 (commutative_binary_cmp0_noout): Remove clobber.
5040 (commutative_binary_cmp0): New pattern.
5041 (noncommutative_binary_cmp0): Likewise.
5042 (noncommutative_binary_cmp0_noout): Likewise.
5043 (noncommutative_binary_comparison_result_used): Removed.
5044 (rsub_cmp0): New pattern.
5045 (rsub_cmp0_noout): Likewise.
5046 (extzvsi): Changed, keep only meaningful variants.
5047 (SQH, SEZ): New iterators.
5048 (SQH_postfix): New mode attribute.
5049 (SEZ_prefix): New code attribute.
5050 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5051 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5052 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5054 (noncommutative_operator): Check the availability of barrel
5057 2018-04-09 Richard Biener <rguenther@suse.de>
5059 PR tree-optimization/85284
5060 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5061 Only use the niter constraining form of simple_iv when the exit
5064 2018-04-09 Tom de Vries <tom@codesourcery.com>
5067 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5068 (define_expand "*memory_barrier"): New define_expand.
5069 (define_insn "memory_barrier"): New insn.
5071 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5073 PR rtl-optimization/80463
5074 PR rtl-optimization/83972
5075 PR rtl-optimization/83480
5077 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5078 correct producer for the insn.
5079 (tidy_control_flow): Fixup seqnos in case of debug insns.
5081 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5083 PR rtl-optimization/83913
5085 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5086 different sched-times when merging exprs.
5088 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5090 PR rtl-optimization/83962
5092 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5093 tidy_fallthru_edge and tidy_control_flow.
5095 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5097 PR rtl-optimization/83530
5099 * sel-sched.c (force_next_insn): New global variable.
5100 (remove_insn_for_debug): When force_next_insn is true, also leave only
5101 next insn in the ready list.
5102 (sel_sched_region): When the region wasn't scheduled, make another pass
5103 over it with force_next_insn set to 1.
5105 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5107 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5109 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5110 for interrupt control.
5111 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5112 functions for interrupt control.
5113 * config/nds32/nds32-intrinsic.md: Likewise.
5114 * config/nds32/nds32_intrinsic.h: Likewise.
5115 * config/nds32/nds32.h (nds32_builtins): Likewise.
5117 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5119 * config/nds32/nds32.c (nds32_init_machine_status,
5120 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5121 strict_aligned_p field.
5122 (nds32_expand_to_rtl_hook): New function.
5123 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5124 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5126 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5127 Chung-Ju Wu <jasonwucj@gmail.com>
5129 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5130 * config/nds32/nds32-n7.md: New file.
5131 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5132 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5134 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5135 * config/nds32/nds32.md (pipeline_model): Add n7.
5136 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5137 * config/nds32/pipelines.md: Include n7 settings.
5139 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5140 Chung-Ju Wu <jasonwucj@gmail.com>
5142 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5143 * config/nds32/nds32-e8.md: New file.
5144 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5145 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5147 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5148 * config/nds32/nds32.md (pipeline_model): Add e8.
5149 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5150 * config/nds32/pipelines.md: Include e8 settings.
5152 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5153 Chung-Ju Wu <jasonwucj@gmail.com>
5155 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5156 * config/nds32/nds32-n8.md: New file.
5157 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5158 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5160 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5161 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5162 * config/nds32/nds32.md (pipeline_model): Add n8.
5163 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5164 * config/nds32/pipelines.md: Include n8 settings.
5166 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5167 Chung-Ju Wu <jasonwucj@gmail.com>
5169 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5170 * config/nds32/nds32-n9-2r1w.md: New file.
5171 * config/nds32/nds32-n9-3r2w.md: New file.
5172 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5173 nds32_register_ports): New or modify for cpu n9.
5174 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5176 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5177 * config/nds32/nds32-utils.c: New file.
5178 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5179 TARGET_MUL_SLOW): Define.
5180 * config/nds32/nds32.md (pipeline_model): New attribute.
5181 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5182 New options that support cpu n9.
5183 * config/nds32/pipelines.md: Include n9 settings.
5184 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5186 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5188 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5189 information if necessary.
5190 (output_cond_branch_compare_zero): Likewise.
5191 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5192 (nds32_target_alignment): Refine for alignment.
5193 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5194 (FUNCTION_BOUNDARY): Modify.
5195 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5197 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5199 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5201 * config/nds32/constants.md (unspec_volatile_element): Add values for
5202 TLB operation and data prefetch.
5203 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5204 functions for TLB operation and data prefetch.
5205 * config/nds32/nds32-intrinsic.md: Likewise.
5206 * config/nds32/nds32_intrinsic.h: Likewise.
5207 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5208 (nds32_print_operand): Likewise.
5209 * config/nds32/nds32.h (nds32_builtins): Likewise.
5211 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5212 Andrew Pinski <pinsika@gcc.gnu.org>
5215 * match.pd: Use constant_boolean_node of correct type instead of
5216 boolean_true_node or boolean_false_node for simplifying
5217 pointer comparisons to zero.
5219 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5221 PR tree-optimization/80021
5222 * tree.c (verify_type_variant): Make error call in verify_variant_match
5223 translatable and remove final full stop.
5225 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5227 * config/nds32/constants.md (unspec_volatile_element): Add
5228 UNSPEC_VOLATILE_EH_RETURN.
5229 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5230 nds32_output_stack_pop): Support dwarf exception handling process.
5231 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5232 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5233 exception handling process.
5234 (nds32_compute_stack_frame): Likewise.
5235 (nds32_return_addr_rtx): Likewise.
5236 (nds32_initial_elimination_offset): Likewise.
5237 (nds32_expand_prologue): Likewise.
5238 (nds32_expand_epilogue): Likewise.
5239 (nds32_dynamic_chain_address): New function.
5240 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5242 (DYNAMIC_CHAIN_ADDRESS): Define.
5243 (EH_RETURN_DATA_REGNO): Define.
5244 (EH_RETURN_STACKADJ_RTX): Define.
5245 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5246 patterns for dwarf exception handling.
5248 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5250 * config/nds32/nds32.h: Clean up obsolete macros.
5252 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5254 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5255 Add enum values for particular instructions.
5256 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5257 particular intrinsic functions.
5258 * config/nds32/nds32-intrinsic.md: Likewise.
5259 * config/nds32/nds32_intrinsic.h: Likewise.
5260 * config/nds32/nds32.h (nds32_builtins): Likewise.
5261 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5262 (btst, ave): New patterns for particular instructions.
5264 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5266 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5267 Add enum values for atomic load/store and memory sync.
5268 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5270 * config/nds32/nds32-intrinsic.md: Likewise.
5271 * config/nds32/nds32_intrinsic.h: Likewise.
5272 * config/nds32/nds32.h (nds32_builtins): Likewise.
5274 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5276 PR tree-optimization/85257
5277 * fold-const.c (native_encode_vector): If not all elts could fit
5278 and off is -1, return 0 rather than offset.
5279 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5280 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5281 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5282 adjust buffer in native_interpret_expr call.
5284 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5286 * config/nds32/constants.md (unspec_volatile_element): Add cache
5287 control enum values.
5288 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5289 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5290 * config/nds32/nds32.c (nds32_cctl_names): New.
5291 (nds32_print_operand): Handle cache control register names.
5292 * config/nds32/nds32.h (nds32_builtins): New enum values.
5293 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5295 * config/nds32/nds32.md (type): Add mmu.
5296 * config/nds32/pipelines.md (simple_insn): Add mmu.
5298 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5300 * config/nds32/nds32.md (type): Remove call.
5301 * config/nds32/pipelines.md (simple_insn): Likewise.
5303 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5305 * config/nds32/constants.md (unspec_volatile_element): Add
5306 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5307 UNSPEC_VOLATILE_FMFCFG.
5308 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5309 description for fmfcfg and fmfcsr.
5310 (bdesc_1arg): Add fmtcsr.
5311 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5312 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5313 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5314 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5315 unspec_fmfcfg): New patterns.
5316 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5317 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5318 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5319 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5320 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5321 __nds32__fmfcfg): Define.
5323 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5325 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5326 intrinsic register names.
5327 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5328 intrinsic register enum values and macros.
5330 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5332 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5333 for load/store addressing form.
5334 (nds32_print_operand_address): Likewise.
5336 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5339 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5340 based on LABEL_REF. Remove useless assertion.
5341 (pic_address_needs_scratch): Fix formatting.
5342 (sparc_legitimize_pic_address): Minor tweaks.
5343 (sparc_delegitimize_address): Adjust assertion accordingly.
5344 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5345 into symbolic_operand.
5346 (movsi_high_pic_label_ref): Likewise.
5347 (movsi_lo_sum_pic_label_ref): Likewise.
5348 (movdi_pic_label_ref): Likewise.
5349 (movdi_high_pic_label_ref): Likewise.
5350 (movdi_lo_sum_pic_label_ref): Likewise.
5352 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5354 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5355 custom LIB_SPEC setup.
5357 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5358 Kito Cheng <kito.cheng@gmail.com>
5360 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5361 * config/riscv/freebsd.h: New.
5363 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5365 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5366 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5369 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5370 Kito Cheng <kito.cheng@gmail.com>
5372 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5373 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5374 * config/nds32/nds32-protos.h (nds32_output_call,
5375 nds32_output_return): Declare.
5376 * config/nds32/nds32.md: Refine all the call and return patterns.
5378 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5381 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5382 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5384 PR rtl-optimization/84872
5385 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5386 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5389 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5391 * expr.c (copy_blkmode_to_reg): Revert 254862.
5392 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5394 2018-04-06 Richard Biener <rguenther@suse.de>
5397 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5398 after seeing a component reference with an adjacent field. Treat
5399 refs to arrays at struct end of external decls similar to
5400 refs to unconstrained commons.
5402 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5405 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5406 look through SAVE_EXPRs with non-side-effects argument. Adjust
5408 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5411 2018-04-06 Richard Biener <rguenther@suse.de>
5414 * alias.c (find_base_term): New wrapper around find_base_term
5415 unwinding CSELIB_VAL_PTR changes.
5416 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5419 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5421 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5423 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5424 constant definitions.
5425 ("nop"): lr 0,0 -> nopr r0
5426 ("nop_lr0", "nop_lr1"): New insn definitions.
5428 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5430 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5431 NDS32_V3PUSH_AVAILABLE_P macro.
5433 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5434 Chung-Ju Wu <jasonwucj@gmail.com>
5436 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5437 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5438 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5439 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5440 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5441 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5442 * config/nds32/constraints.md: New constraints and checking for hard
5443 float configuration.
5444 * config/nds32/iterators.md: New mode iterator and attribute for hard
5445 float configuration.
5446 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5448 * config/nds32/nds32-fpu.md: New file.
5449 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5450 deal with hard float code generation.
5451 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5453 (abi_type, float_reg_number): New enum type.
5454 * config/nds32/nds32-predicates.c: New predicates for hard float.
5455 * config/nds32/nds32-protos.h: Declare functions for hard float.
5456 * config/nds32/nds32.c: Implementation for hard float configuration.
5457 * config/nds32/nds32.h: Definitions for hard float configuration.
5458 * config/nds32/nds32.md: Include hard float machine description and
5459 modify patterns for hard float configuration.
5460 * config/nds32/nds32.opt: New options for hard float configuration.
5461 * config/nds32/predicates.md: New predicates for hard float
5464 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5466 * common/config/nds32/nds32-common.c
5467 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5469 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5472 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5473 CONSTRUCTOR_ELT (ctor, ...)->value.
5475 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5478 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5480 2018-04-05 Tom de Vries <tom@codesourcery.com>
5483 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5486 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5487 Kito Cheng <kito.cheng@gmail.com>
5489 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5490 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5491 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5492 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5494 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5495 Kito Cheng <kito.cheng@gmail.com>
5497 * config/nds32/constraints.md (Ufe): New memory constraint.
5498 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5499 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5500 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5502 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5503 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5505 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5507 * config/nds32/nds32.md: Use optimize_size in the condition for
5508 alu-shift instructions.
5510 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5512 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5514 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5516 * config/nds32/nds32.md (negsi2): Refine pattern.
5518 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5519 Chung-Ju Wu <jasonwucj@gmail.com>
5521 * config/nds32/iterators.md (shift_rotate): New code iterator.
5522 (shift): New code attribute.
5523 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5524 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5525 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5526 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5527 bit-wise operations.
5528 (andsi3, *andsi3): Ditto.
5529 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5530 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5531 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5532 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5533 nds32_ior_operand, nds32_xor_operand): New predicates.
5535 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5537 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5538 (addsi3, subsi3): ... this.
5540 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5542 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5544 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5546 * config/nds32/nds32.md: Adjust indention.
5548 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5550 * config/nds32/nds32.md (feature): New attribute.
5552 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5554 * config/nds32/nds32.md (subtype): New attribute.
5556 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5559 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5560 expansion to perform a bitwise AND of the argument followed by a
5561 boolean negation of the result.
5563 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5565 PR rtl-optimization/84878
5566 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5567 the basic block. Assert the use reference is not artificial and that
5568 it has an associated insn.
5570 2018-04-04 Michael Matz <matz@suse.de>
5572 * builtins.c (compute_objsize): Pass correct operand
5573 to array_at_struct_end_p.
5575 2018-04-04 Richard Biener <rguenther@suse.de>
5578 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5579 from contexts for DINFO_LEVEL_TERSE and below.
5581 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5583 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5584 resiter_operand condition.
5585 * config/nds32/nds32.md (*move<mode>): Ditto.
5587 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5588 Monk Chiang <sh.chiang04@gmail.com>
5590 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5592 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5594 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5596 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5597 Kito Cheng <kito.cheng@gmail.com>
5599 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5600 nds32_cond_code_str, output_cond_branch,
5601 output_cond_branch_compare_zero, nds32_expand_cbranch,
5602 nds32_expand_cstore, nds32_expand_movcc,
5603 nds32_output_cbranchsi4_equality_zero,
5604 nds32_output_cbranchsi4_equality_reg,
5605 nds32_output_cbranchsi4_equality_reg_or_const_int,
5606 nds32_output_cbranchsi4_greater_less_zero: New functions.
5607 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5608 nds32_expand_cstore, nds32_expand_movcc,
5609 nds32_output_cbranchsi4_equality_zero,
5610 nds32_output_cbranchsi4_equality_reg,
5611 nds32_output_cbranchsi4_equality_reg_or_const_int,
5612 nds32_output_cbranchsi4_greater_less_zero): Declare.
5613 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5614 nds32_rimm11s_operand): New predicates.
5615 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5616 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5619 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5621 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5622 * config/nds32/nds32.md: Ditto.
5623 * config/nds32/pipelines.md: Ditto.
5625 2018-04-04 Richard Biener <rguenther@suse.de>
5627 PR tree-optimization/85168
5628 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5629 propagating abnormals.
5631 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5633 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5635 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5636 Kito Cheng <kito.cheng@gmail.com>
5638 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5639 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5640 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5641 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5642 * config/nds32/nds32.md (sibcall_internal): New.
5643 (sibcall_register): Remove.
5644 (sibcall_immediate): Remove.
5645 (sibcall_value_internal): New.
5646 (sibcall_value_register): Remove.
5647 (sibcall_value_immediate): Remove.
5648 * config/nds32/predicates.md (nds32_general_register_operand): New.
5649 (nds32_call_address_operand): New.
5651 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5653 PR rtl-optimization/85167
5654 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5655 bb_defs if *split_p, instead preinitialize it to NULL.
5657 PR tree-optimization/85156
5658 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5659 evaluating the argument multiple times.
5661 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5663 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5665 (_mm_cvtpd_ps): Likewise.
5666 (_mm_cvttpd_epi32): Likewise.
5667 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5668 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5669 vector, pixel, and bool following altivec.h include.
5671 2018-04-03 Martin Sebor <msebor@redhat.com>
5673 * doc/extend.texi (Common Function Attributes): Clarify.
5674 (const attribute): Likewise.
5675 (pure attribute): Likewise.
5677 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5680 * config/i386/i386.c (ix86_expand_vector_set): Use
5681 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5683 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5685 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5686 instructions when changing rounding bits to preserve precision bits
5687 in the x87 control word.
5689 2018-04-03 Martin Liska <mliska@suse.cz>
5691 PR tree-optimization/82491
5692 * rtl.h (strip_offset_and_add): Replace += suboffset with
5693 poly_uint64 () + suboffset.
5695 2018-03-29 Martin Liska <mliska@suse.cz>
5696 Martin Jambor <mjambor@suse.cz>
5699 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5700 param_type is not an integral or pointer type.
5702 2018-04-03 Richard Biener <rguenther@suse.de>
5704 * sese.h (recompute_all_dominators): Remove.
5706 2018-04-02 Martin Sebor <msebor@redhat.com>
5708 * doc/invoke.texi (-Wrestrict): Fix typos.
5710 2018-04-02 Jim Wilson <jimw@sifive.com>
5712 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5713 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5714 (<optab>di3, <optab>si3_extend): Likewise.
5715 (<optab>si3_mask, <optab>si3_mask_1): New.
5716 (<optab>di3_mask, <optab>di3_mask_1): New.
5717 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5718 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5719 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5721 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5723 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5726 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5728 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5729 (nds32_canonicalize_comparison): New function.
5731 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5732 Kito Cheng <kito.cheng@gmail.com>
5733 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5735 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5736 * config/nds32/constants.md (unspec_volatile_element): Add
5737 UNSPEC_VOLATILE_RELAX_GROUP.
5738 * config/nds32/nds32-relax-opt.c: New file.
5739 * config/nds32/nds32-predicates.c
5740 (nds32_symbol_load_store_p): New function.
5741 * config/nds32/nds32-protos.h
5742 (nds32_symbol_load_store_p): Declare function.
5743 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5744 * config/nds32/nds32.c
5745 (nds32_register_pass): New function to register pass.
5746 (nds32_register_passes): New function to register passes.
5747 * config/nds32/nds32.md (relax_group): New pattern.
5748 * config/nds32/nds32.opt (mrelax-hint): New option.
5749 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5751 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5753 * config/nds32/t-nds32: Modify files dependency.
5755 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5757 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5758 (PROFILE_HOOK): Define its implementation.
5760 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5762 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5763 type and 32-bit size.
5765 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5768 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5769 (V_128_256): New mode iterator.
5770 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5771 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5772 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5774 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5775 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5777 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5780 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5781 NaN inputs correctly.
5783 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5786 * config/rs6000/vsx.md (??r): New mode attribute.
5787 (*vsx_mov<mode>_64bit): Use it.
5788 (*vsx_mov<mode>_32bit): Likewise.
5790 2018-03-30 Martin Sebor <msebor@redhat.com>
5792 PR tree-optimization/84818
5793 * builtins.c (check_access): Use warning_n.
5795 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5798 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5800 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5803 2018-03-30 Julia Koval <julia.koval@intel.com>
5806 * x86-tune.def (movx, partial_reg_dependency): Enable for
5809 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5812 * lra-constraints.c (process_alt_operands): Move setting
5813 this_alternative_matches below.
5815 2018-03-29 Martin Liska <mliska@suse.cz>
5818 * doc/invoke.texi: Document how LTO works with debug info.
5819 Describe auto-load support of binutils. Mention 'x86-64'
5820 as valid option value of -march option.
5822 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5824 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5827 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5828 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5829 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5832 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5835 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5836 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5837 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5838 for RS6000_BTM_POWERPC64.
5839 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5840 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5841 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5846 2018-03-28 Carl Love <cel@us.ibm.com>
5849 2017-09-27 Carl Love <cel@us.ibm.com>
5851 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5852 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5853 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5856 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5858 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5859 instead of __vector bool.
5860 (_mm_max_pu8): Likewise.
5861 (_mm_min_pi16): Likewise.
5863 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5866 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5867 (DIVWEUO): Likewise.
5869 (DIVDEUO): Likewise.
5870 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5871 DIVWEUO and DIVDEUO.
5872 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5873 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5874 (div_extend): Likewise.
5875 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5877 (__builtin_divweuo): Likewise.
5878 (__builtin_divdeo): Likewise.
5879 (__builtin_divdeuo): Likewise.
5881 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5884 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5885 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5887 PR tree-optimization/82004
5888 * gimple-match-head.c (optimize_pow_to_exp): New function.
5889 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5890 Don't fold to exp if optimize_pow_to_exp is false.
5892 2018-03-28 Martin Liska <mliska@suse.cz>
5895 * calls.c (initialize_argument_information): Fix trailing space.
5896 * common.opt: Fix typo and provide better explanation for
5897 -fsanitize-coverage option.
5898 * config/i386/i386.opt: Fix typo.
5900 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5901 Martin Liska <mliska@suse.cz>
5904 * gimplify.c (asan_poison_variable): Don't do the check for
5905 gimplify_omp_ctxp here.
5906 (gimplify_decl_expr): Do it here.
5907 (gimplify_target_expr): Likewise.
5909 2018-03-28 Martin Liska <mliska@suse.cz>
5912 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5913 chkp_type_bounds_count if MPX is not enabled.
5915 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5917 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5919 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5922 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5923 function to create the function decl for complex long double
5924 multiply and divide for -mabi=ieeelongdouble.
5925 (init_float128_ieee): Call it.
5927 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5930 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5931 -fcf-protection=branch -mibt.
5932 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5934 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5937 * config/arm/arm.c (arm_valid_symbolic_address): Handle
5938 arm_word_relocations.
5940 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5943 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5944 extern array declarations.
5946 2018-03-27 Richard Biener <rguenther@suse.de>
5949 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5950 explicit single_use checks.
5952 2018-03-27 Richard Biener <rguenther@suse.de>
5954 PR tree-optimization/85082
5955 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5958 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5960 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5961 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5962 Turn on fasynchronous-unwind-tables and funwind-tables.
5964 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5967 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5968 (*bmi_blsr_<mode>_ccz): Ditto.
5970 2018-03-26 Tom de Vries <tom@codesourcery.com>
5972 PR tree-optimization/85063
5973 * omp-general.c (offloading_function_p): New function. Factor out
5975 * omp-offload.c (pass_omp_target_link::gate): ... here.
5976 * omp-general.h (offloading_function_p): Declare.
5977 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5978 with attribute omp declare target for offloading functions.
5980 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5982 PR tree-optimization/84005
5983 * tree-data-ref.h (get_base_for_alignment): Declare.
5984 * tree-data-ref.c (get_base_for_alignment_1): New function.
5985 (get_base_for_alignment): Likewise.
5986 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5987 get_base_for_alignment to find a suitable base object, instead
5988 of always using drb->base_address.
5990 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5993 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5994 known size by default.
5996 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5999 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6000 and non BLKmode operands.
6002 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6005 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6006 Clean up attributes.
6008 2018-03-23 Richard Biener <rguenther@suse.de>
6011 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6012 we are going to emit early debug for LTO.
6014 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6017 * function.c (match_asm_constraints_1): Don't optimize if input
6018 doesn't satisfy general_operand predicate for output's mode.
6021 * alias.c (write_dependence_p): Don't require for x_canonicalized
6022 non-VOIDmode if x has VOIDmode.
6025 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6026 just don't try to optimize it rather than assert it never happens.
6028 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6030 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6031 macro expansions for definition of ST_INTERNAL_<mode> and
6032 LD_INTERNAL_<mode> builtins.
6033 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6035 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6037 (altivec_expand_st_builtin): Likewise.
6038 (altivec_expand_builtin): Remove calls to deleted functions.
6039 (rs6000_address_for_altivec): Delete this function.
6040 * config/rs6000/vector.md: Remove expands for
6041 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6043 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6046 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6047 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6048 re-computing once computed.
6049 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6050 (arm_init_machine_status): Initialize
6051 machine->static_chain_stack_bytes.
6053 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6056 * doc/extend.texi: Add four new prototypes for vec_ld.
6057 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6058 definitions for more logical presentation.
6059 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6060 entries for V1TI variants of __builtin_altivec_ld builtin.
6061 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6062 handling of V1TI variant of LVX icode pattern.
6063 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6064 (rs6000_gimple_fold_builtin): Likewise.
6065 (altivec_init_builtins): Add code to define
6066 __builtin_altivec_lvx_v1ti function.
6068 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6071 * function.c (match_asm_constraints_1): Don't do the optimization
6072 if input isn't a REG, SUBREG, MEM or constant.
6074 2018-03-22 Tom de Vries <tom@codesourcery.com>
6076 PR tree-optimization/84956
6077 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6078 bb_has_abnormal_pred.
6080 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6083 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6084 DECL_INITIAL (decl) to decl at the end.
6085 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6088 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6090 * doc/extend.texi (__builtin_tgmath): Document when complex
6091 integer types are treated as _Complex _Float64.
6093 2018-03-21 Tom de Vries <tom@codesourcery.com>
6095 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6097 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6099 PR tree-optimization/84960
6100 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6101 if it is ENTRY block, move them into single succ of ENTRY in that case.
6103 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6105 PR tree-optimization/84811
6106 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6107 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6108 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6109 (known_subrange_p): Update accordingly. Cast each value involved
6110 in the size comparison, rather than casting the result of the
6113 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6115 PR tree-optimization/84982
6116 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6117 by flipping the least significant bit rather than all bits from
6118 bitpos to bitpos + bitsize - 1.
6120 2018-03-21 Nathan Sidwell <nathan@acm.org>
6122 * doc/extend.texi (Deprecated Features): Remove mention of
6123 long-deleted deprecations.
6125 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6128 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6129 * configure: Regenerate.
6131 2018-03-21 Tom de Vries <tom@codesourcery.com>
6133 PR tree-optimization/83126
6134 * tree-parloops.c (num_phis): New function.
6135 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6137 2018-03-21 Nathan Sidwell <nathan@acm.org>
6139 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6140 mention anon-struct/union members and trailing attributes.
6142 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6144 PR tree-optimization/84969
6145 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6146 builtin memset partitions if they set different rhs values.
6148 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6150 PR rtl-optimization/84989
6151 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6152 VEC_DUPLICATE with scalar result mode.
6154 2018-03-21 Martin Liska <mliska@suse.cz>
6157 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6158 not intended return statement.
6160 2018-03-21 Martin Liska <mliska@suse.cz>
6163 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6164 (chkp_find_bound_slots_1): Limit number of iterations.
6166 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6169 * Minor grammar fixes for x86 options.
6171 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6174 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6175 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6177 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6180 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6181 (altivec_lvx_<mode>_1op): Likewise.
6182 (altivec_stvx_<mode>_2op): Likewise.
6183 (altivec_stvx_<mode>_1op): Likewise.
6184 (altivec_lvx_<VM2:mode>): New define_expand.
6185 (altivec_stvx_<VM2:mode>): Likewise.
6186 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6187 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6188 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6189 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6190 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6191 (rs6000_gen_lvx): Likewise.
6192 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6193 (altivec_expand_stv_builtin): Likewise.
6194 (altivec_expand_builtin): Likewise.
6195 * config/rs6000/vector.md: Likewise.
6197 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6200 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6203 2018-03-20 Richard Biener <rguenther@suse.de>
6206 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6207 sign-conversions as zero, fall back to standard scalar_stmt
6210 2018-03-20 Martin Liska <mliska@suse.cz>
6213 * predict.c (rebuild_frequencies): Handle case when we have
6214 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6216 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6219 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6220 flag_section_anchors.
6221 * varasm.c (use_blocks_for_decl_p): Remove hack for
6222 dw2_force_const_mem.
6225 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6227 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6228 be created, use lowpart_subreg of operands[0] rather than operands[0]
6230 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6231 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6232 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6233 and n constraint instead of aarch64_shift_imm_di and Usd.
6234 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6235 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6237 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6240 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6241 to favor GPR over NEON registers.
6242 (<shift>di3_neon): Likewise.
6244 2018-03-20 Tom de Vries <tom@codesourcery.com>
6247 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6248 (nvptx_process_pars): Emit bar.sync asap and alap.
6250 2018-03-20 Tom de Vries <tom@codesourcery.com>
6253 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6254 seen_label if seen_label is already set.
6256 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6259 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6260 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6261 Use 1U instead of 1. Formatting fixes.
6264 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6265 instead of TREE_TYPE (s1) for the return value.
6267 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6269 PR tree-optimization/84946
6270 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6271 bitsize + bitsize in poly_uint64 rather than poly_int64.
6274 * dwarf2asm.c: Include fold-const.c.
6275 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6276 of decl rather than decl itself.
6278 PR rtl-optimization/84643
6279 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6281 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6284 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6285 calling assemble_variable.
6287 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6290 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6291 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6293 2018-03-19 Jim Wilson <jimw@sifive.com>
6296 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6297 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6298 (riscv_first_stack_step): Likewise.
6299 (riscv_option_override): Use STACK_BOUNDARY instead of
6301 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6303 (BIGGEST_ALIGNMENT): Set to 128.
6304 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6305 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6308 2018-03-19 Richard Biener <rguenther@suse.de>
6310 PR tree-optimization/84933
6311 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6312 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6314 2018-03-19 Richard Biener <rguenther@suse.de>
6316 PR tree-optimization/84859
6317 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6318 (cond_if_else_store_replacement): Perform sinking operation on
6319 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6320 Generalize what a BB with a single eligible store is.
6322 2018-03-19 Richard Biener <rguenther@suse.de>
6324 PR tree-optimization/84929
6325 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6326 chrec_is_positive against non-chrec arg.
6328 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6331 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6333 2018-03-18 Martin Liska <mliska@suse.cz>
6335 PR rtl-optimization/84635
6336 * regrename.c (build_def_use): Use matches_mode only when
6339 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6341 PR tree-optimization/84913
6342 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6343 vectorize chains of COND_EXPRs.
6345 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6347 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6349 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6351 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6353 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6355 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6357 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6358 Kito Cheng <kito.cheng@gmail.com>
6360 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6361 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6362 (nds32_adjust_reg_alloc_order): New function.
6363 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6365 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6367 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6368 nds32_print_operand, nds32_print_operand_address): Use
6369 HOST_WIDE_INT_PRINT_DEC instead.
6371 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6373 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6375 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6378 * config/i386/i386.c (initial_ix86_tune_features,
6379 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6381 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6382 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6383 rather than 1u << ix86_tune. Formatting fix.
6384 (ix86_option_override_internal): Change ix86_arch_mask from
6385 unsigned int to unsigned HOST_WIDE_INT, initialize to
6386 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6387 (ix86_function_specific_restore): Likewise.
6389 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6392 * postreload.c (reload_combine_recognize_pattern): Perform
6393 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6394 truncate_int_for_mode the result for the destination's mode.
6397 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6398 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6400 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6403 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6406 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6407 regno_allocno_class_array and sorted_pseudos.
6408 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6409 insns where regno is used.
6411 2018-03-16 Martin Liska <mliska@suse.cz>
6414 * multiple_target.c (create_dispatcher_calls): Redirect
6415 reference in the symbol table.
6417 2018-03-16 Martin Liska <mliska@suse.cz>
6420 * multiple_target.c (create_dispatcher_calls): Redirect also
6423 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6427 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6428 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6429 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6431 2018-03-16 Julia Koval <julia.koval@intel.com>
6433 * doc/invoke.texi (Skylake Server): Add CLWB.
6434 Cannonlake): Remove CLWB.
6436 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6438 PR tree-optimization/84841
6439 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6441 (FLOAT_ONE_CONST_TYPE): Define.
6442 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6443 (sort_by_operand_rank): Put entries with higher constant_type last
6444 rather than first to match comments.
6446 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6448 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6451 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6454 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6455 check and use address of target always.
6457 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6460 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6461 (indirect_thunk_bnd_needed): Likewise.
6462 (indirect_thunks_used): Likewise.
6463 (indirect_thunks_bnd_used): Likewise.
6464 (indirect_return_needed): New.
6465 (indirect_return_bnd_needed): Likewise.
6466 (output_indirect_thunk_function): Add a bool argument for
6468 (output_indirect_thunk_function): Don't generate alias for
6469 function return thunk.
6470 (ix86_code_end): Call output_indirect_thunk_function to generate
6471 function return thunks.
6472 (ix86_output_function_return): Set indirect_return_bnd_needed
6473 and indirect_return_needed instead of indirect_thunk_bnd_needed
6474 and indirect_thunk_needed.
6476 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6478 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6479 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6480 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6482 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6483 Paul Hua <paul.hua.gm@gmail.com>
6486 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6488 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6490 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6491 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6494 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6497 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6498 instead of GET_MODE_SIZE when comparing Units.
6500 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6503 * varasm.c (hash_section): Return an unchangeble hash value
6504 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6505 Return !aarch64_can_use_per_function_literal_pools_p ().
6507 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6510 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6511 rather than address of cmode as last argument to prepare_cmp_insn.
6513 2018-03-15 Julia Koval <julia.koval@intel.com>
6515 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6516 F_AVX512VNNI, F_AVX512BITALG): New.
6518 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6521 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6522 insn for floating-point loads and stores.
6524 2018-03-14 Carl Love <cel@us.ibm.com>
6526 * config/rs6000/rs6000-c.c: Add macro definitions for
6527 ALTIVEC_BUILTIN_VEC_PERMXOR.
6528 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6529 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6530 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6531 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6533 * config/doc/extend.texi: Add prototypes for vec_permxor.
6535 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6538 * diagnostic-show-locus.c (class layout_point): Convert m_line
6539 from int to linenum_type.
6540 (line_span::comparator): Use linenum "compare" function when
6541 comparing line numbers.
6542 (test_line_span): New function.
6543 (layout_range::contains_point): Convert param "row" from int to
6545 (layout_range::intersects_line_p): Likewise.
6546 (layout::will_show_line_p): Likewise.
6547 (layout::print_source_line): Likewise.
6548 (layout::should_print_annotation_line_p): Likewise.
6549 (layout::print_annotation_line): Likewise.
6550 (layout::print_leading_fixits): Likewise.
6551 (layout::annotation_line_showed_range_p): Likewise.
6552 (struct line_corrections): Likewise for field m_row.
6553 (line_corrections::line_corrections): Likewise for param "row".
6554 (layout::print_trailing_fixits): Likewise.
6555 (layout::get_state_at_point): Likewise.
6556 (layout::get_x_bound_for_row): Likewise.
6557 (layout::print_line): Likewise.
6558 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6559 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6560 * input.c (selftest::test_linenum_comparisons): New function.
6561 (selftest::input_c_tests): Call it.
6562 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6563 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6564 * selftest.h (ASSERT_GT): New macro.
6565 (ASSERT_GT_AT): New macro.
6566 (ASSERT_LT): New macro.
6567 (ASSERT_LT_AT): New macro.
6569 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6571 PR rtl-optimization/84780
6572 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6574 2018-03-14 Martin Liska <mliska@suse.cz>
6576 * tree.c (record_node_allocation_statistics): Use
6577 get_stats_node_kind.
