gcc/ChangeLog

   1 2018-06-05  Martin Liska  <mliska@suse.cz>
   2
   3         PR gcov-profile/47618
   4         * doc/invoke.texi: Document how -fprofile-dir format
   5         is extended.
   6
   7 2018-06-05  Richard Biener  <rguenther@suse.de>
   8
   9         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
  10         removal pretend DOM info isn't available so we do not update
  11         it and only remove edges, not dominated blocks.  Actually free
  12         DOM info in case we removed something.  Remove unreachable blocks.
  13         (mfb_keep_latches): Work with either DOM info or marked backedges.
  14         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
  15         first.  Mark backedges if DOM info isn't available.
  16         (Re-)compute DOM info after cleanup_control_flow_pre.
  17
  18 2018-06-05  Richard Biener  <rguenther@suse.de>
  19
  20         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
  21         (locus_discrim_hasher::hash): Adjust.
  22         (locus_discrim_hasher::equal): Likewise.
  23         (next_discriminator_for_locus): Work on line directly.
  24         (same_line_p): Pass in expanded locus1 as well.
  25         (assign_discriminators): Avoid redundant location expansions.
  26
  27 2018-06-05  Richard Biener  <rguenther@suse.de>
  28
  29         PR tree-optimization/86046
  30         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
  31         if required after clearing TREE_ADDRESSABLE.
  32
  33 2018-06-05  Richard Biener  <rguenther@suse.de>
  34
  35         PR tree-optimization/86047
  36         * tree-ssa-loop.c (for_each_index): Glob handling of all
  37         decls and constants and really handle all of them.
  38
  39 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
  40
  41         PR target/81497
  42         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
  43         qualifier_void_pointer and qualifier_const_void_pointer.
  44         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
  45         (arm_init_builtins): Handle the above.
  46         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
  47         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
  48         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
  49         void intrinsics.
  50
  51 2018-06-05  Martin Liska  <mliska@suse.cz>
  52
  53         * auto-profile.c (read_autofdo_file): Do not use
  54         gcov_ctr_summary struct.
  55         (afdo_callsite_hot_enough_for_early_inline): Likewise.
  56         * coverage.c (struct counts_entry): Likewise.
  57         (read_counts_file): Read just single summary entry.
  58         (get_coverage_counts): Use gcov_summary struct.
  59         * coverage.h (get_coverage_counts): Likewise.
  60         * gcov-dump.c (dump_working_sets): Likewise.
  61         (tag_summary): Dump just single summary.
  62         * gcov-io.c (gcov_write_summary): Write just histogram
  63         summary.
  64         (gcov_read_summary): Read just single summary.
  65         (compute_working_sets): Use gcov_summary struct.
  66         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
  67         of GCOV_COUNTERS_SUMMABLE.
  68         (GCOV_COUNTERS_SUMMABLE): Remove.
  69         (GCOV_FIRST_VALUE_COUNTER): Replace with
  70         GCOV_COUNTER_V_INTERVAL.
  71         (struct gcov_ctr_summary): Remove.
  72         (struct gcov_summary): Directly use fields of former
  73         gcov_ctr_summary.
  74         (compute_working_sets): Use gcov_summary struct.
  75         * gcov.c (read_count_file): Do not use ctrs fields.
  76         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
  77         struct.
  78         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
  79         struct.
  80         * profile.c: Likewise.
  81         * profile.h: Likewise.
  82
  83 2018-06-05  Martin Liska  <mliska@suse.cz>
  84
  85         PR gcov-profile/84846
  86         * gcov.c (output_lines): Print working directory only
  87         in intermediate format.
  88
  89 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
  90
  91         * config/s390/s390-builtin-types.def: Add void function type.
  92         * config/s390/s390-builtins.def: Use the function type for the
  93         tbeginc builtin.
  94
  95 2018-06-04  Jim Wilson  <jimw@sifive.com>
  96
  97         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
  98         to int.
  99         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
 100         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
 101         handle EH_RETURN_DATA_REGNO registers properly.
 102         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
 103         (riscv_expand_epilogue): Update comment.  Change argument name and
 104         type.  Update code to use new name and type.  Pass new args to
 105         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
 106         EXCEPTION_RETURN.
 107         * config/riscv/riscv.md (NORMAL_RETURN): New.
 108         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
 109         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
 110         (eh_return): Call gen_eh_return_internal and emit barrier.
 111         (eh_return_internal): Call riscv_expand_epilogue.
 112
 113 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
 114
 115         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
 116         bit_insertion field and declare can_be_merged_into method.
 117         (merged_store_group::can_be_merged_into): New method.
 118         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
 119         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
 120         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
 121
 122 2018-06-04  Richard Biener  <rguenther@suse.de>
 123
 124         PR tree-optimization/85955
 125         * builtins.c (fold_builtin_sincos): Convert pointers to
 126         destination to appropriate type before dereferencing.
 127
 128 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
 129
 130         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
 131
 132 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
 133
 134         * expr.c (expand_expr_real_1): Force the operand into memory if
 135         its TYPE_MODE is BLKmode and if there is no integer mode for
 136         the number of bits being extracted.
 137
 138 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
 139
 140         PR target/85832
 141         PR target/86036
 142         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
 143         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
 144
 145 2018-06-04  Richard Biener  <rguenther@suse.de>
 146
 147         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
 148         (cleanup_tree_cfg_noloop): ... single caller.  Do
 149         start_recording_case_labels later.
 150
 151 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
 152
 153         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
 154         to _IMMINTRIN_H_INCLUDED.
 155         * config/i386/pconfigintrin.h: Ditto.
 156         * config/i386/waitpkgintrin.h: Ditto.
 157         * config/i386/immintrin.h: Add includes for sgxintrin.h,
 158         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
 159         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
 160         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
 161         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
 162         waitpkgintrin.h and cldemoteintrin.h.
 163
 164 2018-06-04  Richard Biener  <rguenther@suse.de>
 165
 166         PR tree-optimization/86038
 167         * tracer.c (find_best_successor): Check probability for
 168         being initialized, bail out if not.
 169
 170 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
 171
 172         PR target/86003
 173         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
 174         of bits to ignore when comparing architectures.
 175
 176 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
 177
 178         PR tree-optimization/69615
 179         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
 180         maximum or minimum of the type, try to merge it also as if
 181         range1 is + [-, x - 1] or + [x + 1, -].
 182
 183         PR c++/86025
 184         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
 185
 186 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
 187
 188         PR tree-optimization/86034
 189         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
 190         the unsigned bitfield type in a bit insertion sequence if it does not
 191         have a larger precision than the bitfield size.
 192         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
 193
 194 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
 195
 196         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
 197
 198 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
 199
 200         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
 201         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
 202         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
 203         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
 204
 205 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
 206
 207         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
 208         Disable -fdelete-null-pointer-checks for ELF toolchain.
 209
 210 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
 211             Kito Cheng  <kito.cheng@gmail.com>
 212
 213         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
 214         (nds32le-*-*, nds32be-*-*): Integrate checking process.
 215         (nds32*-*-*): Add glibc and uclibc conditions.
 216         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
 217         (TARGET_EXCEPT_UNWIND_INFO): Define.
 218         * config/nds32/elf.h: New file.
 219         * config/nds32/linux.h: New file.
 220         * config/nds32/nds32-elf.opt: New file.
 221         * config/nds32/nds32-linux.opt: New file.
 222         * config/nds32/nds32-fp-as-gp.c
 223         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
 224         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
 225         TARGET_LINUX_ABI.
 226         (nds32_asm_file_end): Ditto.
 227         (nds32_print_operand): Ditto.
 228         (nds32_insert_attributes): Ditto.
 229         (nds32_init_libfuncs): New function.
 230         (TARGET_HAVE_TLS): Define.
 231         (TARGET_INIT_LIBFUNCS): Define.
 232         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
 233         spec content.
 234         (TARGET_ELF): Apply different mcmodel setting.
 235         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
 236         been migrated into elf.h and linux.h files.
 237         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
 238         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
 239         (mcmodel): The content has been migrated into nds32-elf.opt and
 240         nds32-linux.opt files.
 241         * config/nds32/t-elf: New file.
 242         * config/nds32/t-linux: New file.
 243
 244 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
 245             Shiva Chen  <shiva0217@gmail.com>
 246
 247         * config/nds32/constants.md (unspec_volatile_element): Add
 248         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
 249         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
 250         optimization.
 251         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
 252         (make_pass_nds32_fp_as_gp): Declare.
 253         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
 254         optmization pass.
 255         (nds32_asm_function_end_prologue): Remove unused asm output.
 256         (nds32_asm_function_begin_epilogue): Remove unused asm output.
 257         (nds32_asm_file_start): Output necessary fp_as_gp information.
 258         (nds32_option_override): Adjust register usage.
 259         (nds32_expand_prologue): Consider fp_as_gp situation.
 260         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
 261         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
 262         (epilogue): Ditto.
 263         (return): Ditto.
 264         (simple_return): Ditto.
 265         (omit_fp_begin): Output special directive for fp_as_gp.
 266         (omit_fp_end): Output special directive for fp_as_gp.
 267         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
 268         mforbid-fp-as-gp): New options.
 269
 270 2018-06-01  Mark Wielaard  <mark@klomp.org>
 271
 272         * dwarf2out.c (dwarf2out_finish): Remove generation of
 273         DW_AT_loclists_base.
 274
 275 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
 276
 277         * gimple-ssa-store-merging.c: Include gimple-fold.h.
 278         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
 279         (struct merged_store_group): Add bit_insertion field.
 280         (dump_char_array): Use standard hexadecimal format.
 281         (merged_store_group::merged_store_group): Set bit_insertion to false.
 282         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
 283         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
 284         also print the mask in the dump file.
 285         (pass_store_merging::gate): Minor tweak.
 286         (imm_store_chain_info::coalesce_immediate): Fix wrong association
 287         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
 288         stores with INTEGER_CST stores.
 289         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
 290         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
 291         and use it throughout.  Generate bit insertion sequences if need be.
 292         (pass_store_merging::process_store): Remove redundant condition.
 293         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
 294
 295 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
 296
 297         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
 298         the 128-bit floating point types.  Fix function comment.
 299
 300 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 301
 302         * config/aarch64/aarch64-simd.md
 303         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
 304         mnemonics.
 305         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
 306         mnemonics.
 307
 308 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
 309
 310         PR tree-optimization/85989
 311         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
 312         variable.
 313         (backprop::intersect_uses): Check it when deciding whether this
 314         is a backedge reference.
 315         (backprop::process_block): Add each phi to m_visited_phis
 316         after visiting it, then clear it at the end.
 317
 318 2018-06-01  Richard Biener  <rguenther@suse.de>
 319
 320         * tree-vectorizer.h (vect_dr_stmt): New function.
 321         (vect_get_load_cost): Adjust.
 322         (vect_get_store_cost): Likewise.
 323         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
 324         Use vect_dr_stmt instead of DR_SMTT.
 325         (vect_record_base_alignments): Likewise.
 326         (vect_calculate_target_alignment): Likewise.
 327         (vect_compute_data_ref_alignment): Likewise and make static.
 328         (vect_update_misalignment_for_peel): Likewise.
 329         (vect_verify_datarefs_alignment): Likewise.
 330         (vector_alignment_reachable_p): Likewise.
 331         (vect_get_data_access_cost): Likewise.  Pass down
 332         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
 333         (vect_get_peeling_costs_all_drs): Likewise.
 334         (vect_peeling_hash_get_lowest_cost): Likewise.
 335         (vect_enhance_data_refs_alignment): Likewise.
 336         (vect_find_same_alignment_drs): Likewise.
 337         (vect_analyze_data_refs_alignment): Likewise.
 338         (vect_analyze_group_access_1): Likewise.
 339         (vect_analyze_group_access): Likewise.
 340         (vect_analyze_data_ref_access): Likewise.
 341         (vect_analyze_data_ref_accesses): Likewise.
 342         (vect_vfa_segment_size): Likewise.
 343         (vect_small_gap_p): Likewise.
 344         (vectorizable_with_step_bound_p): Likewise.
 345         (vect_prune_runtime_alias_test_list): Likewise.
 346         (vect_analyze_data_refs): Likewise.
 347         (vect_supportable_dr_alignment): Likewise.
 348         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
 349         (vect_gen_prolog_loop_niters): Likewise.
 350         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
 351         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
 352         modify DR_STMT.
 353         (vect_recog_mask_conversion_pattern): Likewise.
 354         (vect_try_gather_scatter_pattern): Likewise.
 355         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
 356         to vect_get_store_cost.
 357         (vect_get_store_cost): Get stmt_info instead of DR.
 358         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
 359         (vect_get_load_cost): Get stmt_info instead of DR.
 360
 361 2018-06-01  Richard Biener  <rguenther@suse.de>
 362
 363         PR middle-end/86017
 364         * gimple-fold.c (var_decl_component_p): Also allow offsetted
 365         vars wrapped in MEM_REFs.
 366
 367 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
 368
 369         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
 370         Fix subreg tests so that we only return a choice between
 371         GENERAL_REGS and FP_REGS if the original classes included both.
 372
 373 2018-06-01  Richard Biener  <rguenther@suse.de>
 374
 375         PR ipa/85960
 376         * tree-ssa-structalias.c (get_function_part_constraint):
 377         Handle NULL fi->decl.
 378         (find_func_aliases_for_call): Properly handle indirect
 379         fi from direct call.
 380         (find_func_clobbers): Likewise.
 381         (ipa_pta_execute): Likewise.
 382         (create_variable_info_for): For functions that are ifunc_resolver
 383         resolve to a varinfo that contains the result of the resolver
 384         call.
 385         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
 386         aliases.
 387
 388 2018-05-31  Michael Collison  <michael.collison@arm.com>
 389
 390         * config/aarch64/aarch64.md:
 391         (*fix_to_zero_extenddfdi2): New pattern.
 392         * gcc.target/aarch64/fix_extend1.c: New testcase.
 393
 394 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
 395
 396         PR middle-end/78809
 397         PR middle-end/83026
 398         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
 399         and BUILT_IN_STRNCMP_EQ.
 400         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
 401         BUILT_IN_STRNCMP_EQ.
 402         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
 403         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
 404         (gimple_fold_builtin): Likewise.
 405         * tree-ssa-strlen.c (compute_string_length): New function.
 406         (determine_min_obsize): New function.
 407         (handle_builtin_string_cmp): New function to handle calls to
 408         string compare functions.