6578 (get_stats_node_kind): New function extracted from
6579 record_node_allocation_statistics.
6580 (free_node): Use get_stats_node_kind.
6582 2018-03-14 Richard Biener <rguenther@suse.de>
6584 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6585 that the value-set of ANTIC_IN doesn't grow.
6588 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6590 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6591 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6593 2018-03-14 Julia Koval <julia.koval@intel.com>
6595 * config.gcc (icelake-client, icelake-server): New.
6597 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6598 (initial_ix86_arch_features): Ditto.
6599 (PTA_SKYLAKE): Add SGX.
6600 (PTA_ICELAKE): Remove.
6601 (PTA_ICELAKE_CLIENT): New.
6602 (PTA_ICELAKE_SERVER): New.
6603 (ix86_option_override_internal): Split up icelake on icelake client and
6605 (get_builtin_code_for_version): Ditto.
6606 (fold_builtin_cpu): Ditto.
6607 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6608 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6609 * config/i386/i386.h (processor_type): Ditto.
6610 * doc/invoke.texi: Ditto.
6612 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6615 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6616 INTEGER_CST offset, add it together with bitpos / 8 and
6617 sign extend based on POINTER_SIZE.
6621 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6624 * config/i386/constraints.md (Yc): New register constraint.
6625 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6626 Use Yc constraint for alternative 2 of operand 0. Remove
6627 preferred_for_speed attribute.
6629 2018-03-14 Richard Biener <rguenther@suse.de>
6631 PR tree-optimization/84830
6632 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6633 with the old one to avoid oscillations.
6635 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6638 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6640 (assign_by_spills): Return a flag of reload assignment failure.
6641 Do not process the reload assignment failures. Do not spill other
6642 reload pseudos if they has the same reg class. Update n if
6644 (lra_assign): Add a return arg. Set up from the result of
6645 assign_by_spills call.
6646 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6647 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6648 usage_insns if it is not NULL.
6649 (spill_hard_reg_in_range): New function.
6650 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6651 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6652 function prototypes.
6653 (lra_assign): Change prototype.
6654 * lra.c (lra): Add code to deal with fails by splitting hard reg
6657 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6659 * config/riscv/riscv.opt (mrelax): New option.
6660 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6661 "norelax" when riscv_mrelax is disabled.
6662 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6664 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6667 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6668 reassociation for int modes.
6670 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6672 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6673 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6675 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6676 * config/aarch64/aarch64-sve.md
6677 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6678 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6679 (*extend<mode><Vwide>2): Rename to...
6680 (aarch64_sve_extend<mode><Vwide>2): ...this.
6681 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6682 renaming the old pattern to...
6683 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6685 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6686 define_expand, renaming the old pattern to...
6687 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6688 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6689 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6690 account when deciding which SVE instruction the optab should use.
6691 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6693 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6695 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6696 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6697 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6698 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6699 (tlsdesc_small_<mode>): Turn a define_expand and use
6700 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6701 (tlsdesc_small_advsimd_<mode>): ...this.
6702 (tlsdesc_small_sve_<mode>): New pattern.
6704 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6706 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6707 (UNSPEC_UMUL_HIGHPART): New constants.
6708 (MUL_HIGHPART): New int iteraor.
6709 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6710 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6712 (*<su>mul<mode>3_highpart): New define_insn.
6714 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6717 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6720 2018-03-13 Martin Liska <mliska@suse.cz>
6723 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6725 (sem_item_optimizer::~sem_item_optimizer): Release it.
6726 (sem_item_optimizer::merge_classes): Register variable aliases.
6727 (sem_item_optimizer::fixup_pt_set): New function.
6728 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6729 * ipa-icf.h: Declare new variables and functions.
6731 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6734 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6735 integer_pow2p@2 and test integer_pow2p in condition.
6736 (A < 0 ? C : 0): Similarly for @1.
6739 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6740 characters starting at p contain '\0' character, don't look beyond
6744 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6745 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6748 * reg-stack.c (change_stack): Change update_end var from int to
6749 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6750 also call set_block_for_insn on the newly added insns and rescan.
6753 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6754 on the last operand.
6757 * tree.c (stabilize_reference_1): Return save_expr (e) for
6758 STATEMENT_LIST even if it doesn't have side-effects.
6760 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6762 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6764 2018-03-12 Renlin Li <renlin.li@arm.com>
6766 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6767 aarch64_output_scalar_simd_mov_immediate.
6769 2018-03-12 Martin Sebor <msebor@redhat.com>
6771 PR tree-optimization/83456
6772 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6773 for perfectly overlapping calls to memcpy.
6774 (gimple_fold_builtin_memory_chk): Same.
6775 (gimple_fold_builtin_strcpy): Handle no-warning.
6776 (gimple_fold_builtin_stxcpy_chk): Same.
6777 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6779 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6781 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6782 parameter. Use it for SFmode.
6783 (rs6000_function_arg_advance_1): Adjust.
6784 (rs6000_function_arg): Adjust.
6785 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6787 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6789 PR rtl-optimization/84169
6790 PR rtl-optimization/84780
6791 * combine.c (can_combine_p): Check for a 2-insn combination whether
6792 the destination register is used between the two insns, too.
6794 2018-03-12 Richard Biener <rguenther@suse.de>
6796 PR tree-optimization/84803
6797 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6798 for refs DR analysis didn't process.
6800 2018-03-12 Richard Biener <rguenther@suse.de>
6802 PR tree-optimization/84777
6803 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6804 force-vectorize loops ignore whether we are optimizing for size.
6806 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6808 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6809 (TARGET_MD_ASM_ADJUST): Define.
6811 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6812 Kito Cheng <kito.cheng@gmail.com>
6813 Chung-Ju Wu <jasonwucj@gmail.com>
6815 * config/nds32/nds32.c (nds32_compute_stack_frame,
6816 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6817 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6818 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6819 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6820 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6821 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6822 * config/nds32/nds32.md (prologue, epilogue): Use macro
6823 NDS32_V3PUSH_AVAILABLE_P to do checking.
6825 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6828 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6829 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6830 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6831 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6832 addition of most attributes on !orig_type_die or the attribute not
6833 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6835 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6836 Chung-Ju Wu <jasonwucj@gmail.com>
6838 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6840 * config/nds32/nds32.opt (mvh): New option.
6842 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6843 Chung-Ju Wu <jasonwucj@gmail.com>
6845 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6847 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6848 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6851 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6852 Chung-Ju Wu <jasonwucj@gmail.com>
6854 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6856 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6857 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6859 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6860 Kito Cheng <kito.cheng@gmail.com>
6861 Chung-Ju Wu <jasonwucj@gmail.com>
6863 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6864 UNSPEC_FFMISM and UNSPEC_FLMISM.
6865 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6866 for ffb, ffmism and flmism.
6867 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6868 (unspec_ffmism): Ditto.
6869 (unspec_flmism): Ditto.
6870 (nds32_expand_builtin_impl): Check if string extension is available.
6871 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6872 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6874 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6877 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6880 * lra-assigns.c (assign_by_spills): Return a flag of reload
6881 assignment failure. Do not process the reload assignment
6882 failures. Do not spill other reload pseudos if they has the same
6884 (lra_assign): Add a return arg. Set up from the result of
6885 assign_by_spills call.
6886 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6887 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6888 usage_insns if it is not NULL.
6889 (spill_hard_reg_in_range): New function.
6890 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6891 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6892 function prototypes.
6893 (lra_assign): Change prototype.
6894 * lra.c (lra): Add code to deal with fails by splitting hard reg
6897 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6900 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6902 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6905 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6906 (dw_val_node): Add val_symbolic_view.
6907 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6908 (symview_upper_bound): New.
6909 (new_line_info_table): Initialize symviews_since_reset.
6910 (dwarf2out_source_line): Count symviews_since_reset and set
6911 symview_upper_bound.
6912 (dw_val_equal_p): Handle symview.
6913 (add_AT_symview): New.
6914 (print_dw_val): Handle symview.
6915 (attr_checksum, attr_checksum_ordered): Likewise.
6916 (same_dw_val_p, size_of_die): Likewise.
6917 (value_format, output_die): Likewise.
6918 (add_high_low_attributes): Use add_AT_symview for entry_view.
6919 (dwarf2out_finish): Reset symview_upper_bound, clear
6922 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6925 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6926 Add strict argument and use it.
6927 (rs6000_split_multireg_move): Update for new strict argument.
6928 (mem_operand_gpr): Disallow all non-offsettable addresses.
6929 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6931 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6934 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6935 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6936 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6939 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6940 decl, use remap_type if we want to use the type.
6942 2018-03-09 Martin Sebor <msebor@redhat.com>
6944 PR tree-optimization/84526
6945 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6947 (builtin_access::generic_overlap): Be prepared to handle non-array
6950 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6952 PR rtl-optimization/84682
6953 * lra-constraints.c (process_address_1): Check is_address flag
6954 for address constraints.
6955 (process_alt_operands): Likewise.
6956 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6957 preprocess_constraints.
6958 * recog.h (preprocess_constraints): Add oploc parameter.
6960 * recog.c (preprocess_constraints): Test address_operand for
6961 CT_ADDRESS constraints.
6963 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6966 * lra-assigns.c (assign_by_spills): Return a flag of reload
6967 assignment failure. Do not process the reload assignment
6968 failures. Do not spill other reload pseudos if they has the same
6970 (lra_assign): Add a return arg. Set up from the result of
6971 assign_by_spills call.
6972 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6973 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6974 usage_insns if it is not NULL.
6975 (spill_hard_reg_in_range): New function.
6976 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6977 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6978 function prototypes.
6979 (lra_assign): Change prototype.
6980 * lra.c (lra): Add code to deal with fails by splitting hard reg
6983 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6986 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6987 Accept complain bool parameter. Only emit errors if it is true.
6988 (arm_parse_cpu_option_name): Likewise.
6989 (arm_target_thumb_only): Adjust callers of the above.
6990 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6991 prototype to take a default true bool parameter.
6992 (arm_parse_arch_option_name): Likewise.
6994 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6995 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6999 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7000 * configure: Regenerate.
7001 * configure.ac ("linker --version-script option"): New.
7002 ("linker soname option"): New.
7004 2018-03-09 Richard Biener <rguenther@suse.de>
7006 PR tree-optimization/84775
7007 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7008 immediate uses of predicate stmts and mark them modified.
7011 PR tree-optimization/84178
7012 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7014 (version_loop_for_if_conversion): Delay update_ssa call.
7015 (tree_if_conversion): Delay update_ssa until after predicate
7018 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7021 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7022 when the function accesses prior frames.
7024 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7027 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7028 gen_llsym, otherwise call maybe_gen_llsym.
7031 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7032 has ',' character inside of it.
7034 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7037 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7038 as clobbering CC_REGNUM.
7040 2018-03-08 Richard Biener <rguenther@suse.de>
7043 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7044 (follow_copies_to_constant): Do not follow SSA names registered
7047 2018-03-08 Richard Biener <rguenther@suse.de>
7049 PR tree-optimization/84178
7050 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7052 (version_loop_for_if_conversion): Delay update_ssa call.
7053 (tree_if_conversion): Delay update_ssa until after predicate
7056 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7058 PR tree-optimization/84178
7059 * tree-if-conv.c (release_bb_predicate): Remove the
7060 the assertion that the stmts have NULL use_ops.
7061 Discard the statements, asserting that they haven't
7062 yet been added to a BB.
7064 2018-03-08 Richard Biener <rguenther@suse.de>
7066 PR tree-optimization/84746
7067 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7068 (phi_translate): Pass in destination ANTIC_OUT set.
7069 (phi_translate_1): Likewise. For a simplified result lookup
7070 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7071 (phi_translate_set): Adjust.
7072 (do_pre_regular_insertion): Likewise.
7073 (do_pre_partial_partial_insertion): Likewise.
7075 2018-03-08 Martin Liska <mliska@suse.cz>
7077 PR gcov-profile/84735
7078 * doc/gcov.texi: Document usage of profile files.
7079 * gcov-io.h: Document changes in the format.
7081 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7085 * dwarf2out.c (struct dw_line_info_table): Update comments for
7087 (FORCE_RESET_NEXT_VIEW): New.
7088 (FORCE_RESETTING_VIEW_P): New.
7089 (RESETTING_VIEW_P): Check for -1 too.
7090 (ZERO_VIEW_P): Likewise.
7091 (new_line_info_table): Force-reset next view.
7092 (dwarf2out_begin_function): Likewise.
7093 (dwarf2out_source_line): Simplify zero_view_p initialization.
7094 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7095 view directly. Omit view when omitting .loc at line 0.
7097 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7099 PR tree-optimization/84740
7100 * tree-switch-conversion.c (process_switch): Call build_constructors
7101 only if info.phi_count is non-zero.
7103 PR tree-optimization/84739
7104 * tree-tailcall.c (find_tail_calls): Check call arguments against
7105 DECL_ARGUMENTS (current_function_decl) rather than
7106 DECL_ARGUMENTS (func) when checking for tail recursion.
7108 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7110 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7111 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7112 Volker Reichelt's entry and add entries for people that perform
7113 GCC fuzzy testing and report numerous bugs.
7115 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7118 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7119 readonly data in sdata, if that is disabled.
7120 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7121 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7122 -mreadonly-in-sdata option.
7124 2018-03-07 Martin Sebor <msebor@redhat.com>
7126 PR tree-optimization/84468
7127 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7128 basic block when looking for nul assignment.
7130 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7133 * except.h (output_function_exception_table): Adjust prototype.
7134 * except.c (output_function_exception_table): Remove FNNAME parameter
7135 and add SECTION parameter. Ouput one part of the table at a time.
7136 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7137 the first part of the exception table and emit unwind directives.
7138 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7139 (i386_pe_seh_cold_init): Likewise.
7140 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7141 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7142 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7143 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7144 * config/i386/winnt.c: Include except.h.
7145 (struct seh_frame_state): Add reg_offset, after_prologue and
7146 in_cold_section fields.
7147 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7148 (i386_pe_seh_cold_init): New function.
7149 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7150 to seh->in_cold_section.
7151 (seh_emit_push): Record the offset of the push.
7152 (seh_emit_save): Record the offet of the save.
7153 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7154 Test seh->after_prologue to disregard the epilogue.
7155 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7156 (i386_pe_end_cold_function): New function.
7158 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7161 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7162 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7165 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7167 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7168 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7171 * multiple_target.c: Include tree-inline.h and intl.h.
7172 (expand_target_clones): Diagnose and fail if node->definition and
7173 !tree_versionable_function_p (node->decl).
7175 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7177 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7179 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7180 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7181 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7183 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7186 * combine.c (try_combine): Use reg_or_subregno instead of handling
7187 just paradoxical SUBREGs and REGs.
7189 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7191 * config/arc/arc.c (arc_finalize_pic): Remove function.
7192 (arc_must_save_register): We use single base PIC register, remove
7193 checks to save/restore the PIC register.
7194 (arc_expand_prologue): Likewise.
7195 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7197 (arc_verify_short): Likewise.
7198 (arc_attr_type): Likewise.
7199 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7200 (walk_stores): Likewise.
7201 (arc_address_cost): Make it static.
7202 (arc_verify_short): Likewise.
7203 (branch_dest): Likewise.
7204 (arc_attr_type): Likewise.
7205 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7206 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7207 (arc_final_prescan_insn): Remove inserting the nops due to
7208 hardware hazards. It is done in reorg step.
7209 (insn_length_variant_t): Remove.
7210 (insn_length_parameters_t): Likewise.
7211 (arc_insn_length_parameters): Likewise.
7212 (arc_get_insn_variants): Likewise.
7213 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7215 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7218 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7221 PR tree-optimization/84687
7222 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7224 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7226 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7228 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7229 Rename to ppc_speculation_barrier.
7230 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7231 __builtin_ppc_speculation_barrier.
7233 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7236 * combine.c (combine_simplify_rtx): Don't try to simplify if
7237 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7240 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7242 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7243 to 32 bytes when compiling for POWER9.
7245 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7248 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7249 regparm >= 3 with no arg reg available also for calls with
7250 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7253 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7255 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7257 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7260 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7262 2018-03-05 Richard Biener <rguenther@suse.de>
7264 PR tree-optimization/84486
7265 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7266 When inserting a __builtin_assume_aligned call set the LHS
7267 SSA name alignment info accordingly.
7269 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7271 PR tree-optimization/84114
7272 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7273 Avoid reassociation of FLOAT_MODE addition.
7275 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7277 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7278 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7279 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7280 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7281 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7282 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7283 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7285 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7286 __builtin_ia32_wbinvd): New builtins.
7287 (SPECIAL_ARGS2): New.
7288 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7289 (SPECIAL_ARGS2): New.
7290 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7291 (ix86_valid_target_attribute_inner_p): Ditto.
7292 (ix86_init_mmx_sse_builtins): Add special_args2.
7293 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7294 TARGET_WBNOINVD_P): New.
7295 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7296 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7297 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7298 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7299 * config/i386/pconfigintrin.h: New file.
7300 * config/i386/wbnoinvdintrin.h: Ditto.
7301 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7303 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7305 2018-03-05 Richard Biener <rguenther@suse.de>
7307 PR tree-optimization/84670
7308 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7310 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7311 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7312 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7313 doesn't grow if all successors (recursively) were visited at least
7316 2018-03-05 Richard Biener <rguenther@suse.de>
7318 PR tree-optimization/84650
7319 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7320 if executed in the loop pipeline.
7322 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7324 * doc/configfiles.texi (Configuration Files): Move info about
7325 conditionalizing $target-protos.h to...
7326 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7327 differs from $target-protos.h.
7329 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7330 Chung-Ju Wu <jasonwucj@gmail.com>
7332 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7333 * config/nds32/nds32-multiple.md (setmemsi): Define.
7334 * config/nds32/nds32-memory-manipulation.c
7335 (nds32_gen_dup_4_byte_to_word_value): New.
7336 (emit_setmem_word_loop): New.
7337 (emit_setmem_byte_loop): New.
7338 (nds32_expand_setmem_loop): New.
7339 (nds32_expand_setmem_loop_v3m): New.
7340 (nds32_expand_setmem_unroll): New.
7341 (nds32_expand_setmem): New.
7343 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7344 Chung-Ju Wu <jasonwucj@gmail.com>
7346 * config/nds32/nds32-memory-manipulation.c
7347 (nds32_emit_load_store): New.
7348 (nds32_emit_post_inc_load_store): New.
7349 (nds32_emit_mem_move): New.
7350 (nds32_emit_mem_move_block): New.
7351 (nds32_expand_movmemsi_loop_unknown_size): New.
7352 (nds32_expand_movmemsi_loop_known_size): New.
7353 (nds32_expand_movmemsi_loop): New.
7354 (nds32_expand_movmemsi_unroll): New.
7355 (nds32_expand_movmemqi): Rename ...
7356 (nds32_expand_movmemsi): ... to this.
7357 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7358 (movmemsi): ... to this.
7359 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7360 (nds32_expand_movmemsi): ... to this.
7362 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7363 Monk Chiang <sh.chiang04@gmail.com>
7364 Chung-Ju Wu <jasonwucj@gmail.com>
7366 * config/nds32/nds32-protos.h
7367 (nds32_expand_load_multiple): New arguments.
7368 (nds32_expand_store_multiple): Ditto.
7369 (nds32_valid_multiple_load_store): Rename ...
7370 (nds32_valid_multiple_load_store_p): ... to this.
7371 * config/nds32/nds32-memory-manipulation.c
7372 (nds32_expand_load_multiple): Refine implementation.
7373 (nds32_expand_store_multiple): Ditto.
7374 * config/nds32/nds32-multiple.md
7375 (load_multiple): Update nds32_expand_load_multiple interface.
7376 (store_multiple): Update nds32_expand_store_multiple interface.
7377 * config/nds32/nds32-predicates.c
7378 (nds32_valid_multiple_load_store): Rename ...
7379 (nds32_valid_multiple_load_store_p): ... to this and refine
7381 * config/nds32/predicates.md
7382 (nds32_load_multiple_and_update_address_operation): New predicate.
7383 (nds32_store_multiple_and_update_address_operation): New predicate.
7385 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7386 Chung-Ju Wu <jasonwucj@gmail.com>
7388 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7389 (combo): New attribute.
7390 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7392 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7394 * config/nds32/nds32.opt: Change -mcmodel= default value.
7396 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7397 Monk Chiang <sh.chiang04@gmail.com>
7398 Chung-Ju Wu <jasonwucj@gmail.com>
7400 * config/nds32/constants.md (unspec_element): New enum.
7401 * config/nds32/constraints.md (Umw): New constraint.
7402 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7403 * config/nds32/nds32-intrinsic.md: Likewise.
7404 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7405 (nds32_valid_smw_lwm_base_p): New.
7406 (nds32_output_smw_single_word): New.
7407 (nds32_output_lmw_single_word): New.
7408 (nds32_expand_unaligned_load): New.
7409 (nds32_expand_unaligned_store): New.
7410 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7411 (nds32_output_smw_single_word): Declare.
7412 (nds32_output_lmw_single_word): Declare.
7413 (nds32_expand_unaligned_load): Declare.
7414 (nds32_expand_unaligned_store): Declare.
7415 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7416 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7417 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7418 NDS32_BUILTIN_UASTORE_DW.
7419 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7422 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7423 Kito Cheng <kito.cheng@gmail.com>
7424 Chung-Ju Wu <jasonwucj@gmail.com>
7426 * config/nds32/nds32-intrinsic.c
7427 (nds32_expand_builtin_null_ftype_reg): Delete.
7428 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7429 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7430 (nds32_read_argument): New.
7431 (nds32_legitimize_target): Ditto.
7432 (nds32_legitimize_argument): Ditto.
7433 (nds32_check_constant_argument): Ditto.
7434 (nds32_expand_unop_builtin): Ditto.
7435 (nds32_expand_unopimm_builtin): Ditto.
7436 (nds32_expand_binop_builtin): Ditto.
7437 (nds32_builtin_decl_impl): Ditto.
7438 (builtin_description): Ditto.
7439 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7440 (nds32_init_builtins_impl): Ditto.
7441 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7442 (nds32_builtin_decl): New.
7443 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7444 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7446 2018-03-02 Jeff Law <law@redhat.com>
7448 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7449 (redundant_insn, fill_simple_delay_slots): Likewise.
7450 (fill_slots_from_thread): Likewise.
7451 * resource.c (mark_referenced_resources): Likewise.
7452 (mark_set_resources, find_dead_or_set_registers): Likewise.
7454 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7456 * substring-locations.h (format_warning_va): Formatting fix for
7458 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7460 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7461 * substring-locations.c: Include intl.h.
7462 (format_warning_va): Turned into small wrapper around
7463 format_warning_n_va, renamed to ...
7464 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7465 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7467 (format_warning_at_substring_n): New function.
7468 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7469 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7470 format_warning_at_substring with just a shorter name instead of
7471 const function pointer.
7472 (fmtwarn_n): New function.
7473 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7474 appropriate, get rid of all the fmtstr temporaries, move conditionals
7475 with G_() wrapped string literals directly into fmtwarn arguments,
7476 cast dir.len to (int), formatting fixes.
7478 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7480 * doc/invoke.texi: Remove "Cilk Plus" references.
7482 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7483 Richard Biener <rguenther@suse.de>
7486 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7487 for error or warning attributes if CALL_FROM_THUNK_P is set.
7490 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7493 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7494 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7497 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7498 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7500 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7501 instead of -1U in last predictors element's probability member.
7503 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7506 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7507 arguments if they are comparable.
7509 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7511 PR tree-optimization/84634
7512 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7513 masks and masked_loop_p with a single loop_masks, making sure it's
7514 null for bb vectorization.
7516 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7518 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7519 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7520 loop->force_vectorize to check whether there is no alias.
7522 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7525 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7527 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7529 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7530 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7531 instead of a loop around prev_real_insn.
7532 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7536 * config/i386/i386.c (ix86_print_operand): Use conditional
7537 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7540 2018-03-02 Richard Biener <rguenther@suse.de>
7542 PR tree-optimization/84427
7543 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7544 (bitmap_set_subtract_values): Rewrite to handle multiple
7547 (prune_clobbered_mems): Likewise.
7548 (phi_translate): Take edge instead of pred/phiblock.
7549 (phi_translate_1): Likewise.
7550 (phi_translate_set): Likewise. Insert all translated
7551 exprs for a value into the set, keeping possibly multiple
7552 expressions per value.
7553 (compute_antic_aux): Adjust for phi_translate changes.
7554 When intersecting union the expressions and prune those
7555 not in the final value set, keeping possibly multiple
7556 expressions per value. Do not use value-insertion
7557 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7558 all expressions. Add verification that the value-sets
7559 only shrink during iteration.
7560 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7561 (do_pre_regular_insertion): Likewise.
7562 (do_pre_partial_partial_insertion): Likewise.
7564 2018-03-02 Richard Biener <rguenther@suse.de>
7567 * config/darwin.c (saved_debug_info_level): New static global.
7568 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7569 (darwin_asm_lto_end): Restore debug info generation settings.
7571 2018-03-01 Martin Liska <mliska@suse.cz>
7574 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7577 2018-03-01 Richard Biener <rguenther@suse.de>
7580 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7581 type annotation with LTO when debug was disabled at compile-time.
7583 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7585 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7587 (mips_final_postscan_insn): Likewise.
7589 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7591 PR rtl-optimization/84528
7592 * alias.c (init_alias_target): Add commentary.
7593 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7594 a unique base value if the frame pointer is not eliminated
7595 to the stack pointer.
7597 2018-03-01 Tom de Vries <tom@codesourcery.com>
7599 PR rtl-optimization/83327
7600 * lra-int.h (hard_regs_spilled_into): Declare.
7601 * lra.c (hard_regs_spilled_into): Define.
7602 (init_reg_info): Init hard_regs_spilled_into.
7603 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7604 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7605 (process_bb_lives): Handle hard_regs_spilled_into.
7606 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7607 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7609 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7611 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7612 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7613 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7614 * config/rs6000/aix72.h: New file.
7616 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7618 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7619 instead of warning_at with conditional singular and plural messages
7623 * stor-layout.c (update_alignment_for_field): For
7624 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7625 && !DECL_PACKED (field), do the alignment update, just use
7626 only desired_align instead of MAX (type_align, desired_align)
7628 (place_field): Don't do known_align < desired_align handling
7629 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7630 is non-NULL, instead do it after rli->prev_field handling and
7631 only if not within a bitfield word. For DECL_PACKED (field)
7632 use type_align of BITS_PER_UNIT.
7634 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7636 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7637 superfluous parentheses and trailing spaces.
7639 2018-02-28 Richard Biener <rguenther@suse.de>
7641 PR tree-optimization/84584
7642 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7643 SCoPs with fake exit edge.
7645 2018-02-28 Martin Liska <mliska@suse.cz>
7648 * timevar.c (timer::print): Fix format to properly print 100%
7651 2018-02-28 Richard Biener <rguenther@suse.de>
7654 * genmatch.c (capture_info::walk_match): Do not mark
7655 captured expressions without operands as expr_p given
7656 they act more like predicates and should be subject to
7657 "lost tail" side-effect preserving.
7659 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7661 PR rtl-optimization/81611
7662 * auto-inc-dec.c (attempt_change): Move dead note from
7663 mem_insn if it's the next use of regno
7664 (find_address): Take address use of reg holding
7665 non-incremented value. Add parm to limit search to the named
7667 (merge_in_block): Attempt to use a mem insn that is the next
7668 use of the original regno.
7670 2018-02-27 Martin Sebor <msebor@redhat.com>
7673 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7674 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7676 2018-02-27 Martin Sebor <msebor@redhat.com>
7678 PR translation/84207
7679 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7680 n argument to unsigned HOST_WIDE_INT.
7681 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7682 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7683 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7684 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7686 2018-02-27 Richard Biener <rguenther@suse.de>
7688 PR tree-optimization/84512
7689 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7690 Do not use the estimate returned from record_stmt_cost for
7691 the scalar iteration cost but sum properly using add_stmt_cost.
7693 2018-02-27 Richard Biener <rguenther@suse.de>
7695 PR tree-optimization/84466
7696 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7697 Adjust last change to less strictly validate use operands.
7699 2018-02-27 Martin Liska <mliska@suse.cz>
7701 PR gcov-profile/84548
7702 * gcov.c (process_file): Allow partial overlap and consider it
7703 also as group functions.
7704 (output_lines): Properly calculate range of lines for a group.
7706 2018-02-27 Martin Liska <mliska@suse.cz>
7708 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7709 'ggc' suffixes. Change first column width.
7710 (timer::print): Fix formatting of the column.
7712 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7714 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7715 preserve inline entry blocks for the sake of debug inline
7716 entry point markers alone.
7717 (remove_unused_locals): Suggest in comments a better place to
7718 force the preservation of inline entry blocks that are
7719 otherwise unused, but do not preserve them.
7721 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7723 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7725 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7728 * config/i386/constraints.md (Bs): Replace
7729 ix86_indirect_branch_register with
7730 TARGET_INDIRECT_BRANCH_REGISTER.
7732 * config/i386/i386.md (indirect_jump): Likewise.
7733 (tablejump): Likewise.
7734 (*sibcall_memory): Likewise.
7735 (*sibcall_value_memory): Likewise.
7736 Peepholes of indirect call and jump via memory: Likewise.
7737 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7738 (*sibcall_value_GOT_32): Likewise.
7739 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7740 (GOT_memory_operand): Likewise.
7741 (call_insn_operand): Likewise.
7742 (sibcall_insn_operand): Likewise.
7743 (GOT32_symbol_operand): Likewise.
7744 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7746 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7748 PR rtl-optimization/83496
7749 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7750 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7751 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7752 redundant insn, if any.
7753 (relax_delay_slots): Likewise.
7754 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7756 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7758 PR tree-optimization/83965
7759 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7760 that grouped statements are part of a reduction chain. Return
7761 true if the statement is not marked as a reduction itself but
7763 (vect_recog_dot_prod_pattern): Don't check whether the statement
7764 is part of a group here.
7765 (vect_recog_sad_pattern): Likewise.
7766 (vect_recog_widen_sum_pattern): Likewise.
7768 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7771 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7773 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7776 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7778 (ix86_output_indirect_function_return): New prototype.
7779 (ix86_split_simple_return_pop_internal): Likewise.
7780 * config/i386/i386.c (indirect_return_via_cx): New.
7781 (indirect_return_via_cx_bnd): Likewise.
7782 (indirect_thunk_name): Handle return va CX_REG.
7783 (output_indirect_thunk_function): Create alias for
7784 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7785 (ix86_output_indirect_jmp): Remove the bool argument.
7786 (ix86_output_indirect_function_return): New function.
7787 (ix86_split_simple_return_pop_internal): Likewise.
7788 * config/i386/i386.md (*indirect_jump): Don't pass false
7789 to ix86_output_indirect_jmp.
7790 (*tablejump_1): Likewise.
7791 (simple_return_pop_internal): Change it to define_insn_and_split.
7792 Call ix86_split_simple_return_pop_internal to split it for
7794 (simple_return_indirect_internal): Call
7795 ix86_output_indirect_function_return instead of
7796 ix86_output_indirect_jmp.
7798 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7801 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7802 memset and value initialization afterwards.
7804 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7806 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7808 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7811 * common/config/aarch64/aarch64-common.c
7812 (aarch_option_optimization_table[]): Switch
7813 off fomit-frame-pointer
7815 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7816 Chung-Ju Wu <jasonwucj@gmail.com>
7818 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7820 (store_multiple): Ditto.
7822 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7824 * config.gcc: Add --with-cpu support for nds32 target.
7825 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7826 * config/nds32/nds32.opt: Add -mcpu= option.
7828 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7830 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7831 isel=yes): Warn for these deprecated options.
7833 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7835 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7836 ISA_2_5_MASKS_EMBEDDED.
7838 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7840 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7841 p->max as pointers rather than using iterative_hash_expr.
7843 2018-02-23 Carl Love <cel@us.ibm.com>
7845 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7846 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7848 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7849 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7850 P8V_BUILTIN_VEC_VUNSIGNED2.
7852 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7855 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7856 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7858 * lra-constraints.c (curr_insn_transform): Set up
7859 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7861 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7862 * lra-eliminations.c (spill_pseudos): Ditto.
7863 (process_insn_for_elimination): Ditto.
7864 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7865 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7866 LRA_NON_CLOBBERED_ALT.
7868 2018-02-22 Martin Sebor <msebor@redhat.com>
7870 PR tree-optimization/84480
7871 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7872 to maybe_diag_stxncpy_trunc. Call it.
7873 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7874 from gimple_fold_builtin_strcpy. Print inlining stack.
7875 (handle_builtin_stxncpy): Print inlining stack.
7876 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7878 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7881 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7882 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7883 and -fcheck-pointer-bounds are used together.
7884 (indirect_thunk_prefix): New enum.
7885 (indirect_thunk_need_prefix): New function.
7886 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7887 "_nt" instead of "_bnd" for NOTRACK prefix.
7888 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7889 (output_indirect_thunk_function): Likewise.
7891 (ix86_code_end): Update output_indirect_thunk_function calls.
7892 (ix86_output_indirect_branch_via_reg): Replace
7893 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7894 (ix86_output_indirect_branch_via_push): Likewise.
7895 (ix86_output_function_return): Likewise.
7896 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7897 incompatible with -fcf-protection=branch and
7898 -fcheck-pointer-bounds.
7900 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7903 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7904 Change gcc_assert call to output_operand_lossage.
7906 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7908 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7910 2018-02-22 DJ Delorie <dj@redhat.com>
7911 Sebastian Perta <sebastian.perta@renesas.com>
7912 Oleg Endo <olegendo@gcc.gnu.org>
7914 * config/rx/rx.c (rx_rtx_costs): New function.
7915 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7917 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7919 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7921 2018-02-22 Martin Liska <mliska@suse.cz>
7924 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7925 Add "native" as a possible value.
7927 2018-02-22 Martin Liska <mliska@suse.cz>
7930 * config/i386/i386.c (ix86_option_override_internal):
7931 Add "native" as a possible value for -march and -mtune.
7933 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7936 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7937 to all type variants.
7939 PR tree-optimization/84503
7940 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7941 width as info->bitpos + info->bitsize - start.
7942 (merged_store_group::merge_overlapping): Simplify width computation.
7943 (check_no_overlap): New function.
7944 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7945 start + width and last_order of the group, fail if check_no_overlap
7947 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7948 to group if check_no_overlap fails.
7950 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7952 * config/rs6000/altivec.md: Delete contraint arguments to
7953 define_expand, define_split, and define_peephole2, and in
7954 define_insn_and_split if always unused.
7955 * config/rs6000/darwin.md: Ditto.
7956 * config/rs6000/dfp.md: Ditto.
7957 * config/rs6000/rs6000.md: Ditto.
7958 * config/rs6000/sync.md: Ditto.
7959 * config/rs6000/vector.md: Ditto.
7960 * config/rs6000/vsx.md: Ditto.
7962 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7964 * config/rs6000/altivec.md: Write output control strings as braced
7965 blocks instead of double-quoted strings.
7966 * config/rs6000/darwin.md: Ditto.
7967 * config/rs6000/rs6000.md: Ditto.
7968 * config/rs6000/vector.md: Ditto.
7969 * config/rs6000/vsx.md: Ditto.
7971 2018-02-21 Jason Merrill <jason@redhat.com>
7973 PR c++/84314 - ICE with templates and fastcall attribute.
7974 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7976 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7978 * ipa-cp.c (determine_versionability): Fix comment typos.
7980 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7983 * ipa-cp.c (determine_versionability): Do not version functions caling
7986 2018-02-21 Martin Liska <mliska@suse.cz>
7989 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7990 Add "native" as a possible value.
7991 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7992 the macro when native cpu detection is available.
7994 2018-02-21 Martin Liska <mliska@suse.cz>
7997 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7998 Add "native" as a possible value.
7999 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8000 when native cpu detection is available.
8002 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8003 Martin Sebor <msebor@redhat.com>
8005 PR tree-optimization/84478
8006 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8008 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8009 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8010 support which is conservatively correct, for 2 only stay conservative
8011 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8012 argument to the 2 argument get_range_strlen, adjust 6 arg
8013 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8015 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8016 (gimple_fold_builtin_strlen): Pass true as last argument to
8019 2018-02-20 Martin Sebor <msebor@redhat.com>
8022 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8023 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8024 (builtin_memref::builtin_memref): Factor out parts into
8025 set_base_and_offset and call it.
8027 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8030 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8031 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8032 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8033 search at the associated MODE_INT.
8035 2018-02-20 Jeff Law <law@redhat.com>
8038 PR tree-optimization/81592
8040 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8041 for range data rather than using global data.
8042 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8043 range data rather than using global data.
8044 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8045 pass it to children as needed.
8046 (struct directive::fmtresult): Similarly.
8047 (struct directive::set_width): Similarly.
8048 (struct directive::set_precision): Similarly.
8049 (format_integer, format_directive, parse_directive): Similarly.
8050 (format_none): Accept unnamed vr_values parameter.
8051 (format_percent, format_floating, format_character): Similarly.
8052 (format_string, format_plain): Similarly.
8053 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8054 the EVRP range analyzer for range data rather than using global data.
8055 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8056 gimple-ssa-evrp-analyze.h
8057 (class sprintf_dom_walker): Add after_dom_children member function.
8058 Add evrp_range_analyzer member.
8059 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8060 range analyzer as needed.
8061 (sprintf_dom_walker::after_dom_children): New member function.
8062 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8064 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8065 (evrp_range_analyzer::pop_to_marker): Likewise.
8067 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8069 PR tree-optimization/84419
8070 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8071 with the required type if its current type is compatible but
8074 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8077 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8078 after vectorization.
8080 2018-02-20 Martin Liska <mliska@suse.cz>
8083 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8084 possible values if we don't have a hint.
8086 2018-02-20 Martin Liska <mliska@suse.cz>
8090 * final.c (shorten_branches): Build align_tab array with one
8092 * opts.c (finish_options): Add alignment option limit check.
8093 (MAX_CODE_ALIGN): Likewise.
8094 (MAX_CODE_ALIGN_VALUE): Likewise.
8095 * doc/invoke.texi: Document maximum allowed option value for
8096 all -falign-* options.
8098 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8101 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8102 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8103 * var-tracking.c (emit_note_insn_var_location): Remove all references
8104 to NOTE_INSN_CALL_ARG_LOCATION.
8105 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8106 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8107 Use copy_rtx_if_shared.
8108 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8109 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8110 (dwarf2out_var_location): Remove handling of
8111 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8113 * final.c (final_scan_insn): Remove all references to
8114 NOTE_INSN_CALL_ARG_LOCATION.
8115 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8116 before dumping final insns.
8117 * except.c (emit_note_eh_region_end): Remove all references to
8118 NOTE_INSN_CALL_ARG_LOCATION.
8119 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8120 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8121 * config/arc/arc.c (hwloop_optimize): Likewise.
8122 * config/arm/arm.c (create_fix_barrier): Likewise.
8123 * config/s390/s390.c (s390_chunkify_start): Likewise.
8124 * config/sh/sh.c (find_barrier): Likewise.
8125 * config/i386/i386.c (rest_of_insert_endbranch,
8126 ix86_seh_fixup_eh_fallthru): Likewise.
8127 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8128 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8129 * config/frv/frv.c (frv_function_prologue): Likewise.
8130 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8132 (note_outside_basic_block_p): Remove all references to
8133 NOTE_INSN_CALL_ARG_LOCATION.
8134 * gengtype.c (adjust_field_rtx_def): Likewise.
8135 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8137 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8138 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8141 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8144 PR tree-optimization/84452
8145 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8146 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8149 2018-02-19 Martin Liska <mliska@suse.cz>
8152 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8154 2018-02-19 Martin Liska <mliska@suse.cz>
8155 Richard Sandiford <richard.sandiford@linaro.org>
8157 PR tree-optimization/82491
8158 * gimple-fold.c (get_base_constructor): Make earlier bail out
8161 2018-02-19 Carl Love <cel@us.ibm.com>
8163 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8164 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8166 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8167 P8V_BUILTIN_VEC_NEG.
8169 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8171 * config/rl78/rl78.md (movdf): New define expand.
8173 2018-02-19 Martin Liska <mliska@suse.cz>
8176 * doc/invoke.texi: Fix typo.
8177 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8179 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8181 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8182 handle rs6000_single_float and rs6000_double_float specially for
8185 2018-02-16 Jeff Law <law@redhat.com>
8187 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8190 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8193 * ipa-inline.c (inline_small_functions): Fix a typo.
8195 2018-02-16 Nathan Sidwell <nathan@acm.org>
8197 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8199 2018-02-16 Carl Love <cel@us.ibm.com>
8201 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8202 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8203 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8204 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8205 expansion to P8V_BUILTIN_VEC_FLOAT2.
8207 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8209 PR rtl-optimization/70023
8210 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8211 src_regno into account.
8213 2018-02-16 Carl Love <cel@us.ibm.com>
8215 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8216 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8217 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8218 * config/rs6000/rs6000.c: Remove case statements for
8219 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8220 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8221 and P9V_BUILTIN_VEC_VINSERT4B.
8222 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8223 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8224 * config/rs6000/vsx.md:
8225 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8228 2018-02-16 Carl Love <cel@us.ibm.com>
8230 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8232 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8234 * config/rs6000/rs6000-c.c: Add the definitions for
8235 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8236 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8237 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8238 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8239 definition for insert4b and define insn *insert3b_internal.
8240 * doc/extend.texi: Add documentation for vec_extract4b.
8242 2018-02-16 Nathan Sidwell <nathan@acm.org>
8244 * doc/extend.texi (Backwards Compatibility): Mention friend
8245 injection. Note for-scope is deprecated.
8246 * doc/invoke.texi (-ffriend-injection): Deprecate.
8248 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8250 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8251 that moved to I2, also allow destinations that are a paradoxical
8252 subreg (instead of a normal reg).
8254 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8257 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8260 2018-02-16 Richard Biener <rguenther@suse.de>
8262 PR tree-optimization/84037
8263 PR tree-optimization/84016
8265 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8266 Adjust vec_construct for the fact we need additional higher latency
8267 128bit inserts for AVX256 and AVX512 vector builds.
8268 (ix86_add_stmt_cost): Scale vector construction cost for
8271 2018-02-16 Richard Biener <rguenther@suse.de>
8273 PR tree-optimization/84417
8274 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8275 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8276 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8278 2018-02-16 Martin Liska <mliska@suse.cz>
8281 * internal-fn.def (ASAN_CHECK): Set proper flags.
8282 (ASAN_MARK): Likewise.
8284 2018-02-16 Julia Koval <julia.koval@intel.com>
8286 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8287 from PTA_CANNONLAKE.
8289 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8292 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8293 Use ++iter rather than iter++ for std::list iterators.
8294 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8295 defer deleting them until all nodes in the forest are processed. Do
8296 free even leaf nodes. Change to_process into auto_vec.
8299 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8300 * vec.h (vec_default_construct): Use memset instead of placement new
8301 if BROKEN_VALUE_INITIALIZATION is defined.
8302 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8303 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8306 PR rtl-optimization/83723
8307 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8308 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8309 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8311 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8313 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8315 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8317 PR rtl-optimization/81443
8318 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8319 from inner REGs to paradoxical SUBREGs.
8321 2018-02-16 Richard Biener <rguenther@suse.de>
8323 PR tree-optimization/84399
8324 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8325 For operands we can analyze at their definition make sure we can
8326 analyze them at each use as well.
8328 2018-02-16 Richard Biener <rguenther@suse.de>
8330 PR tree-optimization/84190
8331 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8332 volatile accesses if the decl isn't volatile.
8334 2018-02-15 Jason Merrill <jason@redhat.com>
8336 PR c++/84314 - ICE with templates and fastcall attribute.
8337 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8338 TYPE_CANONICAL on an existing type.
8340 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8342 PR tree-optimization/84383
8343 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8344 dstoff nor call operand_equal_p if dstbase is NULL.
8346 PR tree-optimization/84334
8347 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8348 also a CONSTANT_CLASS_P, punt.
8350 2018-02-14 Jim Wilson <jimw@sifive.com>
8352 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8353 first SMALL_OPERAND check. New local min_second_step. Move assert
8354 to where locals are set. Add TARGET_RVC support.
8355 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8357 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8359 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8361 2018-02-14 Martin Sebor <msebor@redhat.com>
8363 PR tree-optimization/83698
8364 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8365 arrays constrain the offset range to their bounds.
8366 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8367 (builtin_access::overlap): Avoid setting the size of overlap if it's
8369 (maybe_diag_overlap): Also consider arrays when deciding what values
8370 of offsets to include in diagnostics.
8372 2018-02-14 Martin Sebor <msebor@redhat.com>
8375 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8376 that correspond to the kind of a declaration.
8378 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8381 * config/pa/pa.md: Load address of PIC label using the linkage table
8382 if the label is nonlocal.
8384 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8386 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8387 warning message if user requests -maltivec=be.
8388 * doc/invoke.texi: Document deprecation of -maltivec=be.
8390 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8393 * config/rs6000/rs6000-c.c: Update definitions for
8394 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8395 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8397 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8400 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8401 add _get_ssp intrinsics. Remove argument from
8402 __builtin_ia32_rdssp[d|q].
8403 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8404 * config/i386/i386-builtin.def: Remove argument from
8405 __builtin_ia32_rdssp[d|q].
8406 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8407 ix86_expand_special_args_builtin for _rdssp[d|q].
8408 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8409 Clear register before usage.
8410 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8411 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8413 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8415 PR tree-optimization/84357
8416 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8417 operand 1 of an ARRAY_REF too.
8419 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8422 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8423 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8425 (set_of_reg): New struct.
8426 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8427 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8428 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8430 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8431 Split into bitclr, bitset, bitinvert patterns if appropriate.
8432 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8433 use rx_fuse_in_memory_bitop.
8434 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8435 to named insn, correct maximum insn length.
8437 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8440 * machmode.def: Define a complex mode for PARTIAL_INT.
8441 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8443 * doc/rtl.texi: Document CSPImode.
8444 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8446 (msp430_hard_regno_nregs_with_padding): Likewise.
8448 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8451 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8453 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8455 PR rtl-optimization/84169
8456 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8457 we generated a parallel as new i3 and we split that to new i2 and i3
8458 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8459 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8460 those to i2, not i1. Partially rewrite this scan code.
8462 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8465 * stor-layout.c (place_field): For variable length fields, adjust
8466 offset_align afterwards not just based on the field's alignment,
8467 but also on the size.
8470 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8471 of exps and logs in the use_exp2 case.
8473 2018-02-13 Jeff Law <law@redhat.com>
8475 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8478 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8481 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8485 * common.opt (gas-loc-support, gas-locview-support): New.
8486 (ginline-points, ginternal-reset-location-views): New.
8487 * doc/invoke.texi: Document them. Use @itemx where intended.
8488 (gvariable-location-views): Adjust.
8489 * target.def (reset_location_view): New.
8490 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8491 (TARGET_RESET_LOCATION_VIEW): New.
8492 * doc/tm.texi: Rebuilt.
8493 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8494 (dwarf2out_default_as_locview_support): New.
8495 (output_asm_line_debug_info): Use option variables.
8496 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8497 (output_loc_list): Likewise.
8498 (add_high_low_attributes): Check option variables.
8499 Don't output entry view attribute in strict mode.
8500 (gen_inlined_subroutine_die): Check option variables.
8501 (dwarf2out_inline_entry): Likewise.
8502 (init_sections_and_labels): Likewise.
8503 (dwarf2out_early_finish): Likewise.
8504 (maybe_reset_location_view): New, from...
8505 (dwarf2out_var_location): ... here. Call it.
8506 * debug.h (dwarf2out_default_as_loc_support): Declare.
8507 (dwarf2out_default_as_locview_support): Declare.
8508 * hooks.c (hook_int_rtx_insn_0): New.
8509 * hooks.h (hook_int_rtx_insn_0): Declare.
8510 * toplev.c (process_options): Take -gas-loc-support and
8511 -gas-locview-support from dwarf2out. Enable
8512 -gvariable-location-views by default only with locview
8513 assembler support. Enable -ginternal-reset-location-views by
8514 default only if the target defines the corresponding hook.
8515 Enable -ginline-points by default if location views are
8516 enabled; force it disabled if statement frontiers are
8518 * tree-inline.c (expand_call_inline): Check option variables.
8519 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8521 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8523 PR tree-optimization/84321
8524 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8525 handling. Also check whether the anti-range contains any values
8526 that satisfy the mask; switch to a VR_RANGE if not.
8528 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8531 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8533 2018-02-13 Martin Jambor <mjambor@suse.cz>
8536 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8537 of call statements, also set location of a load to a temporary.
8539 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8541 * config/rl78/rl78.c (add_vector_labels): New function.
8542 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8543 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8544 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8545 which checks that no arguments are passed.
8546 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8547 * doc/extend.texi: Documentation for the new attribute.
8549 2018-02-13 Andreas Schwab <schwab@suse.de>
8551 * config/riscv/linux.h (CPP_SPEC): Define.
8553 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8556 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8557 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8558 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8559 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8560 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8561 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8562 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8563 temporarily for AES and PCLMUL builtins.
8565 PR tree-optimization/84339
8566 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8567 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8571 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8572 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8573 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8575 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8577 * omp-simd-clone.h: New file.
8578 * omp-simd-clone.c: Include omp-simd-clone.h.
8579 (expand_simd_clones): No longer static.
8580 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8581 cgraph.h and omp-simd-clone.h.
8582 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8583 (vect_recog_widen_shift_pattern): Formatting fix.
8584 (vect_pattern_recog_1): Don't check optab for calls.
8587 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8588 operands[2] into a REG before using gen_lowpart on it.
8590 2018-02-12 Jeff Law <law@redhat.com>
8593 * config/sh/sh.c (find_barrier): Consider a sibling call
8596 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8597 successfully back substituting a reg.
8599 2018-02-12 Richard Biener <rguenther@suse.de>
8601 PR tree-optimization/84037
8602 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8603 parameter, move visited init to caller.
8604 (vect_slp_analyze_operations): Separate cost from validity
8605 check, initialize visited once for all instances.
8606 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8608 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8610 (vect_model_promotion_demotion_cost): Likewise.
8611 (vectorizable_bswap): Guard cost modeling with !slp_node
8612 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8614 (vectorizable_call): Likewise.
8615 (vectorizable_conversion): Likewise.
8616 (vectorizable_assignment): Likewise.
8617 (vectorizable_shift): Likewise.
8618 (vectorizable_operation): Likewise.
8619 (vectorizable_store): Likewise.
8620 (vectorizable_load): Likewise.
8621 (vectorizable_condition): Likewise.
8622 (vectorizable_comparison): Likewise.
8624 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8627 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8628 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8629 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8631 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8634 * params.def (inline-min-speedup): Increase from 8 to 15.
8635 (max-inline-insns-auto): Decrease from 40 to 30.
8636 * ipa-split.c (consider_split): Add some buffer for function to
8637 be considered inlining candidate.
8638 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8641 2018-02-12 Richard Biener <rguenther@suse.de>
8643 PR tree-optimization/84037
8644 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8645 matched stmts if we cannot swap the non-matched ones.
8647 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8649 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8650 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8651 _mm_maskz_scalef_round_ss): New intrinsics.
8652 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8653 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8654 __builtin_ia32_scalefss_round): Remove.
8655 (__builtin_ia32_scalefsd_mask_round,
8656 __builtin_ia32_scalefss_mask_round): New intrinsics.
8657 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8658 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8659 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8660 "<round_constraint>")): Changed to ...
8661 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8662 "<round_scalar_constraint>")): ... this.
8663 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8664 %0, %1, %2<round_op3>}"): Changed to ...
8665 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8666 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8667 %2<round_scalar_mask_op3>}"): ... this.
8668 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8670 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8672 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8673 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8674 (_mm_maskz_sqrt_round_ss): New intrinsics.
8675 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8676 (__builtin_ia32_sqrtsd_mask_round)
8677 (__builtin_ia32_sqrtss_mask_round): New builtins.
8678 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8679 (__builtin_ia32_sqrtss_round): Remove.
8680 (__builtin_ia32_sqrtsd_mask_round)
8681 (__builtin_ia32_sqrtss_mask_round): New builtins.
8682 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8683 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8684 ((match_operand:VF_128 1 "vector_operand"
8685 "xBm,<round_constraint>")): Changed to ...
8686 ((match_operand:VF_128 1 "vector_operand"
8687 "xBm,<round_scalar_constraint>")): ... this.
8688 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8689 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8690 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8691 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8692 %<iptr>1<round_scalar_mask_op3>}): ... this.
8693 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8694 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8696 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8699 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8700 Cast vec_cmpeq result to correct type.
8701 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8702 Cast vec_cmpgt result to correct type.
8704 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8706 * final.c (final_scan_insn_1): Renamed from...
8707 (final_scan_insn): ... this. New wrapper, to recover
8708 seen from the outermost call in recursive ones.
8709 * config/sparc/sparc.c (output_return): Drop seen from call.
8710 (output_sibcall): Likewise.
8711 * config/visium/visium.c (output_branch): Likewise.
8713 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8715 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8718 2018-02-10 Alan Modra <amodra@gmail.com>
8721 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8722 Specify LR as an input.
8724 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8727 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8728 remove_member_access_dummy_vars): New functions.
8729 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8730 lower_omp_1, execute_lower_omp): Use them.
8732 PR rtl-optimization/84308
8733 * shrink-wrap.c (spread_components): Release todo vector.
8735 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8737 PR rtl-optimization/57193
8738 * ira-color.c (struct allocno_color_data): Add member
8739 conflict_allocno_hard_prefs.
8740 (update_conflict_allocno_hard_prefs): New.
8741 (bucket_allocno_compare_func): Add a preference based on
8742 conflict_allocno_hard_prefs.
8743 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8744 (color_allocnos): Remove a dead code. Initiate
8745 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8747 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8750 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8751 constraint from =wa to wa. Avoid a subreg on the output operand,
8752 instead use a pseudo and subreg it in a move.
8753 (p9_xxbrd_<mode>): Changed to ...
8754 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8755 (p9_xxbrd_v2df): New expander.
8756 (p9_xxbrw_<mode>): Changed to ...
8757 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8758 (p9_xxbrw_v4sf): New expander.
8760 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8762 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8764 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8767 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8768 multiply in 32-bit mode.
8769 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8770 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8773 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8775 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8776 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8777 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8778 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8780 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8783 * dwarf2out.c (is_trivial_indirect_ref): New function.
8784 (dwarf2out_late_global_decl): Do not generate a location
8785 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8786 and that are not defined in the current unit.
8788 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8790 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8791 instead of a libcall for UNORDERED.
8793 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8796 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8797 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8799 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8802 * config/s390/s390.c (s390_set_current_function): Invoke
8803 s390_indirect_branch_settings also if fndecl didn't change.
8805 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8807 * config/rs6000/rs6000.md (blockage): Set length to zero.
8809 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8811 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8813 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8816 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8817 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8821 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8822 PARALLEL incoming that failed vt_get_decl_and_offset check.
8825 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8826 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8828 (get_variable_section): For decls in named .bss* sections pass true as
8829 second argument to bss_initializer_p.
8831 2018-02-09 Marek Polacek <polacek@redhat.com>
8832 Jakub Jelinek <jakub@redhat.com>
8835 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8836 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8837 Sync some changes from cxx_fold_indirect_ref.
8839 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8841 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8843 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8844 (BLOCK_INLINE_ENTRY_LABEL): New.
8845 (dwarf2out_var_location): Disregard inline entry markers.
8846 (inline_entry_data): New struct.
8847 (inline_entry_data_hasher): New hashtable type.
8848 (inline_entry_data_hasher::hash): New.
8849 (inline_entry_data_hasher::equal): New.
8850 (inline_entry_data_table): New variable.
8851 (add_high_low_attributes): Add DW_AT_entry_pc and
8852 DW_AT_GNU_entry_view attributes if a pending entry is found
8853 in inline_entry_data_table. Add old entry_pc attribute only
8854 if debug nonbinding markers are disabled.
8855 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8856 markers are enabled.
8857 (block_within_block_p, dwarf2out_inline_entry): New.
8858 (dwarf2out_finish): Check that no entries remained in
8859 inline_entry_data_table.
8860 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8861 (final_scan_insn, notice_source_line): Likewise.
8862 (rest_of_clean_state): Skip inline entry markers.
8863 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8865 * gimple.c (gimple_build_debug_inline_entry): New.
8866 * gimple.h (enum gimple_debug_subcode): Add
8867 GIMPLE_DEBUG_INLINE_ENTRY.
8868 (gimple_build_debug_inline_entry): Declare.
8869 (gimple_debug_inline_entry_p): New.
8870 (gimple_debug_nonbind_marker_p): Adjust.
8871 * insn-notes.def (INLINE_ENTRY): New.
8872 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8873 inline entry marker notes.
8874 (print_insn): Likewise.
8875 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8876 (INSN_DEBUG_MARKER_KIND): Likewise.
8877 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8878 * tree-inline.c (expand_call_inline): Build and insert
8879 debug_inline_entry stmt.
8880 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8881 inline entry blocks early, if nonbind markers are enabled.
8882 (dump_scope_block): Dump fragment info.
8883 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8884 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8885 (gimple_build_debug_inline_entry): New.
8886 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8887 Enable/disable inline entry points too.
8888 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8889 (DEBUG_INSN): Describe inline entry markers.
8891 * common.opt (gvariable-location-views): New.
8892 (gvariable-location-views=incompat5): New.
8893 * config.in: Rebuilt.
8894 * configure: Rebuilt.
8895 * configure.ac: Test assembler for view support.
8896 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8897 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8898 * dwarf2out.c (var_loc_view): New typedef.
8899 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8900 (dwarf2out_locviews_in_attribute): New.
8901 (dwarf2out_locviews_in_loclist): New.
8902 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8903 (enum dw_line_info_opcode): Add LI_adv_address.
8904 (struct dw_line_info_table): Add view.
8905 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8906 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8907 (zero_view_p): New variable.
8908 (ZERO_VIEW_P): New macro.
8909 (output_asm_line_debug_info): New.
8910 (struct var_loc_node): Add view.
8911 (add_AT_view_list, AT_loc_list): New.
8912 (add_var_loc_to_decl): Add view param. Test it against last.
8913 (new_loc_list): Add view params. Record them.
8914 (AT_loc_list_ptr): Handle loc and view lists.
8915 (view_list_to_loc_list_val_node): New.
8916 (print_dw_val): Handle dw_val_class_view_list.
8917 (size_of_die): Likewise.
8918 (value_format): Likewise.
8919 (loc_list_has_views): New.
8920 (gen_llsym): Set vl_symbol too.
8921 (maybe_gen_llsym, skip_loc_list_entry): New.
8922 (dwarf2out_maybe_output_loclist_view_pair): New.
8923 (output_loc_list): Output view list or entries too.
8924 (output_view_list_offset): New.
8925 (output_die): Handle dw_val_class_view_list.
8926 (output_dwarf_version): New.
8927 (output_compilation_unit_header): Use it.
8928 (output_skeleton_debug_sections): Likewise.
8929 (output_rnglists, output_line_info): Likewise.
8930 (output_pubnames, output_aranges): Update version comments.
8931 (output_one_line_info_table): Output view numbers in asm comments.
8932 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8933 Call maybe_gen_llsym.
8934 (loc_list_from_tree_1): Adjust.
8935 (add_AT_location_description): Create view list attribute if
8936 needed, check it's absent otherwise.
8937 (convert_cfa_to_fb_loc_list): Adjust.
8938 (maybe_emit_file): Call output_asm_line_debug_info for test.
8939 (dwarf2out_var_location): Reset views as needed. Precompute
8940 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8941 attribute. Set view.
8942 (new_line_info_table): Reset next view.
8943 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8944 (dwarf2out_source_line): Likewise. Output view resets and labels to
8945 the assembler, or select appropriate line info opcodes.
8946 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8947 (optimize_string_length): Catch it. Adjust.
8948 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8949 dw_val_class_view_list, and remove it if no longer needed.
8950 (hash_loc_list): Hash view numbers.
8951 (loc_list_hasher::equal): Compare them.
8952 (optimize_location_lists): Check whether a view list symbol is
8953 needed, and whether the locview attribute is present, and
8954 whether they match. Remove the locview attribute if no longer
8956 (index_location_lists): Call skip_loc_list_entry for test.
8957 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8958 Use output_dwarf_version.
8959 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8960 (struct dw_val_node): Add val_view_list.
8961 * final.c (SEEN_NEXT_VIEW): New.
8962 (set_next_view_needed): New.
8963 (clear_next_view_needed): New.
8964 (maybe_output_next_view): New.
8965 (final_start_function): Rename to...
8966 (final_start_function_1): ... this. Take pointer to FIRST,
8967 add SEEN parameter. Emit param bindings in the initial view.
8968 (final_start_function): Reintroduce SEEN-less interface.
8969 (final): Rename to...
8970 (final_1): ... this. Take SEEN parameter. Output final pending
8971 next view at the end.
8972 (final): Reintroduce seen-less interface.
8973 (final_scan_insn): Output pending next view before switching
8974 sections or ending a block. Mark the next view as needed when
8975 outputting variable locations. Notify debug backend of section
8976 changes, and of location view changes.
8977 (rest_of_handle_final): Adjust.
8978 * toplev.c (process_options): Autodetect value for debug variable
8979 location views option. Warn on incompat5 without -gdwarf-5.
8980 * doc/invoke.texi (gvariable-location-views): New.
8981 (gvariable-location-views=incompat5): New.
8982 (gno-variable-location-views): New.
8984 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8986 PR tree-optimization/84136
8987 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8988 that the result of find_edge is non-NULL.
8990 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8993 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8994 storing integer register in SImode. Fix cost of 256 and 512
8995 byte aligned SSE register store.
8997 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8999 * config/i386/i386.c (ix86_multiplication_cost): Fix
9000 multiplication cost for TARGET_AVX512DQ.
9002 2018-02-08 Marek Polacek <polacek@redhat.com>
9004 PR tree-optimization/84238
9005 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9008 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9010 PR tree-optimization/84265
9011 * tree-vect-stmts.c (vectorizable_store): Don't treat
9012 VMAT_CONTIGUOUS accesses as grouped.
9013 (vectorizable_load): Likewise.
9015 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9017 PR tree-optimization/81635
9018 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9019 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9020 (test_round_for_mask): New functions.
9021 (wide_int_cc_tests): Call test_round_for_mask.
9022 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9023 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9024 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9025 range returned by get_range_info.
9027 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9030 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9031 * symtab.c: Include builtins.h
9032 (symtab_node::output_to_lto_symbol_table_p): Move here
9033 from lto-streamer-out.c:output_symbol_p.
9034 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9035 (output_symbol_p): Move all logic to symtab.c
9036 (produce_symtab): Update.
9038 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9040 * config/s390/s390-opts.h (enum indirect_branch): Define.
9041 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9042 (s390_indirect_branch_via_thunk)
9043 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9044 (enum s390_indirect_branch_type): Define.
9045 * config/s390/s390.c (struct s390_frame_layout, struct
9046 machine_function): Remove.
9047 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9048 (indirect_branch_table_label_no, indirect_branch_table_name):
9050 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9051 (enum s390_indirect_branch_option): Define.
9052 (s390_return_addr_from_memory): New function.
9053 (s390_handle_string_attribute): New function.
9054 (s390_attribute_table): Add new attribute handler.
9055 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9056 (s390_indirect_branch_via_thunk): New function.
9057 (s390_indirect_branch_via_inline_thunk): New function.
9058 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9059 sibling call optimization for non z10 compiles.
9060 (s390_emit_call): Force indirect branch target to be a single
9061 register. Add r1 clobber for non-z10 compiles.
9062 (s390_emit_epilogue): Emit return jump via return_use expander.
9063 (s390_reorg): Handle JUMP_INSNs as execute targets.
9064 (s390_option_override_internal): Perform validity checks for the
9065 new command line options.
9066 (s390_indirect_branch_attrvalue): New function.
9067 (s390_indirect_branch_settings): New function.
9068 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9069 (s390_output_indirect_thunk_function): New function.
9070 (s390_code_end): Implement target hook.
9071 (s390_case_values_threshold): Implement target hook.
9072 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9074 * config/s390/s390.h (struct s390_frame_layout)
9075 (struct machine_function): Move here from s390.c.
9076 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9077 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9078 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9079 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9080 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9081 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9082 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9083 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9084 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9085 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9086 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9087 (mnemonic attribute): Add values which aren't recognized
9089 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9090 pattern for branch conversion. Fix mnemonic attribute.
9091 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9092 indirect branch via thunk if requested.
9093 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9094 ("*indirect_jump"): Disable for branch conversion using out of
9096 ("indirect_jump_via_thunk<mode>_z10")
9097 ("indirect_jump_via_thunk<mode>")
9098 ("indirect_jump_via_inlinethunk<mode>_z10")
9099 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9100 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9101 ("casesi_jump_via_inlinethunk<mode>_z10")
9102 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9103 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9104 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9105 ("*indirect2_jump"): Disable for branch conversion.
9106 ("casesi_jump"): Turn into expander and expand patterns for branch
9108 ("return_use"): New expander.
9109 ("*return"): Emit return via thunk and rename it to ...
9110 ("*return<mode>"): ... this one.
9111 * config/s390/s390.opt: Add new options and and enum for the
9114 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9116 * lra-constraints.c (match_reload): Unconditionally use
9117 gen_lowpart_SUBREG, rather than selecting between that
9118 and equivalent gen_rtx_SUBREG code.
9120 2018-02-08 Richard Biener <rguenther@suse.de>
9122 PR tree-optimization/84233
9123 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9124 changed flag instead of boguously re-using phi_inserted.
9126 2018-02-08 Martin Jambor <mjambor@suse.cz>
9128 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9129 static local variables.
9131 2018-02-08 Richard Biener <rguenther@suse.de>
9133 PR tree-optimization/84278
9134 * tree-vect-stmts.c (vectorizable_store): When looking for
9135 smaller vector types to perform grouped strided loads/stores
9136 make sure the mode is supported by the target.
9137 (vectorizable_load): Likewise.
9139 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9141 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9142 Increase LDP/STP opportunities by adding adjacent callee-saves.
9144 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9146 PR rtl-optimization/84068
9147 PR rtl-optimization/83459
9148 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9150 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9152 PR tree-optimization/84224
9153 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9154 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9157 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9160 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9161 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9162 count, remove one USE.
9164 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9166 * doc/install.texi (Configuration): Document the
9167 --with-long-double-format={ibm,ieee} PowerPC configuration
9171 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9172 Convert from define_expand to be define_insn_and_split. Rework
9173 float/double/_Float128 conversions to QI/HI/SImode to work with
9174 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9175 conversions to QI/HImode types did a store and then a load to
9176 truncate the value. For conversions to VSX registers, don't split
9177 the insn, instead emit the code directly. Use the code iterator
9178 any_fix to combine signed and unsigned conversions.
9179 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9180 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9181 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9182 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9183 (fix_<mode>di2_hw): Likewise.
9184 (fixuns_<mode>di2_hw): Likewise.
9185 (fix_<mode>si2_hw): Likewise.
9186 (fixuns_<mode>si2_hw): Likewise.
9187 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9188 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9189 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9190 fix<uns>_trunc<SFDF:mode>si2_p8.
9191 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9192 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9193 (fix<uns>_<mode>_mem): Likewise.
9194 (fctiw<u>z_<mode>_mem): Likewise.
9195 (fix<uns>_<mode>_mem): Likewise.
9196 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9197 the register allocator from doing a direct move to the GPRs to do
9198 a store, and instead use the ISA 3.0 store byte/half-word from
9199 vector register instruction. For IEEE 128-bit floating point,
9200 also optimize stores of 32-bit ints.
9201 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9203 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9205 * genextract.c (push_pathstr_operand): New function to support
9207 (walk_rtx): Call push_pathstr_operand.
9208 (print_path): Support [a-zA-Z].
9210 2018-02-07 Richard Biener <rguenther@suse.de>
9212 PR tree-optimization/84037
9213 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9214 (cse_and_gimplify_to_preheader): Declare.
9215 (vect_get_place_in_interleaving_chain): Likewise.
9216 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9218 (_loop_vec_info::~_loop_vec_info): Delete it.
9219 (cse_and_gimplify_to_preheader): New function.
9220 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9221 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9222 (vectorizable_load): Likewise. For grouped stores always base
9223 the IV on the first element.
9224 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9225 condition before gimplifying.
9227 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9229 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9230 *DIV_EXPR and *MOD_EXPR.
9232 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9235 * config/i386/i386.c (ix86_option_override_internal): Mask out
9236 the CF_SET bit when checking -fcf-protection.
9238 2018-02-07 Tom de Vries <tom@codesourcery.com>
9241 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9244 2018-02-07 Richard Biener <rguenther@suse.de>
9246 PR tree-optimization/84204
9247 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9250 PR tree-optimization/84205
9251 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9252 special-case isl_ast_op_zdiv_r.
9254 PR tree-optimization/84223
9255 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9256 Only add conditions from within the region.
9257 (gather_bbs::after_dom_children): Adjust.
9259 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9262 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9263 * config/avr/avr.md: Only post-reload split REG-REG moves if
9264 either register is GENERAL_REG_P.