 409         (strlen_optimize_stmt): Add handling to builtin string compare
 410         calls.
 411         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
 412         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
 413         * tree.c (build_common_builtin_nodes): Add new defines of
 414         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
 415
 416 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
 417
 418         PR target/85984
 419         * bb-reorder.c (pass_partition_blocks::gate): Return false for
 420         functions with naked attribute.
 421
 422 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
 423
 424         * config/i386/sse.md (avx_vec_concat<mode>):
 425         Substitute concat_tg_mode mode attribute with xtg_mode.
 426         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
 427         (concat_tg_mode): Remove mode attribute.
 428
 429 2018-05-31  Martin Sebor  <msebor@redhat.com>
 430
 431         PR c/82063
 432         * calls.c (alloc_max_size): Correct a logic error/typo.
 433         Treat excessive arguments as infinite.  Warn for invalid arguments.
 434         * doc/invoke.texi (-Walloc-size-larger-than): Update.
 435
 436 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
 437
 438         PR target/85829
 439         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
 440         and movx for Haswell.
 441
 442 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
 443             Cesar Philippidis  <cesar@codesourcery.com>
 444
 445         PR middle-end/85879
 446         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
 447         when emitting error on private/firstprivate reductions.
 448         * omp-low.c (lower_omp_target): Avoid reference-type processing
 449         on pointers for firstprivate clause.
 450
 451 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
 452
 453         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
 454         (st1x2): Likewise.
 455         (st1x3): Likewise.
 456         * config/aarch64/aarch64-simd.md
 457         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
 458         (aarch64_ld1_x3_<mode>): Likewise
 459         (aarch64_st1x2<VALLDIF:mode>): Likewise
 460         (aarch64_st1_x2_<mode>): Likewise
 461         (aarch64_st1x3<VALLDIF:mode>): Likewise
 462         (aarch64_st1_x3_<mode>): Likewise
 463         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
 464         (vld1_s8_x3): Likewise.
 465         (vld1_u16_x3): Likewise.
 466         (vld1_s16_x3): Likewise.
 467         (vld1_u32_x3): Likewise.
 468         (vld1_s32_x3): Likewise.
 469         (vld1_u64_x3): Likewise.
 470         (vld1_s64_x3): Likewise.
 471         (vld1_f16_x3): Likewise.
 472         (vld1_f32_x3): Likewise.
 473         (vld1_f64_x3): Likewise.
 474         (vld1_p8_x3): Likewise.
 475         (vld1_p16_x3): Likewise.
 476         (vld1_p64_x3): Likewise.
 477         (vld1q_u8_x3): Likewise.
 478         (vld1q_s8_x3): Likewise.
 479         (vld1q_u16_x3): Likewise.
 480         (vld1q_s16_x3): Likewise.
 481         (vld1q_u32_x3): Likewise.
 482         (vld1q_s32_x3): Likewise.
 483         (vld1q_u64_x3): Likewise.
 484         (vld1q_s64_x3): Likewise.
 485         (vld1q_f16_x3): Likewise.
 486         (vld1q_f32_x3): Likewise.
 487         (vld1q_f64_x3): Likewise.
 488         (vld1q_p8_x3): Likewise.
 489         (vld1q_p16_x3): Likewise.
 490         (vld1q_p64_x3): Likewise.
 491         (vst1_s64_x2): Likewise.
 492         (vst1_u64_x2): Likewise.
 493         (vst1_f64_x2): Likewise.
 494         (vst1_s8_x2): Likewise.
 495         (vst1_p8_x2): Likewise.
 496         (vst1_s16_x2): Likewise.
 497         (vst1_p16_x2): Likewise.
 498         (vst1_s32_x2): Likewise.
 499         (vst1_u8_x2): Likewise.
 500         (vst1_u16_x2): Likewise.
 501         (vst1_u32_x2): Likewise.
 502         (vst1_f16_x2): Likewise.
 503         (vst1_f32_x2): Likewise.
 504         (vst1_p64_x2): Likewise.
 505         (vst1q_s8_x2): Likewise.
 506         (vst1q_p8_x2): Likewise.
 507         (vst1q_s16_x2): Likewise.
 508         (vst1q_p16_x2): Likewise.
 509         (vst1q_s32_x2): Likewise.
 510         (vst1q_s64_x2): Likewise.
 511         (vst1q_u8_x2): Likewise.
 512         (vst1q_u16_x2): Likewise.
 513         (vst1q_u32_x2): Likewise.
 514         (vst1q_u64_x2): Likewise.
 515         (vst1q_f16_x2): Likewise.
 516         (vst1q_f32_x2): Likewise.
 517         (vst1q_f64_x2): Likewise.
 518         (vst1q_p64_x2): Likewise.
 519         (vst1_s64_x3): Likewise.
 520         (vst1_u64_x3): Likewise.
 521         (vst1_f64_x3): Likewise.
 522         (vst1_s8_x3): Likewise.
 523         (vst1_p8_x3): Likewise.
 524         (vst1_s16_x3): Likewise.
 525         (vst1_p16_x3): Likewise.
 526         (vst1_s32_x3): Likewise.
 527         (vst1_u8_x3): Likewise.
 528         (vst1_u16_x3): Likewise.
 529         (vst1_u32_x3): Likewise.
 530         (vst1_f16_x3): Likewise.
 531         (vst1_f32_x3): Likewise.
 532         (vst1_p64_x3): Likewise.
 533         (vst1q_s8_x3): Likewise.
 534         (vst1q_p8_x3): Likewise.
 535         (vst1q_s16_x3): Likewise.
 536         (vst1q_p16_x3): Likewise.
 537         (vst1q_s32_x3): Likewise.
 538         (vst1q_s64_x3): Likewise.
 539         (vst1q_u8_x3): Likewise.
 540         (vst1q_u16_x3): Likewise.
 541         (vst1q_u32_x3): Likewise.
 542         (vst1q_u64_x3): Likewise.
 543         (vst1q_f16_x3): Likewise.
 544         (vst1q_f32_x3): Likewise.
 545         (vst1q_f64_x3): Likewise.
 546         (vst1q_p64_x3): Likewise.
 547
 548 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
 549
 550         * config/msp430/msp430.c (msp430_output_labelref): Prepend
 551         user_label_prefix to name.
 552
 553         * tree-core.h: Update comment about the format of NAME string
 554         passed to handler in attribute_spec.
 555
 556         * config/msp430/msp430.md: Remove erroneous subreg expression from
 557         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
 558         zero_extend{q,h}isi2.
 559
 560 2018-05-30  Borislav Petkov  <bp@suse.de>
 561
 562         * doc/extend.texi: Document some architecture specific
 563         constraints and sort entries.
 564
 565 2018-05-30  Martin Sebor  <msebor@redhat.com>
 566
 567         PR middle-end/85369
 568         * builtins.c (expand_builtin_stpcpy_1): New function.
 569         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
 570         only if the former succeeds.
 571
 572 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
 573
 574         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
 575         in saphira.
 576
 577 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 578
 579         * doc/invoke.texi (-flinker-output): Document
 580
 581 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 582
 583         * passes.c (ipa_write_summaries): Only modify statements if body
 584         is in memory.
 585         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
 586         incrementally linking.
 587         (ipa_passes): Likewise.
 588         * lto-cgraph.c (lto_output_node): When incrementally linking do not
 589         pass down resolution info.
 590         * common.opt (flag_incremental_link): Update info.
 591         * gcc.c (plugin specs): Turn flinker-output=* to
 592         -plugin-opt=-linker-output-known
 593         * toplev.c (compile_file): Also cut compilation when doing incremental
 594         link.
 595         * flag-types. (enum lto_partition_model): Add
 596         LTO_LINKER_OUTPUT_NOLTOREL.
 597         (invoke.texi): Add -flinker-output docs.
 598         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
 599         link same way as WPA; do not stream in dead initializers.
 600
 601         * dwarf2out.c (dwarf2out_die_ref_for_decl,
 602         darf2out_register_external_decl): Support incremental link.
 603
 604 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 605
 606         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
 607
 608 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 609
 610         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
 611         it down to simple_object_copy_lto_debug_sections.
 612         (run_gcc): Determine incremental LTO link time and configure
 613         lto1 into non-wpa mode, disable renaming of debug sections.
 614
 615 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 616
 617         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
 618         descriptions of various incorrectly documented functions.
 619
 620 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
 621
 622         Revert:
 623         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
 624         address check not strict.
 625
 626 2018-05-30  Richard Biener  <rguenther@suse.de>
 627
 628         PR tree-optimization/85964
 629         * tracer.c (better_p): Drop initialized count check, we only
 630         call the function with initialized counts now.
 631         (find_best_successor): Do find a best edge if one
 632         has uninitialized count.
 633         (find_best_predecessor): Likewise.  Do BB frequency check only
 634         if count is initialized.
 635
 636 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
 637
 638         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
 639         (aarch64_ldrstr_offset_compare): New.
 640         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
 641         load/store orderings.
 642         (aarch64_gen_adjusted_ldpstp): Likewise.
 643
 644 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
 645
 646         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
 647         Check for subset of GENERAL_REGS and FP_REGS.
 648         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
 649         r=w alternative.
 650
 651 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
 652
 653         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
 654         and wi::to_poly_offset.  Add the current offset and then check
 655         whether the sum fits, rather than using an unchecked addition of
 656         a checked term.  Check for a shwi rather than a uhwi.
 657         * expr.c (get_bit_range): Use tree_to_poly_uint64.
 658         (store_constructor): Use poly_int_tree_p.
 659         (expand_expr_real_1): Likewise.
 660         * function.c (assign_temp): Likewise.
 661         * fold-const.c (const_binop): Use poly_int_tree_p and
 662         wi::to_poly_offset.
 663         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
 664         division.
 665         * ipa-icf-gimple.c (func_checker::compare_operand): Use
 666         to_poly_offset for MEM offsets.
 667         * ipa-icf.c (sem_variable::equals): Likewise.
 668         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
 669         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
 670         wi::to_poly_offset for BIT_FIELD_REF offsets.
 671         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
 672         wi::to_poly_offset.
 673         * var-tracking.c (emit_note_insn_var_location): Use
 674         tree_to_poly_uint64.
 675
 676 2018-05-29  Jim Wilson  <jimw@sifive.com>
 677
 678         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
 679
 680 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
 681
 682         PR target/85950
 683         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
 684         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
 685         sequence.
 686         (sse4_1_round<mode>2): Use nonimmediate_operand
 687         for operand 1 predicate.
 688
 689 2018-05-29  Martin Sebor  <msebor@redhat.com>
 690             Richard Biener  <rguenther@suse.de>
 691
 692         PR testsuite/85888
 693         * calls.c (get_size_range): Call determine_value_range instead
 694         of get_value_range..
 695         * tree-vrp.h (determine_value_range): Declared new function.
 696         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
 697
 698 2018-05-29  Richard Biener  <rguenther@suse.de>
 699
 700         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
 701         sure to use non-pattern stmts for get_earlier_stmt arguments.
 702         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
 703         called on pattern stmts.
 704         (get_later_stmt): Likewise.
 705
 706 2018-05-29  Martin Liska  <mliska@suse.cz>
 707
 708         PR gcov-profile/85759
 709         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
 710         env variables.
 711
 712 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
 713
 714         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
 715         VEC_UNPACK_*_EXPR.
 716         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
 717         VEC_PACK_*_EXPR.
 718
 719         PR target/85918
 720         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
 721         VEC_PACK_FLOAT_EXPR): New tree codes.
 722         * tree-pretty-print.c (op_code_prio): Handle
 723         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
 724         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
 725         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
 726         * tree-inline.c (estimate_operator_cost): Likewise.
 727         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
 728         * fold-const.c (const_binop): Likewise.
 729         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
 730         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
 731         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
 732         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
 733         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
 734         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
 735         * expr.c (expand_expr_real_2): Likewise.
 736         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
 737         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
 738         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
 739         optabs.
 740         * optabs.c (expand_widen_pattern_expr): For
 741         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
 742         sign from result type rather than operand's type.
 743         (expand_binop_directly): For vec_packu_float_optab and
 744         vec_packs_float_optab allow result type to be different from operand's
 745         type.
 746         * optabs-tree.c (optab_for_tree_code): Handle
 747         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
 748         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
 749         * tree-vect-generic.c (expand_vector_operations_1):  Handle
 750         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
 751         VEC_PACK_FLOAT_EXPR.
 752         * tree-vect-stmts.c (supportable_widening_operation): Handle
 753         FIX_TRUNC_EXPR.
 754         (supportable_narrowing_operation): Handle FLOAT_EXPR.
 755         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
 756         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
 757         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
 758         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
 759         mode attributes.
 760         (vec_pack<floatprefix>_float_<mode>): New expander.
 761         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
 762         attributes.
 763         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
 764         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
 765         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
 766         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
 767         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
 768         Document.
 769         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
 770         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
 771         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
 772         VEC_PACK_FLOAT_EXPR): Document.
 773
 774 2018-05-29  Richard Biener  <rguenther@suse.de>
 775
 776         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
 777         member.
 778         (stmt_vec_info_vec): Make pointer.
 779         (init_stmt_vec_info_vec): Remove.
 780         (free_stmt_vec_info_vec): Likewise.
 781         (set_stmt_vec_info_vec): New function.
 782         (free_stmt_vec_infos): Likewise.
 783         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
 784         (set_vinfo_for_stmt): Likewise.
 785         (get_earlier_stmt): Likewise.
 786         (get_later_stmt): Likewise.
 787         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
 788         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
 789         (vec_info::~vec_info): Free stmt_vec_infos.
 790         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
 791         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
 792         (pass_slp_vectorize::execute): Likewise.
 793         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
 794         (free_stmt_vec_info_vec): Likewise.
 795         (set_stmt_vec_info_vec): New function.
 796         (free_stmt_vec_infos): Likewise.
 797         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
 798         the global stmt_vec_info_vec.
 799         * tree-parloops.c (gather_scalar_reductions): Use
 800         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
 801         vector.
 802
 803 2018-05-29  Richard Biener  <rguenther@suse.de>
 804
 805         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
 806
 807 2018-05-29  Martin Liska  <mliska@suse.cz>
 808             David Malcolm  <dmalcolm@redhat.com>
 809
 810         * vec.c (test_reverse): New.
 811         (vec_c_tests): Add new test.
 812         * vec.h (vl_ptr>::reverse): New function.
 813
 814 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
 815
 816         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
 817
 818         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
 819         and later.