9266 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9268 PR tree-optimization/84235
9269 * tree-ssa-scopedtables.c
9270 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9271 if the subtraction is performed in floating point type where NaNs are
9272 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9273 build 1. Formatting fix.
9275 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9278 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9279 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9280 and skip it regardless of bb boundaries. Use CALL_P macro,
9281 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9282 unnecessarily, formatting fix.
9284 2018-02-06 Michael Collison <michael.collison@arm.com>
9286 * config/arm/thumb2.md:
9287 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9288 (*thumb_mov_notscc): Ditto.
9290 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9293 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9294 unsigned_fix, not "s".
9296 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9298 * configure.ac (gcc_fn_eh_frame_ro): New function.
9299 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9300 correct .eh_frame permissions.
9301 * configure: Regenerate.
9303 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9305 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9308 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9310 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9311 Display warning message for -mno-speculate-indirect-jumps.
9313 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9315 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9317 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9319 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9321 PR tree-optimization/84225
9322 * tree-eh.c (find_trapping_overflow): Only call
9323 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9325 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9328 * config/i386/i386.c: Reimplement the check of possible options
9329 -mibt/-mshstk conbination. Change error messages.
9330 * doc/invoke.texi: Fix a typo: remove extra '='.
9332 2018-02-06 Marek Polacek <polacek@redhat.com>
9334 PR tree-optimization/84228
9335 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9337 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9340 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9341 emitted arch directives.
9342 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9343 __ARM_FEATURE_COPROC before changing architectures.
9345 2018-02-06 Richard Biener <rguenther@suse.de>
9347 * config/i386/i386.c (print_reg): Fix typo.
9348 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9350 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9352 * configure: Regenerate.
9354 2018-02-05 Martin Sebor <msebor@redhat.com>
9356 PR tree-optimization/83369
9357 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9360 2018-02-05 Martin Liska <mliska@suse.cz>
9362 * doc/invoke.texi: Cherry-pick upstream r323995.
9364 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9366 * ira.c (ira_init_register_move_cost): Adjust comment.
9368 2018-02-05 Martin Liska <mliska@suse.cz>
9370 PR gcov-profile/84137
9371 * doc/gcov.texi: Fix typo in documentation.
9373 2018-02-05 Martin Liska <mliska@suse.cz>
9375 PR gcov-profile/83879
9376 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9377 using dlopen functionality.
9379 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9381 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9382 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9383 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9384 _mm_maskz_range_round_ss): New intrinsics.
9385 (__builtin_ia32_rangesd128_round)
9386 (__builtin_ia32_rangess128_round): Remove.
9387 (__builtin_ia32_rangesd128_mask_round,
9388 __builtin_ia32_rangess128_mask_round): New builtins.
9389 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9390 __builtin_ia32_rangess128_round): Remove.
9391 (__builtin_ia32_rangesd128_mask_round,
9392 __builtin_ia32_rangess128_mask_round): New builtins.
9393 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9394 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9395 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9396 "<round_saeonly_constraint>")): Changed to ...
9397 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9398 "<round_saeonly_scalar_constraint>")): ... this.
9399 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9400 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9401 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9402 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9403 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9405 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9407 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9409 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9410 Remove all values except native, 8540 and 8548.
9412 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9414 * config/i386/i386.c (ix86_output_function_return): Pass
9415 INVALID_REGNUM, instead of -1, as invalid register number to
9416 indirect_thunk_name and output_indirect_thunk.
9418 2018-02-02 Julia Koval <julia.koval@intel.com>
9420 * config.gcc: Add -march=icelake.
9421 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9422 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9423 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9424 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9425 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9426 (processor_target_table): Add icelake.
9427 (ix86_option_override_internal): Handle new PTAs.
9428 (get_builtin_code_for_version): Handle icelake.
9429 (M_INTEL_COREI7_ICELAKE): New.
9430 (fold_builtin_cpu): Handle icelake.
9431 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9432 * doc/invoke.texi: Add -march=icelake.
9434 2018-02-02 Julia Koval <julia.koval@intel.com>
9436 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9437 to wide_int_bitmask.
9438 * wide-int-bitmask.h: New.
9440 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9443 * config/i386/i386.md: Replace Pmode with word_mode in
9444 builtin_setjmp_setup and builtin_longjmp to support x32.
9446 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9450 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9452 (rs6000_supported_cpu_names): New static variable.
9453 (linux_cpu_translation_table): Likewise.
9454 (elf_platform) <cpu>: Define new static variable and use it.
9455 Translate kernel AT_PLATFORM name to canonical name if needed.
9456 Error if platform name is unknown.
9458 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9461 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9463 2018-02-01 Jeff Law <law@redhat.com>
9466 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9467 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9468 the scratch if RELEASE_VIA_POP is false.
9469 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9470 If we have to save a temporary register, decrement SIZE appropriately.
9471 Pass new arguments to release_scratch_register_on_entry.
9472 (ix86_adjust_stack_and_probe): Likewise.
9473 (ix86_emit_probe_stack_range): Pass new arguments to
9474 release_scratch_register_on_entry.
9476 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9478 PR rtl-optimization/84157
9479 * combine.c (change_zero_ext): Use REG_P predicate in
9480 front of HARD_REGISTER_P predicate.
9482 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9484 * config/avr/avr.c (avr_option_override): Move disabling of
9485 -fdelete-null-pointer-checks to...
9486 * common/config/avr/avr-common.c (avr_option_optimization_table):
9489 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9491 PR tree-optimization/81635
9492 * tree-data-ref.c (split_constant_offset_1): For types that
9493 wrap on overflow, try to use range info to prove that wrapping
9496 2018-02-01 Renlin Li <renlin.li@arm.com>
9499 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9501 (aarch64_register_move_cost): Likewise.
9502 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9504 (REG_CLASS_NAMES): Likewise.
9505 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9506 TAILCALL_ADDR_REGS. Remove IP registers.
9507 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9509 2018-02-01 Richard Biener <rguenther@suse.de>
9511 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9512 for specifying RPO order and allow NULL for that.
9513 * domwalk.c (dom_walker::dom_walker): Likewise.
9514 (dom_walker::walk): Handle NULL RPO order.
9515 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9517 (rewrite_update_dom_walker): Likewise.
9518 (mark_def_dom_walker): Likewise.
9520 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9522 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9523 (aarch64_maybe_expand_sve_subreg_move): Declare.
9524 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9525 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9527 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9528 that are semantically a reverse operation.
9529 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9530 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9531 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9533 (aarch64_can_change_mode_class): For big-endian, forbid changes
9534 between two SVE modes if they have different element sizes.
9536 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9538 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9539 the TImode handling for big-endian targets.
9541 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9543 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9544 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9546 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9547 Remove BSWAP handing for big-endian targets and use the form of
9548 LD1RQ appropariate for the mode.
9550 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9552 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9553 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9556 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9559 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9560 check for operands that need to go through aarch64_sve_reload_be.
9562 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9564 PR tree-optimization/81661
9565 PR tree-optimization/84117
9566 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9567 * tree-eh.c: Include gimplify.h.
9568 (find_trapping_overflow, replace_trapping_overflow,
9569 rewrite_to_non_trapping_overflow): New functions.
9570 * tree-vect-loop.c: Include tree-eh.h.
9571 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9572 * tree-data-ref.c: Include tree-eh.h.
9573 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9575 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9577 PR rtl-optimization/84123
9578 * combine.c (change_zero_ext): Check if hard register satisfies
9579 can_change_dest_mode before calling gen_lowpart_SUBREG.
9581 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9584 * ira.c (ira_init_register_move_cost): Remove assert.
9586 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9588 PR rtl-optimization/84071
9589 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9590 * doc/tm.texi: Regenerate.
9592 2018-01-31 Richard Biener <rguenther@suse.de>
9594 PR tree-optimization/84132
9595 * tree-data-ref.c (analyze_miv_subscript): Properly
9596 check whether evolution_function_is_affine_multivariate_p
9597 before calling gcd_of_steps_may_divide_p.
9599 2018-01-31 Julia Koval <julia.koval@intel.com>
9602 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9603 * config/i386/i386.md (rdpid_rex64) New.
9604 (rdpid): Make 32bit only.
9606 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9609 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9610 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9612 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9615 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9617 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9619 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9621 PR rtl-optimization/84071
9622 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9623 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9625 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9627 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9628 (arc_attribute_table): Add 'aux' attribute.
9629 (arc_in_small_data_p): Consider aux like variables.
9630 (arc_is_aux_reg_p): New function.
9631 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9632 (arc_get_aux_arg): New function.
9633 (prepare_move_operands): Handle aux-register access.
9634 (arc_handle_aux_attribute): New function.
9635 * doc/extend.texi (ARC Variable attributes): Add subsection.
9637 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9639 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9640 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9641 (arc_attribute_table): Add 'uncached' attribute.
9642 (arc_print_operand): Print '.di' flag for uncached memory
9644 (arc_in_small_data_p): Do not consider for small data the uncached
9646 (arc_is_uncached_mem_p): New function.
9647 * config/arc/predicates.md (compact_store_memory_operand): Check
9648 for uncached memory accesses.
9649 (nonvol_nonimm_operand): Likewise.
9650 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9652 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9655 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9656 falign-loops=): Add Optimization flag.
9658 2018-01-30 Jeff Law <law@redhat.com>
9661 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9662 INT_REGISTERS_SAVED. Check it prior to calling
9663 get_scratch_register_on_entry.
9664 (ix86_adjust_stack_and_probe): Similarly.
9665 (ix86_emit_probe_stack_range): Similarly.
9666 (ix86_expand_prologue): Corresponding changes.
9668 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9671 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9672 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9674 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9677 * lra-constraints.c (curr_insn_transform): Process AND in the
9680 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9682 PR rtl-optimization/83986
9683 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9684 dependence against last_pending_memory_flush in addition to
9687 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9689 PR tree-optimization/81611
9690 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9693 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9696 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9699 2018-01-30 Richard Biener <rguenther@suse.de>
9700 Jakub Jelinek <jakub@redhat.com>
9702 PR tree-optimization/84111
9703 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9704 inner loops added during recursion, as they don't have up-to-date
9707 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9710 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9711 (can_inline_edge_by_limits_p): ... here.
9712 (can_early_inline_edge_p, check_callers,
9713 update_caller_keys, update_callee_keys, recursive_inlining,
9714 add_new_edges_to_heap, speculation_useful_p,
9715 inline_small_functions,
9716 inline_small_functions, flatten_function,
9717 inline_to_all_callers_1): Update.
9719 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9721 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9724 2018-01-30 Richard Biener <rguenther@suse.de>
9726 PR tree-optimization/83008
9727 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9728 invariant and constant vector uses in stmts when they need
9731 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9734 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9735 * configure: Regenerate.
9737 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9739 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9741 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9742 Use gen_rtx_REG rather than gen_lowpart.
9744 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9746 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9747 rather than 0 when creating partial subregs.
9749 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9751 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9754 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9757 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9758 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9759 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9760 flags. This restores the settings used before the 2017-07-24.
9761 Turning off pre increment/decrement/modify allows IVOPTS to
9762 optimize DF/SF loops where the index is an int.
9764 2018-01-29 Richard Biener <rguenther@suse.de>
9765 Kelvin Nilsen <kelvin@gcc.gnu.org>
9768 * tree-vect-stmts.c (vectorizable_call): Don't call
9769 targetm.vectorize_builtin_md_vectorized_function if callee is
9772 2018-01-22 Carl Love <cel@us.ibm.com>
9774 * doc/extend.tex: Fix typo in second arg in
9775 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9777 2018-01-29 Richard Biener <rguenther@suse.de>
9779 PR tree-optimization/84086
9780 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9781 (flush_ssaname_freelist): When SSA names were released reset
9782 the SCEV hash table.
9784 2018-01-29 Richard Biener <rguenther@suse.de>
9786 PR tree-optimization/84057
9787 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9788 removed paths when removing edges.
9790 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9792 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9793 -mfunction-return=@var{choice}.
9795 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9798 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9800 (layout::print_source_line): Likewise.
9801 (test_get_line_width_without_trailing_whitespace): Add test cases.
9803 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9806 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9809 2018-01-26 Jim Wilson <jimw@sifive.com>
9811 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9813 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9816 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9818 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9819 and CMP + SUB-immediate -> SUBS.
9821 2018-01-26 Martin Sebor <msebor@redhat.com>
9823 PR tree-optimization/83896
9824 * tree-ssa-strlen.c (get_string_len): Rename...
9825 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9826 Avoid assuming length is constant.
9827 (handle_char_store): Use HOST_WIDE_INT for string length.
9829 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9832 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9833 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9835 2018-01-26 Richard Biener <rguenther@suse.de>
9837 PR rtl-optimization/84003
9838 * dse.c (record_store): Only record redundant stores when
9839 the earlier store aliases at least all accesses the later one does.
9841 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9843 PR rtl-optimization/83985
9844 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9845 REG_CFA_RESTORE insns.
9846 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9847 REG_CFA_RESTORE insns here.
9850 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9851 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9853 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9855 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9856 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9857 (arc_init): Likewise.
9858 (arc_override_options): Likewise.
9859 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9861 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9863 * config/arc/arc.h (TARGET_DBNZ): Define.
9864 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9865 properly set the tune attribute.
9866 (dbnz): Use TARGET_DBNZ guard.
9867 * config/arc/arc.opt (mtune): Add core3 option.
9869 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9871 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9872 recognize new pic like addresses.
9873 (arc_delegitimize_address): Clean up.
9875 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9877 * config/arc/arc-arches.def: Option mrf16 valid for all
9879 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9880 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9881 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9882 * config/arc/arc-tables.opt: Regenerate.
9883 * config/arc/arc.c (arc_conditional_register_usage): Handle
9884 reduced register file case.
9885 (arc_file_start): Set must have build attributes.
9886 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9888 * config/arc/arc.opt (mrf16): Add new option.
9889 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9890 * config/arc/genmultilib.awk: Handle new mrf16 option.
9891 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9892 * config/arc/t-multilib: Regenerate.
9893 * doc/invoke.texi (ARC Options): Document mrf16 option.
9895 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9897 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9898 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9899 (arc_attribute_table): Add 'secure_call' attribute.
9900 (arc_print_operand): Print secure call operand.
9901 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9903 (arc_is_secure_call_p): New function. * config/arc/arc.md
9904 (call_i): Add support for sjli instruction.
9905 (call_value_i): Likewise.
9906 * config/arc/constraints.md (Csc): New constraint.
9908 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9909 John Eric Martin <John.Martin@emmicro-us.com>
9911 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9912 * config/arc/arc.c (_arc_jli_section): New struct.
9913 (arc_jli_section): New type.
9914 (rc_jli_sections): New static variable.
9915 (arc_handle_jli_attribute): New function.
9916 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9917 (arc_file_end): New function.
9918 (TARGET_ASM_FILE_END): Define.
9919 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9920 (arc_add_jli_section): New function.
9921 (jli_call_scan): Likewise.
9922 (arc_reorg): Call jli_call_scan.
9923 (arc_output_addsi): Remove 'S' from printing asm operand.
9924 (arc_is_jli_call_p): New function.
9925 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9927 (movhi_insn): Likewise.
9928 (movsi_insn): Likewise.
9929 (movsi_set_cc_insn): Likewise.
9930 (loadqi_update): Likewise.
9931 (load_zeroextendqisi_update): Likewise.
9932 (load_signextendqisi_update): Likewise.
9933 (loadhi_update): Likewise.
9934 (load_zeroextendhisi_update): Likewise.
9935 (load_signextendhisi_update): Likewise.
9936 (loadsi_update): Likewise.
9937 (loadsf_update): Likewise.
9938 (movsicc_insn): Likewise.
9939 (bset_insn): Likewise.
9940 (bxor_insn): Likewise.
9941 (bclr_insn): Likewise.
9942 (bmsk_insn): Likewise.
9943 (bicsi3_insn): Likewise.
9944 (cmpsi_cc_c_insn): Likewise.
9945 (movsi_ne): Likewise.
9946 (movsi_cond_exec): Likewise.
9947 (clrsbsi2): Likewise.
9958 (call_i): Remove 'S' asm letter, add jli instruction.
9959 (call_value_i): Likewise.
9960 * config/arc/arc.op (mjli-always): New option.
9961 * config/arc/constraints.md (Cji): New constraint.
9962 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9964 (subsf3_fpx): Likewise.
9965 (mulsf3_fpx): Likewise.
9966 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9968 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9970 * doc/invoke.texi (ARC): Document mjli-always option.
9972 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9974 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9975 avoid addition with 0 and use incw and decw where possible.
9977 2018-01-26 Richard Biener <rguenther@suse.de>
9979 PR tree-optimization/81082
9980 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9981 association if it requires casting to unsigned.
9982 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9983 from fold_plusminus_mult_expr to catch important cases late when
9984 range info is available.
9986 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9988 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9989 * configure.ac (hidden_linkonce): New test.
9990 * configure: Regenerate.
9991 * config.in: Regenerate.
9993 2018-01-26 Julia Koval <julia.koval@intel.com>
9995 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9996 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9997 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9998 _mm_mask_bitshuffle_epi64_mask): Fix type.
9999 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10000 USI_FTYPE_V4DI_V4DI_USI): Remove.
10001 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10002 __builtin_ia32_vpshufbitqmb256_mask,
10003 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10004 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10005 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10007 2018-01-26 Alan Modra <amodra@gmail.com>
10010 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10011 UNSPEC_VBPERMQ. Sort other unspecs.
10013 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10015 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10017 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10019 PR middle-end/83055
10020 * predict.c (drop_profile): Do not push/pop cfun; update also
10022 (handle_missing_profiles): Fix logic looking for zero profiles.
10024 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10026 PR middle-end/83977
10027 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10028 on functions with #pragma omp declare simd or functions with simd
10030 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10031 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10032 Remove trailing \n from warning_at calls.
10034 2018-01-25 Tom de Vries <tom@codesourcery.com>
10037 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10038 for neutered workers.
10040 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10043 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10044 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10046 2018-01-24 Jeff Law <law@redhat.com>
10049 * i386.c (get_probe_interval): Move to earlier point.
10050 (ix86_compute_frame_layout): If -fstack-clash-protection and
10051 the frame is larger than the probe interval, then use pushes
10052 to save registers rather than reg->mem moves.
10053 (ix86_expand_prologue): Remove conditional for int_registers_saved
10056 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10059 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10060 min/max for never referenced object.
10062 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10064 PR middle-end/83977
10065 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10067 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10068 attributes from DECL_ATTRIBUTES (decl) without affecting
10069 DECL_ATTRIBUTES (current_function_decl).
10070 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10071 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10073 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10075 PR tree-optimization/83979
10076 * fold-const.c (fold_comparison): Use constant_boolean_node
10077 instead of boolean_{true,false}_node.
10079 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10081 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10084 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10086 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10087 Simplify the clause that sets the length attribute.
10088 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10089 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10090 clause that sets the length attribute.
10091 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10093 2018-01-24 Tom de Vries <tom@codesourcery.com>
10096 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10097 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10098 Add strict parameter.
10099 (prevent_branch_around_nothing): Insert dummy insn between branch to
10100 label and label with no ptx insn inbetween.
10101 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10103 2018-01-24 Tom de Vries <tom@codesourcery.com>
10106 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10107 for neutered threads in warp.
10108 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10110 2018-01-24 Richard Biener <rguenther@suse.de>
10112 PR tree-optimization/83176
10113 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10116 2018-01-24 Richard Biener <rguenther@suse.de>
10118 PR tree-optimization/82819
10119 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10120 code generating pluses that are no-ops in the target precision.
10122 2018-01-24 Richard Biener <rguenther@suse.de>
10124 PR middle-end/84000
10125 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10127 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10129 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10130 to merge probabilities.
10131 * predict.c (probably_never_executed): Also mark as cold functions
10132 with global 0 profile and guessed local profile.
10133 * profile-count.c (profile_probability::combine_with_count): New
10135 * profile-count.h (profile_probability::operator*,
10136 profile_probability::operator*=, profile_probability::operator/,
10137 profile_probability::operator/=): Reduce precision to adjusted
10138 and set value to guessed on contradictory divisions.
10139 (profile_probability::combine_with_freq): Remove.
10140 (profile_probability::combine_wiht_count): Declare.
10141 (profile_count::force_nonzero):: Set to adjusted.
10142 (profile_count::probability_in):: Set quality to adjusted.
10143 * tree-ssa-tail-merge.c (replace_block_by): Use
10144 combine_with_count.
10146 2018-01-23 Andrew Waterman <andrew@sifive.com>
10147 Jim Wilson <jimw@sifive.com>
10149 * config/riscv/riscv.c (riscv_stack_boundary): New.
10150 (riscv_option_override): Set riscv_stack_boundary. Handle
10151 riscv_preferred_stack_boundary_arg.
10152 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10153 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10154 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10155 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10156 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10157 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10159 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10162 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10163 of struct ix86_frame.
10164 (ix86_expand_epilogue): Likewise. Add a local variable for
10165 the reg_save_offset field in struct ix86_frame.
10167 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10169 PR tree-optimization/82604
10170 * tree-loop-distribution.c (enum partition_kind): New enum item
10171 PKIND_PARTIAL_MEMSET.
10172 (partition_builtin_p): Support above new enum item.
10173 (generate_code_for_partition): Ditto.
10174 (compute_access_range): Differentiate cases that equality can be
10175 proven at all loops, the innermost loops or no loops.
10176 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10177 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10178 (finalize_partitions, distribute_loop): Don't fuse partition of
10179 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10180 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10181 parloop is enabled.
10183 2018-01-23 Martin Liska <mliska@suse.cz>
10185 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10186 order to ignore the predictor.
10187 (PRED_POLYMORPHIC_CALL): Likewise.
10188 (PRED_RECURSIVE_CALL): Likewise.
10190 2018-01-23 Martin Liska <mliska@suse.cz>
10192 * tree-profile.c (tree_profiling): Print function header to
10193 aware reader which function we are working on.
10194 * value-prof.c (gimple_find_values_to_profile): Do not print
10195 not interesting value histograms.
10197 2018-01-23 Martin Liska <mliska@suse.cz>
10199 * profile-count.h (enum profile_quality): Add
10200 profile_uninitialized as the first value. Do not number values
10201 as they are zero based.
10202 (profile_count::verify): Update sanity check.
10203 (profile_probability::verify): Likewise.
10205 2018-01-23 Nathan Sidwell <nathan@acm.org>
10207 * doc/invoke.texi (ffor-scope): Deprecate.
10209 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10211 PR tree-optimization/83510
10212 * domwalk.c (set_all_edges_as_executable): New function.
10213 (dom_walker::dom_walker): Convert bool param
10214 "skip_unreachable_blocks" to enum reachability. Move setup of
10215 edge flags to set_all_edges_as_executable and only do it when
10216 reachability is REACHABLE_BLOCKS.
10217 * domwalk.h (enum dom_walker::reachability): New enum.
10218 (dom_walker::dom_walker): Convert bool param
10219 "skip_unreachable_blocks" to enum reachability.
10220 (set_all_edges_as_executable): New decl.
10221 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10222 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10224 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10225 but converting true to REACHABLE_BLOCKS.
10226 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10228 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10229 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10230 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10232 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10233 if check_all_array_refs will be called.
10235 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10237 * tree.c (selftest::test_location_wrappers): Add more test
10240 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10242 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10243 (selftest::test_bit_in_range): Likewise.
10245 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10248 * doc/sourcebuild.texi (vect_float): Say that the selector
10249 only describes the situation when -funsafe-math-optimizations is on.
10250 (vect_float_strict): Document.
10252 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10254 PR tree-optimization/83965
10255 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10256 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10257 instead of checking only for a reduction.
10258 (vect_recog_widen_sum_pattern): Likewise.
10260 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10262 * predict.c (probably_never_executed): Only use precise profile info.
10263 (compute_function_frequency): Skip after inlining hack since we now
10264 have quality checking.
10266 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10268 * profile-count.h (profile_probability::very_unlikely,
10269 profile_probability::unlikely, profile_probability::even): Set
10270 precision to guessed.
10272 2018-01-23 Richard Biener <rguenther@suse.de>
10274 PR tree-optimization/83963
10275 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10276 Properly terminate dominator walk when crossing the exit edge not
10277 when visiting its source block.
10279 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10282 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10283 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10285 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10287 PR tree-optimization/83957
10288 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10289 semicolon after for body surrounded by braces.
10291 PR tree-optimization/83081
10292 * profile-count.h (profile_probability::split): New method.
10293 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10294 Use profile_probability::split.
10295 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10296 when splitting a single conditional jump into 2.
10298 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10300 PR tree-optimization/69452
10301 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10304 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10306 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10307 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10308 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10310 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10312 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10314 * config/rl78/rl78.md (movdi): New define_expand.
10315 * config/rl78/rl78.c (rl78_split_movdi): New function.
10317 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10320 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10322 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10323 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10324 128-bit to produce an UNSPEC move to get the double word with the
10325 signbit and then a shift directly to do signbit.
10326 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10327 implementation with a new version that just does either a direct
10328 move or a regular move. Move memory interface to separate insns.
10329 Move insns so they are next to the expander.
10330 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10331 with signbit move. Split big and little endian case.
10332 (signbit<mode>2_dm_mem_le): Likewise.
10333 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10334 (signbit<mode>2_dm2): Likewise.
10336 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10338 * config/rl78/rl78.md (anddi3): New define_expand.
10340 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10342 * config/rl78/rl78.md (umindi3): New define_expand.
10344 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10346 * config/rl78/rl78.md (smindi3): New define_expand.
10348 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10350 * config/rl78/rl78.md (smaxdi3): New define_expand.
10352 2018-01-22 Carl Love <cel@us.ibm.com>
10354 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10355 LVX_V1TI): Add macro expansion.
10356 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10357 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10358 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10359 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10360 Change check to determine if the instruction is a byte reversing
10361 entry. Fix typo in comment.
10362 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10363 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10364 Add def_builtin calls for new builtins.
10365 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10366 Add define_insn expansion.
10368 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10370 * config/rl78/rl78.md (umaxdi3): New define_expand.
10372 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10374 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10375 for non-QImode registers.
10377 2018-01-22 Richard Biener <rguenther@suse.de>
10379 PR tree-optimization/83963
10380 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10381 including the loop exit block.
10382 (scop_detection::merge_sese): Likewise.
10383 (scop_detection::add_scop): Do it here instead.
10385 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10387 * doc/sourcebuild.texi (arm_softfloat): Document.
10389 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10392 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10393 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10394 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10396 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10397 David Edelsohn <dje.gcc@gmail.com>
10400 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10401 Change "crset eq" to "crset 2".
10402 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10403 (*call_indirect_aix<mode>_nospec): Likewise.
10404 (*call_value_indirect_aix<mode>_nospec): Likewise.
10405 (*call_indirect_elfv2<mode>_nospec): Likewise.
10406 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10407 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10408 change assembly output from . to $.
10409 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10410 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10411 (*tablejump<mode>_internal1_nospec): Likewise.
10413 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10416 * config/sh/sh_optimize_sett_clrt.cc:
10417 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10419 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10421 PR tree-optimization/83940
10422 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10423 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10424 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10425 use it to pass back the definition type.
10426 (vect_check_store_rhs): Likewise rhs_dt_out.
10427 (vect_build_gather_load_calls): Add a mask_dt argument and use
10428 it instead of a call to vect_is_simple_use.
10429 (vectorizable_store): Update calls to vect_check_load_store_mask
10430 and vect_check_store_rhs. Use the dt returned by the latter instead
10431 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10432 instead of calls to vect_is_simple_use. Pass the scalar rather
10433 than the vector operand to vect_is_simple_use when handling
10434 second and subsequent copies of an rhs value.
10435 (vectorizable_load): Update calls to vect_check_load_store_mask
10436 and vect_build_gather_load_calls. Use the cached mask_dt and
10437 gs_info.offset_dt instead of calls to vect_is_simple_use.
10439 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10441 PR middle-end/83945
10442 * tree-emutls.c: Include gimplify.h.
10443 (lower_emutls_2): New function.
10444 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10445 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10446 it before further processing.
10449 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10450 UINTVAL (trueop1) instead of INTVAL (op1).
10452 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10456 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10457 INCOMING_FRAME_SP_OFFSET if not defined.
10458 (scan_trace): Add ENTRY argument. If true and
10459 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10460 emit a note to adjust the CFA offset.
10461 (create_cfi_notes): Adjust scan_trace callers.
10462 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10463 INCOMING_FRAME_SP_OFFSET in the CIE.
10464 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10465 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10467 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10468 * doc/tm.texi: Regenerated.
10470 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10472 PR rtl-optimization/83147
10473 * lra-constraints.c (remove_inheritance_pseudos): Use
10474 lra_substitute_pseudo_within_insn.
10476 2018-01-19 Tom de Vries <tom@codesourcery.com>
10477 Cesar Philippidis <cesar@codesourcery.com>
10480 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10482 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10485 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10486 spaces for function labels.
10488 2018-01-19 Martin Liska <mliska@suse.cz>
10490 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10491 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10492 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10493 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10494 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10495 (PRED_CONST_RETURN): Change from 69 to 65.
10496 (PRED_NULL_RETURN): Change from 91 to 71.
10497 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10498 (PRED_LOOP_GUARD): Change from 66 to 73.
10500 2018-01-19 Martin Liska <mliska@suse.cz>
10502 * predict.c (predict_insn_def): Add new assert.
10503 (struct branch_predictor): Change type to signed integer.
10504 (test_prediction_value_range): Amend test to cover
10505 PROB_UNINITIALIZED.
10506 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10507 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10508 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10509 (PRED_LOOP_IV_COMPARE): Likewise.
10510 * predict.h (PROB_UNINITIALIZED): Define new constant.
10512 2018-01-19 Martin Liska <mliska@suse.cz>
10514 * predict.c (dump_prediction): Add new format for
10515 analyze_brprob.py script which is enabled with -details
10517 * profile-count.h (precise_p): New function.
10519 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10521 PR tree-optimization/83922
10522 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10523 there are no statements that need masking.
10524 (vect_active_double_reduction_p): New function.
10525 (vect_analyze_loop_operations): Use it when handling phis that
10526 are not in the loop header.
10528 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10530 PR tree-optimization/83914
10531 * tree-vect-loop.c (vectorizable_induction): Don't convert
10532 init_expr or apply the peeling adjustment for inductions
10533 that are nested within the vectorized loop.
10535 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10537 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10540 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10544 * function.h (gimplify_parameters): Add gimple_seq * argument.
10545 * function.c: Include gimple.h and options.h.
10546 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10547 for the added local temporaries if needed.
10548 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10549 if there are any parameter cleanups, wrap whole body into a
10550 try/finally with the cleanups.
10552 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10555 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10556 Use GET_MODE_CLASS for scalar floating point.
10558 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10562 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10563 Fix call of call_cgraph_insertion_hooks.
10565 2018-01-18 Martin Sebor <msebor@redhat.com>
10567 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10569 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10572 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10575 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10578 * common.opt: (-ffile-prefix-map): New option.
10579 * opts.c (common_handle_option): Defer it.
10580 * opts-global.c (handle_common_deferred_options): Handle it.
10581 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10582 * file-prefix-map.h: New file.
10583 (remap_debug_filename, add_debug_prefix_map): ...here.
10584 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10585 * final.c (debug_prefix_map, add_debug_prefix_map
10586 remap_debug_filename): Move to...
10587 * file-prefix-map.c: New file.
10588 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10589 generalize, get rid of alloca(), use strrchr() instead of strchr().
10590 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10591 Implement in terms of add_prefix_map().
10592 (remap_macro_filename, remap_debug_filename): Implement in term of
10594 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10595 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10596 * dbxout.c: Include file-prefix-map.h.
10597 * varasm.c: Likewise.
10598 * vmsdbgout.c: Likewise.
10599 * xcoffout.c: Likewise.
10600 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10601 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10602 * doc/invoke.texi (-ffile-prefix-map): Document.
10603 (-fdebug-prefix-map): Update description.
10605 2018-01-18 Martin Liska <mliska@suse.cz>
10607 * config/i386/i386.c (indirect_thunk_name): Document that also
10609 (output_indirect_thunk): Document why both instructions
10610 (pause and lfence) are generated.
10612 2018-01-18 Richard Biener <rguenther@suse.de>
10614 PR tree-optimization/83887
10615 * graphite-scop-detection.c
10616 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10617 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10618 (scop_detection::merge_sese): Re-implement with a flood-fill
10619 algorithm that properly finds a SESE region if it exists.
10621 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10624 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10625 pointer_diff optimizations use view_convert instead of convert.
10627 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10629 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10630 Generate different code for -mno-speculate-indirect-jumps.
10631 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10632 (*call_indirect_aix<mode>): Disable for
10633 -mno-speculate-indirect-jumps.
10634 (*call_indirect_aix<mode>_nospec): New define_insn.
10635 (*call_value_indirect_aix<mode>): Disable for
10636 -mno-speculate-indirect-jumps.
10637 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10638 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10639 -mno-speculate-indirect-jumps.
10640 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10642 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10644 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10645 long double type, set the flags for noting the default long double
10646 type, even if we don't pass or return a long double type.
10648 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10651 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10654 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10656 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10657 support for merge[hl].
10658 (fold_mergehl_helper): New helper function.
10659 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10660 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10661 (altivec_vmrglw_direct): Add xxmrglw insn.
10663 2018-01-17 Andrew Waterman <andrew@sifive.com>
10665 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10666 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10668 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10671 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10672 call the lto_location_cache before reading the
10673 DECL_SOURCE_LOCATION of the types.
10675 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10676 Richard Sandiford <richard.sandiford@linaro.org>
10678 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10679 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10680 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10681 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10682 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10684 * config/aarch64/constraints.md (aarch64_movti_operand):
10686 * config/aarch64/predicates.md (Uti): Add new constraint.
10688 2018-01-17 Carl Love <cel@us.ibm.com>
10690 * config/rs6000/vsx.md (define_expand xl_len_r,
10691 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10692 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10694 (define_expand, define_insn): Move the shift left from the
10695 define_insn to the define_expand for lxvl and stxvl instructions.
10696 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10697 and XL_LEN_R definitions to PURE.
10699 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10701 * config/i386/i386.c (indirect_thunk_name): Declare regno
10702 as unsigned int. Compare regno with INVALID_REGNUM.
10703 (output_indirect_thunk): Ditto.
10704 (output_indirect_thunk_function): Ditto.
10705 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10706 in the call to output_indirect_thunk_function.
10708 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10710 PR middle-end/83884
10711 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10712 rather than the size of inner_type to determine the stack slot size
10713 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10715 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10718 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10721 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10723 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10724 endian Linux systems to optionally enable multilibs for selecting
10725 the long double type if the user configured an explicit type.
10726 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10727 have no long double multilibs if not defined.
10728 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10729 warn if the user used -mabi={ieee,ibm}longdouble and we built
10730 multilibs for long double.
10731 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10732 appropriate multilib option.
10733 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10735 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10736 for building long double multilibs.
10737 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10739 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10741 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10744 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10746 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10749 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10752 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10755 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10757 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10758 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10760 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10762 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10763 different rtl trees depending on TARGET_64BIT.
10764 (rs6000_gen_lvx): Likewise.
10766 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10768 * config/visium/visium.md (nop): Tweak comment.
10769 (hazard_nop): Likewise.
10771 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10773 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10774 -mspeculate-indirect-jumps.
10775 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10776 for -mno-speculate-indirect-jumps.
10777 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10778 (*call_value_indirect_elfv2<mode>): Disable for
10779 -mno-speculate-indirect-jumps.
10780 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10781 (indirect_jump): Emit different RTL for
10782 -mno-speculate-indirect-jumps.
10783 (*indirect_jump<mode>): Disable for
10784 -mno-speculate-indirect-jumps.
10785 (*indirect_jump<mode>_nospec): New define_insn.
10786 (tablejump): Emit different RTL for
10787 -mno-speculate-indirect-jumps.
10788 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10789 (tablejumpsi_nospec): New define_expand.
10790 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10791 (tablejumpdi_nospec): New define_expand.
10792 (*tablejump<mode>_internal1): Disable for
10793 -mno-speculate-indirect-jumps.
10794 (*tablejump<mode>_internal1_nospec): New define_insn.
10795 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10798 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10800 * caller-save.c (insert_save): Drop unnecessary parameter. All
10803 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10804 Richard Biener <rguenth@suse.de>
10807 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10808 return early, inline manually is_gimple_sizepos. Make sure if we
10809 call gimplify_expr we don't end up with a gimple constant.
10810 * tree.c (variably_modified_type_p): Don't return true for
10811 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10812 * gimplify.h (is_gimple_sizepos): Remove.
10814 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10816 PR tree-optimization/83857
10817 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10818 vectorizable_live_operation for pure SLP statements.
10819 (vectorizable_live_operation): Handle PHIs.
10821 2018-01-16 Richard Biener <rguenther@suse.de>
10823 PR tree-optimization/83867
10824 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10825 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10827 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10830 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10831 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10832 If off is not INTEGER_CST, issue a may not be aligned warning
10833 rather than isn't aligned. Use isn%'t rather than isn't.
10834 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10836 <case MULT_EXPR>: Improve the case when bottom and one of the
10837 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10838 operand, in that case check if the other operand is multiple of
10839 bottom divided by the INTEGER_CST operand.
10841 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10844 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10845 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10846 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10847 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10848 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10849 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10850 (pa_function_arg_size): New function.
10852 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10854 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10855 in a separate statement.
10857 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10859 PR tree-optimization/83847
10860 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10861 group gathers and scatters.
10863 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10865 PR rtl-optimization/86620
10866 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10868 PR rtl-optimization/83213
10869 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10870 to last if both are JUMP_INSNs.
10872 PR tree-optimization/83843
10873 * gimple-ssa-store-merging.c
10874 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10875 store_immediate_info for bswap/nop orig_stores.
10877 2018-01-15 Andrew Waterman <andrew@sifive.com>
10879 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10881 <UDIV>: Increase cost if !TARGET_DIV.
10883 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10885 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10886 (define_attr "cr_logical_3op"): New.
10887 (cceq_ior_compare): Adjust.
10888 (cceq_ior_compare_complement): Adjust.
10889 (*cceq_rev_compare): Adjust.
10890 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10891 (is_cracked_insn): Adjust.
10892 (insn_must_be_first_in_group): Adjust.
10893 * config/rs6000/40x.md: Adjust.
10894 * config/rs6000/440.md: Adjust.
10895 * config/rs6000/476.md: Adjust.
10896 * config/rs6000/601.md: Adjust.
10897 * config/rs6000/603.md: Adjust.
10898 * config/rs6000/6xx.md: Adjust.
10899 * config/rs6000/7450.md: Adjust.
10900 * config/rs6000/7xx.md: Adjust.
10901 * config/rs6000/8540.md: Adjust.
10902 * config/rs6000/cell.md: Adjust.
10903 * config/rs6000/e300c2c3.md: Adjust.
10904 * config/rs6000/e500mc.md: Adjust.
10905 * config/rs6000/e500mc64.md: Adjust.
10906 * config/rs6000/e5500.md: Adjust.
10907 * config/rs6000/e6500.md: Adjust.
10908 * config/rs6000/mpc.md: Adjust.
10909 * config/rs6000/power4.md: Adjust.
10910 * config/rs6000/power5.md: Adjust.
10911 * config/rs6000/power6.md: Adjust.
10912 * config/rs6000/power7.md: Adjust.
10913 * config/rs6000/power8.md: Adjust.
10914 * config/rs6000/power9.md: Adjust.
10915 * config/rs6000/rs64.md: Adjust.
10916 * config/rs6000/titan.md: Adjust.
10918 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10920 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10921 ix86_indirect_branch_register logic.
10923 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10925 * config/i386/constraints.md (Bs): Update
10926 ix86_indirect_branch_register check. Don't check
10927 ix86_indirect_branch_register with GOT_memory_operand.
10929 * config/i386/predicates.md (GOT_memory_operand): Don't check
10930 ix86_indirect_branch_register here.
10931 (GOT32_symbol_operand): Likewise.
10933 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10935 * config/i386/predicates.md (constant_call_address_operand):
10936 Rewrite ix86_indirect_branch_register logic.
10937 (sibcall_insn_operand): Likewise.
10939 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10941 * config/i386/constraints.md (Bs): Replace
10942 ix86_indirect_branch_thunk_register with
10943 ix86_indirect_branch_register.
10945 * config/i386/i386.md (indirect_jump): Likewise.
10946 (tablejump): Likewise.
10947 (*sibcall_memory): Likewise.
10948 (*sibcall_value_memory): Likewise.
10949 Peepholes of indirect call and jump via memory: Likewise.
10950 * config/i386/i386.opt: Likewise.
10951 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10952 (GOT_memory_operand): Likewise.
10953 (call_insn_operand): Likewise.
10954 (sibcall_insn_operand): Likewise.
10955 (GOT32_symbol_operand): Likewise.
10957 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10959 PR middle-end/83837
10960 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10961 type rather than type addr's type points to.
10962 (expand_omp_atomic_mutex): Likewise.
10963 (expand_omp_atomic): Likewise.
10965 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10968 * config/i386/i386.c (output_indirect_thunk_function): Use
10969 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10970 for __x86_return_thunk.
10972 2018-01-15 Richard Biener <rguenther@suse.de>
10974 PR middle-end/83850
10975 * expmed.c (extract_bit_field_1): Fix typo.
10977 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10980 * config/arm/iterators.md (VF): New mode iterator.
10981 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10982 Remove integer-related logic from pattern.
10983 (neon_vabd<mode>_3): Likewise.
10985 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10987 PR middle-end/82694
10988 * common.opt (fstrict-overflow): No longer an alias.
10989 (fwrapv-pointer): New option.
10990 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10991 also for pointer types based on flag_wrapv_pointer.
10992 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10993 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10994 opts->x_flag_wrapv got set.
10995 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10996 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10997 POINTER_TYPE_OVERFLOW_UNDEFINED.
10998 * match.pd: Likewise in address comparison pattern.
10999 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11001 2018-01-15 Richard Biener <rguenther@suse.de>
11004 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11005 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11006 Reset type names to their identifier if their TYPE_DECL doesn't
11007 have linkage (and thus is used for ODR and devirt).
11008 (save_debug_info_for_decl): Remove.
11009 (save_debug_info_for_type): Likewise.
11010 (add_tree_to_fld_list): Adjust.
11011 * tree-pretty-print.c (dump_generic_node): Make dumping of
11012 type names more robust.
11014 2018-01-15 Richard Biener <rguenther@suse.de>
11016 * BASE-VER: Bump to 8.0.1.
11018 2018-01-14 Martin Sebor <msebor@redhat.com>
11021 * builtins.c (check_access): Avoid warning when the no-warning bit
11024 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11026 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11027 * ira-color (allocno_hard_regs_compare): Likewise.
11029 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11032 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11033 Use .pushsection/.popsection.
11035 2018-01-14 Martin Sebor <msebor@redhat.com>
11038 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11040 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11042 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11043 entry from extra_headers.
11044 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11045 extra_headers, make the list bitwise identical to the i?86-*-* one.
11047 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11049 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11050 -mcmodel=large with -mindirect-branch=thunk,
11051 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11052 -mfunction-return=thunk-extern.
11053 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11054 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11055 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11057 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11059 * config/i386/i386.c (print_reg): Print the name of the full
11060 integer register without '%'.
11061 (ix86_print_operand): Handle 'V'.
11062 * doc/extend.texi: Document 'V' modifier.
11064 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11066 * config/i386/constraints.md (Bs): Disallow memory operand for
11067 -mindirect-branch-register.
11069 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11070 (GOT_memory_operand): Likewise.
11071 (call_insn_operand): Likewise.
11072 (sibcall_insn_operand): Likewise.
11073 (GOT32_symbol_operand): Likewise.
11074 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11075 for -mindirect-branch-register.
11076 (tablejump): Likewise.
11077 (*sibcall_memory): Likewise.
11078 (*sibcall_value_memory): Likewise.
11079 Disallow peepholes of indirect call and jump via memory for
11080 -mindirect-branch-register.
11081 (*call_pop): Replace m with Bw.
11082 (*call_value_pop): Likewise.
11083 (*sibcall_pop_memory): Replace m with Bs.
11084 * config/i386/i386.opt (mindirect-branch-register): New option.
11085 * doc/invoke.texi: Document -mindirect-branch-register option.
11087 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11089 * config/i386/i386-protos.h (ix86_output_function_return): New.
11090 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11091 set function_return_type.
11092 (indirect_thunk_name): Add ret_p to indicate thunk for function
11094 (output_indirect_thunk_function): Pass false to
11095 indirect_thunk_name.
11096 (ix86_output_indirect_branch_via_reg): Likewise.
11097 (ix86_output_indirect_branch_via_push): Likewise.
11098 (output_indirect_thunk_function): Create alias for function
11099 return thunk if regno < 0.
11100 (ix86_output_function_return): New function.
11101 (ix86_handle_fndecl_attribute): Handle function_return.
11102 (ix86_attribute_table): Add function_return.
11103 * config/i386/i386.h (machine_function): Add
11104 function_return_type.
11105 * config/i386/i386.md (simple_return_internal): Use
11106 ix86_output_function_return.
11107 (simple_return_internal_long): Likewise.
11108 * config/i386/i386.opt (mfunction-return=): New option.
11109 (indirect_branch): Mention -mfunction-return=.
11110 * doc/extend.texi: Document function_return function attribute.
11111 * doc/invoke.texi: Document -mfunction-return= option.
11113 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11115 * config/i386/i386-opts.h (indirect_branch): New.
11116 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11117 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11118 with local indirect jump when converting indirect call and jump.
11119 (ix86_set_indirect_branch_type): New.
11120 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11121 (indirectlabelno): New.
11122 (indirect_thunk_needed): Likewise.
11123 (indirect_thunk_bnd_needed): Likewise.
11124 (indirect_thunks_used): Likewise.
11125 (indirect_thunks_bnd_used): Likewise.
11126 (INDIRECT_LABEL): Likewise.
11127 (indirect_thunk_name): Likewise.
11128 (output_indirect_thunk): Likewise.
11129 (output_indirect_thunk_function): Likewise.
11130 (ix86_output_indirect_branch_via_reg): Likewise.
11131 (ix86_output_indirect_branch_via_push): Likewise.
11132 (ix86_output_indirect_branch): Likewise.
11133 (ix86_output_indirect_jmp): Likewise.
11134 (ix86_code_end): Call output_indirect_thunk_function if needed.
11135 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11137 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11138 (ix86_attribute_table): Add indirect_branch.
11139 * config/i386/i386.h (machine_function): Add indirect_branch_type
11140 and has_local_indirect_jump.
11141 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11143 (tablejump): Likewise.
11144 (*indirect_jump): Use ix86_output_indirect_jmp.
11145 (*tablejump_1): Likewise.
11146 (simple_return_indirect_internal): Likewise.
11147 * config/i386/i386.opt (mindirect-branch=): New option.
11148 (indirect_branch): New.
11151 (thunk-inline): Likewise.
11152 (thunk-extern): Likewise.
11153 * doc/extend.texi: Document indirect_branch function attribute.
11154 * doc/invoke.texi: Document -mindirect-branch= option.
11156 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11159 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11161 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11163 * ipa-inline.c (want_inline_small_function_p): Return false if
11164 inlining has already failed with CIF_FINAL_ERROR.
11165 (update_caller_keys): Call want_inline_small_function_p before
11167 (update_callee_keys): Likewise.
11169 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11171 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11173 (rs6000_quadword_masked_address_p): Likewise.
11174 (quad_aligned_load_p): Likewise.
11175 (quad_aligned_store_p): Likewise.
11176 (const_load_sequence_p): Add comment to describe the outer-most loop.
11177 (mimic_memory_attributes_and_flags): New function.
11178 (rs6000_gen_stvx): Likewise.
11179 (replace_swapped_aligned_store): Likewise.
11180 (rs6000_gen_lvx): Likewise.
11181 (replace_swapped_aligned_load): Likewise.
11182 (replace_swapped_load_constant): Capitalize argument name in
11183 comment describing this function.
11184 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11185 and stores that access quad-word aligned addresses and replace
11186 with stvx or lvx instructions when appropriate.
11187 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11188 New function prototype.
11189 (rs6000_quadword_masked_address_p): Likewise.
11190 (rs6000_gen_lvx): Likewise.
11191 (rs6000_gen_stvx): Likewise.
11192 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11193 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11194 when memory address is aligned.
11195 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11196 this split to select lvx instruction when memory address is aligned.
11197 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11198 instruction when memory address is aligned.
11199 (*vsx_le_perm_load_v16qi): Likewise.
11200 (four unnamed splitters): Modify to select the stvx instruction
11201 when memory is aligned.
11203 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11205 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11206 which appears in the queue multiple times.
11208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11209 Alan Hayward <alan.hayward@arm.com>
11210 David Sherwood <david.sherwood@arm.com>
11212 * tree-vectorizer.h (vec_lower_bound): New structure.
11213 (_loop_vec_info): Add check_nonzero and lower_bounds.
11214 (LOOP_VINFO_CHECK_NONZERO): New macro.
11215 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11216 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11217 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11218 fields. Make seg_len the distance travelled, not including the
11220 (dr_direction_indicator): Declare.
11221 (dr_zero_step_indicator): Likewise.
11222 (dr_known_forward_stride_p): Likewise.
11223 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11225 (runtime_alias_check_p): Allow runtime alias checks with
11227 (operator ==): Compare access_size and align.
11228 (prune_runtime_alias_test_list): Rework for new distinction between
11229 the access_size and seg_len.
11230 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11232 (get_segment_min_max): New function.
11233 (create_intersect_range_checks): Use it.
11234 (dr_step_indicator): New function.
11235 (dr_direction_indicator): Likewise.
11236 (dr_zero_step_indicator): Likewise.
11237 (dr_known_forward_stride_p): Likewise.
11238 * tree-loop-distribution.c (data_ref_segment_size): Return
11239 DR_STEP * (niters - 1).
11240 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11242 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11243 (vect_preserves_scalar_order_p): New function, split out from...
11244 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11245 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11246 (vect_vfa_access_size): New function.
11247 (vect_vfa_align): Likewise.
11248 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11249 (dump_lower_bound): New function.
11250 (vect_check_lower_bound): Likewise.
11251 (vect_small_gap_p): Likewise.
11252 (vectorizable_with_step_bound_p): Likewise.
11253 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11254 depencies if the vectorization factor is 1. Convert the checks
11255 for nonzero steps into checks on the bounds of DR_STEP. Try using
11256 a bunds check for variable steps if the minimum required step is
11257 relatively small. Update calls to the dr_with_seg_len
11258 constructor and to vect_compile_time_alias.
11259 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11261 (vect_loop_versioning): Call it.
11262 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11264 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11267 Alan Hayward <alan.hayward@arm.com>
11268 David Sherwood <david.sherwood@arm.com>
11270 * doc/sourcebuild.texi (vect_scatter_store): Document.
11271 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11273 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11275 * genopinit.c (main): Add supports_vec_scatter_store and
11276 supports_vec_scatter_store_cached to target_optabs.
11277 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11278 IFN_MASK_SCATTER_STORE.
11279 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11281 * internal-fn.h (internal_store_fn_p): Declare.
11282 (internal_fn_stored_value_index): Likewise.
11283 * internal-fn.c (scatter_store_direct): New macro.
11284 (expand_scatter_store_optab_fn): New function.
11285 (direct_scatter_store_optab_supported_p): New macro.
11286 (internal_store_fn_p): New function.
11287 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11288 IFN_MASK_SCATTER_STORE.
11289 (internal_fn_mask_index): Likewise.
11290 (internal_fn_stored_value_index): New function.
11291 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11292 for scatter stores.
11293 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11294 * optabs-query.c (supports_vec_scatter_store_p): New function.
11295 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11296 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11297 true for scatter stores.
11298 (vect_gather_scatter_fn_p): Handle scatter stores too.
11299 (vect_check_gather_scatter): Consider using scatter stores if
11300 supports_vec_scatter_store_p.
11301 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11302 scatter stores too.
11303 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11304 internal_fn_stored_value_index.
11305 (check_load_store_masking): Handle scatter stores too.
11306 (vect_get_store_rhs): Make public.
11307 (vectorizable_call): Use internal_store_fn_p.
11308 (vectorizable_store): Handle scatter store internal functions.
11309 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11310 when deciding whether the end of the group has been reached.
11311 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11312 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11313 (mask_scatter_store<mode>): New insns.
11315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11316 Alan Hayward <alan.hayward@arm.com>
11317 David Sherwood <david.sherwood@arm.com>
11319 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11320 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11321 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11323 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11324 Use vect_truncate_gather_scatter_offset if we can't treat the
11325 operation as a normal gather load or scatter store.
11326 (get_group_load_store_type): Take the gather_scatter_info
11327 as argument. Try using a gather load or scatter store for
11328 single-element groups.
11329 (get_load_store_type): Update calls to get_group_load_store_type
11330 and vect_use_strided_gather_scatters_p.
11332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11333 Alan Hayward <alan.hayward@arm.com>
11334 David Sherwood <david.sherwood@arm.com>
11336 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11337 optional tree argument.
11338 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11340 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11341 but continue to use the current value as a fallback.
11342 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11343 to compare the updates.
11344 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11345 (get_load_store_type): Use it when handling a strided access.
11346 (vect_get_strided_load_store_ops): New function.
11347 (vect_get_data_ptr_increment): Likewise.
11348 (vectorizable_load): Handle strided gather loads. Always pass
11349 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11352 Alan Hayward <alan.hayward@arm.com>
11353 David Sherwood <david.sherwood@arm.com>
11355 * doc/md.texi (gather_load@var{m}): Document.
11356 (mask_gather_load@var{m}): Likewise.
11357 * genopinit.c (main): Add supports_vec_gather_load and
11358 supports_vec_gather_load_cached to target_optabs.
11359 * optabs-tree.c (init_tree_optimization_optabs): Use
11360 ggc_cleared_alloc to allocate target_optabs.
11361 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11362 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11364 * internal-fn.h (internal_load_fn_p): Declare.
11365 (internal_gather_scatter_fn_p): Likewise.
11366 (internal_fn_mask_index): Likewise.
11367 (internal_gather_scatter_fn_supported_p): Likewise.
11368 * internal-fn.c (gather_load_direct): New macro.
11369 (expand_gather_load_optab_fn): New function.
11370 (direct_gather_load_optab_supported_p): New macro.
11371 (direct_internal_fn_optab): New function.
11372 (internal_load_fn_p): Likewise.
11373 (internal_gather_scatter_fn_p): Likewise.
11374 (internal_fn_mask_index): Likewise.
11375 (internal_gather_scatter_fn_supported_p): Likewise.
11376 * optabs-query.c (supports_at_least_one_mode_p): New function.
11377 (supports_vec_gather_load_p): Likewise.
11378 * optabs-query.h (supports_vec_gather_load_p): Declare.
11379 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11380 and memory_type field.
11381 (NUM_PATTERNS): Bump to 15.
11382 * tree-vect-data-refs.c: Include internal-fn.h.
11383 (vect_gather_scatter_fn_p): New function.
11384 (vect_describe_gather_scatter_call): Likewise.
11385 (vect_check_gather_scatter): Try using internal functions for
11386 gather loads. Recognize existing calls to a gather load function.
11387 (vect_analyze_data_refs): Consider using gather loads if
11388 supports_vec_gather_load_p.
11389 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11390 (vect_get_gather_scatter_offset_type): Likewise.
11391 (vect_convert_mask_for_vectype): Likewise.
11392 (vect_add_conversion_to_patterm): Likewise.
11393 (vect_try_gather_scatter_pattern): Likewise.
11394 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11395 (vect_vect_recog_func_ptrs): Add it.
11396 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11397 internal_fn_mask_index and internal_gather_scatter_fn_p.
11398 (check_load_store_masking): Take the gather_scatter_info as an
11399 argument and handle gather loads.
11400 (vect_get_gather_scatter_ops): New function.
11401 (vectorizable_call): Check internal_load_fn_p.
11402 (vectorizable_load): Likewise. Handle gather load internal
11404 (vectorizable_store): Update call to check_load_store_masking.
11405 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11406 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11407 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11408 (aarch64_gather_scale_operand_d): New predicates.
11409 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11410 (mask_gather_load<mode>): New insns.
11412 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11413 Alan Hayward <alan.hayward@arm.com>
11414 David Sherwood <david.sherwood@arm.com>
11416 * optabs.def (fold_left_plus_optab): New optab.
11417 * doc/md.texi (fold_left_plus_@var{m}): Document.
11418 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11419 * internal-fn.c (fold_left_direct): Define.
11420 (expand_fold_left_optab_fn): Likewise.
11421 (direct_fold_left_optab_supported_p): Likewise.
11422 * fold-const-call.c (fold_const_fold_left): New function.
11423 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11424 * tree-parloops.c (valid_reduction_p): New function.
11425 (gather_scalar_reductions): Use it.
11426 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11427 (vect_finish_replace_stmt): Declare.
11428 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11429 (needs_fold_left_reduction_p): New function, split out from...
11430 (vect_is_simple_reduction): ...here. Accept reductions that
11431 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11432 (vect_force_simple_reduction): Also store the reduction type in
11433 the assignment's STMT_VINFO_REDUC_TYPE.
11434 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11435 (merge_with_identity): New function.
11436 (vect_expand_fold_left): Likewise.
11437 (vectorize_fold_left_reduction): Likewise.
11438 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11439 scalar phi in place for it. Check for target support and reject
11440 cases that would reassociate the operation. Defer the transform
11441 phase to vectorize_fold_left_reduction.
11442 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11443 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11444 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11448 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11449 call to ifc_temp_var.
11451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11452 Alan Hayward <alan.hayward@arm.com>
11453 David Sherwood <david.sherwood@arm.com>
11455 * target.def (legitimize_address_displacement): Take the original
11456 offset as a poly_int.
11457 * targhooks.h (default_legitimize_address_displacement): Update
11459 * targhooks.c (default_legitimize_address_displacement): Likewise.
11460 * doc/tm.texi: Regenerate.
11461 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11462 as an argument, moving assert of ad->disp == ad->disp_term to...
11463 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11464 Try calling targetm.legitimize_address_displacement before expanding
11465 the address rather than afterwards, and adjust for the new interface.
11466 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11467 Match the new hook interface. Handle SVE addresses.
11468 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11469 new hook interface.
11471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11473 * Makefile.in (OBJS): Add early-remat.o.
11474 * target.def (select_early_remat_modes): New hook.
11475 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11476 * doc/tm.texi: Regenerate.
11477 * targhooks.h (default_select_early_remat_modes): Declare.
11478 * targhooks.c (default_select_early_remat_modes): New function.
11479 * timevar.def (TV_EARLY_REMAT): New timevar.
11480 * passes.def (pass_early_remat): New pass.
11481 * tree-pass.h (make_pass_early_remat): Declare.
11482 * early-remat.c: New file.
11483 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11485 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11488 Alan Hayward <alan.hayward@arm.com>
11489 David Sherwood <david.sherwood@arm.com>
11491 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11492 vfm1 with a bound_epilog parameter.
11493 (vect_do_peeling): Update calls accordingly, and move the prologue
11494 call earlier in the function. Treat the base bound_epilog as 0 for
11495 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11496 this base when peeling for gaps.
11497 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11498 with fully-masked loops.
11499 (vect_estimate_min_profitable_iters): Handle the single peeled
11500 iteration in that case.
11502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11503 Alan Hayward <alan.hayward@arm.com>
11504 David Sherwood <david.sherwood@arm.com>
11506 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11507 single-element interleaving even if the size is not a power of 2.
11508 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11509 accesses for single-element interleaving if the group size is
11512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11513 Alan Hayward <alan.hayward@arm.com>
11514 David Sherwood <david.sherwood@arm.com>
11516 * doc/md.texi (fold_extract_last_@var{m}): Document.
11517 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11518 * optabs.def (fold_extract_last_optab): New optab.
11519 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11520 * internal-fn.c (fold_extract_direct): New macro.
11521 (expand_fold_extract_optab_fn): Likewise.
11522 (direct_fold_extract_optab_supported_p): Likewise.
11523 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11524 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11525 EXTRACT_LAST_REDUCTION.
11526 (get_initial_def_for_reduction): Do not create an initial vector
11527 for EXTRACT_LAST_REDUCTION reductions.
11528 (vectorizable_reduction): Leave the scalar phi in place for
11529 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11530 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11531 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11532 transform phase to vectorizable_condition.
11533 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11535 (vect_finish_stmt_generation): ...here.
11536 (vect_finish_replace_stmt): New function.
11537 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11538 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11540 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11543 Alan Hayward <alan.hayward@arm.com>
11544 David Sherwood <david.sherwood@arm.com>
11546 * doc/md.texi (extract_last_@var{m}): Document.
11547 * optabs.def (extract_last_optab): New optab.
11548 * internal-fn.def (EXTRACT_LAST): New internal function.
11549 * internal-fn.c (cond_unary_direct): New macro.
11550 (expand_cond_unary_optab_fn): Likewise.
11551 (direct_cond_unary_optab_supported_p): Likewise.
11552 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11553 loops using EXTRACT_LAST.
11554 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11555 (extract_last_<mode>): ...this optab.
11556 (vec_extract<mode><Vel>): Update accordingly.
11558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11559 Alan Hayward <alan.hayward@arm.com>
11560 David Sherwood <david.sherwood@arm.com>
11562 * target.def (empty_mask_is_expensive): New hook.
11563 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11564 * doc/tm.texi: Regenerate.
11565 * targhooks.h (default_empty_mask_is_expensive): Declare.
11566 * targhooks.c (default_empty_mask_is_expensive): New function.
11567 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11568 if the target says that empty masks are expensive.
11569 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11571 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11574 Alan Hayward <alan.hayward@arm.com>
11575 David Sherwood <david.sherwood@arm.com>
11577 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11578 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11579 (vect_use_loop_mask_for_alignment_p): New function.
11580 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11581 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11582 niters_skip argument. Make sure that the first niters_skip elements
11583 of the first iteration are inactive.
11584 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11585 Update call to vect_set_loop_masks_directly.
11586 (get_misalign_in_elems): New function, split out from...
11587 (vect_gen_prolog_loop_niters): ...here.
11588 (vect_update_init_of_dr): Take a code argument that specifies whether
11589 the adjustment should be added or subtracted.
11590 (vect_update_init_of_drs): Likewise.
11591 (vect_prepare_for_masked_peels): New function.
11592 (vect_do_peeling): Skip prologue peeling if we're using a mask
11593 instead. Update call to vect_update_inits_of_drs.
11594 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11596 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11597 alignment. Do not include the number of peeled iterations in
11598 the minimum threshold in that case.
11599 (vectorizable_induction): Adjust the start value down by
11600 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11601 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11602 Take the number of skipped iterations into account when calculating
11604 * tree-vect-stmts.c (vect_gen_while_not): New function.
11606 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11607 Alan Hayward <alan.hayward@arm.com>
11608 David Sherwood <david.sherwood@arm.com>
11610 * doc/sourcebuild.texi (vect_fully_masked): Document.
11611 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11612 default value to 0.
11613 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11615 (vect_analyze_loop_2): ...here. Don't check the vectorization
11616 factor against the number of loop iterations if the loop is
11619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11620 Alan Hayward <alan.hayward@arm.com>
11621 David Sherwood <david.sherwood@arm.com>
11623 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11624 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11625 (dump_groups): Update accordingly.
11626 (iv_use::mem_type): New member variable.
11627 (address_p): New function.
11628 (record_use): Add a mem_type argument and initialize the new
11630 (record_group_use): Add a mem_type argument. Use address_p.
11631 Remove obsolete null checks of base_object. Update call to record_use.
11632 (find_interesting_uses_op): Update call to record_group_use.
11633 (find_interesting_uses_cond): Likewise.
11634 (find_interesting_uses_address): Likewise.
11635 (get_mem_type_for_internal_fn): New function.
11636 (find_address_like_use): Likewise.
11637 (find_interesting_uses_stmt): Try find_address_like_use before
11638 calling find_interesting_uses_op.
11639 (addr_offset_valid_p): Use the iv mem_type field as the type
11640 of the addressed memory.
11641 (add_autoinc_candidates): Likewise.
11642 (get_address_cost): Likewise.
11643 (split_small_address_groups_p): Use address_p.
11644 (split_address_groups): Likewise.
11645 (add_iv_candidate_for_use): Likewise.
11646 (autoinc_possible_for_pair): Likewise.
11647 (rewrite_groups): Likewise.
11648 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11649 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11650 (get_alias_ptr_type_for_ptr_address): New function.
11651 (rewrite_use_address): Rewrite address uses in calls that were
11652 identified by find_address_like_use.
11654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11655 Alan Hayward <alan.hayward@arm.com>
11656 David Sherwood <david.sherwood@arm.com>
11658 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11660 * gimple-expr.h (is_gimple_addressable: Likewise.
11661 * gimple-expr.c (is_gimple_address): Likewise.
11662 * internal-fn.c (expand_call_mem_ref): New function.
11663 (expand_mask_load_optab_fn): Use it.
11664 (expand_mask_store_optab_fn): Likewise.
11666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11667 Alan Hayward <alan.hayward@arm.com>
11668 David Sherwood <david.sherwood@arm.com>
11670 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11671 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11672 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11673 (cond_umax@var{mode}): Document.
11674 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11675 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11676 (cond_umin_optab, cond_umax_optab): New optabs.
11677 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11678 (COND_IOR, COND_XOR): New internal functions.
11679 * internal-fn.h (get_conditional_internal_fn): Declare.
11680 * internal-fn.c (cond_binary_direct): New macro.
11681 (expand_cond_binary_optab_fn): Likewise.
11682 (direct_cond_binary_optab_supported_p): Likewise.
11683 (get_conditional_internal_fn): New function.
11684 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11685 Cope with reduction statements that are vectorized as calls rather
11687 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11688 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11689 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11690 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11691 (UNSPEC_COND_EOR): New unspecs.
11692 (optab): Add mappings for them.
11693 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11694 (sve_int_op, sve_fp_op): New int attributes.
11696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11697 Alan Hayward <alan.hayward@arm.com>
11698 David Sherwood <david.sherwood@arm.com>
11700 * optabs.def (while_ult_optab): New optab.
11701 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11702 * internal-fn.def (WHILE_ULT): New internal function.
11703 * internal-fn.h (direct_internal_fn_supported_p): New override
11704 that takes two types as argument.
11705 * internal-fn.c (while_direct): New macro.
11706 (expand_while_optab_fn): New function.
11707 (convert_optab_supported_p): Likewise.
11708 (direct_while_optab_supported_p): New macro.
11709 * wide-int.h (wi::udiv_ceil): New function.
11710 * tree-vectorizer.h (rgroup_masks): New structure.
11711 (vec_loop_masks): New typedef.
11712 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11713 and fully_masked_p.
11714 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11715 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11716 (vect_max_vf): New function.
11717 (slpeel_make_loop_iterate_ntimes): Delete.
11718 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11719 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11720 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11721 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11722 internal-fn.h, stor-layout.h and optabs-query.h.
11723 (vect_set_loop_mask): New function.
11724 (add_preheader_seq): Likewise.
11725 (add_header_seq): Likewise.
11726 (interleave_supported_p): Likewise.
11727 (vect_maybe_permute_loop_masks): Likewise.
11728 (vect_set_loop_masks_directly): Likewise.
11729 (vect_set_loop_condition_masked): Likewise.
11730 (vect_set_loop_condition_unmasked): New function, split out from
11731 slpeel_make_loop_iterate_ntimes.
11732 (slpeel_make_loop_iterate_ntimes): Rename to..
11733 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11734 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11735 (vect_do_peeling): Update call accordingly.
11736 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11738 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11739 mask_compare_type, can_fully_mask_p and fully_masked_p.
11740 (release_vec_loop_masks): New function.
11741 (_loop_vec_info): Use it to free the loop masks.
11742 (can_produce_all_loop_masks_p): New function.
11743 (vect_get_max_nscalars_per_iter): Likewise.
11744 (vect_verify_full_masking): Likewise.
11745 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11746 retries, and free the mask rgroups before retrying. Check loop-wide
11747 reasons for disallowing fully-masked loops. Make the final decision
11748 about whether use a fully-masked loop or not.
11749 (vect_estimate_min_profitable_iters): Do not assume that peeling
11750 for the number of iterations will be needed for fully-masked loops.
11751 (vectorizable_reduction): Disable fully-masked loops.
11752 (vectorizable_live_operation): Likewise.
11753 (vect_halve_mask_nunits): New function.
11754 (vect_double_mask_nunits): Likewise.
11755 (vect_record_loop_mask): Likewise.
11756 (vect_get_loop_mask): Likewise.
11757 (vect_transform_loop): Handle the case in which the final loop
11758 iteration might handle a partial vector. Call vect_set_loop_condition
11759 instead of slpeel_make_loop_iterate_ntimes.
11760 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11761 (check_load_store_masking): New function.
11762 (prepare_load_store_mask): Likewise.
11763 (vectorizable_store): Handle fully-masked loops.
11764 (vectorizable_load): Likewise.