 820
 821 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
 822
 823         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
 824
 825 2018-05-28  Richard Biener  <rguenther@suse.de>
 826
 827         PR tree-optimization/85933
 828         * tree-vect-data-refs.c (vect_record_base_alignments): Only
 829         look at stmts marked as vectorizable.
 830
 831 2018-05-28  Richard Biener  <rguenther@suse.de>
 832
 833         PR tree-optimization/85934
 834         * tree-vect-generic.c (expand_vector_operations_1): Hoist
 835         vector boolean check before scalar optimization.
 836
 837 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
 838
 839         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
 840         for armv5te.
 841
 842 2018-05-28  Mark Wielaard  <mark@klomp.org>
 843
 844         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
 845         if it is an expression containing a minus sign.
 846
 847 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
 848
 849         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
 850
 851 2018-05-27  Paul Koning  <ni1d@arrl.net>
 852
 853         * config/pdp11/pdp11.md (truncsihi2): Remove.
 854
 855 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
 856             Chung-Ju Wu  <jasonwucj@gmail.com>
 857
 858         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
 859         implementation.
 860         (unaligned_store_dw): Ditto.
 861         * config/nds32/nds32-memory-manipulation.c
 862         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
 863         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
 864         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
 865         (emit_setmem_word_loop): Rename to ...
 866         (emit_setmem_doubleword_loop): ... this.
 867         (nds32_gen_dup_4_byte_to_word_value): New function.
 868         (nds32_gen_dup_8_byte_to_double_word_value): New function.
 869         (nds32_expand_setmem_loop): Refine implementation.
 870         (nds32_expand_setmem_loop_v3m): Ditto.
 871         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
 872         pattern.
 873
 874 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
 875
 876         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
 877
 878 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
 879
 880         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
 881         (nds32_init_machine_status): Initialize machine->attr_naked_p and
 882         machine->attr_no_prologue_p.
 883         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
 884         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
 885         (nds32_expand_epilogue): Consider attr_naked_p.
 886         (nds32_expand_epilogue_v3pop): Likewise.
 887         (nds32_can_use_return_insn): Likewise.
 888         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
 889         attr_no_prologue_p fields.
 890         * config/nds32/nds32.opt (mret-in-naked-func): New option.
 891
 892 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
 893
 894         PR target/85918
 895         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
 896         attributes.
 897         * config/i386/sse.md
 898         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
 899         Rename to ...
 900         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
 901         ... this.
 902         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
 903         Rename to ...
 904         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
 905         ... this.
 906         (*<floatsuffix>floatv2div2sf2): Rename to ...
 907         (*float<floatunssuffix>v2div2sf2): ... this.
 908         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
 909         (float<floatunssuffix>v2div2sf2_mask): ... this.
 910         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
 911         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
 912         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
 913         to ...
 914         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
 915         ... this.
 916         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
 917         Rename to ...
 918         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
 919         ... this.
 920         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
 921         Rename to ...
 922         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
 923         ... this.
 924         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
 925         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
 926         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
 927         gen_ufix_truncv8dfv8si2.
 928         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
 929         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
 930         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
 931         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
 932         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
 933         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
 934         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
 935         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
 936
 937 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
 938
 939         PR target/85900
 940         PR target/85345
 941         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
 942
 943 2018-05-25  Jim Wilson  <jimw@sifive.com>
 944
 945         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
 946         * config/riscv/riscv.c (struct machine_function): Add
 947         interrupt_handler_p and attribute_checked_p fields.
 948         (riscv_attribute_table): Add interrupt.
 949         (riscv_interrupt_type_p): New.
 950         (riscv_save_reg_p): Save extra regs for interrupt handler.
 951         (riscv_use_save_libcall): Return false  for interrupt handler.
 952         (riscv_first_stack_step): Add forward declaration.
 953         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
 954         for interrupt handler with large frame.  Use it for saved reg list.
 955         (riscv_expand_prologue): Move flag_stack_usage_info support to
 956         eliminate duplication.
 957         (riscv_expand_epilogue): Generate mret for interrupt handler.
 958         (riscv_epilogue_uses): New.
 959         (riscv_can_use_return_insn): Return false for interrupt handler.
 960         (riscv_function_ok_for_sibcall): Likewise.
 961         (riscv_set_current_function): Add interrupt handler support.
 962         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
 963         * config/riscv/riscv.md (UNSPECV_MRET): New.
 964         (GP_REGNUM): New.
 965         (riscv_frflags, riscv_fsflags): Use tab after opcode.
 966         (riscv_mret): New.
 967         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
 968
 969 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
 970
 971         PR tree-optimization/85712
 972         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
 973         this candidate has already been replaced in-situ by a copy.
 974
 975 2018-05-25  Jason Merrill  <jason@redhat.com>
 976
 977         PR c++/80485 - inline function non-zero address.
 978         * symtab.c (nonzero_address): Check DECL_COMDAT.
 979
 980 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
 981
 982         PR target/83628
 983         * config/alpha/alpha.md (ashlsi3): New insn pattern.
 984         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
 985         extension of SImode operation.  Use const123_operand predicate.
 986         (*saddsi_1): Remove.
 987         (*saddl_se_1): Ditto.
 988         (*ssubsi_1): Ditto.
 989         (*ssubl_se_1): Ditto.
 990         * config/alpha/predicates.md (const123_operand): New predicate.
 991         * config/alpha/constraints.md (P): Use IN_RANGE.
 992
 993 2018-05-25  Richard Biener  <rguenther@suse.de>
 994
 995         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
 996         defaulted to true.
 997         (ref_maybe_used_by_stmt_p): Likewise.
 998         (stmt_may_clobber_ref_p): Likewise.
 999         (stmt_may_clobber_ref_p_1): Likewise.
1000         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1001         and pass it along.
1002         (ref_maybe_used_by_stmt_p): Likewise.
1003         (stmt_may_clobber_ref_p): Likewise.
1004         (stmt_may_clobber_ref_p_1): Likewise.
1005         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1006         the alias oracle to disambiguate DRs with stmts DR analysis
1007         couldn't handle.
1008         (vect_analyze_data_refs): Do not give up on not analyzable
1009         DRs for BB vectorization.  Remove code truncating the dataref
1010         vector.
1011
1012 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
1013
1014         PR target/85832
1015         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1016         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
1017         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1018
1019 2018-05-25  Richard Biener  <rguenther@suse.de>
1020
1021         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1022         function, combining stmt data ref gathering and fatal analysis
1023         parts.
1024         (vect_analyze_data_refs): Remove now redudnant code and simplify.
1025         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1026         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1027         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1028         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1029
1030 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
1031
1032         PR tree-optimization/85720
1033         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1034         SCC if all partitions are builtins.
1035         (version_loop_by_alias_check): New parameter.  Generate cancelable
1036         runtime alias check if all partitions are builtins.
1037         (distribute_loop): Update call to above function.
1038
1039 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
1040
1041         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1042         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1043         (parm_default_def_partition_arg): Ditto.
1044         (set_parm_default_def_partition): Ditto.
1045         (get_parm_default_def_partitions): Ditto and make it static.
1046         (get_undefined_value_partitions): Ditto and make it static.
1047         (remove_ssa_form): Refactor call to init_var_map here.
1048         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1049         computation for loop region.
1050         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1051         (register_default_def): Delete.
1052         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1053         (parm_default_def_partition_arg): Ditto.
1054         (set_parm_default_def_partition): Ditto.
1055         (get_parm_default_def_partitions): Ditto and make it static.
1056         (get_undefined_value_partitions): Ditto and make it static.
1057         (coalesce_with_default, coalesce_with_default): Update comment.
1058         (create_coalesce_list_for_region): New func factored out from
1059         create_outofssa_var_map.
1060         (populate_coalesce_list_for_outofssa): New func factored out from
1061         create_outofssa_var_map and coalesce_ssa_name.
1062         (create_outofssa_var_map): Delete.
1063         (coalesce_ssa_name): Refactor to support live range computation.
1064         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1065         (get_parm_default_def_partitions): Delete.
1066         (get_undefined_value_partitions): Ditto.
1067         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1068         computation for loop region.
1069         (new_tree_live_info, loe_visit_block): Ditto.
1070         (live_worklist, set_var_live_on_entry): Ditto.
1071         (calculate_live_on_exit, verify_live_on_entry): Ditto.
1072         * tree-ssa-live.h (struct _var_map): New fields.
1073         (init_var_map): Change decl.
1074         (region_contains_p): New.
1075
1076 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
1077
1078         * tree-ssa-live.h (live_merge_and_clear): Delete.
1079
1080 2018-05-25  Richard Biener  <rguenther@suse.de>
1081
1082         PR c++/85912
1083         * tree-dump.c (dequeue_and_dump): Remove access to removed
1084         operand 2 of a SWITCH_EXPR.
1085
1086 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1087
1088         * doc/sourcebuild.texi (vect_double_cond_arith): Include
1089         multiplication and division.
1090         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1091         (cond_udiv@var{m}, cond_umod@var{m}): Document.
1092         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1093         (cond_udiv_optab, cond_umod_optab): New optabs.
1094         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1095         (IFN_COND_RDIV): New internal functions.
1096         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1097         TRUNC_MOD_EXPR and RDIV_EXPR.
1098         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1099         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1100         New unspecs.
1101         (SVE_INT_BINARY): Include mult.
1102         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1103         (optab, sve_int_op): Handle mult.
1104         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1105         UNSPEC_COND_DIV.
1106         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1107         for SVE_INT_BINARY_SD.
1108
1109 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1110
1111         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1112         (optab, sve_int_op): Handle div and udiv.
1113         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1114         for SVE_INT_BINARY_SD.
1115         (*<optab><mode>3): New insn for the same.
1116
1117 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1118
1119         * tree-vect-patterns.c: Include predict.h.
1120         (vect_recog_divmod_pattern): Restrict check for division support
1121         to when optimizing for size.
1122
1123 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1124
1125         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1126         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1127         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1128         (gimple_match_op::set_op): Likewise.
1129         (gimple_resimplify4): Declare.
1130         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1131         (expr::gen_transform): Likewise.
1132         (decision_tree::gen): Generate a simplification routine for 4 operands.
1133         * gimple-match-head.c (gimple_simplify): Add an overload for
1134         4 operands.  In the top-level function, handle up to 4 call
1135         arguments and call gimple_resimplify4.
1136         (gimple_resimplify4): New function.
1137         (build_call_internal): Pass a fourth operand.
1138         (maybe_push_to_seq): Likewise.
1139         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1140         Fold VEC_COND_EXPRs of an operation and a default value into
1141         an IFN_COND_* function if possible.
1142         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1143         New unspecs.
1144         (SVE_COND_FP_BINARY): Include them.
1145         (optab, sve_fp_op): Handle them.
1146         (SVE_INT_BINARY_REV): New code iterator.
1147         (SVE_COND_FP_BINARY_REV): New int iterator.
1148         (commutative): New int attribute.
1149         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1150         Declare.
1151         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1152         function.
1153         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1154         (*cond_<optab><mode>): New patterns for reversed operands.
1155
1156 2018-05-25  Richard Biener  <rguenther@suse.de>
1157
1158         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1159         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1160         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1161         (STMT_VINFO_GROUPED_ACCESS): Adjust.
1162         * tree-vect-data-refs.c (everywhere): Adjust users.
1163         * tree-vect-loop.c (everywhere): Likewise.
1164         * tree-vect-slp.c (everywhere): Likewise.
1165         * tree-vect-stmts.c (everywhere): Likewise.
1166         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1167
1168 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1169
1170         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1171         Rename to...
1172         (gcc_cv_as_section_exclude): ... this.
1173         Try Solaris as #exclude syntax.
1174         * configure: Regenerate.
1175         * config.in: Regenerate.
1176         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1177         SECTION_EXCLUDE.
1178         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1179         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1180
1181         * varasm.c (default_elf_asm_named_section): Don't check if
1182         HAVE_GAS_SECTION_EXCLUDE is defined.
1183
1184 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1185
1186         * doc/md.texi: Update the documentation of the cond_* optabs
1187         to mention the new final operand.  Fix GET_MODE_NUNITS call.
1188         Describe the scalar case too.
1189         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1190         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1191         instead of 2.
1192         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1193         (get_conditional_internal_fn): Update comment.
1194         * tree-vect-loop.c (vectorizable_reduction): Pass the original
1195         accumulator value as a final argument to conditional functions.
1196         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1197         a define_expand and add an "else" operand.  Assert for now that
1198         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
1199         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1200         (*cond_<optab><mode>): New patterns.
1201         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1202         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1203         (UNSPEC_COND_EOR): Delete.
1204         (optab): Remove associated mappings.
1205         (SVE_INT_BINARY): New code iterator.
1206         (sve_int_op): Remove int attribute and add "minus" to the code
1207         attribute.
1208         (SVE_COND_INT_OP): Delete.
1209         (SVE_COND_FP_OP): Rename to...
1210         (SVE_COND_FP_BINARY): ...this.
1211
1212 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1213
1214         * optabs.c (can_reuse_operands_p): New function.
1215         (maybe_legitimize_operands): Try to reuse the results for
1216         earlier operands.
1217
1218 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
1219
1220         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1221         Add {q} suffix to insn mnemonic.
1222
1223 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1224
1225         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1226         (msp430_warn_func_return): New.
1227
1228 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
1229
1230         * fold-const.c (tree_nonzero_bits): New function.
1231         * fold-const.h (tree_nonzero_bits): Likewise.
1232         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1233         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1234
1235 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
1236
1237         PR target/85900
1238         PR target/85345
1239         * varasm.c (assemble_alias): Check ifunc_resolver only on
1240         FUNCTION_DECL.
1241
1242 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
1243
1244         PR target/85903
1245         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1246         when memory input operand is handled.
1247
1248 2018-05-24  Luis Machado  <luis.machado@linaro.org>
1249
1250         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1251         global.
1252         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1253
1254 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
1255
1256         * match.pd: Delay FMA folds until after vectorization.
1257
1258 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1259
1260         PR target/83009
1261         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1262         address check not strict.
1263
1264 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
1265
1266         * gimple-match.h (gimple_match_op): New class.
1267         (mprts_hook): Replace parameters with a gimple_match_op *.
1268         (maybe_build_generic_op): Likewise.
1269         (gimple_simplified_result_is_gimple_val): Replace parameters with
1270         a const gimple_match_op *.