11765 (supportable_widening_operation): Use vect_halve_mask_nunits for
11767 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11768 (vect_gen_while): New function.
11769 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11770 (aarch64_uqdec<mode>): New insn.
11772 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11773 Alan Hayward <alan.hayward@arm.com>
11774 David Sherwood <david.sherwood@arm.com>
11776 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11777 (reduc_xor_scal_optab): New optabs.
11778 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11779 (reduc_xor_scal_@var{m}): Document.
11780 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11781 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11782 internal functions.
11783 * fold-const-call.c (fold_const_call): Handle them.
11784 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11785 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11786 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11787 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11788 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11789 (UNSPEC_XORV): New unspecs.
11790 (optab): Add entries for them.
11791 (BITWISEV): New int iterator.
11792 (bit_reduc_op): New int attributes.
11794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11795 Alan Hayward <alan.hayward@arm.com>
11796 David Sherwood <david.sherwood@arm.com>
11798 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11799 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11800 * optabs.def (vec_shl_insert_optab): New optab.
11801 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11802 (duplicate_and_interleave): Likewise.
11803 * tree-vect-loop.c: Include internal-fn.h.
11804 (neutral_op_for_slp_reduction): New function, split out from
11805 get_initial_defs_for_reduction.
11806 (get_initial_def_for_reduction): Handle option 2 for variable-length
11807 vectors by loading the neutral value into a vector and then shifting
11808 the initial value into element 0.
11809 (get_initial_defs_for_reduction): Replace the code argument with
11810 the neutral value calculated by neutral_op_for_slp_reduction.
11811 Use gimple_build_vector for constant-length vectors.
11812 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11813 but the first group_size elements have a neutral value.
11814 Use duplicate_and_interleave otherwise.
11815 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11816 Update call to get_initial_defs_for_reduction. Handle SLP
11817 reductions for variable-length vectors by creating one vector
11818 result for each scalar result, with the elements associated
11819 with other scalar results stubbed out with the neutral value.
11820 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11821 Require IFN_VEC_SHL_INSERT for double reductions on
11822 variable-length vectors, or SLP reductions that have
11823 a neutral value. Require can_duplicate_and_interleave_p
11824 support for variable-length unchained SLP reductions if there
11825 is no neutral value, such as for MIN/MAX reductions. Also require
11826 the number of vector elements to be a multiple of the number of
11827 SLP statements when doing variable-length unchained SLP reductions.
11828 Update call to vect_create_epilog_for_reduction.
11829 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11830 and remove initial values.
11831 (duplicate_and_interleave): Make public.
11832 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11833 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11836 Alan Hayward <alan.hayward@arm.com>
11837 David Sherwood <david.sherwood@arm.com>
11839 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11840 (can_duplicate_and_interleave_p): New function.
11841 (vect_get_and_check_slp_defs): Take the vector of statements
11842 rather than just the current one. Remove excess parentheses.
11843 Restriction rejectinon of vect_constant_def and vect_external_def
11844 for variable-length vectors to boolean types, or types for which
11845 can_duplicate_and_interleave_p is false.
11846 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11847 (duplicate_and_interleave): New function.
11848 (vect_get_constant_vectors): Use gimple_build_vector for
11849 constant-length vectors and suitable variable-length constant
11850 vectors. Use duplicate_and_interleave for other variable-length
11851 vectors. Don't defer the update when inserting new statements.
11853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11854 Alan Hayward <alan.hayward@arm.com>
11855 David Sherwood <david.sherwood@arm.com>
11857 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11858 min_profitable_iters doesn't go negative.
11860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11861 Alan Hayward <alan.hayward@arm.com>
11862 David Sherwood <david.sherwood@arm.com>
11864 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11865 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11866 * optabs.def (vec_mask_load_lanes_optab): New optab.
11867 (vec_mask_store_lanes_optab): Likewise.
11868 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11869 (MASK_STORE_LANES): Likewise.
11870 * internal-fn.c (mask_load_lanes_direct): New macro.
11871 (mask_store_lanes_direct): Likewise.
11872 (expand_mask_load_optab_fn): Handle masked operations.
11873 (expand_mask_load_lanes_optab_fn): New macro.
11874 (expand_mask_store_optab_fn): Handle masked operations.
11875 (expand_mask_store_lanes_optab_fn): New macro.
11876 (direct_mask_load_lanes_optab_supported_p): Likewise.
11877 (direct_mask_store_lanes_optab_supported_p): Likewise.
11878 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11880 (vect_load_lanes_supported): Likewise.
11881 * tree-vect-data-refs.c (strip_conversion): New function.
11882 (can_group_stmts_p): Likewise.
11883 (vect_analyze_data_ref_accesses): Use it instead of checking
11884 for a pair of assignments.
11885 (vect_store_lanes_supported): Take a masked_p parameter.
11886 (vect_load_lanes_supported): Likewise.
11887 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11888 vect_store_lanes_supported and vect_load_lanes_supported.
11889 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11890 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11891 parameter. Don't allow gaps for masked accesses.
11892 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11893 and vect_load_lanes_supported.
11894 (get_load_store_type): Take a masked_p parameter and update
11895 call to get_group_load_store_type.
11896 (vectorizable_store): Update call to get_load_store_type.
11897 Handle IFN_MASK_STORE_LANES.
11898 (vectorizable_load): Update call to get_load_store_type.
11899 Handle IFN_MASK_LOAD_LANES.
11901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11902 Alan Hayward <alan.hayward@arm.com>
11903 David Sherwood <david.sherwood@arm.com>
11905 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11907 * config/aarch64/aarch64-protos.h
11908 (aarch64_sve_struct_memory_operand_p): Declare.
11909 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11910 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11911 (VPRED, vpred): Handle SVE structure modes.
11912 * config/aarch64/constraints.md (Utx): New constraint.
11913 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11914 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11915 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11916 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11917 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11918 structure modes. Split into pieces after RA.
11919 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11920 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11922 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11923 SVE structure modes.
11924 (aarch64_classify_address): Likewise.
11925 (sizetochar): Move earlier in file.
11926 (aarch64_print_operand): Handle SVE register lists.
11927 (aarch64_array_mode): New function.
11928 (aarch64_sve_struct_memory_operand_p): Likewise.
11929 (TARGET_ARRAY_MODE): Redefine.
11931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11932 Alan Hayward <alan.hayward@arm.com>
11933 David Sherwood <david.sherwood@arm.com>
11935 * target.def (array_mode): New target hook.
11936 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11937 * doc/tm.texi: Regenerate.
11938 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11939 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11940 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11941 targetm.array_mode.
11942 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11946 Alan Hayward <alan.hayward@arm.com>
11947 David Sherwood <david.sherwood@arm.com>
11949 * fold-const.c (fold_binary_loc): Check the argument types
11950 rather than the result type when testing for a vector operation.
11952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11954 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11955 * doc/tm.texi: Regenerate.
11957 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11958 Alan Hayward <alan.hayward@arm.com>
11959 David Sherwood <david.sherwood@arm.com>
11961 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11962 (sve): Document new AArch64 extension.
11963 * doc/md.texi (w): Extend the description of the AArch64
11964 constraint to include SVE vectors.
11965 (Upl, Upa): Document new AArch64 predicate constraints.
11966 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11968 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11969 (msve-vector-bits=): New option.
11970 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11971 SVE when these are disabled.
11972 (sve): New extension.
11973 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11974 modes. Adjust their number of units based on aarch64_sve_vg.
11975 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11976 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11977 aarch64_addr_query_type.
11978 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11979 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11980 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11981 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11982 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11983 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11984 (aarch64_simd_imm_zero_p): Delete.
11985 (aarch64_check_zero_based_sve_index_immediate): Declare.
11986 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11987 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11988 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11989 (aarch64_sve_float_mul_immediate_p): Likewise.
11990 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11991 rather than an rtx.
11992 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11993 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11994 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11995 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11996 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11997 (aarch64_regmode_natural_size): Likewise.
11998 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11999 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12001 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12002 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12003 for VG and the SVE predicate registers.
12004 (V_ALIASES): Add a "z"-prefixed alias.
12005 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12006 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12007 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12008 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12009 (REG_CLASS_NAMES): Add entries for them.
12010 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12011 and the predicate registers.
12012 (aarch64_sve_vg): Declare.
12013 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12014 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12015 (REGMODE_NATURAL_SIZE): Define.
12016 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12018 * config/aarch64/aarch64.c: Include cfgrtl.h.
12019 (simd_immediate_info): Add a constructor for series vectors,
12020 and an associated step field.
12021 (aarch64_sve_vg): New variable.
12022 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12023 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12024 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12025 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12026 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12027 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12028 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12029 (aarch64_get_mask_mode): New functions.
12030 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12031 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12032 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12033 predicate modes and predicate registers. Explicitly restrict
12034 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12035 to store a vector mode if it is recognized by
12036 aarch64_classify_vector_mode.
12037 (aarch64_regmode_natural_size): New function.
12038 (aarch64_hard_regno_caller_save_mode): Return the original mode
12040 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12041 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12042 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12043 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12045 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12046 does not overlap dest if the function is frame-related. Handle
12048 (aarch64_split_add_offset): New function.
12049 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12050 them aarch64_add_offset.
12051 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12052 and update call to aarch64_sub_sp.
12053 (aarch64_add_cfa_expression): New function.
12054 (aarch64_expand_prologue): Pass extra temporary registers to the
12055 functions above. Handle the case in which we need to emit new
12056 DW_CFA_expressions for registers that were originally saved
12057 relative to the stack pointer, but now have to be expressed
12058 relative to the frame pointer.
12059 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12061 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12062 IP0 and IP1 values for SVE frames.
12063 (aarch64_expand_vec_series): New function.
12064 (aarch64_expand_sve_widened_duplicate): Likewise.
12065 (aarch64_expand_sve_const_vector): Likewise.
12066 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12067 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12068 into the register, rather than emitting a SET directly.
12069 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12070 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12071 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12072 (offset_9bit_signed_scaled_p): New functions.
12073 (aarch64_replicate_bitmask_imm): New function.
12074 (aarch64_bitmask_imm): Use it.
12075 (aarch64_cannot_force_const_mem): Reject expressions involving
12076 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12077 (aarch64_classify_index): Handle SVE indices, by requiring
12078 a plain register index with a scale that matches the element size.
12079 (aarch64_classify_address): Handle SVE addresses. Assert that
12080 the mode of the address is VOIDmode or an integer mode.
12081 Update call to aarch64_classify_symbol.
12082 (aarch64_classify_symbolic_expression): Update call to
12083 aarch64_classify_symbol.
12084 (aarch64_const_vec_all_in_range_p): New function.
12085 (aarch64_print_vector_float_operand): Likewise.
12086 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12087 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12088 and the FP immediates 1.0 and 0.5.
12089 (aarch64_print_address_internal): Handle SVE addresses.
12090 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12091 (aarch64_regno_regclass): Handle predicate registers.
12092 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12094 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12095 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12096 (aarch64_convert_sve_vector_bits): New function.
12097 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12098 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12099 rather than an rtx.
12100 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12101 Handle SVE vector and predicate modes. Accept VL-based constants
12102 that need only one temporary register, and VL offsets that require
12103 no temporary registers.
12104 (aarch64_conditional_register_usage): Mark the predicate registers
12105 as fixed if SVE isn't available.
12106 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12107 Return true for SVE vector and predicate modes.
12108 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12109 rather than an unsigned int. Handle SVE modes.
12110 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12112 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12114 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12115 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12116 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12117 (aarch64_sve_float_mul_immediate_p): New functions.
12118 (aarch64_sve_valid_immediate): New function.
12119 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12120 Explicitly reject structure modes. Check for INDEX constants.
12121 Handle PTRUE and PFALSE constants.
12122 (aarch64_check_zero_based_sve_index_immediate): New function.
12123 (aarch64_simd_imm_zero_p): Delete.
12124 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12125 vector modes. Accept constants in the range of CNT[BHWD].
12126 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12127 ask for an Advanced SIMD mode.
12128 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12129 (aarch64_simd_vector_alignment): Handle SVE predicates.
12130 (aarch64_vectorize_preferred_vector_alignment): New function.
12131 (aarch64_simd_vector_alignment_reachable): Use it instead of
12133 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12134 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12136 (MAX_VECT_LEN): Delete.
12137 (expand_vec_perm_d): Add a vec_flags field.
12138 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12139 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12140 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12142 (aarch64_evpc_rev): Rename to...
12143 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12144 (aarch64_evpc_rev_global): New function.
12145 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12146 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12148 (aarch64_evpc_sve_tbl): New function.
12149 (aarch64_expand_vec_perm_const_1): Update after rename of
12150 aarch64_evpc_rev. Handle SVE permutes too, trying
12151 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12152 than aarch64_evpc_tbl.
12153 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12154 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12155 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12156 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12157 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12158 (aarch64_expand_sve_vcond): New functions.
12159 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12160 of aarch64_vector_mode_p.
12161 (aarch64_dwarf_poly_indeterminate_value): New function.
12162 (aarch64_compute_pressure_classes): Likewise.
12163 (aarch64_can_change_mode_class): Likewise.
12164 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12165 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12166 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12167 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12168 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12169 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12170 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12171 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12173 (Dn, Dl, Dr): Accept const as well as const_vector.
12174 (Dz): Likewise. Compare against CONST0_RTX.
12175 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12176 of "vector" where appropriate.
12177 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12178 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12179 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12180 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12181 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12182 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12183 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12184 (v_int_equiv): Extend to SVE modes.
12185 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12187 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12188 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12189 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12190 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12191 (SVE_COND_FP_CMP): New int iterators.
12192 (perm_hilo): Handle the new unpack unspecs.
12193 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12195 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12196 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12197 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12198 (aarch64_equality_operator, aarch64_constant_vector_operand)
12199 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12200 (aarch64_sve_nonimmediate_operand): Likewise.
12201 (aarch64_sve_general_operand): Likewise.
12202 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12203 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12204 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12205 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12206 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12207 (aarch64_sve_float_arith_immediate): Likewise.
12208 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12209 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12210 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12211 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12212 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12213 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12214 (aarch64_sve_float_arith_operand): Likewise.
12215 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12216 (aarch64_sve_float_mul_operand): Likewise.
12217 (aarch64_sve_vec_perm_operand): Likewise.
12218 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12219 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12220 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12221 as well as const_vector.
12222 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12223 in file. Use CONST0_RTX and CONSTM1_RTX.
12224 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12225 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12226 Use aarch64_simd_imm_zero.
12227 * config/aarch64/aarch64-sve.md: New file.
12228 * config/aarch64/aarch64.md: Include it.
12229 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12230 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12231 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12232 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12233 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12234 (sve): New attribute.
12235 (enabled): Disable instructions with the sve attribute unless
12237 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12238 aarch64_expand_mov_immediate.
12239 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12240 CNT[BHSD] immediates.
12241 (movti): Split CONST_POLY_INT moves into two halves.
12242 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12243 Split additions that need a temporary here if the destination
12244 is the stack pointer.
12245 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12246 (*add<mode>3_poly_1): New instruction.
12247 (set_clobber_cc): New expander.
12249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12251 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12252 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12253 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12254 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12255 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12256 Change innermode from fixed_mode_size to machine_mode.
12257 (simplify_subreg): Update call accordingly. Handle a constant-sized
12258 subreg of a variable-length CONST_VECTOR.
12260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12261 Alan Hayward <alan.hayward@arm.com>
12262 David Sherwood <david.sherwood@arm.com>
12264 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12265 (add_offset_to_base): New function, split out from...
12266 (create_mem_ref): ...here. When handling a scale other than 1,
12267 check first whether the address is valid without the offset.
12268 Add it into the base if so, leaving the index and scale as-is.
12270 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12273 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12274 fold_for_warn before checking if arg2 is INTEGER_CST.
12276 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12278 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12279 (store_multiple_operation): Delete.
12280 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12281 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12282 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12283 guarded by TARGET_STRING.
12284 (rs6000_output_load_multiple): Delete.
12285 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12286 OPTION_MASK_STRING / TARGET_STRING handling.
12287 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12288 (const rs6000_opt_masks) <"string">: Change mask to 0.
12289 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12290 (MASK_STRING): Delete.
12291 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12293 (load_multiple): Delete.
12300 (store_multiple): Delete.
12307 (movmemsi_8reg): Delete.
12308 (corresponding unnamed define_insn): Delete.
12309 (movmemsi_6reg): Delete.
12310 (corresponding unnamed define_insn): Delete.
12311 (movmemsi_4reg): Delete.
12312 (corresponding unnamed define_insn): Delete.
12313 (movmemsi_2reg): Delete.
12314 (corresponding unnamed define_insn): Delete.
12315 (movmemsi_1reg): Delete.
12316 (corresponding unnamed define_insn): Delete.
12317 * config/rs6000/rs6000.opt (mno-string): New.
12318 (mstring): Replace by deprecation warning stub.
12319 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12321 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12323 * regrename.c (regrename_do_replace): If replacing the same
12324 reg multiple times, try to reuse last created gen_raw_REG.
12327 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12328 main to workaround a bug in GDB.
12330 2018-01-12 Tom de Vries <tom@codesourcery.com>
12333 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12335 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12337 PR rtl-optimization/80481
12338 * ira-color.c (get_cap_member): New function.
12339 (allocnos_conflict_by_live_ranges_p): Use it.
12340 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12341 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12343 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12346 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12347 (*saddl_se_1): Ditto.
12348 (*ssubsi_1): Ditto.
12349 (*ssubl_se_1): Ditto.
12351 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12353 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12354 rather than wi::to_widest for DR_INITs.
12355 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12356 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12357 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12359 (vect_analyze_group_access_1): Note that here.
12361 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12363 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12364 polynomial type sizes.
12366 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12368 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12369 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12370 (gimple_add_tmp_var): Likewise.
12372 2018-01-12 Martin Liska <mliska@suse.cz>
12374 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12375 (gimple_alloc_sizes): Likewise.
12376 (dump_gimple_statistics): Use PRIu64 in printf format.
12377 * gimple.h: Change uint64_t to int.
12379 2018-01-12 Martin Liska <mliska@suse.cz>
12381 * tree-core.h: Use uint64_t instead of int.
12382 * tree.c (tree_node_counts): Likewise.
12383 (tree_node_sizes): Likewise.
12384 (dump_tree_statistics): Use PRIu64 in printf format.
12386 2018-01-12 Martin Liska <mliska@suse.cz>
12388 * Makefile.in: As qsort_chk is implemented in vec.c, add
12389 vec.o to linkage of gencfn-macros.
12390 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12391 passing the info to record_node_allocation_statistics.
12392 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12394 * ggc-common.c (struct ggc_usage): Add operator== and use
12395 it in operator< and compare function.
12396 * mem-stats.h (struct mem_usage): Likewise.
12397 * vec.c (struct vec_usage): Remove operator< and compare
12398 function. Can be simply inherited.
12400 2018-01-12 Martin Jambor <mjambor@suse.cz>
12403 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12404 * tree-ssa-math-opts.c: Include domwalk.h.
12405 (convert_mult_to_fma_1): New function.
12406 (fma_transformation_info): New type.
12407 (fma_deferring_state): Likewise.
12408 (cancel_fma_deferring): New function.
12409 (result_of_phi): Likewise.
12410 (last_fma_candidate_feeds_initial_phi): Likewise.
12411 (convert_mult_to_fma): Added deferring logic, split actual
12412 transformation to convert_mult_to_fma_1.
12413 (math_opts_dom_walker): New type.
12414 (math_opts_dom_walker::after_dom_children): New method, body moved
12415 here from pass_optimize_widening_mul::execute, added deferring logic
12417 (pass_optimize_widening_mul::execute): Moved most of code to
12418 math_opts_dom_walker::after_dom_children.
12419 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12420 * config/i386/i386.c (ix86_option_override_internal): Added
12421 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12423 2018-01-12 Richard Biener <rguenther@suse.de>
12426 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12427 inline instance vars.
12429 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12432 * config/rx/rx.c (rx_is_restricted_memory_address):
12433 Handle SUBREG case.
12435 2018-01-12 Richard Biener <rguenther@suse.de>
12437 PR tree-optimization/80846
12438 * target.def (split_reduction): New target hook.
12439 * targhooks.c (default_split_reduction): New function.
12440 * targhooks.h (default_split_reduction): Declare.
12441 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12442 target requests first reduce vectors by combining low and high
12444 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12445 (get_vectype_for_scalar_type_and_size): Export.
12446 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12447 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12448 * doc/tm.texi: Regenerate.
12449 * config/i386/i386.c (ix86_split_reduction): Implement
12450 TARGET_VECTORIZE_SPLIT_REDUCTION.
12452 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12455 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12456 in PIC mode except for TARGET_VXWORKS_RTP.
12457 * config/sparc/sparc.c: Include cfgrtl.h.
12458 (TARGET_INIT_PIC_REG): Define.
12459 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12460 (sparc_pic_register_p): New predicate.
12461 (sparc_legitimate_address_p): Use it.
12462 (sparc_legitimize_pic_address): Likewise.
12463 (sparc_delegitimize_address): Likewise.
12464 (sparc_mode_dependent_address_p): Likewise.
12465 (gen_load_pcrel_sym): Remove 4th parameter.
12466 (load_got_register): Adjust call to above. Remove obsolete stuff.
12467 (sparc_expand_prologue): Do not call load_got_register here.
12468 (sparc_flat_expand_prologue): Likewise.
12469 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12470 (sparc_use_pseudo_pic_reg): New function.
12471 (sparc_init_pic_reg): Likewise.
12472 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12473 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12475 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12477 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12478 Add item for branch_cost.
12480 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12482 PR rtl-optimization/83565
12483 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12484 not extend the result to a larger mode for rotate operations.
12485 (num_sign_bit_copies1): Likewise.
12487 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12490 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12492 Use values-Xc.o for -pedantic.
12493 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12495 2018-01-12 Martin Liska <mliska@suse.cz>
12498 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12500 (possible_polymorphic_call_targets): Use it.
12501 (ipa_devirt): Likewise.
12503 2018-01-12 Martin Liska <mliska@suse.cz>
12505 * profile-count.h (enum profile_quality): Use 0 as invalid
12506 enum value of profile_quality.
12508 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12510 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12511 -mext-string options.
12513 2018-01-12 Richard Biener <rguenther@suse.de>
12515 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12516 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12517 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12519 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12521 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12523 * configure.ac (--with-long-double-format): Add support for the
12524 configuration option to change the default long double format on
12526 * config.gcc (powerpc*-linux*-*): Likewise.
12527 * configure: Regenerate.
12528 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12529 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12530 used without modification.
12532 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12534 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12535 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12536 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12537 MISC_BUILTIN_SPEC_BARRIER.
12538 (rs6000_init_builtins): Likewise.
12539 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12541 (speculation_barrier): New define_insn.
12542 * doc/extend.texi: Document __builtin_speculation_barrier.
12544 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12547 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12548 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12549 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12551 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12552 integral modes instead of "ss" and "sd".
12553 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12554 vectors with 32-bit and 64-bit elements.
12555 (vecdupssescalarmodesuffix): New mode attribute.
12556 (vec_dup<mode>): Use it.
12558 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12561 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12562 frame if argument is passed on stack.
12564 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12567 * ree.c (combine_reaching_defs): Optimize also
12568 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12569 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12571 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12573 PR middle-end/83189
12574 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12576 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12578 PR middle-end/83718
12579 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12580 after they are computed.
12582 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12584 PR tree-optimization/83695
12585 * gimple-loop-linterchange.cc
12586 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12587 reset cached scev information after interchange.
12588 (pass_linterchange::execute): Remove call to scev_reset_htab.
12590 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12592 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12593 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12594 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12595 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12596 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12597 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12598 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12599 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12600 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12601 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12602 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12603 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12604 (V_lane_reg): Likewise.
12605 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12607 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12608 (vfmal_lane_low<mode>_intrinsic,
12609 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12610 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12611 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12612 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12613 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12614 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12616 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12618 * config/arm/arm-cpus.in (fp16fml): New feature.
12619 (ALL_SIMD): Add fp16fml.
12620 (armv8.2-a): Add fp16fml as an option.
12621 (armv8.3-a): Likewise.
12622 (armv8.4-a): Add fp16fml as part of fp16.
12623 * config/arm/arm.h (TARGET_FP16FML): Define.
12624 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12626 * config/arm/arm-modes.def (V2HF): Define.
12627 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12628 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12629 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12630 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12631 vfmsl_low, vfmsl_high): New set of builtins.
12632 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12633 (vfml_op): New code attribute.
12634 (VFMLHALVES): New int iterator.
12635 (VFML, VFMLSEL): New mode attributes.
12636 (V_reg): Define mapping for V2HF.
12637 (V_hi, V_lo): New mode attributes.
12638 (VF_constraint): Likewise.
12639 (vfml_half, vfml_half_selector): New int attributes.
12640 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12642 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12643 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12645 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12646 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12647 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12648 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12650 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12651 Document new effective target and option set.
12653 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12655 * config/arm/arm-cpus.in (armv8_4): New feature.
12656 (ARMv8_4a): New fgroup.
12657 (armv8.4-a): New arch.
12658 * config/arm/arm-tables.opt: Regenerate.
12659 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12660 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12661 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12662 Add matching rules for -march=armv8.4-a and extensions.
12663 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12665 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12668 * config/rx/rx.md (BW): New mode attribute.
12669 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12671 2018-01-11 Richard Biener <rguenther@suse.de>
12673 PR tree-optimization/83435
12674 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12675 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12676 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12678 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12679 Alan Hayward <alan.hayward@arm.com>
12680 David Sherwood <david.sherwood@arm.com>
12682 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12684 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12685 (aarch64_print_address_internal): Use it to check for a zero offset.
12687 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12688 Alan Hayward <alan.hayward@arm.com>
12689 David Sherwood <david.sherwood@arm.com>
12691 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12692 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12693 Return a poly_int64 rather than a HOST_WIDE_INT.
12694 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12695 rather than a HOST_WIDE_INT.
12696 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12697 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12698 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12699 final_offset from HOST_WIDE_INT to poly_int64.
12700 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12701 to_constant when getting the number of units in an Advanced SIMD
12703 (aarch64_builtin_vectorized_function): Check for a constant number
12705 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12707 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12708 attribute instead of GET_MODE_NUNITS.
12709 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12710 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12711 GET_MODE_SIZE for fixed-size registers.
12712 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12713 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12714 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12715 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12716 (aarch64_print_operand, aarch64_print_address_internal)
12717 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12718 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12719 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12720 Handle polynomial GET_MODE_SIZE.
12721 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12722 wider than SImode without modification.
12723 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12724 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12725 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12726 passing and returning SVE modes.
12727 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12728 rather than GEN_INT.
12729 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12730 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12731 (aarch64_allocate_and_probe_stack_space): Likewise.
12732 (aarch64_layout_frame): Cope with polynomial offsets.
12733 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12734 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12735 polynomial offsets.
12736 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12737 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12738 poly_int64 rather than a HOST_WIDE_INT.
12739 (aarch64_get_separate_components, aarch64_process_components)
12740 (aarch64_expand_prologue, aarch64_expand_epilogue)
12741 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12742 (aarch64_anchor_offset): New function, split out from...
12743 (aarch64_legitimize_address): ...here.
12744 (aarch64_builtin_vectorization_cost): Handle polynomial
12745 TYPE_VECTOR_SUBPARTS.
12746 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12748 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12749 number of elements from the PARALLEL rather than the mode.
12750 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12751 rather than GET_MODE_BITSIZE.
12752 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12753 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12754 (aarch64_expand_vec_perm_const_1): Handle polynomial
12755 d->perm.length () and d->perm elements.
12756 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12757 Apply to_constant to d->perm elements.
12758 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12759 polynomial CONST_VECTOR_NUNITS.
12760 (aarch64_move_pointer): Take amount as a poly_int64 rather
12762 (aarch64_progress_pointer): Avoid temporary variable.
12763 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12764 the mode attribute instead of GET_MODE.
12766 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12767 Alan Hayward <alan.hayward@arm.com>
12768 David Sherwood <david.sherwood@arm.com>
12770 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12771 x exists before using it.
12772 (aarch64_add_constant_internal): Rename to...
12773 (aarch64_add_offset_1): ...this. Replace regnum with separate
12774 src and dest rtxes. Handle the case in which they're different,
12775 including when the offset is zero. Replace scratchreg with an rtx.
12776 Use 2 additions if there is no spare register into which we can
12777 move a 16-bit constant.
12778 (aarch64_add_constant): Delete.
12779 (aarch64_add_offset): Replace reg with separate src and dest
12780 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12781 Use aarch64_add_offset_1.
12782 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12783 an rtx rather than an int. Take the delta as a poly_int64
12784 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12785 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12786 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12787 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12788 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12789 and aarch64_add_sp.
12790 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12791 aarch64_add_constant.
12793 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12795 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12796 Use scalar_float_mode.
12798 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12800 * config/aarch64/aarch64-simd.md
12801 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12802 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12803 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12804 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12805 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12806 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12807 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12808 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12809 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12810 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12812 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12815 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12816 targ_options->x_arm_arch_string is non NULL.
12818 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12820 * config/aarch64/aarch64.h
12821 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12823 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12826 * expmed.c (emit_store_flag_force): Swap if const op0
12827 and change VOIDmode to mode of op0.
12829 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12831 PR rtl-optimization/83761
12832 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12833 than bytes to mode_for_size.
12835 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12837 PR middle-end/83189
12838 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12839 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12842 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12844 PR middle-end/83575
12845 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12846 when in layout mode.
12847 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12848 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12851 2018-01-10 Michael Collison <michael.collison@arm.com>
12853 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12854 * config/aarch64/aarch64-option-extension.def: Add
12855 AARCH64_OPT_EXTENSION of 'fp16fml'.
12856 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12857 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12858 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12859 * config/aarch64/constraints.md (Ui7): New constraint.
12860 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12861 (VFMLA_SEL_W): Ditto.
12864 (VFMLA16_LOW): New int iterator.
12865 (VFMLA16_HIGH): Ditto.
12866 (UNSPEC_FMLAL): New unspec.
12867 (UNSPEC_FMLSL): Ditto.
12868 (UNSPEC_FMLAL2): Ditto.
12869 (UNSPEC_FMLSL2): Ditto.
12870 (f16mac): New code attribute.
12871 * config/aarch64/aarch64-simd-builtins.def
12872 (aarch64_fmlal_lowv2sf): Ditto.
12873 (aarch64_fmlsl_lowv2sf): Ditto.
12874 (aarch64_fmlalq_lowv4sf): Ditto.
12875 (aarch64_fmlslq_lowv4sf): Ditto.
12876 (aarch64_fmlal_highv2sf): Ditto.
12877 (aarch64_fmlsl_highv2sf): Ditto.
12878 (aarch64_fmlalq_highv4sf): Ditto.
12879 (aarch64_fmlslq_highv4sf): Ditto.
12880 (aarch64_fmlal_lane_lowv2sf): Ditto.
12881 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12882 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12883 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12884 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12885 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12886 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12887 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12888 (aarch64_fmlal_lane_highv2sf): Ditto.
12889 (aarch64_fmlsl_lane_highv2sf): Ditto.
12890 (aarch64_fmlal_laneq_highv2sf): Ditto.
12891 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12892 (aarch64_fmlalq_lane_highv4sf): Ditto.
12893 (aarch64_fmlsl_lane_highv4sf): Ditto.
12894 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12895 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12896 * config/aarch64/aarch64-simd.md:
12897 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12898 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12899 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12900 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12901 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12902 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12903 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12904 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12905 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12906 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12907 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12908 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12909 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12910 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12911 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12912 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12913 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12914 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12915 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12916 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12917 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12918 (vfmlsl_low_u32): Ditto.
12919 (vfmlalq_low_u32): Ditto.
12920 (vfmlslq_low_u32): Ditto.
12921 (vfmlal_high_u32): Ditto.
12922 (vfmlsl_high_u32): Ditto.
12923 (vfmlalq_high_u32): Ditto.
12924 (vfmlslq_high_u32): Ditto.
12925 (vfmlal_lane_low_u32): Ditto.
12926 (vfmlsl_lane_low_u32): Ditto.
12927 (vfmlal_laneq_low_u32): Ditto.
12928 (vfmlsl_laneq_low_u32): Ditto.
12929 (vfmlalq_lane_low_u32): Ditto.
12930 (vfmlslq_lane_low_u32): Ditto.
12931 (vfmlalq_laneq_low_u32): Ditto.
12932 (vfmlslq_laneq_low_u32): Ditto.
12933 (vfmlal_lane_high_u32): Ditto.
12934 (vfmlsl_lane_high_u32): Ditto.
12935 (vfmlal_laneq_high_u32): Ditto.
12936 (vfmlsl_laneq_high_u32): Ditto.
12937 (vfmlalq_lane_high_u32): Ditto.
12938 (vfmlslq_lane_high_u32): Ditto.
12939 (vfmlalq_laneq_high_u32): Ditto.
12940 (vfmlslq_laneq_high_u32): Ditto.
12941 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12942 (AARCH64_FL_FOR_ARCH8_4): New.
12943 (AARCH64_ISA_F16FML): New ISA flag.
12944 (TARGET_F16FML): New feature flag for fp16fml.
12945 (doc/invoke.texi): Document new fp16fml option.
12947 2018-01-10 Michael Collison <michael.collison@arm.com>
12949 * config/aarch64/aarch64-builtins.c:
12950 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12951 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12952 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12953 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12954 (AARCH64_ISA_SHA3): New ISA flag.
12955 (TARGET_SHA3): New feature flag for sha3.
12956 * config/aarch64/iterators.md (sha512_op): New int attribute.
12957 (CRYPTO_SHA512): New int iterator.
12958 (UNSPEC_SHA512H): New unspec.
12959 (UNSPEC_SHA512H2): Ditto.
12960 (UNSPEC_SHA512SU0): Ditto.
12961 (UNSPEC_SHA512SU1): Ditto.
12962 * config/aarch64/aarch64-simd-builtins.def
12963 (aarch64_crypto_sha512hqv2di): New builtin.
12964 (aarch64_crypto_sha512h2qv2di): Ditto.
12965 (aarch64_crypto_sha512su0qv2di): Ditto.
12966 (aarch64_crypto_sha512su1qv2di): Ditto.
12967 (aarch64_eor3qv8hi): Ditto.
12968 (aarch64_rax1qv2di): Ditto.
12969 (aarch64_xarqv2di): Ditto.
12970 (aarch64_bcaxqv8hi): Ditto.
12971 * config/aarch64/aarch64-simd.md:
12972 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12973 (aarch64_crypto_sha512su0qv2di): Ditto.
12974 (aarch64_crypto_sha512su1qv2di): Ditto.
12975 (aarch64_eor3qv8hi): Ditto.
12976 (aarch64_rax1qv2di): Ditto.