1271         (gimple_simplify): Replace code_helper * and tree * parameters with
1272         a gimple_match_op * parameter.
1273         (gimple_resimplify1): Replace code_helper *, tree and tree *
1274         parameters with a gimple_match_op * parameter.
1275         (gimple_resimplify2): Likewise.
1276         (gimple_resimplify3): Likewise.
1277         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1278         parameters with a gimple_match_op * parameter.
1279         * gimple-match-head.c (gimple_simplify): Change prototypes of
1280         auto-generated functions to take a gimple_match_op * instead of
1281         separate code_helper * and tree * parameters.  Make the same
1282         change in the top-level overload and update calls to the
1283         gimple_resimplify routines.  Update calls to the auto-generated
1284         functions and to maybe_push_res_to_seq in the publicly-facing
1285         operation-specific gimple_simplify overloads.
1286         (gimple_match_op::MAX_NUM_OPS): Define.
1287         (gimple_resimplify1): Replace rcode and ops with a single res_op
1288         parameter.  Update call to gimple_simplify.
1289         (gimple_resimplify2): Likewise.
1290         (gimple_resimplify3): Likewise.
1291         (mprts_hook): Replace parameters with a gimple_match_op *.
1292         (maybe_build_generic_op): Likewise.
1293         (build_call_internal): Replace type, nargs and ops with
1294         a gimple_match_op *.
1295         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1296         with a single gimple_match_op *.  Update calls to mprts_hook,
1297         build_call_internal and gimple_simplified_result_is_gimple_val.
1298         Factor out code that is common to the tree_code and combined_fn cases.
1299         * genmatch.c (expr::gen_transform): Replace tem_code and
1300         tem_ops with a gimple_match_op called tem_op.  Update calls
1301         to the gimple_resimplify functions and maybe_push_res_to_seq.
1302         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1303         res_ops.  Update call to the gimple_resimplify functions.
1304         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1305         (decision_tree::gen): Make the functions take a gimple_match_op *
1306         called res_op instead of separate res_code and res_ops parameters.
1307         Update call accordingly.
1308         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1309         and ops with a single res_op parameter.  Update calls to
1310         maybe_build_generic_op and maybe_push_res_to_seq.
1311         (fold_stmt_1): Update calls to gimple_simplify and
1312         replace_stmt_with_simplification.
1313         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1314         and gimple_simplified_result_is_gimple_val.
1315         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1316         gimple_simplify.
1317         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1318         with a gimple_match_op *.
1319         (vn_nary_build_or_lookup): Likewise.  Update call to
1320         vn_nary_build_or_lookup_1.
1321         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1322         gimple_match_op *.  Update calls to the gimple_resimplify routines
1323         and to gimple_simplified_result_is_gimple_val.
1324         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1325         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1326         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1327         (visit_nary_op): Likewise.
1328         (visit_reference_op_load): Likewise.
1329
1330 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1331
1332         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1333         modifier for printing the step amount.
1334
1335 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
1336
1337         PR target/78849
1338         * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1339         types.
1340
1341 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
1342
1343         * doc/sourcebuild.texi (Endianness): New subsubsection.
1344
1345 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1346
1347         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1348         <prefetch_dynamic_strides>: New const bool field.
1349         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1350         prefetch_dynamic_strides.
1351         (exynosm1_prefetch_tune): Likewise.
1352         (thunderxt88_prefetch_tune): Likewise.
1353         (thunderx_prefetch_tune): Likewise.
1354         (thunderx2t99_prefetch_tune): Likewise.
1355         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1356         false.
1357         (aarch64_override_options_internal): Update to set
1358         PARAM_PREFETCH_DYNAMIC_STRIDES.
1359         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1360         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1361         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1362         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1363         prefetch-dynamic-strides setting.
1364
1365 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1366
1367         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1368         <minimum_stride>: New const int field.
1369         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1370         minimum_stride field defaulting to -1.
1371         (exynosm1_prefetch_tune): Likewise.
1372         (thunderxt88_prefetch_tune): Likewise.
1373         (thunderx_prefetch_tune): Likewise.
1374         (thunderx2t99_prefetch_tune): Likewise.
1375         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1376         <default_opt_level>: Set to 3.
1377         (aarch64_override_options_internal): Update to set
1378         PARAM_PREFETCH_MINIMUM_STRIDE.
1379         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1380         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1381         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1382         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1383         stride is constant and is below the minimum stride threshold.
1384
1385 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1386
1387         * config/arm/arm-cpus.in (mode26): Delete.
1388         (armv4): Delete mode26 reference.
1389         * config/arm/arm.c (arm_configure_build_target): Delete use of
1390         isa_bit_mode26.
1391
1392 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
1393
1394         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1395         New insn pattern.
1396         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1397         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
1398         for non-SSE modes.
1399         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
1400         (floatunsdidf2): Ditto.
1401
1402 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
1403
1404         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1405         (fixuns_trunc<mode>si2_avx512f): Ditto.
1406         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1407         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1408         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1409
1410 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
1411
1412         PR rtl-optimization/79985
1413         * df-scan.c (df_insn_refs_collect): Remove special case for
1414         global registers and asm statements.
1415
1416 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
1417
1418         * extend.texi (Global Register Variables): Rewrite the bullet list.
1419         Note that the register is available for allocation. Note that access
1420         via inline asm must use constraints. Add note about async-signal
1421         handlers. Remove paragraph about automagic register selection.
1422
1423 2018-05-23  Richard Biener  <rguenther@suse.de>
1424
1425         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1426         of fixed offset from memset VN.
1427
1428 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
1429
1430         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1431         first_interp field.
1432         (alloc_cand_and_find_basis): Initialize first_interp field.
1433         (slsr_process_mul): Modify first_interp field.
1434         (slsr_process_add): Likewise.
1435         (slsr_process_cast): Modify first_interp field for each new
1436         interpretation.
1437         (slsr_process_copy): Likewise.
1438         (dump_candidate): Dump first_interp field.
1439         (replace_mult_candidate): Process all interpretations, not just
1440         subsequent ones.
1441         (replace_rhs_if_not_dup): Likewise.
1442         (replace_one_candidate): Likewise.
1443
1444 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
1445
1446         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1447         Add new boolean.
1448         (aarch64_needs_frame_chain): New function.
1449         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1450
1451 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
1452
1453         PR target/84882
1454         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1455         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1456         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1457         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1458         as true for strict-align.
1459         (aarch64_can_inline_p): Perform checks even when callee has no
1460         attributes to check for strict alignment.
1461         * doc/extend.texi (AArch64 Function Attributes): Document
1462         no-strict-align.
1463         * doc/invoke.texi: (AArch64 Options): Likewise.
1464
1465 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
1466
1467         PR tree-optimization/85853
1468         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1469         the handling of the root of the node to...
1470         (vect_slp_analyze_node_operations_1): ...this new function,
1471         and run the whole thing with the child nodes' def types
1472         set according to their SLP node's def type.
1473
1474 2018-05-23  Richard Biener  <rguenther@suse.de>
1475
1476         PR middle-end/85874
1477         * tree-data-ref.c (create_runtime_alias_checks): Defer
1478         and ignore overflow warnings.
1479
1480 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
1481
1482         PR tree-optimization/85822
1483         * tree-vrp.c (is_masked_range_test): Fix handling of negative
1484         constants.
1485
1486 2018-05-23  Richard Biener  <rguenther@suse.de>
1487
1488         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1489         memset constants via native_interpret_expr.
1490
1491 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
1492
1493         PR target/85345
1494         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1495         attribute.
1496         (cgraph_node::create_alias): Likewise.
1497         (cgraph_node::get_availability): Check ifunc_resolver instead
1498         of looking up ifunc attribute.
1499         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1500         * varasm.c (do_assemble_alias): Likewise.
1501         (assemble_alias): Likewise.
1502         (default_binds_local_p_3): Likewise.
1503         * cgraph.h (cgraph_node): Add ifunc_resolver.
1504         (cgraph_node::only_called_directly_or_aliased_p): Return false
1505         for IFUNC resolver.
1506         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1507         attribute.
1508         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1509         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1510         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1511         instead of looking up ifunc attribute.
1512
1513 2018-05-22  Luis Machado  <luis.machado@linaro.org>
1514
1515         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1516
1517 2018-05-22  Martin Sebor  <msebor@redhat.com>
1518
1519         PR middle-end/85359
1520         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1521         only when expasion succeeds.
1522         (expand_builtin_strcmp): Same.
1523         (expand_builtin_strncmp): Same.
1524
1525 2018-05-22  Martin Sebor  <msebor@redhat.com>
1526
1527         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1528
1529 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
1530             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1531
1532         * config/aarch64/aarch64-ldpstp.md: Replace uses of
1533         aarch64_mem_pair_operand with memory_operand and delete operand swapping
1534         code.
1535         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1536         Add check for legitimate_address.
1537         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1538         (aarch64_swap_ldrstr_operands): New.
1539         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1540         Define prototype.
1541
1542 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
1543             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1544
1545         * config/aarch64/aarch64.md: New patterns to generate stp
1546         and ldp.
1547         (store_pair_sw, store_pair_dw): New patterns to generate stp for
1548         single words and double words.
1549         (load_pair_sw, load_pair_dw): Likewise.
1550         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1551         Delete.
1552         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1553         Delete.
1554         * config/aarch64/aarch64-ldpstp.md: Modify peephole
1555         for different mode ldpstp and add peephole for merged zero stores.
1556         Likewise for loads.
1557         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1558         Add size check.
1559         (aarch64_gen_store_pair): Rename calls to match new patterns.
1560         (aarch64_gen_load_pair): Rename calls to match new patterns.
1561         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1562         (load_pair<DREG:mode><DREG2:mode>): ... This.
1563         (store_pair<mode>): Rename to...
1564         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1565         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1566         New mode iterators.
1567         (V_INT_EQUIV): Handle SImode.
1568         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1569         New predicate.
1570
1571 2018-05-22  Martin Sebor  <msebor@redhat.com>
1572
1573         PR c/85623
1574         * calls.c (maybe_warn_nonstring_arg): Use string length to set
1575         or ajust the presumed bound on an operation to avoid unnecessary
1576         warnings.
1577
1578 2018-05-22  Martin Sebor  <msebor@redhat.com>
1579
1580         PR tree-optimization/85826
1581         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1582         assuming that a DECL necesarily has a constant size.
1583
1584 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
1585
1586         PR middle-end/85862
1587         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1588
1589 2018-05-22  Richard Biener  <rguenther@suse.de>
1590
1591         PR tree-optimization/85834
1592         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1593         non-constant and non-zero memset arguments.
1594
1595 2018-05-22  Martin Liska  <mliska@suse.cz>
1596
1597         PR ipa/85607
1598         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1599
1600 2018-05-22  Richard Biener  <rguenther@suse.de>
1601
1602         PR tree-optimization/85863
1603         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1604         comparisons when vectype is specified.
1605         (vectorizable_condition): Do not specify vectype for
1606         vect_is_simple_cond when SLP vectorizing.
1607
1608 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
1609
1610         PR target/85657
1611         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1612         define __ibm128 as long double.
1613         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1614         as a distinct type when IEEE 128-bit support is enabled.
1615         (init_float128_ieee): Fix up conversions between IFmode and IEEE
1616         128-bit types to use the correct functions.
1617         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1618         convert between 128-bit floating point types that have different
1619         modes but the same representation, instead of using gen_lowpart to
1620         makean alias.
1621         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1622         KFmode.
1623         (IFKF_reg): New attributes to give the register constraints for
1624         IFmode and KFmode.
1625         (extend<mode>tf2_internal): New insns to mark an explicit
1626         conversion between 128-bit floating point types that have a
1627         different mode but share the same representation.
1628
1629 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
1630
1631         PR tree-optimization/85814
1632         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1633         a null return from get_strinfo when unsharing the next
1634         strinfo in the chain.
1635
1636 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
1637
1638         PR gcc/84923
1639         * varasm.c (weak_finish): Clean up weak_decls.
1640
1641 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1642
1643         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1644         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1645         UNSPEC_UADALP values.
1646         * config/aarch64/iterators.md (ABAL): New int iterator.
1647         (ABDL2): Likewise.
1648         (ADALP): Likewise.
1649         (sur): Add mappings for the above.
1650         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1651         New define_insn.
1652         (aarch64_<sur>abal<mode>_4): Likewise.
1653         (aarch64_<sur>adalp<mode>_3): Likewise.
1654         (<sur>sadv16qi): New define_expand.
1655
1656 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
1657
1658         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1659         (*movdf_internal): Ditto.
1660         (*rcpsf2_sse): Ditto.
1661         (*rsqrtsf2_sse): Ditto.
1662         (*sqrt<mode>2_sse): Ditto.
1663
1664 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
1665
1666         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1667         eor3q<mode>4.
1668         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1669         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1670         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1671         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1672         vbcaxq_s64): New.
1673         * config/aarch64/arm_neon.h: Likewise.
1674         * config/aarch64/iterators.md (VQ_I): New.
1675
1676 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
1677
1678         * config.gcc: Add arc/t-multilib-linux to tmake_file for
1679         arc*-*-linux*.
1680         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1681         MULTILIB_DIRNAMES
1682
1683 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
1684
1685         * config/nds32/constraints.md (S): New constraint.
1686         * config/nds32/nds32.md (call_internal): Use constraint S.
1687         (call_value_internal): Likewise.
1688         (sibcall_internal): Likewise.
1689         (sibcall_value_internal): Likewise.
1690
1691 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
1692             Chung-Ju Wu  <jasonwucj@gmail.com>
1693
1694         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1695         into consideration.
1696
1697 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
1698             Chung-Ju Wu  <jasonwucj@gmail.com>
1699
1700         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1701         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1702         (nds32_rtx_costs_impl): Simplify.
1703         (nds32_address_cost_impl): Simplify.
1704         (nds32_init_rtx_costs): New function.
1705         (nds32_rtx_costs_speed_prefer): Likewise.
1706         (nds32_rtx_costs_size_prefer): Likewise.
1707         (nds32_address_cost_speed_prefer): Likewise.
1708         (nds32_address_cost_speed_fwprop): Likewise.
1709         (nds32_address_cost_size_prefer): Likewise.
1710         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1711         * config/nds32/nds32.c (nds32_option_override): Use
1712         nds32_init_rtx_costs function.
1713
1714 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
1715
1716         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1717         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1718         (TARGET_PIPELINE_N8): Likewise.
1719         (TARGET_PIPELINE_N10): Likewise.
1720         (TARGET_PIPELINE_N13): Likewise.
1721         (TARGET_PIPELINE_GRAYWOLF): Likewise.
1722
1723 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
1724
1725         * config/nds32/nds32-fpu.md: Update copyright year.
1726
1727 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1728
1729         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1730
1731 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1732
1733         * config/nds32/nds32.c
1734         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1735         * config/nds32/nds32.opt (minline-asm-r15): New option.
1736
1737 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1738
1739         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1740         MASK_HW_ABS.
1741         * config/nds32/nds32.md (abssi2): New pattern.
1742
1743 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
1744
1745         * config/i386/i386.md (rex64namesuffix): New mode attribute.
1746         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1747         Merge insn pattern from sse_cvtsi2ss<round_name> and
1748         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1749         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1750         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1751         using SWI48 mode iterator.
1752         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1753         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1754         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1755         pattern from sse_cvttss2si<round_saeonly_name>
1756         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1757         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1758         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1759         using SWI48 mode iterator.
1760         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1761         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1762         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1763         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1764         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1765         using SWI48 mode iterator.
1766         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1767         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1768         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1769         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1770         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1771         SWI48 mode iterator.
1772         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1773         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1774         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1775         pattern from sse_cvttsd2si<round_saeonly_name>
1776         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1777
1778 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1779
1780         * config/nds32/nds32-md-auxiliary.c
1781         (nds32_valid_smw_lwm_base_p): Refine.
1782         (nds32_output_smw_single_word): Refine.
1783         (nds32_output_smw_double_word): New.
1784         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1785
1786 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1787
1788         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1789         (nds32_output_stack_pop): Refine.
1790         (nds32_expand_unaligned_load): Refine.
1791         (nds32_expand_unaligned_store): Refine.
1792
1793 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
1794             Chung-Ju Wu  <jasonwucj@gmail.com>
1795
1796         * config/nds32/constants.md: Add TP_REGNUM constant.
1797         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1798         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1799         UNSPEC_ADD32.
1800         * config/nds32/nds32-doubleword.md: Consider flag_pic.
1801         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1802         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1803         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1804         and PIC code generation.
1805         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1806         code generation.
1807         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1808         optimization.
1809         * config/nds32/nds32.md: Support TLS and PIC.
1810         * config/nds32/nds32.c: Support TLS and PIC.
1811         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1812         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1813         predicate.
1814
1815 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1816
1817         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1818         mode with E_ prefix.
1819
1820 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
1821             Chung-Ju Wu  <jasonwucj@gmail.com>
1822
1823         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1824         * config/nds32/nds32-md-auxiliary.c
1825         (symbolic_reference_mentioned_p): New.
1826         (nds32_legitimize_ict_address): New.
1827         (nds32_expand_ict_move): New.
1828         (nds32_indirect_call_referenced_p): New.
1829         (nds32_symbol_binds_local_p): Delete.
1830         (nds32_long_call_p): Modify.
1831         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1832         * config/nds32/nds32-protos.h
1833         (symbolic_reference_mentioned_p): Declare.
1834         (nds32_legitimize_ict_address): Declare.
1835         (nds32_expand_ict_move): Declare.
1836         (nds32_indirect_call_referenced_p): Declare.
1837         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1838         (nds32_relax_group): Use nds32_ict_const_p as condition.
1839         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1840         (nds32_asm_file_start): Output ict_model directive in asm code.
1841         (nds32_legitimate_address_p): Consider indirect call.
1842         (nds32_print_operand): Consider indirect call.
1843         (nds32_print_operand_address): Consider indirect call.
1844         (nds32_insert_attributes): Handle "indirect_call" attribute.
1845         (TARGET_LEGITIMATE_ADDRESS_P): Define.
1846         (TARGET_LEGITIMATE_CONSTANT_P): Define.
1847         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1848         (TARGET_DELEGITIMIZE_ADDRESS): Define.
1849         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1850         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1851         (TARGET_ICT_MODEL_SMALL): Define.
1852         (TARGET_ICT_MODEL_LARGE): Define.
1853         * config/nds32/nds32.md (movsi): Consider ict model.
1854         (call, call_value): Consider ict model.
1855         (sibcall, sibcall_value): Consider ict model.
1856         * config/nds32/nds32.opt (mict-model): New option.
1857         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1858         model.
1859
1860 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
1861             Monk Chiang  <sh.chiang04@gmail.com>
1862             Jim Wilson <jimw@sifive.com>
1863
1864         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1865         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
1866         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1867         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1868         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
1869         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1870         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1871         compute save_libcall_adjustment properly.
1872         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1873         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1874         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1875         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1876         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1877         (ABI_SPEC): Handle mabi=ilp32e.
1878         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1879         (RVE): Add RVE mask.
1880         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1881         <-march>: Add rv32e as an example.
1882
1883 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
1884
1885         PR c++/82899
1886         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1887         (intra_create_variable_infos): Handle C++ constructors.
1888
1889 2018-05-18  Martin Liska  <mliska@suse.cz>
1890
1891         * passes.def: Remove a redundant pass.
1892
1893 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
1894
1895         PR bootstrap/85838
1896         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1897
1898 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1899
1900         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1901         (ARMv4): Update.
1902         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1903         (ARMv6m): Update.
1904         (armv2, armv2a, armv3, armv3m): Delete architectures.
1905         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1906         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1907         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1908         Delete cpus.
1909         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1910         (*mulsidi3adddi): Likewise.
1911         (mulsidi3): Likewise.
1912         (*mulsidi3_nov6): Likewise.
1913         (umulsidi3): Likewise.
1914         (umulsidi3_nov6): Likewise.
1915         (umaddsidi4): Likewise.
1916         (*umulsidi3adddi): Likewise.
1917         (smulsi3_highpart): Likewise.
1918         (*smulsi3_highpart_nov6): Likewise.
1919         (umulsi3_highpart): Likewise.
1920         (*umulsi3_highpart_nov6): Likewise.
1921         * config/arm/arm.h (arm_arch3m): Delete.
1922         * config/arm/arm.c (arm_arch3m): Delete.
1923         (arm_option_override_internal): Update armv3-related comment.
1924         (arm_configure_build_target): Delete use of isa_bit_mode32.
1925         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1926         (arm_rtx_costs_internal): Delete check of arm_arch3m.
1927         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1928         (mulsa3): Likewise.
1929         (mulusa3): Likewise.
1930         * config/arm/arm-protos.h (arm_arch3m): Delete.
1931         * config/arm/arm-tables.opt: Regenerate.
1932         * config/arm/arm-tune.md: Likewise.
1933         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1934         deleted architectures.
1935
1936 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1937
1938         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1939         (armv5t, armv5te): New features.
1940         (ARMv5, ARMv5e): Delete fgroups.
1941         (ARMv5t, ARMv5te): Adjust for above changes.
1942         (ARMv6m): Likewise.
1943         (armv5, armv5e): Delete arches.
1944         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1945         arm_arch5.
1946         (*call_reg_arm): Likewise.
1947         (*call_value_reg_armv5): Likewise.
1948         (*call_value_reg_arm): Likewise.
1949         (*call_symbol): Likewise.
1950         (*call_value_symbol): Likewise.
1951         (*sibcall_insn): Likewise.
1952         (*sibcall_value_insn): Likewise.
1953         (clzsi2): Likewise.
1954         (prefetch): Likewise.
1955         (define_split and define_peephole2 dependent on arm_arch5):
1956         Likewise.
1957         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1958         arm_arch5e.
1959         (TARGET_ARM_QBIT): Likewise.
1960         (TARGET_DSP_MULTIPLY): Likewise.
1961         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1962         (arm_arch5, arm_arch5e): Delete.
1963         (arm_arch5t, arm_arch5te): Declare.
1964         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1965         (arm_arch5t): Declare.
1966         (arm_option_reconfigure_globals): Update for the above.
1967         (arm_options_perform_arch_sanity_checks): Update comment, replace
1968         use of arm_arch5 with arm_arch5t.
1969         (use_return_insn): Likewise.
1970         (arm_emit_call_insn): Likewise.
1971         (output_return_instruction): Likewise.
1972         (arm_final_prescan_insn): Likewise.
1973         (arm_coproc_builtin_available): Likewise.
1974         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1975         arm_arch5e with arm_arch5t and arm_arch5te.
1976         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1977         (arm_arch5t, arm_arch5te): Declare.
1978         * config/arm/arm-tables.opt: Regenerate.
1979         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1980         * config/arm/t-multilib: Likewise.
1981         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1982         instead of arm_arch5.
1983         (*call_reg_thumb1): Likewise.
1984         (*call_value_reg_thumb1_v5): Likewise.
1985         (*call_value_reg_thumb1): Likewise.
1986         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1987         unreachable path.
1988         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1989
1990 2018-05-18  Martin Liska  <mliska@suse.cz>
1991
1992         PR gcov-profile/84846
1993         * doc/gcov.texi: Document -t option of gcov tool.
1994
1995 2018-05-18  Martin Liska  <mliska@suse.cz>
1996
1997         PR gcov-profile/84846
1998         * gcov.c (print_usage): Add new -t option.
1999         (process_args): Handle the option.
2000         (generate_results): Use stdout as output when requested by
2001         the option.
2002
2003 2018-05-18  Martin Liska  <mliska@suse.cz>
2004
2005         PR gcov-profile/84846
2006         * coverage.c (coverage_init): Write PWD to .gcno file.
2007         * doc/gcov.texi: Document how working directory is printed.
2008         * gcov-dump.c (dump_gcov_file): Print PWD.
2009         * gcov.c (output_intermediate_file): Likewise.
2010         (read_graph_file): Read PWD string.
2011         (output_lines): Print PWD.
2012
2013 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2014
2015         PR middle-end/85817
2016         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2017         for retval and return false if all args to phi are zero.
2018
2019 2018-05-18  Richard Biener  <rguenther@suse.de>
2020
2021         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2022         method.
2023         (evrp_dom_walker::before_dom_children): Call it.
2024
2025 2018-05-18  Richard Biener  <rguenther@suse.de>
2026
2027         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2028         results when processing array refs with variable index.
2029
2030 2018-05-18  Toon Moene  <toon@moene.org>
2031
2032         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2033         directly after that of -floop-interchange. Indicate that both
2034         options are enabled by default when specifying -O3.
2035
2036 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2037
2038         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2039         iterator.  Delete separate integer-mode vec_set<mode> expander.
2040         (aarch64_simd_vec_setv2di): Delete.
2041         (vec_setv2di): Delete.
2042         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2043         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
2044         the "w, r" alternative.
2045
2046 2018-05-18  Martin Liska  <mliska@suse.cz>
2047
2048         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2049         * tree-pass.h (make_pass_lower_switch_O0): New function.
2050         * tree-switch-conversion.c (node_has_low_bound): Remove.
2051         (node_has_high_bound): Likewise.
2052         (node_is_bounded): Likewise.
2053         (class pass_lower_switch): Make it a template type and create
2054         two instances.
2055         (pass_lower_switch::execute): Add template argument.
2056         (make_pass_lower_switch): New function.
2057         (make_pass_lower_switch_O0): New function.
2058         (do_jump_if_equal): Remove.
2059         (emit_case_nodes): Simplify to just handle all 3 cases and leave
2060         all the hard work to tree optimization passes.
2061
2062 2018-05-18  Martin Liska  <mliska@suse.cz>
2063
2064         * dbgcnt.c (limit_low): Renamed from limit.
2065         (limit_high): New variable.
2066         (dbg_cnt_is_enabled): Check for upper limit.
2067         (dbg_cnt): Adjust dumping.
2068         (dbg_cnt_set_limit_by_index): Add new argument for high
2069         value.
2070         (dbg_cnt_set_limit_by_name): Likewise.
2071         (dbg_cnt_process_single_pair): Parse new format.
2072         (dbg_cnt_process_opt): Use strtok.
2073         (dbg_cnt_list_all_counters): Remove 'value' and add
2074         'limit_high'.
2075         * doc/invoke.texi: Document changes.
2076
2077 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
2078
2079         * doc/sourcebuild.texi (scalar_all_fma): Document.
2080         * tree.def (FMA_EXPR): Delete.
2081         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2082         * internal-fn.c (ternary_direct): New macro.
2083         (expand_ternary_optab_fn): Likewise.
2084         (direct_ternary_optab_supported_p): Likewise.
2085         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2086         * builtins.c (fold_builtin_fma): Delete.
2087         (fold_builtin_3): Don't call it.
2088         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2089         * expr.c (expand_expr_real_2): Likewise.
2090         * fold-const.c (operand_equal_p): Likewise.
2091         (fold_ternary_loc): Likewise.
2092         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2093         * gimple.c (DEFTREECODE): Likewise.
2094         * gimplify.c (gimplify_expr): Likewise.
2095         * optabs-tree.c (optab_for_tree_code): Likewise.
2096         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2097         * tree-eh.c (operation_could_trap_p): Likewise.
2098         (stmt_could_throw_1_p): Likewise.
2099         * tree-inline.c (estimate_operator_cost): Likewise.
2100         * tree-pretty-print.c (dump_generic_node): Likewise.
2101         (op_code_prio): Likewise.
2102         * tree-ssa-loop-im.c (stmt_cost): Likewise.
2103         * tree-ssa-operands.c (get_expr_operands): Likewise.
2104         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2105         * fold-const-call.h (fold_fma): Delete.
2106         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2107         CFN_FNMA and CFN_FNMS.
2108         (fold_fma): Delete.
2109         * genmatch.c (combined_fn): New enum.
2110         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2111         (commutative_op): New function.
2112         (commutate): Use it.  Handle more than 2 operands.
2113         (dt_operand::gen_gimple_expr): Use commutative_op.
2114         (parser::parse_expr): Allow :c to be used with non-binary
2115         operators if the commutative operand is known.
2116         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2117         CFN_FMS, CFN_FNMA and CFN_FNMS.
2118         (backprop::process_assign_use): Remove FMA_EXPR handling.
2119         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2120         (gen_hsa_fma): New function.
2121         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2122         IFN_FNMA and IFN_FNMS.
2123         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2124         * gimple-fold.h (follow_all_ssa_edges): Declare.
2125         * gimple-fold.c (follow_all_ssa_edges): New function.
2126         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2127         gimple_build interface and use follow_all_ssa_edges to fold the result.
2128         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2129         instead of checking for optabs directly.
2130         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2131         rather than FMA_EXPRs.