12977 (aarch64_xarqv2di): Ditto.
12978 (aarch64_bcaxqv8hi): Ditto.
12979 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12980 (vsha512h2q_u64): Ditto.
12981 (vsha512su0q_u64): Ditto.
12982 (vsha512su1q_u64): Ditto.
12983 (veor3q_u16): Ditto.
12984 (vrax1q_u64): Ditto.
12985 (vxarq_u64): Ditto.
12986 (vbcaxq_u16): Ditto.
12987 * config/arm/types.md (crypto_sha512): New type attribute.
12988 (crypto_sha3): Ditto.
12989 (doc/invoke.texi): Document new sha3 option.
12991 2018-01-10 Michael Collison <michael.collison@arm.com>
12993 * config/aarch64/aarch64-builtins.c:
12994 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12995 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12996 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12997 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12998 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12999 (AARCH64_ISA_SM4): New ISA flag.
13000 (TARGET_SM4): New feature flag for sm4.
13001 * config/aarch64/aarch64-simd-builtins.def
13002 (aarch64_sm3ss1qv4si): Ditto.
13003 (aarch64_sm3tt1aq4si): Ditto.
13004 (aarch64_sm3tt1bq4si): Ditto.
13005 (aarch64_sm3tt2aq4si): Ditto.
13006 (aarch64_sm3tt2bq4si): Ditto.
13007 (aarch64_sm3partw1qv4si): Ditto.
13008 (aarch64_sm3partw2qv4si): Ditto.
13009 (aarch64_sm4eqv4si): Ditto.
13010 (aarch64_sm4ekeyqv4si): Ditto.
13011 * config/aarch64/aarch64-simd.md:
13012 (aarch64_sm3ss1qv4si): Ditto.
13013 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13014 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13015 (aarch64_sm4eqv4si): Ditto.
13016 (aarch64_sm4ekeyqv4si): Ditto.
13017 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13018 (sm3part_op): Ditto.
13019 (CRYPTO_SM3TT): Ditto.
13020 (CRYPTO_SM3PART): Ditto.
13021 (UNSPEC_SM3SS1): New unspec.
13022 (UNSPEC_SM3TT1A): Ditto.
13023 (UNSPEC_SM3TT1B): Ditto.
13024 (UNSPEC_SM3TT2A): Ditto.
13025 (UNSPEC_SM3TT2B): Ditto.
13026 (UNSPEC_SM3PARTW1): Ditto.
13027 (UNSPEC_SM3PARTW2): Ditto.
13028 (UNSPEC_SM4E): Ditto.
13029 (UNSPEC_SM4EKEY): Ditto.
13030 * config/aarch64/constraints.md (Ui2): New constraint.
13031 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13032 * config/arm/types.md (crypto_sm3): New type attribute.
13033 (crypto_sm4): Ditto.
13034 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13035 (vsm3tt1aq_u32): Ditto.
13036 (vsm3tt1bq_u32): Ditto.
13037 (vsm3tt2aq_u32): Ditto.
13038 (vsm3tt2bq_u32): Ditto.
13039 (vsm3partw1q_u32): Ditto.
13040 (vsm3partw2q_u32): Ditto.
13041 (vsm4eq_u32): Ditto.
13042 (vsm4ekeyq_u32): Ditto.
13043 (doc/invoke.texi): Document new sm4 option.
13045 2018-01-10 Michael Collison <michael.collison@arm.com>
13047 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13048 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13049 (AARCH64_FL_FOR_ARCH8_4): New.
13050 (AARCH64_FL_V8_4): New flag.
13051 (doc/invoke.texi): Document new armv8.4-a option.
13053 2018-01-10 Michael Collison <michael.collison@arm.com>
13055 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13056 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13057 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13058 * config/aarch64/aarch64-option-extension.def: Add
13059 AARCH64_OPT_EXTENSION of 'sha2'.
13060 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13061 (crypto): Disable sha2 and aes if crypto disabled.
13062 (crypto): Enable aes and sha2 if enabled.
13063 (simd): Disable sha2 and aes if simd disabled.
13064 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13066 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13067 (TARGET_SHA2): New feature flag for sha2.
13068 (TARGET_AES): New feature flag for aes.
13069 * config/aarch64/aarch64-simd.md:
13070 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13071 conditional on TARGET_AES.
13072 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13073 (aarch64_crypto_sha1hsi): Make pattern conditional
13075 (aarch64_crypto_sha1hv4si): Ditto.
13076 (aarch64_be_crypto_sha1hv4si): Ditto.
13077 (aarch64_crypto_sha1su1v4si): Ditto.
13078 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13079 (aarch64_crypto_sha1su0v4si): Ditto.
13080 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13081 (aarch64_crypto_sha256su0v4si): Ditto.
13082 (aarch64_crypto_sha256su1v4si): Ditto.
13083 (doc/invoke.texi): Document new aes and sha2 options.
13085 2018-01-10 Martin Sebor <msebor@redhat.com>
13087 PR tree-optimization/83781
13088 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13091 2018-01-11 Martin Sebor <msebor@gmail.com>
13092 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13094 PR tree-optimization/83501
13095 PR tree-optimization/81703
13097 * tree-ssa-strlen.c (get_string_cst): Rename...
13098 (get_string_len): ...to this. Handle global constants.
13099 (handle_char_store): Adjust.
13101 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13102 Jim Wilson <jimw@sifive.com>
13104 * config/riscv/riscv-protos.h (riscv_output_return): New.
13105 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13106 (riscv_attribute_table, riscv_output_return),
13107 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13108 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13109 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13111 (riscv_expand_prologue): Add early return for naked function.
13112 (riscv_expand_epilogue): Likewise.
13113 (riscv_function_ok_for_sibcall): Return false for naked function.
13114 (riscv_set_current_function): New.
13115 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13116 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13117 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13118 * doc/extend.texi (RISC-V Function Attributes): New.
13120 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13122 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13123 check for 128-bit long double before checking TCmode.
13124 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13125 128-bit long doubles before checking TFmode or TCmode.
13126 (FLOAT128_IBM_P): Likewise.
13128 2018-01-10 Martin Sebor <msebor@redhat.com>
13130 PR tree-optimization/83671
13131 * builtins.c (c_strlen): Unconditionally return zero for the empty
13133 Use -Warray-bounds for warnings.
13134 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13135 for non-constant array indices with COMPONENT_REF, arrays of
13136 arrays, and pointers to arrays.
13137 (gimple_fold_builtin_strlen): Determine and set length range for
13138 non-constant character arrays.
13140 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13142 PR middle-end/81897
13143 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13146 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13148 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13150 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13153 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13154 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13155 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13156 indexed_or_indirect_operand predicate.
13157 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13158 (*vsx_le_perm_load_v8hi): Likewise.
13159 (*vsx_le_perm_load_v16qi): Likewise.
13160 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13161 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13162 (*vsx_le_perm_store_v8hi): Likewise.
13163 (*vsx_le_perm_store_v16qi): Likewise.
13164 (eight unnamed splitters): Likewise.
13166 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13168 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13169 * config/rs6000/emmintrin.h: Likewise.
13170 * config/rs6000/mmintrin.h: Likewise.
13171 * config/rs6000/xmmintrin.h: Likewise.
13173 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13176 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13178 * tree.c (tree_nop_conversion): Return true for location wrapper
13180 (maybe_wrap_with_location): New function.
13181 (selftest::check_strip_nops): New function.
13182 (selftest::test_location_wrappers): New function.
13183 (selftest::tree_c_tests): Call it.
13184 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13185 (maybe_wrap_with_location): New decl.
13186 (EXPR_LOCATION_WRAPPER_P): New macro.
13187 (location_wrapper_p): New inline function.
13188 (tree_strip_any_location_wrapper): New inline function.
13190 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13193 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13194 stack_realign_offset for the largest alignment of stack slot
13196 (ix86_find_max_used_stack_alignment): New function.
13197 (ix86_finalize_stack_frame_flags): Use it. Set
13198 max_used_stack_alignment if we don't realign stack.
13199 * config/i386/i386.h (machine_function): Add
13200 max_used_stack_alignment.
13202 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13204 * config/arm/arm.opt (-mbranch-cost): New option.
13205 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13208 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13211 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13212 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13214 2018-01-10 Richard Biener <rguenther@suse.de>
13217 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13218 early out so it also covers the case where we have a non-NULL
13221 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13223 PR tree-optimization/83753
13224 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13225 for non-strided grouped accesses if the number of elements is 1.
13227 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13230 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13231 * i386.h (TARGET_USE_GATHER): Define.
13232 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13234 2018-01-10 Martin Liska <mliska@suse.cz>
13237 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13238 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13240 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13241 CLEANUP_NO_PARTITIONING is not set.
13243 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13245 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13246 for vectors, as a partial revert of r254296.
13247 * rtl.h (const_vec_p): Delete.
13248 (const_vec_duplicate_p): Don't test for vector CONSTs.
13249 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13250 * expmed.c (make_tree): Likewise.
13253 * common.md (E, F): Use CONSTANT_P instead of checking for
13255 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13256 checking for CONST_VECTOR.
13258 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13260 PR middle-end/83575
13261 * predict.c (force_edge_cold): Handle in more sane way edges
13262 with no prediction.
13264 2018-01-09 Carl Love <cel@us.ibm.com>
13266 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13268 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13269 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13270 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13271 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13272 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13273 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13274 * config/rs6000/rs6000-protos.h: Add extern defition for
13275 rs6000_generate_float2_double_code.
13276 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13278 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13279 (float2_v2df): Add define_expand.
13281 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13284 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13285 op_mode in the force_to_mode call.
13287 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13289 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13290 instead of checking each element individually.
13291 (aarch64_evpc_uzp): Likewise.
13292 (aarch64_evpc_zip): Likewise.
13293 (aarch64_evpc_ext): Likewise.
13294 (aarch64_evpc_rev): Likewise.
13295 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13296 instead of checking each element individually. Return true without
13298 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13299 whether all selected elements come from the same input, instead of
13300 checking each element individually. Remove calls to gen_rtx_REG,
13301 start_sequence and end_sequence and instead assert that no rtl is
13304 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13306 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13307 order of HIGH and CONST checks.
13309 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13311 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13312 if the destination isn't an SSA_NAME.
13314 2018-01-09 Richard Biener <rguenther@suse.de>
13316 PR tree-optimization/83668
13317 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13319 (canonicalize_loop_form): ... here, renamed from ...
13320 (canonicalize_loop_closed_ssa_form): ... this and amended to
13321 swap successor edges for loop exit blocks to make us use
13322 the RPO order we need for initial schedule generation.
13324 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13326 PR tree-optimization/64811
13327 * match.pd: When optimizing comparisons with Inf, avoid
13328 introducing or losing exceptions from comparisons with NaN.
13330 2018-01-09 Martin Liska <mliska@suse.cz>
13333 * asan.c (shadow_mem_size): Add gcc_assert.
13335 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13337 Don't save registers in main().
13340 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13341 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13342 * config/avr/avr.c (avr_set_current_function): Don't error if
13343 naked, OS_task or OS_main are specified at the same time.
13344 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13346 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13348 * common/config/avr/avr-common.c (avr_option_optimization_table):
13349 Switch on -mmain-is-OS_task for optimizing compilations.
13351 2018-01-09 Richard Biener <rguenther@suse.de>
13353 PR tree-optimization/83572
13354 * graphite.c: Include cfganal.h.
13355 (graphite_transform_loops): Connect infinite loops to exit
13356 and remove fake edges at the end.
13358 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13360 * ipa-inline.c (edge_badness): Revert accidental checkin.
13362 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13365 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13366 symbols; not inline clones.
13368 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13371 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13372 hard registers. Formatting fixes.
13374 PR preprocessor/83722
13375 * gcc.c (try_generate_repro): Pass
13376 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13377 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13380 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13381 Kito Cheng <kito.cheng@gmail.com>
13383 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13384 (riscv_leaf_function_p): Delete.
13385 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13387 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13389 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13391 (do_ifelse): New function.
13392 (do_isel): New function.
13393 (do_sub3): New function.
13394 (do_add3): New function.
13395 (do_load_mask_compare): New function.
13396 (do_overlap_load_compare): New function.
13397 (expand_compare_loop): New function.
13398 (expand_block_compare): Call expand_compare_loop() when appropriate.
13399 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13400 option description.
13401 (-mblock-compare-inline-loop-limit): New option.
13403 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13406 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13407 Reverse order of second and third operands in first alternative.
13408 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13409 of first and second elements in UNSPEC_VPERMR vector.
13410 (altivec_expand_vec_perm_le): Likewise.
13412 2018-01-08 Jeff Law <law@redhat.com>
13414 PR rtl-optimizatin/81308
13415 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13416 (process_switch): If group_case_labels makes a change, then set
13418 (pass_convert_switch::execute): If a switch is converted, then
13419 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13421 PR rtl-optimization/81308
13422 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13425 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13427 PR target/83663 - Revert r255946
13428 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13429 generation for cases where splatting a value is not useful.
13430 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13431 across a vec_duplicate and a paradoxical subreg forming a vector
13432 mode to a vec_concat.
13434 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13436 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13437 -march=armv8.3-a variants.
13438 * config/arm/t-multilib: Likewise.
13439 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13441 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13443 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13445 (cceq_ior_compare_complement): Give it a name so I can use it, and
13446 change boolean_or_operator predicate to boolean_operator so it can
13447 be used to generate a crand.
13448 (eqne): New code iterator.
13449 (bd/bd_neg): New code_attrs.
13450 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13451 a single define_insn.
13452 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13453 decrement (bdnzt/bdnzf/bdzt/bdzf).
13454 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13455 with the new names of the branch decrement patterns, and added the
13456 names of the branch decrement conditional patterns.
13458 2018-01-08 Richard Biener <rguenther@suse.de>
13460 PR tree-optimization/83563
13461 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13464 2018-01-08 Richard Biener <rguenther@suse.de>
13466 PR middle-end/83713
13467 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13469 2018-01-08 Richard Biener <rguenther@suse.de>
13471 PR tree-optimization/83685
13472 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13473 references to abnormals.
13475 2018-01-08 Richard Biener <rguenther@suse.de>
13478 * dwarf2out.c (output_indirect_strings): Handle empty
13479 skeleton_debug_str_hash.
13480 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13482 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13484 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13485 (emit_store_direct): Likewise.
13486 (arc_trampoline_adjust_address): Likewise.
13487 (arc_asm_trampoline_template): New function.
13488 (arc_initialize_trampoline): Use asm_trampoline_template.
13489 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13490 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13491 * config/arc/arc.md (flush_icache): Delete pattern.
13493 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13495 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13496 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13499 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13502 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13503 by not USED_FOR_TARGET.
13504 (make_pass_resolve_sw_modes): Likewise.
13506 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13508 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13511 2018-01-08 Richard Biener <rguenther@suse.de>
13513 PR middle-end/83580
13514 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13516 2018-01-08 Richard Biener <rguenther@suse.de>
13518 PR middle-end/83517
13519 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13521 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13523 PR middle-end/81897
13524 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13525 basic blocks with a small number of successors.
13526 (convert_control_dep_chain_into_preds): Improve handling of
13528 (dump_predicates): Split apart into...
13529 (dump_pred_chain): ...here...
13530 (dump_pred_info): ...and here.
13531 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13532 (can_chain_union_be_invalidated_p): Improve check for invalidation
13534 (uninit_uses_cannot_happen): Avoid unnecessary if
13535 convert_control_dep_chain_into_preds yielded nothing.
13537 2018-01-06 Martin Sebor <msebor@redhat.com>
13539 PR tree-optimization/83640
13540 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13541 subtracting negative offset from size.
13542 (builtin_access::overlap): Adjust offset bounds of the access to fall
13543 within the size of the object if possible.
13545 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13547 PR rtl-optimization/83699
13548 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13549 extract_bit_field_as_subreg to cases in which the extracted
13550 value is also a vector.
13552 * lra-constraints.c (process_alt_operands): Test for the equivalence
13553 substitutions when detecting a possible reload cycle.
13555 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13558 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13559 by default if flag_selective_schedling{,2}. Formatting fixes.
13561 PR rtl-optimization/83682
13562 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13563 if it has non-VECTOR_MODE element mode.
13564 (vec_duplicate_p): Likewise.
13566 PR middle-end/83694
13567 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13568 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13570 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13573 * config/i386/i386-builtin.def
13574 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13575 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13576 Require also OPTION_MASK_ISA_AVX512F in addition to
13577 OPTION_MASK_ISA_GFNI.
13578 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13579 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13580 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13581 to OPTION_MASK_ISA_GFNI.
13582 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13583 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13584 OPTION_MASK_ISA_AVX512BW.
13585 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13586 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13587 addition to OPTION_MASK_ISA_GFNI.
13588 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13589 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13590 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13591 to OPTION_MASK_ISA_GFNI.
13592 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13593 a requirement for all ISAs rather than any of them with a few
13595 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13597 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13598 bitmasks to be enabled with 3 exceptions, instead of requiring any
13599 enabled ISA with lots of exceptions.
13600 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13601 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13602 Change avx512bw in isa attribute to avx512f.
13603 * config/i386/sgxintrin.h: Add license boilerplate.
13604 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13605 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13606 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13607 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13609 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13610 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13611 temporarily sse2 rather than sse if not enabled already.
13614 * config/i386/sse.md (VI248_VLBW): Rename to ...
13615 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13616 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13617 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13618 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13619 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13620 mode iterator instead of VI248_VLBW.
13622 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13624 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13625 (record_modified): Skip clobbers; add debug output.
13626 (param_change_prob): Use sreal frequencies.
13628 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13630 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13631 punt for user-aligned variables.
13633 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13635 * tree-chrec.c (chrec_contains_symbols): Return true for
13638 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13641 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13642 of (x|y) == x for BICS pattern.
13644 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13646 PR tree-optimization/83605
13647 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13648 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13651 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13653 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13654 * config/epiphany/rtems.h: New file.
13656 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13657 Uros Bizjak <ubizjak@gmail.com>
13660 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13661 QIreg_operand instead of register_operand predicate.
13662 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13663 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13664 comments instead of -fmitigate[-_]rop.
13666 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13669 * cgraphunit.c (symbol_table::compile): Switch to text_section
13670 before calling assembly_start debug hook.
13671 * run-rtl-passes.c (run_rtl_passes): Likewise.
13674 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13676 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13677 range_int_cst_p rather than !symbolic_range_p before calling
13678 extract_range_from_multiplicative_op_1.
13680 2018-01-04 Jeff Law <law@redhat.com>
13682 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13683 redundant test in assertion.
13685 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13687 * doc/rtl.texi: Document machine_mode wrapper classes.
13689 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13691 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13692 using tree_to_uhwi.
13694 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13696 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13697 the VEC_PERM_EXPR fold to fail.
13699 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13702 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13703 to switched_sections.
13705 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13708 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13709 test for d.testing.
13711 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13714 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13715 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13717 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13720 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13721 is BLKmode and bitpos not zero or mode change is needed.
13723 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13726 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13729 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13732 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13733 instead of MULT rtx. Update all corresponding splitters.
13734 (*saddl_se): Ditto.
13735 (*ssub<modesuffix>): Ditto.
13736 (*ssubl_se): Ditto.
13737 (*cmp_sadd_di): Update split patterns.
13738 (*cmp_sadd_si): Ditto.
13739 (*cmp_sadd_sidi): Ditto.
13740 (*cmp_ssub_di): Ditto.
13741 (*cmp_ssub_si): Ditto.
13742 (*cmp_ssub_sidi): Ditto.
13743 * config/alpha/predicates.md (const23_operand): New predicate.
13744 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13745 Look for ASHIFT, not MULT inner operand.
13746 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13748 2018-01-04 Martin Liska <mliska@suse.cz>
13750 PR gcov-profile/83669
13751 * gcov.c (output_intermediate_file): Add version to intermediate
13753 * doc/gcov.texi: Document new field 'version' in intermediate
13754 file format. Fix location of '-k' option of gcov command.
13756 2018-01-04 Martin Liska <mliska@suse.cz>
13759 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13761 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13763 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13765 2018-01-03 Martin Sebor <msebor@redhat.com>
13767 PR tree-optimization/83655
13768 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13769 checking calls with invalid arguments.
13771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13773 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13774 (vectorizable_mask_load_store): Delete.
13775 (vectorizable_call): Return false for masked loads and stores.
13776 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13777 instead of gimple_assign_rhs1.
13778 (vectorizable_load): Handle IFN_MASK_LOAD.
13779 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13783 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13785 (vectorizable_mask_load_store): ...here.
13786 (vectorizable_load): ...and here.
13788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13790 * tree-vect-stmts.c (vect_build_all_ones_mask)
13791 (vect_build_zero_merge_argument): New functions, split out from...
13792 (vectorizable_load): ...here.
13794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13796 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13798 (vectorizable_mask_load_store): ...here.
13799 (vectorizable_store): ...and here.
13801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13803 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13805 (vectorizable_mask_load_store): ...here.
13807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13809 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13810 (vect_model_store_cost): Take a vec_load_store_type instead of a
13812 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13813 (vect_model_store_cost): Take a vec_load_store_type instead of a
13815 (vectorizable_mask_load_store): Update accordingly.
13816 (vectorizable_store): Likewise.
13817 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13821 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13822 IFN_MASK_LOAD calls here rather than...
13823 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13826 Alan Hayward <alan.hayward@arm.com>
13827 David Sherwood <david.sherwood@arm.com>
13829 * expmed.c (extract_bit_field_1): For vector extracts,
13830 fall back to extract_bit_field_as_subreg if vec_extract
13833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13834 Alan Hayward <alan.hayward@arm.com>
13835 David Sherwood <david.sherwood@arm.com>
13837 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13838 they are variable or constant sized.
13839 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13840 slots for constant-sized data.
13842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13843 Alan Hayward <alan.hayward@arm.com>
13844 David Sherwood <david.sherwood@arm.com>
13846 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13847 handling COND_EXPRs with boolean comparisons, try to find a better
13848 basis for the mask type than the boolean itself.
13850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13852 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13853 is calculated and how it can be overridden.
13854 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13855 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13857 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13861 Alan Hayward <alan.hayward@arm.com>
13862 David Sherwood <david.sherwood@arm.com>
13864 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13865 Remove the mode argument.
13866 (aarch64_simd_valid_immediate): Remove the mode and inverse
13868 * config/aarch64/iterators.md (bitsize): New iterator.
13869 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13870 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13871 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13872 aarch64_simd_valid_immediate.
13873 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13874 (aarch64_reg_or_bic_imm): Likewise.
13875 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13876 with an insn_type enum and msl with a modifier_type enum.
13877 Replace element_width with a scalar_mode. Change the shift
13878 to unsigned int. Add constructors for scalar_float_mode and
13879 scalar_int_mode elements.
13880 (aarch64_vect_float_const_representable_p): Delete.
13881 (aarch64_can_const_movi_rtx_p)
13882 (aarch64_simd_scalar_immediate_valid_for_move)
13883 (aarch64_simd_make_constant): Update call to
13884 aarch64_simd_valid_immediate.
13885 (aarch64_advsimd_valid_immediate_hs): New function.
13886 (aarch64_advsimd_valid_immediate): Likewise.
13887 (aarch64_simd_valid_immediate): Remove mode and inverse
13888 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13889 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13890 and aarch64_float_const_representable_p on the result.
13891 (aarch64_output_simd_mov_immediate): Remove mode argument.
13892 Update call to aarch64_simd_valid_immediate and use of
13893 simd_immediate_info.
13894 (aarch64_output_scalar_simd_mov_immediate): Update call
13897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13898 Alan Hayward <alan.hayward@arm.com>
13899 David Sherwood <david.sherwood@arm.com>
13901 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13902 (mode_nunits): Likewise CONST_MODE_NUNITS.
13903 * machmode.def (ADJUST_NUNITS): Document.
13904 * genmodes.c (mode_data::need_nunits_adj): New field.
13905 (blank_mode): Update accordingly.
13906 (adj_nunits): New variable.
13907 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13909 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13910 listed in adj_nunits.
13911 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13912 listed in adj_nunits. Don't emit case statements for such modes.
13913 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13914 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13915 nothing if adj_nunits is nonnull.
13916 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13917 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13918 (emit_mode_fbit): Update use of print_maybe_const_decl.
13919 (emit_move_size): Likewise. Treat the array as non-const
13921 (emit_mode_adjustments): Handle adj_nunits.
13923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13925 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13926 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13927 (VECTOR_MODES): Use it.
13928 (make_vector_modes): Take the prefix as an argument.
13930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13931 Alan Hayward <alan.hayward@arm.com>
13932 David Sherwood <david.sherwood@arm.com>
13934 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13935 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13936 for MODE_VECTOR_BOOL.
13937 * machmode.def (VECTOR_BOOL_MODE): Document.
13938 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13939 (make_vector_bool_mode): New function.
13940 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13942 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13943 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13945 * stor-layout.c (int_mode_for_mode): Likewise.
13946 * tree.c (build_vector_type_for_mode): Likewise.
13947 * varasm.c (output_constant_pool_2): Likewise.
13948 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13949 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13950 for MODE_VECTOR_BOOL.
13951 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13952 of mode class checks.
13953 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13954 instead of a list of mode class checks.
13955 (expand_vector_scalar_condition): Likewise.
13956 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13959 Alan Hayward <alan.hayward@arm.com>
13960 David Sherwood <david.sherwood@arm.com>
13962 * machmode.h (mode_size): Change from unsigned short to
13964 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13965 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13966 or if measurement_type is not polynomial.
13967 (fixed_size_mode::includes_p): Check for constant-sized modes.
13968 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13969 return a poly_uint16 rather than an unsigned short.
13970 (emit_mode_size): Change the type of mode_size from unsigned short
13971 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13972 (emit_mode_adjustments): Cope with polynomial vector sizes.
13973 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13975 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13977 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13978 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13979 * caller-save.c (setup_save_areas): Likewise.
13980 (replace_reg_with_saved_mem): Likewise.
13981 * calls.c (emit_library_call_value_1): Likewise.
13982 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13983 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13984 (gen_lowpart_for_combine): Likewise.
13985 * convert.c (convert_to_integer_1): Likewise.
13986 * cse.c (equiv_constant, cse_insn): Likewise.
13987 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13988 (cselib_subst_to_values): Likewise.
13989 * dce.c (word_dce_process_block): Likewise.
13990 * df-problems.c (df_word_lr_mark_ref): Likewise.
13991 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13992 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13993 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13994 (rtl_for_decl_location): Likewise.
13995 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13996 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13997 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13998 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13999 (expand_expr_real_1): Likewise.
14000 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14001 (pad_below): Likewise.
14002 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14003 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14004 * ira.c (get_subreg_tracking_sizes): Likewise.
14005 * ira-build.c (ira_create_allocno_objects): Likewise.
14006 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14007 (ira_sort_regnos_for_alter_reg): Likewise.
14008 * ira-costs.c (record_operand_costs): Likewise.
14009 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14010 (resolve_simple_move): Likewise.
14011 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14012 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14013 (lra_constraints): Likewise.
14014 (CONST_POOL_OK_P): Reject variable-sized modes.
14015 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14016 (add_pseudo_to_slot, lra_spill): Likewise.
14017 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14018 * optabs-query.c (get_best_extraction_insn): Likewise.
14019 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14020 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14021 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14022 * recog.c (offsettable_address_addr_space_p): Likewise.
14023 * regcprop.c (maybe_mode_change): Likewise.
14024 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14025 * regrename.c (build_def_use): Likewise.
14026 * regstat.c (dump_reg_info): Likewise.
14027 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14028 (find_reloads, find_reloads_subreg_address): Likewise.
14029 * reload1.c (eliminate_regs_1): Likewise.
14030 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14031 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14032 (simplify_binary_operation_1, simplify_subreg): Likewise.
14033 * targhooks.c (default_function_arg_padding): Likewise.
14034 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14035 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14036 (verify_gimple_assign_ternary): Likewise.
14037 * tree-inline.c (estimate_move_cost): Likewise.
14038 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14039 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14040 (get_address_cost_ainc): Likewise.
14041 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14042 (vect_supportable_dr_alignment): Likewise.
14043 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14044 (vectorizable_reduction): Likewise.
14045 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14046 (vectorizable_operation, vectorizable_load): Likewise.
14047 * tree.c (build_same_sized_truth_vector_type): Likewise.
14048 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14049 * var-tracking.c (emit_note_insn_var_location): Likewise.
14050 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14051 (ADDR_VEC_ALIGN): Likewise.
14053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14054 Alan Hayward <alan.hayward@arm.com>
14055 David Sherwood <david.sherwood@arm.com>
14057 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14059 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14060 or if measurement_type is polynomial.
14061 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14062 * combine.c (make_extraction): Likewise.
14063 * dse.c (find_shift_sequence): Likewise.
14064 * dwarf2out.c (mem_loc_descriptor): Likewise.
14065 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14066 (extract_bit_field, extract_low_bits): Likewise.
14067 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14068 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14069 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14070 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14071 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14072 * reload.c (find_reloads): Likewise.
14073 * reload1.c (alter_reg): Likewise.
14074 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14075 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14076 * tree-if-conv.c (predicate_mem_writes): Likewise.
14077 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14078 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14079 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14080 * valtrack.c (dead_debug_insert_temp): Likewise.
14081 * varasm.c (mergeable_constant_section): Likewise.
14082 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14085 Alan Hayward <alan.hayward@arm.com>
14086 David Sherwood <david.sherwood@arm.com>
14088 * expr.c (expand_assignment): Cope with polynomial mode sizes
14089 when assigning to a CONCAT.
14091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14092 Alan Hayward <alan.hayward@arm.com>
14093 David Sherwood <david.sherwood@arm.com>
14095 * machmode.h (mode_precision): Change from unsigned short to
14097 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14099 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14100 or if measurement_type is not polynomial.
14101 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14102 in which the mode is already known to be a scalar_int_mode.
14103 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14104 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14106 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14107 for GET_MODE_PRECISION.
14108 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14109 for GET_MODE_PRECISION.
14110 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14112 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14113 (expand_field_assignment, make_extraction): Likewise.
14114 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14115 (get_last_value): Likewise.
14116 * convert.c (convert_to_integer_1): Likewise.
14117 * cse.c (cse_insn): Likewise.
14118 * expr.c (expand_expr_real_1): Likewise.
14119 * lra-constraints.c (simplify_operand_subreg): Likewise.
14120 * optabs-query.c (can_atomic_load_p): Likewise.
14121 * optabs.c (expand_atomic_load): Likewise.
14122 (expand_atomic_store): Likewise.
14123 * ree.c (combine_reaching_defs): Likewise.
14124 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14125 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14126 * tree.h (type_has_mode_precision_p): Likewise.
14127 * ubsan.c (instrument_si_overflow): Likewise.
14129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14130 Alan Hayward <alan.hayward@arm.com>
14131 David Sherwood <david.sherwood@arm.com>
14133 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14134 polynomial numbers of units.
14135 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14136 (valid_vector_subparts_p): New function.
14137 (build_vector_type): Remove temporary shim and take the number
14138 of units as a poly_uint64 rather than an int.
14139 (build_opaque_vector_type): Take the number of units as a
14140 poly_uint64 rather than an int.
14141 * tree.c (build_vector_from_ctor): Handle polynomial
14142 TYPE_VECTOR_SUBPARTS.
14143 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14144 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14145 (build_vector_from_val): If the number of units is variable,
14146 use build_vec_duplicate_cst for constant operands and
14147 VEC_DUPLICATE_EXPR otherwise.
14148 (make_vector_type): Remove temporary is_constant ().
14149 (build_vector_type, build_opaque_vector_type): Take the number of
14150 units as a poly_uint64 rather than an int.
14151 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14153 * cfgexpand.c (expand_debug_expr): Likewise.
14154 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14155 (store_constructor, expand_expr_real_1): Likewise.
14156 (const_scalar_mask_from_tree): Likewise.
14157 * fold-const-call.c (fold_const_reduction): Likewise.
14158 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14159 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14160 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14161 (fold_relational_const): Likewise.
14162 (native_interpret_vector): Likewise. Change the size from an
14163 int to an unsigned int.
14164 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14165 TYPE_VECTOR_SUBPARTS.
14166 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14167 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14168 duplicating a non-constant operand into a variable-length vector.
14169 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14170 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14171 * ipa-icf.c (sem_variable::equals): Likewise.
14172 * match.pd: Likewise.
14173 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14174 * print-tree.c (print_node): Likewise.
14175 * stor-layout.c (layout_type): Likewise.
14176 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14177 * tree-cfg.c (verify_gimple_comparison): Likewise.
14178 (verify_gimple_assign_binary): Likewise.
14179 (verify_gimple_assign_ternary): Likewise.
14180 (verify_gimple_assign_single): Likewise.
14181 * tree-pretty-print.c (dump_generic_node): Likewise.
14182 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14183 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14184 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14185 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14186 (vect_shift_permute_load_chain): Likewise.
14187 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14188 (expand_vector_condition, optimize_vector_constructor): Likewise.
14189 (lower_vec_perm, get_compute_type): Likewise.
14190 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14191 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14192 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14193 (vect_recog_mask_conversion_pattern): Likewise.
14194 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14195 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14196 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14197 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14198 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14199 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14200 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14201 (supportable_widening_operation): Likewise.
14202 (supportable_narrowing_operation): Likewise.
14203 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14205 * varasm.c (output_constant): Likewise.
14207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14208 Alan Hayward <alan.hayward@arm.com>
14209 David Sherwood <david.sherwood@arm.com>
14211 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14212 so that both the length == 3 and length != 3 cases set up their
14213 own permute vectors. Add comments explaining why we know the
14214 number of elements is constant.
14215 (vect_permute_load_chain): Likewise.
14217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14218 Alan Hayward <alan.hayward@arm.com>
14219 David Sherwood <david.sherwood@arm.com>
14221 * machmode.h (mode_nunits): Change from unsigned char to
14223 (ONLY_FIXED_SIZE_MODES): New macro.
14224 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14225 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14226 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14228 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14229 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14230 or if measurement_type is not polynomial.
14231 * genmodes.c (ZERO_COEFFS): New macro.
14232 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14234 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14235 Use ZERO_COEFFS when emitting initializers.
14236 * data-streamer.h (bp_pack_poly_value): New function.
14237 (bp_unpack_poly_value): Likewise.
14238 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14239 for GET_MODE_NUNITS.
14240 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14241 for GET_MODE_NUNITS.
14242 * tree.c (make_vector_type): Remove temporary shim and make
14243 the real function take the number of units as a poly_uint64
14244 rather than an int.
14245 (build_vector_type_for_mode): Handle polynomial nunits.
14246 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14247 * emit-rtl.c (const_vec_series_p_1): Likewise.