2132         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2133         call to IFN_FMA instead of an FMA_EXPR.
2134
2135 2018-05-17  Jim Wilson  <jimw@sifive.com>
2136
2137         * expr.c (do_tablejump): When converting index to Pmode, if we have a
2138         sign extended promoted subreg, and the range does not have the sign bit
2139         set, then do a sign extend.
2140
2141         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2142         test, check for sign extended subreg and/or constant operands, and
2143         do a sign extend in that case.
2144
2145 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
2146
2147         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2148         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2149         Add untyped.
2150         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2151         Change logics_shift_reg to logics_shift_imm.
2152         (thunderx2t99_fp_loadpair_basic): Delete.
2153         (thunderx2t99_fp_storepair_basic): Delete.
2154         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2155         (thunderx2t99_asimd_polynomial): Delete.
2156         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2157         and neon_fp_mul_d_scalar_q.
2158         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2159         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2160         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2161         (thunderx2t99_asimd_lut): Add missing tbl types.
2162         (thunderx2t99_asimd_ext): Delete.
2163         (thunderx2t99_asimd_load1_1_mult): Delete.
2164         (thunderx2t99_asimd_load1_2_mult): Delete.
2165         (thunderx2t99_asimd_load1_ldp): New.
2166         (thunderx2t99_asimd_load1): New.
2167         (thunderx2t99_asimd_load2): Add missing *load2* types.
2168         (thunderx2t99_asimd_load3): New.
2169         (thunderx2t99_asimd_load4): New.
2170         (thunderx2t99_asimd_store1_1_mult): Delete.
2171         (thunderx2t99_asimd_store1_2_mult): Delete.
2172         (thunderx2t99_asimd_store2_mult): Delete.
2173         (thunderx2t99_asimd_store2_onelane): Delete.
2174         (thunderx2t99_asimd_store_stp): New.
2175         (thunderx2t99_asimd_store1): New.
2176         (thunderx2t99_asimd_store2): New.
2177         (thunderx2t99_asimd_store3): New.
2178         (thunderx2t99_asimd_store4): New.
2179
2180 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
2181
2182         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2183         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
2184
2185 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
2186             Segher Boessenkool  <segher@kernel.crashing.org>
2187
2188         PR target/85698
2189         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2190         operand.
2191
2192 2018-05-17  Richard Biener  <rguenther@suse.de>
2193
2194         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2195         for pruning loop and prune defs feeding only already visited PHIs.
2196
2197 2018-05-17  Richard Biener  <rguenther@suse.de>
2198
2199         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2200
2201 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
2202             Richard Biener  <rguenther@suse.de>
2203
2204         PR tree-optimization/85793
2205         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2206         for VMAT_ELEMENTWISE.
2207
2208 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
2209
2210         * internal-fn.h (lookup_internal_fn): Declare
2211         * internal-fn.c (lookup_internal_fn): New function.
2212         * gimple.c (gimple_build_call_from_tree): Handle calls to
2213         internal functions.
2214         * gimple-pretty-print.c (dump_gimple_call): Print "." before
2215         internal function names.
2216         * tree-pretty-print.c (dump_generic_node): Likewise.
2217         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2218
2219 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
2220
2221         * gimple-fold.h (gimple_build): Make the function forms take
2222         combined_fn rather than built_in_function.
2223         (gimple_simplify): Likewise.
2224         * gimple-match-head.c (gimple_simplify): Likewise.
2225         * gimple-fold.c (gimple_build): Likewise.
2226         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2227         rather than gimple_build_call_internal.
2228         (get_initial_defs_for_reduction): Likewise.
2229         (vect_create_epilog_for_reduction): Likewise.
2230         (vectorizable_live_operation): Likewise.
2231
2232 2018-05-17  Martin Liska  <mliska@suse.cz>
2233
2234         * gimple-ssa-sprintf.c (format_directive): Do not use
2235         space in between 'G_' and '('.
2236
2237 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
2238
2239         PR target/85323
2240         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2241         even if the mask is not all ones.
2242
2243         PR target/85323
2244         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2245         vector.
2246         (ix86_gimple_fold_builtin): Likewise.
2247
2248         PR target/85323
2249         * config/i386/i386.c: Include tree-vector-builder.h.
2250         (ix86_vector_shift_count): New function.
2251         (ix86_fold_builtin): Fold shift builtins by scalar count.
2252         (ix86_gimple_fold_builtin): Likewise.
2253
2254         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2255         _mm512_setzero): New intrinsics.
2256
2257 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
2258             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2259
2260         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2261         code generation for cases where splatting a value is not useful.
2262         * simplify-rtx.c (simplify_ternary_operation): Simplify
2263         vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2264         mode to a vec_concat.
2265
2266 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
2267
2268         * config.gcc: Support "goldmont-plus".
2269         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2270         "goldmont-plus".
2271         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2272         PROCESSOR_GOLDMONT_PLUS.
2273         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2274         (processor_target_table): Add "goldmont-plus".
2275         (PTA_GOLDMONT_PLUS): Define.
2276         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2277         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2278         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2279         (fold_builtin_cpu): Add "goldmont-plus".
2280         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2281         (ix86_option_override_internal): Add "goldmont-plus".
2282         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2283         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2284         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2285         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2286
2287 2018-05-17  Richard Biener  <rguenther@suse.de>
2288
2289         PR tree-optimization/85757
2290         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2291         remove defs that only feed that PHI from further processing.
2292
2293 2018-05-16  Jim Wilson  <jimw@sifive.com>
2294
2295         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2296         asterisk to name.
2297         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2298
2299 2018-05-16  Mark Wielaard  <mark@klomp.org>
2300
2301         * dwarf2out.c (count_index_strings): New function.
2302         (output_indirect_strings): Call count_index_strings and generate
2303         header for dwarf_version >= 5.
2304
2305 2018-05-16  Mark Wielaard  <mark@klomp.org>
2306
2307         * dwarf2out.c (dwarf_FORM): New function.
2308         (set_indirect_string): Use dwarf_FORM.
2309         (reset_indirect_string): Likewise.
2310         (size_of_die): Likewise.
2311         (value_format): Likewise.
2312         (output_die): Likewise.
2313         (add_skeleton_AT_string): Likewise.
2314         (output_macinfo_op): Likewise.
2315         (index_string): Likewise.
2316         (output_index_string_offset): Likewise.
2317         (output_index_string): Likewise.
2318         (count_index_strings): Likewise.
2319
2320 2018-05-16  Carl Love  <cel@us.ibm.com>
2321
2322         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2323         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2324
2325 2018-05-16  Martin Jambor  <mjambor@suse.cz>
2326
2327         * ipa-prop.c (ipa_free_all_edge_args): Remove.
2328         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2329
2330 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
2331
2332         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2333         (fnma<mode>4): Likewise.
2334         (fms<mode>4): Likewise.
2335         (fnms<mode>4): Likewise.
2336         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2337         (aarch64_fnma<mode>4): Likewise.
2338         (aarch64_fms<mode>4): Likewise.
2339         (aarch64_fnms<mode>4): Likewise.
2340         (aarch64_fnmadd<mode>4): Likewise.
2341
2342 2018-05-16  Jason Merrill  <jason@redhat.com>
2343
2344         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
2345
2346 2018-05-16  Richard Biener  <rguenther@suse.de>
2347
2348         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2349         (dump_stmt_cost): Declare.
2350         (add_stmt_cost): Dump cost we add.
2351         (add_stmt_costs): New function.
2352         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2353         No longer exported.
2354         (vect_analyze_stmt): Adjust prototype.
2355         (vectorizable_condition): Likewise.
2356         (vectorizable_live_operation): Likewise.
2357         (vectorizable_reduction): Likewise.
2358         (vectorizable_induction): Likewise.
2359         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2360         cost vector to pass to vectorizable_ and record afterwards.
2361         (vect_model_reduction_cost): Take cost vector argument and adjust.
2362         (vect_model_induction_cost): Likewise.
2363         (vectorizable_reduction): Likewise.
2364         (vectorizable_induction): Likewise.
2365         (vectorizable_live_operation): Likewise.
2366         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2367         SLP_TREE_NUMBER_OF_VEC_STMTS.
2368         (vect_analyze_slp_cost_1): Remove.
2369         (vect_analyze_slp_cost): Likewise.
2370         (vect_slp_analyze_node_operations): Take visited args and
2371         a target cost vector.  Avoid processing already visited stmt sets.
2372         (vect_slp_analyze_operations): Use a local cost vector to gather
2373         costs and register those of non-discarded instances.
2374         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2375         (vect_schedule_slp_instance): Remove copying of
2376         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
2377         zero.
2378         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2379         adding cost.  Record cost entry location.
2380         (vect_prologue_cost_for_slp_op): Function to compute cost of
2381         a constant or invariant generated for SLP vect in the prologue,
2382         split out from vect_analyze_slp_cost_1.
2383         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
2384         (vect_model_promotion_demotion_cost): Likewise.
2385         (vect_model_store_cost): Likewise, make static.
2386         (vect_model_load_cost): Likewise.
2387         (vectorizable_bswap): Add cost vector arg and adjust.
2388         (vectorizable_call): Likewise.
2389         (vectorizable_simd_clone_call): Likewise.
2390         (vectorizable_conversion): Likewise.
2391         (vectorizable_assignment): Likewise.
2392         (vectorizable_shift): Likewise.
2393         (vectorizable_operation): Likewise.
2394         (vectorizable_store): Likewise.
2395         (vectorizable_load): Likewise.
2396         (vectorizable_condition): Likewise.
2397         (vectorizable_comparison): Likewise.
2398         (can_vectorize_live_stmts): Likewise.
2399         (vect_analyze_stmt): Likewise.
2400         (vect_transform_stmt): Adjust calls to vectorizable_*.
2401         * tree-vectorizer.c: Include gimple-pretty-print.h.
2402         (dump_stmt_cost): New function.
2403
2404 2018-05-16  Richard Biener  <rguenther@suse.de>
2405
2406         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2407         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2408         * tree-ssa-dse.c: Include tree-ssa-loop.h.
2409         (check_name): New callback.
2410         (dse_classify_store): Track cycles via a visited bitmap of PHI
2411         defs and simplify handling of in-loop and across loop dead stores
2412         and properly fail for loop-variant refs.  Handle byte-tracking with
2413         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2414         limiting the walk.
2415
2416 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
2417
2418         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2419         (vect_get_mask_type_for_stmt): Likewise.
2420         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2421         split out from...
2422         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
2423         to determine the statement's vector type and the vector type that
2424         should be used for calculating nunits.  Deal with cases in which
2425         the type has to be deferred.
2426         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2427         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2428         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2429         (vect_determine_vf_for_stmt): New functions, split out from...
2430         (vect_determine_vectorization_factor): ...here.
2431         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2432         (vect_get_mask_type_for_stmt): New functions, split out from
2433         vect_determine_vectorization_factor.
2434
2435 2018-05-16  Richard Biener  <rguenther@suse.de>
2436
2437         * tree-cfg.c (verify_gimple_assign_ternary): Properly
2438         verify the [VEC_]COND_EXPR embedded comparison.
2439
2440 2018-05-15  Martin Sebor  <msebor@redhat.com>
2441
2442         PR tree-optimization/85753
2443         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2444         RECORD_TYPE in addition to ARRAY_TYPE.
2445
2446 2018-05-15  Martin Sebor  <msebor@redhat.com>
2447
2448         PR middle-end/85643
2449         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2450
2451 2018-05-15  Richard Biener  <rguenther@suse.de>
2452
2453         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2454         add by_clobber_p one.  Change algorithm to collect all defs
2455         representing uses we need to walk and try reducing them to
2456         a single one before failing.
2457         (dse_dom_walker::dse_optimize_stmt): Adjust.
2458
2459 2018-05-13  Mark Wielaard  <mark@klomp.org>
2460
2461         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2462         (size_of_loc_descr): Likewise.
2463         (output_loc_operands): Likewise.
2464         (output_loc_operands_raw): Likewise.
2465         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2466         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2467         (hash_loc_operands): Likewise.
2468         (compare_loc_operands): Likewise.
2469
2470 2018-05-14  Mark Wielaard  <mark@klomp.org>
2471
2472         * dwarf2out.c (count_index_addrs): New function.
2473         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2474
2475 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2476
2477         PR tree-optimization/83648
2478         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2479         return value as malloc candidate.
2480
2481 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2482
2483         PR ipa/85734
2484         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2485         as true in call to suggest_attribute.
2486
2487 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
2488
2489         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2490         -mreadonly-in-sdata.
2491
2492 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2493
2494         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2495         New pattern.
2496         (aarch64_crypto_aesd_fused): Likewise.
2497
2498 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
2499
2500         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2501         (movsi_aarch64): Likewise.
2502         (load_pairsi): Likewise.
2503         (load_pairdi): Likewise.
2504         (store_pairsi): Likewise.
2505         (store_pairdi): Likewise.
2506         (load_pairsf): Likewise.
2507         (load_pairdf): Likewise.
2508         (store_pairsf): Likewise.
2509         (store_pairdf): Likewise.
2510         (zero_extend): Likewise.
2511         (trunc): Swap alternatives.
2512         (fcvt_target): Add '?' to prefer w over r.
2513
2514 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
2515
2516         PR target/85756
2517         * config/i386/i386.md: Disallow non-commutative arithmetics in
2518         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2519         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
2520         in the peephole2 before it.
2521
2522 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
2523
2524         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2525         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2526         (ix86_handle_option): Handle -mcldemote.
2527         * config.gcc: New header.
2528         * config/i386/cldemoteintrin.h: New file.
2529         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2530         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2531         -mcldemote.
2532         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2533         OPTION_MASK_ISA_CLDEMOTE.
2534         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2535         (ix86_valid_target_attribute_inner_p): Ditto.
2536         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2537         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2538         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2539         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2540         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2541         (cldemote): New.
2542         * config/i386/i386.opt: Add -mcldemote.
2543         * config/i386/x86intrin.h: New header.
2544         * doc/invoke.texi: Add -mcldemote.
2545
2546 2018-05-14  Richard Biener  <rguenther@suse.de>
2547
2548         * doc/match-and-simplify.texi: Adjust :s documentation.
2549
2550 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
2551
2552         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2553         intended memcpy size.
2554         (REORDER_45): Likewise.
2555
2556 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
2557
2558         * sort.cc: New file.
2559         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2560         * vec.c (qsort_chk): Use gcc_qsort.
2561         * Makefile.in (OBJS-libcommon): Add sort.o.
2562         (build/sort.o): New target.  Use it...
2563         (BUILD_RTL): ... here, and...
2564         (build/gencfn-macros): ... here, and...
2565         (build/genmatch): ... here.
2566
2567 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2568             Chung-Ju Wu  <jasonwucj@gmail.com>
2569
2570         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2571         * config/nds32/nds32-graywolf.md: New file.
2572         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2573         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2574         pipeline.
2575         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2576         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2577         * config/nds32/nds32.md (pipeline_model): Add graywolf.
2578         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2579         * config/nds32/pipelines.md: Include n15 settings.
2580
2581 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2582             Chung-Ju Wu  <jasonwucj@gmail.com>
2583
2584         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2585         * config/nds32/nds32-n13.md: New file.
2586         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2587         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2588         pipeline.
2589         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2590         * config/nds32/nds32.md (pipeline_model): Add n13.
2591         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2592         * config/nds32/pipelines.md: Include n13 settings.
2593
2594 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2595             Chung-Ju Wu  <jasonwucj@gmail.com>
2596
2597         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2598         * config/nds32/nds32-n10.md: New file.
2599         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2600         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2601         pipeline.
2602         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2603         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2604         * config/nds32/nds32.md (pipeline_model): Add n10.
2605         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2606         * config/nds32/pipelines.md: Include n10 settings.
2607
2608 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
2609             Kito Cheng  <kito.cheng@gmail.com>
2610             Chung-Ju Wu  <jasonwucj@gmail.com>
2611
2612         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2613         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2614         Add enum values for DSP extension instructions.
2615         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2616         New constraints.
2617         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2618         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2619         New code iterators.
2620         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2621         * config/nds32/nds32-dspext.md: New file for DSP implementation.
2622         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2623         * config/nds32/nds32-intrinsic.md: Likewise.
2624         * config/nds32/nds32_intrinsic.h: Likewise.
2625         * config/nds32/nds32-md-auxiliary.c: Likewise.
2626         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2627         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2628         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2629         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2630         * config/nds32/nds32-protos.h: New declarations for DSP extension.
2631         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2632         TYPE_DMAC in switch statement.
2633         * config/nds32/nds32.c: New checking and implementation for DSP
2634         extension instructions.
2635         * config/nds32/nds32.h: Likewise.
2636         * config/nds32/nds32.md: Likewise.
2637         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2638         * config/nds32/predicates.md: Implement new predicates for DSP
2639         extension.
2640
2641 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2642
2643         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2644         Reformat alternatives and attributes so it is easier to identify
2645         which constraints/attributes go with which instruction.
2646         (mov<mode>_hardfloat32, FMOVE64): Likewise.
2647         (mov<mode>_softfloat32, FMOVE64): Likewise.
2648         (mov<mode>_hardfloat64, FMOVE64): Likewise.
2649         (mov<mode>_softfloat64, FMOVE64): Likewise.
2650
2651 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2652
2653         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2654         subsection.
2655         (Basic PowerPC Built-in Functions): The new name of the
2656         subsection previously known as "PowerPC Built-in Functions".
2657         (Basic PowerPC Built-in Functions Available on all Configurations):
2658         New subsubsection.
2659         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2660         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2661         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2662         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2663
2664 2018-05-11  Martin Jambor  <mjambor@suse.cz>
2665
2666         PR ipa/85655
2667         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2668         single const.
2669
2670 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
2671
2672         PR target/85733
2673         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2674
2675 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
2676
2677         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2678         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2679         (ix86_handle_option): Handle -mwaitpkg.
2680         * config.gcc: New header.
2681         * config/i386/cpuid.h (bit_WAITPKG): New bit.
2682         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2683         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2684         function type.
2685         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2686         OPTION_MASK_ISA_WAITPKG.
2687         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2688         (ix86_option_override_internal): Add PTA_WAITPKG.
2689         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2690         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2691         IX86_BUILTIN_TPAUSE.
2692         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2693         __builtin_ia32_umwait and __builtin_ia32_tpause.
2694         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2695         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2696         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2697         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2698         UNSPECV_TPAUSE): New.
2699         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2700         * config/i386/i386.opt: Add -mwaitpkg.
2701         * config/i386/waitpkgintrin.h: New file.
2702         * config/i386/x86intrin.h: New header.
2703         * doc/invoke.texi: Add -mwaitpkg.
2704
2705 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
2706
2707         PR target/85606
2708         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2709         equivalent.
2710         (cortex-m0): Use armv6s-m isa.
2711         (cortex-m0plus): Likewise.
2712         (cortex-m1): Likewise.
2713         (cortex-m0.small-multiply): Likewise.
2714         (cortex-m0plus.small-multiply): Likewise.
2715         (cortex-m1.small-multiply): Likewise.
2716
2717 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
2718             Jakub Jelinek  <jakub@redhat.com>
2719
2720         PR tree-optimization/85692
2721         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2722         source permute as well.
2723
2724 2018-05-11  Martin Liska  <mliska@suse.cz>
2725
2726         PR sanitizer/85556
2727         * doc/extend.texi: Document LLVM style format for no_sanitize
2728         attribute.
2729
2730 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
2731
2732         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2733         mode_supports_vsx_dform_quad to mode_supports_dq_form.
2734         (mode_supports_vsx_dform_quad): Likewise.
2735         (mode_supports_vmx_dform): Move these functions to be next to the
2736         other mode_supports functions.
2737         (mode_supports_dq_form): Likewise.
2738         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2739         mode_supports_dq_form.
2740         (reg_offset_addressing_ok_p): Likewise.
2741         (offsettable_ok_by_alignment): Likewise.
2742         (rs6000_legitimate_offset_address_p): Likewise.
2743         (legitimate_lo_sum_address_p): Likewise.
2744         (rs6000_legitimize_address): Likewise.
2745         (rs6000_legitimize_reload_address): Likewise.
2746         (rs6000_secondary_reload_inner): Likewise.
2747         (rs6000_preferred_reload_class): Likewise.
2748         (rs6000_output_move_128bit): Likewise.
2749
2750 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
2751
2752         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2753         Generate SImode target register for null target.
2754         <case IX86_BUILTIN_XGETBV>: Ditto.
2755         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2756         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2757
2758 2018-05-10  Carl Love  <cel@us.ibm.com>
2759
2760         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2761         dcbtt and dcbtstt if operands[2] is 0.
2762
2763 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
2764
2765         PR target/85693
2766         * config/i386/sse.md (usadv64qi): New expander.
2767
2768 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
2769
2770         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2771         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2772         -maltivec=be support.
2773         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2774         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2775         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2776         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2777         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2778         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2779         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2780         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2781         altivec_vsumsws): Adjust.
2782         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2783         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2784         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2785         support.
2786         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2787         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2788         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2789         (altivec_lve<VI_char>x): Delete expand.
2790         (*altivec_lve<VI_char>x_internal): Rename to...
2791         (altivec_lve<VI_char>x): ... this.
2792         (altivec_lvxl_<mode>): Delete expand.
2793         (*altivec_lvxl_<mode>_internal): Rename to ...
2794         (altivec_lvxl_<mode>): ... this.
2795         (altivec_stvxl_<mode>): Delete expand.
2796         (*altivec_stvxl_<mode>_internal): Rename to ...
2797         (altivec_stvxl_<mode>): ... this.
2798         (altivec_stve<VI_char>x): Delete expand.
2799         (*altivec_stve<VI_char>x_internal): Rename to ...
2800         (altivec_stve<VI_char>x): ... this.
2801         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2802         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2803         reduc_plus_scal_<mode>): Adjust.
2804         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2805         comment.
2806         (rs6000_cpu_cpp_builtins): Adjust.
2807         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2808         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2809         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2810         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2811         -maltivec=be support.
2812         (rs6000_split_vec_extract_var): Adjust.
2813         (rs6000_split_v4si_init): Adjust.
2814         (swap_selector_for_mode): Delete.
2815         (altivec_expand_lvx_be, altivec_expand_stvx_be,
2816         altivec_expand_stvex_be): Delete.
2817         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2818         -maltivec=be support.
2819         (rs6000_gimple_fold_builtin): Ditto.
2820         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2821         Adjust.
2822         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2823         (TARGET_DIRECT_MOVE_64BIT): Adjust.
2824         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2825         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2826         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2827         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2828         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2829         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2830         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2831         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2832         anonymous split): Adjust.
2833         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2834         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2835
2836 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
2837
2838         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2839         when --with-gxx-include-dir is also specified.
2840         * configure: Regenerate.
2841
2842 2018-05-09  Jim Wilson  <jimw@sifive.com>
2843
2844         PR target/84797
2845         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2846         * config/riscv/t-withmultilib: New.
2847         * config/riscv/withmultilib.h: New.
2848         * doc/install.texi: Document RISC-V --with-multilib-list support.
2849
2850 2018-05-09  Richard Biener  <rguenther@suse.de>
2851
2852         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2853         vector.
2854         (vect_bb_vectorization_profitable_p): Adjust.  Compute
2855         actual scalar cost using the cost vector and the add_stmt_cost
2856         machinery.
2857
2858 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2859
2860         PR rtl-optimization/85645
2861         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2862         in the REG_CFA_REGISTER note for LR, don't leave it empty.
2863
2864 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2865
2866         PR rtl-optimization/85645
2867         * shrink-wrap.c (spread_components): Return a boolean saying if
2868         anything was changed.
2869         (try_shrink_wrapping_separate): Iterate spread_components until
2870         nothing changes anymore.
2871
2872 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2873
2874         PR rtl-optimization/85645
2875         * regrename.c (build_def_use): Also kill the chains that include the
2876         destination of a REG_CFA_REGISTER note.
2877
2878 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2879
2880         PR rtl-optimization/85645
2881         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2882         insn that has a REG_CFA_REGISTER note.
2883
2884 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
2885
2886         * cfgexpand.c (expand_clobber): New function.
2887         (expand_gimple_stmt_1): Use it.
2888         * tree-vect-stmts.c (vect_clobber_variable): New function,
2889         split out from...
2890         (vectorizable_simd_clone_call): ...here.
2891         (vectorizable_store): Emit a clobber either side of an
2892         IFN_STORE_LANES sequence.
2893         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2894
2895 2018-05-09  Tom de Vries  <tom@codesourcery.com>
2896
2897         PR target/85626
2898         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2899         (define_insn "trap_if_false"): Add exit after trap.
2900
2901 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
2902
2903         PR rtl-optimization/85638
2904         * bb-reorder.c: Include common/common-target.h.
2905         (create_forwarder_block): New function extracted from...
2906         (fix_up_crossing_landing_pad): ...here.  Rename into...
2907         (dw2_fix_up_crossing_landing_pad): ...this.
2908         (sjlj_fix_up_crossing_landing_pad): New function.
2909         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2910         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2911         from both partitions and exit the loop after one iteration.
2912
2913 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2914
2915         Revert:
2916         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2917         subsection.
2918         (Basic PowerPC Built-in Functions): The new name of the
2919         subsection previously known as "PowerPC Built-in Functions".
2920         (Basic PowerPC Built-in Functions Available on all Configurations):
2921         New subsubsection.
2922         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2923         subsubsection.
2924         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2925         subsubsection.
2926         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2927         subsubsection.
2928         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2929         subsubsection.
2930
2931 2018-05-08  Jim Wilson  <jimw@sifive.com>
2932
2933         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2934         (LD_EMUL_SUFFIX): New.
2935         (LINK_SPEC): Use it.
2936
2937 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2938
2939         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2940         subsection.
2941         (Basic PowerPC Built-in Functions): The new name of the
2942         subsection previously known as "PowerPC Built-in Functions".
2943         (Basic PowerPC Built-in Functions Available on all Configurations):
2944         New subsubsection.
2945         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2946         subsubsection.
2947         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2948         subsubsection.
2949         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2950         subsubsection.
2951         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2952         subsubsection.
2953
2954 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
2955
2956         PR target/85683
2957         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2958         after cmpelim optimization.
2959
2960 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
2961
2962         * config.gcc: Support "goldmont".
2963         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2964         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2965         PROCESSOR_GOLDMONT.
2966         * config/i386/i386.c (m_GOLDMONT): Define.
2967         (processor_target_table): Add "goldmont".
2968         (PTA_GOLDMONT): Define.
2969         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2970         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2971         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2972         (fold_builtin_cpu): Add "goldmont".
2973         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2974         (ix86_option_override_internal): Add "goldmont".
2975         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2976         (processor_type): Add PROCESSOR_GOLDMONT.
2977         * config/i386/i386.md: Add CPU "glm".
2978         * config/i386/glm.md: New file.
2979         * config/i386/x86-tune.def: Add m_GOLDMONT.
2980         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2981
2982 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
2983
2984         PR target/85572
2985         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2986         E_V4DImode.
2987         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2988         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
2989         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
2990
2991         PR target/85317
2992         * config/i386/i386.c (ix86_fold_builtin): Handle
2993         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2994
2995         PR target/85480
2996         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2997         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2998
2999 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
3000
3001         PR target/85658
3002         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3003         (check_arch): Likewise.
3004         (check_fpu): Return the result rather than printing it.
3005         (end arch): Fix operator precedence.
3006         (end cpu): Likewise.
3007         (END): Print the result from check_fpu.
3008
3009 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
3010             Alan Hayward  <alan.hayward@arm.com>
3011             David Sherwood  <david.sherwood@arm.com>
3012
3013         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3014         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3015         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3016         (*fcmuo<mode>_and): New patterns.
3017
3018 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
3019
3020         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3021         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3022         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3023         (cmp_op, sve_imm_con): New code attributes.
3024         (SVE_COND_INT_CMP, imm_con): Delete.
3025         (cmp_op): Remove above unspecs from int attribute.
3026         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3027         to...
3028         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
3029         comparison-specific unspecs.
3030         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3031         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3032         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3033         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3034         (*vec_fcm<cmp_op><mode>): Rename to...
3035         (*fcm<cmp_op><mode>): ...this and adjust likewise.
3036         (*vec_fcmuo<mode>): Rename to...
3037         (*fcmuo<mode>): ...this and adjust likewise.
3038         (*pred_fcm<cmp_op><mode>): New pattern.
3039         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3040         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3041         functions.
3042         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3043         and UNORDERED.
3044         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3045         (aarch64_emit_sve_predicated_cond): New function.
3046         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3047         (aarch64_emit_unspec_cond_or): Replace with...