14248 (gen_rtx_CONST_VECTOR): Likewise.
14249 * fold-const.c (test_vec_duplicate_folding): Likewise.
14250 * genrecog.c (validate_pattern): Likewise.
14251 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14252 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14253 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14254 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14255 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14256 * rtlanal.c (subreg_get_info): Likewise.
14257 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14258 (vect_grouped_load_supported): Likewise.
14259 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14260 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14261 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14262 (simplify_const_unary_operation, simplify_binary_operation_1)
14263 (simplify_const_binary_operation, simplify_ternary_operation)
14264 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14265 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14266 instead of CONST_VECTOR_NUNITS.
14267 * varasm.c (output_constant_pool_2): Likewise.
14268 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14269 explicit-encoded elements in the XVEC for variable-length vectors.
14271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14273 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14276 Alan Hayward <alan.hayward@arm.com>
14277 David Sherwood <david.sherwood@arm.com>
14279 * coretypes.h (fixed_size_mode): Declare.
14280 (fixed_size_mode_pod): New typedef.
14281 * builtins.h (target_builtins::x_apply_args_mode)
14282 (target_builtins::x_apply_result_mode): Change type to
14283 fixed_size_mode_pod.
14284 * builtins.c (apply_args_size, apply_result_size, result_vector)
14285 (expand_builtin_apply_args_1, expand_builtin_apply)
14286 (expand_builtin_return): Update accordingly.
14288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14290 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14291 * cselib.c (cselib_hash_rtx): Likewise.
14292 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14293 CONST_VECTOR encoding.
14295 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14296 Jeff Law <law@redhat.com>
14299 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14300 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14301 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14302 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14305 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14306 explicitly probe *sp in a noreturn function if there were any callee
14307 register saves or frame pointer is needed.
14309 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14312 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14313 BLKmode for ternary, binary or unary expressions.
14316 * var-tracking.c (delete_vta_debug_insn): New inline function.
14317 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14318 insns from get_insns () to NULL instead of each bb separately.
14319 Use delete_vta_debug_insn. No longer static.
14320 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14321 delete_vta_debug_insns callers.
14322 * rtl.h (delete_vta_debug_insns): Declare.
14323 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14324 instead of variable_tracking_main.
14326 2018-01-03 Martin Sebor <msebor@redhat.com>
14328 PR tree-optimization/83603
14329 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14330 arguments past the endof the argument list in functions declared
14331 without a prototype.
14332 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14333 Avoid checking when arguments are null.
14335 2018-01-03 Martin Sebor <msebor@redhat.com>
14338 * doc/extend.texi (attribute const): Fix a typo.
14339 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14340 issuing -Wsuggest-attribute for void functions.
14342 2018-01-03 Martin Sebor <msebor@redhat.com>
14344 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14345 offset_int::from instead of wide_int::to_shwi.
14346 (maybe_diag_overlap): Remove assertion.
14347 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14348 * gimple-ssa-sprintf.c (format_directive): Same.
14349 (parse_directive): Same.
14350 (sprintf_dom_walker::compute_format_length): Same.
14351 (try_substitute_return_value): Same.
14353 2018-01-03 Jeff Law <law@redhat.com>
14355 PR middle-end/83654
14356 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14357 non-constant residual for zero at runtime and avoid probing in
14358 that case. Reorganize code for trailing problem to mirror handling
14361 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14363 PR tree-optimization/83501
14364 * tree-ssa-strlen.c (get_string_cst): New.
14365 (handle_char_store): Call get_string_cst.
14367 2018-01-03 Martin Liska <mliska@suse.cz>
14369 PR tree-optimization/83593
14370 * tree-ssa-strlen.c: Include tree-cfg.h.
14371 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14372 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14373 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14375 (strlen_dom_walker::before_dom_children): Call
14376 gimple_purge_dead_eh_edges. Dump tranformation with details
14378 (strlen_dom_walker::before_dom_children): Update call by adding
14379 new argument cleanup_eh.
14380 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14382 2018-01-03 Martin Liska <mliska@suse.cz>
14385 * cif-code.def (VARIADIC_THUNK): New enum value.
14386 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14389 2018-01-03 Jan Beulich <jbeulich@suse.com>
14391 * sse.md (mov<mode>_internal): Tighten condition for when to use
14392 vmovdqu<ssescalarsize> for TI and OI modes.
14394 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14396 Update copyright years.
14398 2018-01-03 Martin Liska <mliska@suse.cz>
14401 * ipa-visibility.c (function_and_variable_visibility): Skip
14402 functions with noipa attribure.
14404 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14406 * gcc.c (process_command): Update copyright notice dates.
14407 * gcov-dump.c (print_version): Ditto.
14408 * gcov.c (print_version): Ditto.
14409 * gcov-tool.c (print_version): Ditto.
14410 * gengtype.c (create_file): Ditto.
14411 * doc/cpp.texi: Bump @copying's copyright year.
14412 * doc/cppinternals.texi: Ditto.
14413 * doc/gcc.texi: Ditto.
14414 * doc/gccint.texi: Ditto.
14415 * doc/gcov.texi: Ditto.
14416 * doc/install.texi: Ditto.
14417 * doc/invoke.texi: Ditto.
14419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14421 * vector-builder.h (vector_builder::m_full_nelts): Change from
14422 unsigned int to poly_uint64.
14423 (vector_builder::full_nelts): Update prototype accordingly.
14424 (vector_builder::new_vector): Likewise.
14425 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14426 (vector_builder::operator ==): Likewise.
14427 (vector_builder::finalize): Likewise.
14428 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14429 Take the number of elements as a poly_uint64 rather than an
14431 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14432 from unsigned int to poly_uint64.
14433 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14434 (vec_perm_indices::new_vector): Likewise.
14435 (vec_perm_indices::length): Likewise.
14436 (vec_perm_indices::nelts_per_input): Likewise.
14437 (vec_perm_indices::input_nelts): Likewise.
14438 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14439 number of elements per input as a poly_uint64 rather than an
14440 unsigned int. Use the original encoding for variable-length
14441 vectors, rather than clamping each individual element.
14442 For the second and subsequent elements in each pattern,
14443 clamp the step and base before clamping their sum.
14444 (vec_perm_indices::series_p): Handle polynomial element counts.
14445 (vec_perm_indices::all_in_range_p): Likewise.
14446 (vec_perm_indices_to_tree): Likewise.
14447 (vec_perm_indices_to_rtx): Likewise.
14448 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14449 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14450 (tree_vector_builder::new_binary_operation): Handle polynomial
14451 element counts. Return false if we need to know the number
14452 of elements at compile time.
14453 * fold-const.c (fold_vec_perm): Punt if the number of elements
14454 isn't known at compile time.
14456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14458 * vec-perm-indices.h (vec_perm_builder): Change element type
14459 from HOST_WIDE_INT to poly_int64.
14460 (vec_perm_indices::element_type): Update accordingly.
14461 (vec_perm_indices::clamp): Handle polynomial element_types.
14462 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14463 (vec_perm_indices::all_in_range_p): Likewise.
14464 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14466 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14467 polynomial vec_perm_indices element types.
14468 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14469 * fold-const.c (fold_vec_perm): Likewise.
14470 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14471 * tree-vect-generic.c (lower_vec_perm): Likewise.
14472 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14473 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14474 element type to HOST_WIDE_INT.
14476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14477 Alan Hayward <alan.hayward@arm.com>
14478 David Sherwood <david.sherwood@arm.com>
14480 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14481 rather than an int. Use plus_constant.
14482 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14483 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14486 Alan Hayward <alan.hayward@arm.com>
14487 David Sherwood <david.sherwood@arm.com>
14489 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14490 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
14497 mode sizes. Require a constant size for BLKmode parameters
14498 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14499 forces a parameter to be padded at the lsb end in order to
14500 fill a complete number of words, require the parameter size
14501 to be ordered wrt UNITS_PER_WORD.
14503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14504 Alan Hayward <alan.hayward@arm.com>
14505 David Sherwood <david.sherwood@arm.com>
14507 * reload1.c (spill_stack_slot_width): Change element type
14508 from unsigned int to poly_uint64_pod.
14509 (alter_reg): Treat mode sizes as polynomial.
14511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14512 Alan Hayward <alan.hayward@arm.com>
14513 David Sherwood <david.sherwood@arm.com>
14515 * reload.c (complex_word_subreg_p): New function.
14516 (reload_inner_reg_of_subreg, push_reload): Use it.
14518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14519 Alan Hayward <alan.hayward@arm.com>
14520 David Sherwood <david.sherwood@arm.com>
14522 * lra-constraints.c (process_alt_operands): Reject matched
14523 operands whose sizes aren't ordered.
14524 (match_reload): Refer to this check here.
14526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14527 Alan Hayward <alan.hayward@arm.com>
14528 David Sherwood <david.sherwood@arm.com>
14530 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14531 that the mode size is in the set {1, 2, 4, 8, 16}.
14533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14534 Alan Hayward <alan.hayward@arm.com>
14535 David Sherwood <david.sherwood@arm.com>
14537 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14538 Use plus_constant instead of gen_rtx_PLUS.
14540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14541 Alan Hayward <alan.hayward@arm.com>
14542 David Sherwood <david.sherwood@arm.com>
14544 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14545 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14546 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14547 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14548 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14549 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14550 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14551 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14552 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14553 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14555 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14556 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14557 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14558 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14559 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14560 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14561 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14562 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14563 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14564 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14566 * expr.c (emit_move_resolve_push): Treat the input and result
14567 of PUSH_ROUNDING as a poly_int64.
14568 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14569 (emit_push_insn): Likewise.
14570 * lra-eliminations.c (mark_not_eliminable): Likewise.
14571 * recog.c (push_operand): Likewise.
14572 * reload1.c (elimination_effects): Likewise.
14573 * rtlanal.c (nonzero_bits1): Likewise.
14574 * calls.c (store_one_arg): Likewise. Require the padding to be
14575 known at compile time.
14577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14578 Alan Hayward <alan.hayward@arm.com>
14579 David Sherwood <david.sherwood@arm.com>
14581 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14582 Use plus_constant instead of gen_rtx_PLUS.
14584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14585 Alan Hayward <alan.hayward@arm.com>
14586 David Sherwood <david.sherwood@arm.com>
14588 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14589 rather than an int.
14591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14592 Alan Hayward <alan.hayward@arm.com>
14593 David Sherwood <david.sherwood@arm.com>
14595 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14596 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14597 via stack temporaries. Treat the mode size as polynomial too.
14599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14600 Alan Hayward <alan.hayward@arm.com>
14601 David Sherwood <david.sherwood@arm.com>
14603 * expr.c (expand_expr_real_2): When handling conversions involving
14604 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14605 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14606 as a poly_uint64 too.
14608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14609 Alan Hayward <alan.hayward@arm.com>
14610 David Sherwood <david.sherwood@arm.com>
14612 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
14619 REGMODE_NATURAL_SIZE.
14620 * expmed.c (store_bit_field_1): Likewise.
14621 * expr.c (store_constructor): Likewise.
14622 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14623 and polynomial REGMODE_NATURAL_SIZE.
14624 (gen_lowpart_common): Likewise.
14625 * reginfo.c (record_subregs_of_mode): Likewise.
14626 * rtlanal.c (read_modify_subreg_p): Likewise.
14628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14629 Alan Hayward <alan.hayward@arm.com>
14630 David Sherwood <david.sherwood@arm.com>
14632 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14633 numbers of elements.
14635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14636 Alan Hayward <alan.hayward@arm.com>
14637 David Sherwood <david.sherwood@arm.com>
14639 * match.pd: Cope with polynomial numbers of vector elements.
14641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14642 Alan Hayward <alan.hayward@arm.com>
14643 David Sherwood <david.sherwood@arm.com>
14645 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14646 in a POINTER_PLUS_EXPR.
14648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14649 Alan Hayward <alan.hayward@arm.com>
14650 David Sherwood <david.sherwood@arm.com>
14652 * omp-simd-clone.c (simd_clone_subparts): New function.
14653 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14654 (ipa_simd_modify_function_body): Likewise.
14656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14657 Alan Hayward <alan.hayward@arm.com>
14658 David Sherwood <david.sherwood@arm.com>
14660 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14661 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14662 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14663 (expand_vector_condition, vector_element): Likewise.
14664 (subparts_gt): New function.
14665 (get_compute_type): Use subparts_gt.
14666 (count_type_subparts): Delete.
14667 (expand_vector_operations_1): Use subparts_gt instead of
14668 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
14675 (vect_compile_time_alias): ...this new function. Do the calculation
14676 on poly_ints rather than trees.
14677 (vect_prune_runtime_alias_test_list): Update call accordingly.
14679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14680 Alan Hayward <alan.hayward@arm.com>
14681 David Sherwood <david.sherwood@arm.com>
14683 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14685 (vect_schedule_slp_instance): Likewise.
14687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14688 Alan Hayward <alan.hayward@arm.com>
14689 David Sherwood <david.sherwood@arm.com>
14691 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14692 constant and extern definitions for variable-length vectors.
14693 (vect_get_constant_vectors): Note that the number of units
14694 is known to be constant.
14696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14697 Alan Hayward <alan.hayward@arm.com>
14698 David Sherwood <david.sherwood@arm.com>
14700 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14701 of units as polynomial. Choose between WIDE and NARROW based
14704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14705 Alan Hayward <alan.hayward@arm.com>
14706 David Sherwood <david.sherwood@arm.com>
14708 * tree-vect-stmts.c (simd_clone_subparts): New function.
14709 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14712 Alan Hayward <alan.hayward@arm.com>
14713 David Sherwood <david.sherwood@arm.com>
14715 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14716 vectors as polynomial. Use build_index_vector for
14717 IFN_GOMP_SIMD_LANE.
14719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14720 Alan Hayward <alan.hayward@arm.com>
14721 David Sherwood <david.sherwood@arm.com>
14723 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14724 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14725 for variable-length vectors.
14726 (vectorizable_mask_load_store): Treat the number of units as
14727 polynomial, asserting that it is constant if the condition has
14728 already been enforced.
14729 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
14736 of units as polynomial. Punt if we can't tell at compile time
14737 which vector contains the final result.
14739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14740 Alan Hayward <alan.hayward@arm.com>
14741 David Sherwood <david.sherwood@arm.com>
14743 * tree-vect-loop.c (vectorizable_induction): Treat the number
14744 of units as polynomial. Punt on SLP inductions. Use an integer
14745 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14746 cast of such a series for variable-length floating-point
14749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14750 Alan Hayward <alan.hayward@arm.com>
14751 David Sherwood <david.sherwood@arm.com>
14753 * tree.h (build_index_vector): Declare.
14754 * tree.c (build_index_vector): New function.
14755 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14756 of units as polynomial, forcibly converting it to a constant if
14757 vectorizable_reduction has already enforced the condition.
14758 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14759 to create a {1,2,3,...} vector.
14760 (vectorizable_reduction): Treat the number of units as polynomial.
14761 Choose vectype_in based on the largest scalar element size rather
14762 than the smallest number of units. Enforce the restrictions
14765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14766 Alan Hayward <alan.hayward@arm.com>
14767 David Sherwood <david.sherwood@arm.com>
14769 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14770 number of units as polynomial.
14772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14773 Alan Hayward <alan.hayward@arm.com>
14774 David Sherwood <david.sherwood@arm.com>
14776 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14777 * target.def (autovectorize_vector_sizes): Return the vector sizes
14778 by pointer, using vector_sizes rather than a bitmask.
14779 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14780 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14781 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14783 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14784 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14785 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14786 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14787 * omp-general.c (omp_max_vf): Likewise.
14788 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14789 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14790 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14791 * tree-vect-slp.c (vect_slp_bb): Likewise.
14792 * doc/tm.texi: Regenerate.
14793 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14795 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14796 the vector size as a poly_uint64 rather than an unsigned int.
14797 (current_vector_size): Change from an unsigned int to a poly_uint64.
14798 (get_vectype_for_scalar_type): Update accordingly.
14799 * tree.h (build_truth_vector_type): Take the size and number of
14800 units as a poly_uint64 rather than an unsigned int.
14801 (build_vector_type): Add a temporary overload that takes
14802 the number of units as a poly_uint64 rather than an unsigned int.
14803 * tree.c (make_vector_type): Likewise.
14804 (build_truth_vector_type): Take the number of units as a poly_uint64
14805 rather than an unsigned int.
14807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14808 Alan Hayward <alan.hayward@arm.com>
14809 David Sherwood <david.sherwood@arm.com>
14811 * target.def (get_mask_mode): Take the number of units and length
14812 as poly_uint64s rather than unsigned ints.
14813 * targhooks.h (default_get_mask_mode): Update accordingly.
14814 * targhooks.c (default_get_mask_mode): Likewise.
14815 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14816 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14823 * omp-general.c (omp_max_vf): Likewise.
14824 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14825 (expand_omp_simd): Handle polynomial safelen.
14826 * omp-low.c (omplow_simd_context): Add a default constructor.
14827 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14828 (lower_rec_simd_input_clauses): Update accordingly.
14829 (lower_rec_input_clauses): Likewise.
14831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14832 Alan Hayward <alan.hayward@arm.com>
14833 David Sherwood <david.sherwood@arm.com>
14835 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14836 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14837 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14838 (vect_analyze_slp_cost): Likewise.
14839 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14840 (vect_model_load_cost): Likewise.
14842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14843 Alan Hayward <alan.hayward@arm.com>
14844 David Sherwood <david.sherwood@arm.com>
14846 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14847 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14848 from an unsigned int * to a poly_uint64_pod *.
14849 (calculate_unrolling_factor): New function.
14850 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14853 Alan Hayward <alan.hayward@arm.com>
14854 David Sherwood <david.sherwood@arm.com>
14856 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14857 from an unsigned int to a poly_uint64.
14858 (_loop_vec_info::slp_unrolling_factor): Likewise.
14859 (_loop_vec_info::vectorization_factor): Change from an int
14861 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14862 (vect_get_num_vectors): New function.
14863 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14864 (vect_get_num_copies): Use vect_get_num_vectors.
14865 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14866 to an unsigned int *.
14867 (vect_analyze_data_refs): Change min_vf from an int * to a
14869 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14870 than an unsigned HOST_WIDE_INT.
14871 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14872 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14873 to an unsigned int *.
14874 (vect_analyze_data_ref_dependences): Likewise.
14875 (vect_compute_data_ref_alignment): Handle polynomial vf.
14876 (vect_enhance_data_refs_alignment): Likewise.
14877 (vect_prune_runtime_alias_test_list): Likewise.
14878 (vect_shift_permute_load_chain): Likewise.
14879 (vect_supportable_dr_alignment): Likewise.
14880 (dependence_distance_ge_vf): Take the vectorization factor as a
14881 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14882 (vect_analyze_data_refs): Change min_vf from an int * to a
14884 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14885 vfm1 as a poly_uint64 rather than an int. Make the same change
14886 for the returned bound_scalar.
14887 (vect_gen_vector_loop_niters): Handle polynomial vf.
14888 (vect_do_peeling): Likewise. Update call to
14889 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14890 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14892 * tree-vect-loop.c (vect_determine_vectorization_factor)
14893 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14894 (vect_get_known_peeling_cost): Likewise.
14895 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14896 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14897 (vect_transform_loop): Likewise. Use the lowest possible VF when
14898 updating the upper bounds of the loop.
14899 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14900 rather than an int.
14901 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14902 polynomial unroll factors.
14903 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14904 (vect_make_slp_decision): Likewise.
14905 (vect_supported_load_permutation_p): Likewise, and polynomial
14907 (vect_analyze_slp_cost): Handle polynomial vf.
14908 (vect_slp_analyze_node_operations): Likewise.
14909 (vect_slp_analyze_bb_1): Likewise.
14910 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14911 than an unsigned HOST_WIDE_INT.
14912 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14913 (vectorizable_load): Handle polynomial vf.
14914 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14916 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14919 Alan Hayward <alan.hayward@arm.com>
14920 David Sherwood <david.sherwood@arm.com>
14922 * match.pd: Handle bit operations involving three constants
14923 and try to fold one pair.
14925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14927 * tree-vect-loop-manip.c: Include gimple-fold.h.
14928 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14929 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14930 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14931 Add a path that uses a step of VF instead of 1, but disable it
14933 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14934 and niters_no_overflow parameters. Update calls to
14935 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14936 Create a new SSA name if the latter choses to use a ste other
14937 than zero, and return it via niters_vector_mult_vf_var.
14938 * tree-vect-loop.c (vect_transform_loop): Update calls to
14939 vect_do_peeling, vect_gen_vector_loop_niters and
14940 slpeel_make_loop_iterate_ntimes.
14941 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14942 (vect_gen_vector_loop_niters): Update declarations after above changes.
14944 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14946 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14947 128-bit round to integer instructions.
14948 (ceil<mode>2): Likewise.
14949 (btrunc<mode>2): Likewise.
14950 (round<mode>2): Likewise.
14952 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14954 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14955 unaligned VSX load/store on P8/P9.
14956 (expand_block_clear): Allow the use of unaligned VSX
14957 load/store on P8/P9.
14959 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14961 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14963 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14964 swap associated with both a load and a store.
14966 2018-01-02 Andrew Waterman <andrew@sifive.com>
14968 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14969 * config/riscv/riscv.md (clear_cache): Use it.
14971 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14973 * web.c: Remove out-of-date comment.
14975 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14977 * expr.c (fixup_args_size_notes): Check that any existing
14978 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14979 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14980 (emit_single_push_insn): ...here.
14982 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14984 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14985 (const_vector_encoded_nelts): New function.
14986 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14987 (const_vector_int_elt, const_vector_elt): Declare.
14988 * emit-rtl.c (const_vector_int_elt_1): New function.
14989 (const_vector_elt): Likewise.
14990 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14991 of CONST_VECTOR_ELT.
14993 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14995 * expr.c: Include rtx-vector-builder.h.
14996 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14997 directly on the tree encoding.
14998 (const_vector_from_tree): Likewise.
14999 * optabs.c: Include rtx-vector-builder.h.
15000 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15001 sequence of "u" values.
15002 * vec-perm-indices.c: Include rtx-vector-builder.h.
15003 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15004 directly on the vec_perm_indices encoding.
15006 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15008 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15009 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15010 * rtx-vector-builder.h: New file.
15011 * rtx-vector-builder.c: Likewise.
15012 * rtl.h (rtx_def::u2): Add a const_vector field.
15013 (CONST_VECTOR_NPATTERNS): New macro.
15014 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15015 (CONST_VECTOR_DUPLICATE_P): Likewise.
15016 (CONST_VECTOR_STEPPED_P): Likewise.
15017 (CONST_VECTOR_ENCODED_ELT): Likewise.
15018 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15019 (unwrap_const_vec_duplicate): Likewise.
15020 (const_vec_series_p): Check for a non-duplicated vector encoding.
15021 Say that the function only returns true for integer vectors.
15022 * emit-rtl.c: Include rtx-vector-builder.h.
15023 (gen_const_vec_duplicate_1): Delete.
15024 (gen_const_vector): Call gen_const_vec_duplicate instead of
15025 gen_const_vec_duplicate_1.
15026 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15027 (gen_const_vec_duplicate): Use rtx_vector_builder.
15028 (gen_const_vec_series): Likewise.
15029 (gen_rtx_CONST_VECTOR): Likewise.
15030 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15031 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15032 Build a new vector rather than modifying a CONST_VECTOR in-place.
15033 (handle_special_swappables): Update call accordingly.
15034 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15035 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15036 Build a new vector rather than modifying a CONST_VECTOR in-place.
15037 (handle_special_swappables): Update call accordingly.
15039 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15041 * simplify-rtx.c (simplify_const_binary_operation): Use
15042 CONST_VECTOR_ELT instead of XVECEXP.
15044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15046 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15047 the selector elements to be different from the data elements
15048 if the selector is a VECTOR_CST.
15049 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15050 ssizetype for the selector.
15052 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15054 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15055 before testing each element individually.
15056 * tree-vect-generic.c (lower_vec_perm): Likewise.
15058 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15060 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15061 * selftest-run-tests.c (selftest::run_tests): Call it.
15062 * vector-builder.h (vector_builder::operator ==): New function.
15063 (vector_builder::operator !=): Likewise.
15064 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15065 (vec_perm_indices::all_from_input_p): New function.
15066 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15067 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15068 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15069 instead of reading the VECTOR_CST directly. Detect whether both
15070 vector inputs are the same before constructing the vec_perm_indices,
15071 and update the number of inputs argument accordingly. Use the
15072 utility functions added above. Only construct sel2 if we need to.
15074 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15076 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15077 the broadcast of the low byte.
15078 (expand_mult_highpart): Use an explicit encoding for the permutes.
15079 * optabs-query.c (can_mult_highpart_p): Likewise.
15080 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15081 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15082 (vectorizable_bswap): Likewise.
15083 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15084 explicit encoding for the power-of-2 permutes.
15085 (vect_permute_store_chain): Likewise.
15086 (vect_grouped_load_supported): Likewise.
15087 (vect_permute_load_chain): Likewise.
15089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15091 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15092 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15093 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15094 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15095 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15096 (vect_gen_perm_mask_any): Likewise.
15098 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15100 * int-vector-builder.h: New file.
15101 * vec-perm-indices.h: Include int-vector-builder.h.
15102 (vec_perm_indices): Redefine as an int_vector_builder.
15103 (auto_vec_perm_indices): Delete.
15104 (vec_perm_builder): Redefine as a stand-alone class.
15105 (vec_perm_indices::vec_perm_indices): New function.
15106 (vec_perm_indices::clamp): Likewise.
15107 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15108 (vec_perm_indices::new_vector): New function.
15109 (vec_perm_indices::new_expanded_vector): Update for new
15110 vec_perm_indices class.
15111 (vec_perm_indices::rotate_inputs): New function.
15112 (vec_perm_indices::all_in_range_p): Operate directly on the
15113 encoded form, without computing elided elements.
15114 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15115 encoding. Update for new vec_perm_indices class.
15116 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15117 the given vec_perm_builder.
15118 (expand_vec_perm_var): Update vec_perm_builder constructor.
15119 (expand_mult_highpart): Use vec_perm_builder instead of
15120 auto_vec_perm_indices.
15121 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15122 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15123 or double series encoding as appropriate.
15124 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15125 vec_perm_indices instead of auto_vec_perm_indices.
15126 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15127 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15128 (vect_permute_store_chain): Likewise.
15129 (vect_grouped_load_supported): Likewise.
15130 (vect_permute_load_chain): Likewise.
15131 (vect_shift_permute_load_chain): Likewise.
15132 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15133 (vect_transform_slp_perm_load): Likewise.
15134 (vect_schedule_slp_instance): Likewise.
15135 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15136 (vectorizable_mask_load_store): Likewise.
15137 (vectorizable_bswap): Likewise.
15138 (vectorizable_store): Likewise.
15139 (vectorizable_load): Likewise.
15140 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15141 vec_perm_indices instead of auto_vec_perm_indices. Use
15142 tree_to_vec_perm_builder to read the vector from a tree.
15143 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15144 vec_perm_builder instead of a vec_perm_indices.
15145 (have_whole_vector_shift): Use vec_perm_builder and
15146 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15147 truncation to calc_vec_perm_mask_for_shift.
15148 (vect_create_epilog_for_reduction): Likewise.
15149 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15150 from auto_vec_perm_indices to vec_perm_indices.
15151 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15152 instead of changing individual elements.
15153 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15154 the vector in d.perm.
15155 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15156 from auto_vec_perm_indices to vec_perm_indices.
15157 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15158 instead of changing individual elements.
15159 (arm_vectorize_vec_perm_const): Use new_vector to install
15160 the vector in d.perm.
15161 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15162 Update vec_perm_builder constructor.
15163 (rs6000_expand_interleave): Likewise.
15164 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15165 (rs6000_expand_interleave): Likewise.
15167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15169 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15170 to qimode could truncate the indices.
15171 * optabs.c (expand_vec_perm_var): Likewise.
15173 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15175 * Makefile.in (OBJS): Add vec-perm-indices.o.
15176 * vec-perm-indices.h: New file.
15177 * vec-perm-indices.c: Likewise.
15178 * target.h (vec_perm_indices): Replace with a forward class
15180 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15181 * optabs.h: Include vec-perm-indices.h.
15182 (expand_vec_perm): Delete.
15183 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15184 (expand_vec_perm_const): Declare.
15185 * target.def (vec_perm_const_ok): Replace with...
15186 (vec_perm_const): ...this new hook.
15187 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15188 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15189 * doc/tm.texi: Regenerate.
15190 * optabs.def (vec_perm_const): Delete.
15191 * doc/md.texi (vec_perm_const): Likewise.
15192 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15193 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15194 expand_vec_perm for constant permutation vectors. Assert that
15195 the mode of variable permutation vectors is the integer equivalent
15196 of the mode that is being permuted.
15197 * optabs-query.h (selector_fits_mode_p): Declare.
15198 * optabs-query.c: Include vec-perm-indices.h.
15199 (selector_fits_mode_p): New function.
15200 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15201 is defined, instead of checking whether the vec_perm_const_optab
15202 exists. Use targetm.vectorize.vec_perm_const instead of
15203 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15204 fit in the vector mode before using a variable permute.
15205 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15206 vec_perm_indices instead of an rtx.
15207 (expand_vec_perm): Replace with...
15208 (expand_vec_perm_const): ...this new function. Take the selector
15209 as a vec_perm_indices rather than an rtx. Also take the mode of
15210 the selector. Update call to shift_amt_for_vec_perm_mask.
15211 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15212 Use vec_perm_indices::new_expanded_vector to expand the original
15213 selector into bytes. Check whether the indices fit in the vector
15214 mode before using a variable permute.
15215 (expand_vec_perm_var): Make global.
15216 (expand_mult_highpart): Use expand_vec_perm_const.
15217 * fold-const.c: Includes vec-perm-indices.h.
15218 * tree-ssa-forwprop.c: Likewise.
15219 * tree-vect-data-refs.c: Likewise.
15220 * tree-vect-generic.c: Likewise.
15221 * tree-vect-loop.c: Likewise.
15222 * tree-vect-slp.c: Likewise.
15223 * tree-vect-stmts.c: Likewise.
15224 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15226 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15227 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15228 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15229 (aarch64_vectorize_vec_perm_const): ...this new function.
15230 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15231 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15232 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15233 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15234 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15235 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15236 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15238 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15239 check for NEON modes.
15240 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15241 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15242 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15243 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15245 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15246 the old VEC_PERM_CONST conditions.
15247 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15248 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15249 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15250 (ia64_vectorize_vec_perm_const_ok): Merge into...
15251 (ia64_vectorize_vec_perm_const): ...this new function.
15252 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15253 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15254 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15255 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15256 * config/mips/mips.c (mips_expand_vec_perm_const)
15257 (mips_vectorize_vec_perm_const_ok): Merge into...
15258 (mips_vectorize_vec_perm_const): ...this new function.
15259 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15260 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15261 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15262 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15263 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15264 (rs6000_expand_vec_perm_const): Delete.
15265 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15267 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15268 (altivec_expand_vec_perm_const_le): Take each operand individually.
15269 Operate on constant selectors rather than rtxes.
15270 (altivec_expand_vec_perm_const): Likewise. Update call to
15271 altivec_expand_vec_perm_const_le.
15272 (rs6000_expand_vec_perm_const): Delete.
15273 (rs6000_vectorize_vec_perm_const_ok): Delete.
15274 (rs6000_vectorize_vec_perm_const): New function.
15275 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15276 an element count and rtx array.
15277 (rs6000_expand_extract_even): Update call accordingly.
15278 (rs6000_expand_interleave): Likewise.
15279 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15280 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15281 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15282 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15283 (rs6000_expand_vec_perm_const): Delete.
15284 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15285 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15286 (altivec_expand_vec_perm_const_le): Take each operand individually.
15287 Operate on constant selectors rather than rtxes.
15288 (altivec_expand_vec_perm_const): Likewise. Update call to
15289 altivec_expand_vec_perm_const_le.
15290 (rs6000_expand_vec_perm_const): Delete.
15291 (rs6000_vectorize_vec_perm_const_ok): Delete.
15292 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15293 reference to the SPE evmerge intructions.
15294 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15295 an element count and rtx array.
15296 (rs6000_expand_extract_even): Update call accordingly.
15297 (rs6000_expand_interleave): Likewise.
15298 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15299 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15301 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15303 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15305 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15306 vector mode and that that mode matches the mode of the data
15308 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15309 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15310 directly using expand_vec_perm_1 when forcing selectors into
15312 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15314 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15316 * optabs-query.h (can_vec_perm_p): Delete.
15317 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15318 * optabs-query.c (can_vec_perm_p): Split into...
15319 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15320 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15321 particular selector is valid.
15322 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15323 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15324 (vect_grouped_load_supported): Likewise.
15325 (vect_shift_permute_load_chain): Likewise.
15326 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15327 (vect_transform_slp_perm_load): Likewise.
15328 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15329 (vectorizable_bswap): Likewise.
15330 (vect_gen_perm_mask_checked): Likewise.
15331 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15332 implementations of variable permutation vectors into account
15333 when deciding which selector to use.
15334 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15335 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15336 with a false third argument.
15337 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15338 to test whether the constant selector is valid and can_vec_perm_var_p
15339 to test whether a variable selector is valid.
15341 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15343 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15344 * optabs-query.c (can_vec_perm_p): Likewise.
15345 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15346 instead of vec_perm_indices.
15347 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15348 (vect_gen_perm_mask_checked): Likewise,
15349 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15350 (vect_gen_perm_mask_checked): Likewise,
15352 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15354 * optabs-query.h (qimode_for_vec_perm): Declare.
15355 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15356 (qimode_for_vec_perm): ...this new function.
15357 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15359 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15361 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15362 does not have a conditional at the top.
15364 2018-01-02 Richard Biener <rguenther@suse.de>
15366 * ipa-inline.c (big_speedup_p): Fix expression.
15368 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15371 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15374 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15378 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15379 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15380 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15381 cond_taken_branch_cost 3->4.
15383 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15385 PR tree-optimization/83581
15386 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15387 TODO_cleanup_cfg if any changes have been made.
15389 PR middle-end/83608
15390 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15391 convert_modes if target mode has the right side, but different mode
15394 PR middle-end/83609
15395 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15396 last argument when extracting from CONCAT. If either from_real or
15397 from_imag is NULL, use expansion through memory. If result is not
15398 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15399 the parts directly to inner mode, if even that fails, use expansion
15402 PR middle-end/83623
15403 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15404 check for bswap in mode rather than HImode and use that in expand_unop
15407 Copyright (C) 2018 Free Software Foundation, Inc.
15409 Copying and distribution of this file, with or without modification,
15410 are permitted in any medium without royalty provided the copyright
15411 notice and this notice are preserved.