3048         (aarch64_emit_sve_or_conds): ...this new function.  Use
3049         aarch64_emit_sve_ptrue_op for the individual comparisons and
3050         aarch64_emit_binop to OR them together.
3051         (aarch64_emit_inverted_unspec_cond): Replace with...
3052         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
3053         aarch64_emit_sve_ptrue_op for the comparison and
3054         aarch64_emit_unop to invert the result.
3055         (aarch64_expand_sve_vec_cmp_float): Update after the above
3056         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
3057
3058 2018-05-07  Nathan Sidwell  <nathan@acm.org>
3059
3060         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3061         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3062         (Backwards Compatibility): Likewise.
3063
3064 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3065
3066         PR bootstrap/85681
3067         Revert:
3068         2018-05-07  Luis Machado  <luis.machado@linaro.org>
3069
3070         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3071         <prefetch_dynamic_strides>: New const bool field.
3072         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3073         prefetch_dynamic_strides.
3074         (exynosm1_prefetch_tune): Likewise.
3075         (thunderxt88_prefetch_tune): Likewise.
3076         (thunderx_prefetch_tune): Likewise.
3077         (thunderx2t99_prefetch_tune): Likewise.
3078         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3079         (aarch64_override_options_internal): Update to set
3080         PARAM_PREFETCH_DYNAMIC_STRIDES.
3081         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3082         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3083         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3084         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3085         prefetch-dynamic-strides setting.
3086
3087         2018-05-07  Luis Machado  <luis.machado@linaro.org>
3088
3089         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3090         <minimum_stride>: New const int field.
3091         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3092         minimum_stride field.
3093         (exynosm1_prefetch_tune): Likewise.
3094         (thunderxt88_prefetch_tune): Likewise.
3095         (thunderx_prefetch_tune): Likewise.
3096         (thunderx2t99_prefetch_tune): Likewise.
3097         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3098         (aarch64_override_options_internal): Update to set
3099         PARAM_PREFETCH_MINIMUM_STRIDE.
3100         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3101         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3102         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3103         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3104         stride is constant and is below the minimum stride threshold.
3105
3106 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3107
3108         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3109         to 512.
3110
3111 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3112
3113         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3114         <prefetch_dynamic_strides>: New const bool field.
3115         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3116         prefetch_dynamic_strides.
3117         (exynosm1_prefetch_tune): Likewise.
3118         (thunderxt88_prefetch_tune): Likewise.
3119         (thunderx_prefetch_tune): Likewise.
3120         (thunderx2t99_prefetch_tune): Likewise.
3121         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3122         (aarch64_override_options_internal): Update to set
3123         PARAM_PREFETCH_DYNAMIC_STRIDES.
3124         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3125         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3126         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3127         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3128         prefetch-dynamic-strides setting.
3129
3130 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3131
3132         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3133         <minimum_stride>: New const int field.
3134         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3135         minimum_stride field.
3136         (exynosm1_prefetch_tune): Likewise.
3137         (thunderxt88_prefetch_tune): Likewise.
3138         (thunderx_prefetch_tune): Likewise.
3139         (thunderx2t99_prefetch_tune): Likewise.
3140         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3141         (aarch64_override_options_internal): Update to set
3142         PARAM_PREFETCH_MINIMUM_STRIDE.
3143         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3144         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3145         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3146         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3147         stride is constant and is below the minimum stride threshold.
3148
3149 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
3150
3151         PR c++/85659
3152         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3153         the type is addressable.  Don't force op into register if it has
3154         BLKmode.
3155
3156 2018-05-05  Roland McGrath  <mcgrathr@google.com>
3157
3158         PR other/77609
3159         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3160         any section for which we don't know a specific type it should have,
3161         regardless of name.  Previously this was done only for the exact
3162         names ".init_array", ".fini_array", and ".preinit_array".
3163         (default_elf_asm_named_section): Add comment about
3164         relationship with default_section_type_flags and SECTION_NOTYPE.
3165         (get_section): Don't consider it a type conflict if one side has
3166         SECTION_NOTYPE and the other doesn't, as long as neither has the
3167         SECTION_BSS et al used in the default_section_type_flags logic.
3168
3169 2018-05-05  Tom de Vries  <tom@codesourcery.com>
3170
3171         PR target/85653
3172         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3173         (workaround_barsyncs): New function.
3174         (nvptx_reorg): Use workaround_barsyncs.
3175         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3176         (define_expand "nvptx_membar_cta"): New define_expand.
3177         (define_insn "*nvptx_membar_cta"): New insn.
3178
3179 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
3180
3181         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3182         To improve optimization opportunities.
3183         * builtin-types.def: The new needed builtin types for the above.
3184
3185 2018-05-04  Richard Biener  <rguenther@suse.de>
3186
3187         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3188         * gimple-ssa-store-merging.c
3189         (imm_store_chain_info::output_merged_store): Remove redundant create,
3190         release split_store vector contents on failure.
3191         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3192         scalar stmt vector on cache hit.
3193
3194 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
3195
3196         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3197         Xilinx FP support.
3198         * config.gcc (powerpc-xilinx-eabi*): Remove.
3199         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3200         support.
3201         (fusion_addis_mem_combo_load): Ditto.
3202         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3203         FP support.
3204         (rs6000_cpu_cpp_builtins): Ditto.
3205         * config/rs6000/rs6000-linux.c
3206         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3207         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3208         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3209         support.
3210         (rs6000_setup_reg_addr_masks): Ditto.
3211         (rs6000_init_hard_regno_mode_ok): Ditto.
3212         (rs6000_option_override_internal): Ditto.
3213         (legitimate_lo_sum_address_p): Ditto.
3214         (rs6000_legitimize_address): Ditto.
3215         (rs6000_legitimize_reload_address): Ditto.
3216         (rs6000_legitimate_address_p): Ditto.
3217         (abi_v4_pass_in_fpr): Ditto.
3218         (setup_incoming_varargs): Ditto.
3219         (rs6000_gimplify_va_arg): Ditto.
3220         (rs6000_split_multireg_move): Ditto.
3221         (rs6000_savres_strategy): Ditto.
3222         (rs6000_emit_prologue_components): Ditto.
3223         (rs6000_emit_epilogue_components): Ditto.
3224         (rs6000_emit_prologue): Ditto.
3225         (rs6000_emit_epilogue): Ditto.
3226         (rs6000_elf_file_end): Ditto.
3227         (rs6000_function_value): Ditto.
3228         (rs6000_libcall_value): Ditto.
3229         * config/rs6000/rs6000.h: Ditto.
3230         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3231         (TARGET_MINMAX): ... this.  New.
3232         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3233         * config/rs6000/rs6000.md: Remove Xilinx FP support.
3234         (*movsi_internal1_single): Delete.
3235         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3236         mfpu=, mxilinx-fpu): Delete.
3237         * config/rs6000/singlefp.h: Delete.
3238         * config/rs6000/sysv4.h: Remove Xilinx FP support.
3239         * config/rs6000/t-rs6000: Ditto.
3240         * config/rs6000/t-xilinx: Delete.
3241         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3242         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3243         (VStype_simple): Delete.
3244         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3245         * config/rs6000/xfpu.h: Delete.
3246         * config/rs6000/xfpu.md: Delete.
3247         * config/rs6000/xilinx.h: Delete.
3248         * config/rs6000/xilinx.opt: Delete.
3249         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3250         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3251
3252 2018-05-04  Tom de Vries  <tom@codesourcery.com>
3253
3254         PR libgomp/85639
3255         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3256         if ignore == 0.
3257
3258 2018-05-04  Richard Biener  <rguenther@suse.de>
3259
3260         PR middle-end/85627
3261         * tree-complex.c (update_complex_assignment): We are always in SSA form.
3262         (expand_complex_div_wide): Likewise.
3263         (expand_complex_operations_1): Likewise.
3264         (expand_complex_libcall): Preserve EH info of the original stmt.
3265         (tree_lower_complex): Handle removed blocks.
3266         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3267         on complex multiplication and division libcall builtins.
3268
3269 2018-05-04  Richard Biener  <rguenther@suse.de>
3270
3271         PR middle-end/85574
3272         * fold-const.c (negate_expr_p): Restrict negation of operand
3273         zero of a division to when we know that can happen without
3274         overflow.
3275         (fold_negate_expr_1): Likewise.
3276
3277 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
3278
3279         PR libstdc++/85466
3280         * real.h (real_nextafter): Declare.
3281         * real.c (real_nextafter): New function.
3282         * fold-const-call.c (fold_const_nextafter): New function.
3283         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3284         CASE_CFN_NEXTTOWARD.
3285         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3286         even when arg1_mode is different from arg0_mode.
3287
3288 2018-05-03  Nathan Sidwell  <nathan@acm.org>
3289
3290         * doc/extend.texi (Deprecated Features): Remove
3291         -ffriend-injection.
3292         (Backwards Compatibility): Likewise.
3293         * doc/invoke.texi (C++ Language Options): Likewise.
3294         (C++ Dialect Options): Likewise.
3295
3296 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
3297
3298         PR target/85530
3299         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3300         _mm512_mask_mullox_epi64): New intrinsics.
3301
3302 2018-05-03  Tom de Vries  <tom@codesourcery.com>
3303
3304         PR testsuite/85106
3305         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3306         dump files): Add offload-tree.
3307
3308 2018-05-03  Richard Biener  <rguenther@suse.de>
3309
3310         PR tree-optimization/85615
3311         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3312         to loops not nested in BBs loop father to avoid creating multi-entry
3313         loops.
3314
3315 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3316
3317         PR tree-optimization/70291
3318         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3319         arguments.  Change return type to tree.  Emit libcall as a new
3320         statement rather than replacing existing one when inplace_p is true.
3321         (expand_complex_multiplication_components): New function.
3322         (expand_complex_multiplication): Expand floating-point complex
3323         multiplication using the above.
3324         (expand_complex_division): Rename inner_type parameter to type.
3325         Update expand_complex_libcall call-site.
3326         (expand_complex_operations_1): Update expand_complex_multiplication
3327         and expand_complex_division call-sites.
3328
3329 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
3330
3331         PR target/85582
3332         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3333         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3334         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3335         the highest significant bit of the shift count mask is clear.  In
3336         check whether and[sq]i3 is needed verify that all significant bits
3337         of the shift count other than the highest are set.
3338
3339 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3340
3341         PR libgomp/82428
3342         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3343         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3344         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3345         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3346         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3347         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3348         __builtin_goacc_parlevel_size.
3349
3350 2018-05-02  Richard Biener  <rguenther@suse.de>
3351
3352         PR tree-optimization/85597
3353         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3354         do not use split vect_get_vec_defs call but call vect_get_slp_defs
3355         directly.
3356
3357 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3358
3359         PR testsuite/85106
3360         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3361         dump files): Add ltrans-tree.
3362
3363 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3364
3365         PR testsuite/85106
3366         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3367         dump files): Add wpa-ipa.
3368
3369 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
3370
3371         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
3372         powerpc*-*-linux*paired* target.
3373         * config/rs6000/750cl.h: Delete.
3374         * config/rs6000/paired.h: Delete.
3375         * config/rs6000/paired.md: Delete.
3376         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3377         float support.
3378         * config/rs6000/rs6000-builtin.def: Remove paired float support.
3379         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3380         comment.  Remove paired float support.
3381         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3382         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3383         VECTOR_PAIRED.
3384         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3385         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3386         declarations.
3387         * config/rs6000/rs6000.c: Remove paired float support.
3388         (paired_expand_vector_init, paired_expand_vector_move,
3389         paired_emit_vector_compare, paired_emit_vector_cond_expr,
3390         (paired_expand_lv_builtin, paired_expand_stv_builtin,
3391         paired_expand_builtin, paired_expand_predicate_builtin,
3392         paired_init_builtins): Delete.
3393         * config/rs6000/rs6000.h: Remove paired float support.
3394         * config/rs6000/rs6000.md: Remove paired float support.
3395         (move_from_CR_ov_bit): Delete.
3396         * config/rs6000/rs6000.opt (mpaired): Delete.
3397         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3398         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3399
3400 2018-05-02  Richard Biener  <rguenther@suse.de>
3401
3402         PR middle-end/85567
3403         * gimplify.c (gimplify_save_expr): When in SSA form allow
3404         SAVE_EXPRs to compute to SSA vars.
3405
3406 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
3407
3408         PR target/85582
3409         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3410         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3411         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3412         clobber operands[2], instead use a new pseudo.  Formatting fixes.
3413
3414 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
3415
3416         PR tree-optimization/85586
3417         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3418         exit early for statements in the same group if the accesses are
3419         not strided.
3420
3421 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3422
3423         PR lto/85451
3424         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3425         error message.
3426
3427 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
3428
3429         PR tree-optimization/85143
3430         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3431
3432 2018-05-01  Tom de Vries  <tom@codesourcery.com>
3433
3434         PR lto/85451
3435         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3436         not found" error message.
3437
3438 2018-05-01  Tom de Vries  <tom@codesourcery.com>
3439
3440         PR other/83786
3441         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3442         * vec.c (test_ordered_remove_if): New function.
3443         (vec_c_tests): Call test_ordered_remove_if.
3444         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3445         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3446         * tree-vect-patterns.c (vect_pattern_recog_1): Use
3447         VEC_ORDERED_REMOVE_IF.
3448
3449 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3450
3451         PR tree-optimization/82665
3452         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3453         pointer subtraction where arguments come from a memchr call.
3454
3455 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
3456
3457         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3458         --push-state --as-needed and --pop-state instead of --as-needed and
3459         --no-as-needed if ld supports it.
3460         * configure: Regenerated.
3461
3462         PR web/85578
3463         * doc/install.texi2html: Replace _002d with - and _002a with * in
3464         generated html files using sed.
3465
3466 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3467
3468         PR c++/85523
3469         * gcc-rich-location.c (blank_line_before_p): New function.
3470         (use_new_line): New function.
3471         (gcc_rich_location::add_fixit_insert_formatted): New function.
3472         * gcc-rich-location.h
3473         (gcc_rich_location::add_fixit_insert_formatted): New function.
3474
3475 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3476
3477         * selftest.c (assert_streq): Rename "expected" and "actual" to
3478         "val1" and "val2".  Extend NULL-handling to cover both inputs
3479         symmetrically, while still requiring both to be non-NULL for a pass.
3480         * selftest.h (assert_streq): Rename "expected" and "actual" to
3481         "val1" and "val2".
3482         (ASSERT_EQ): Likewise.
3483         (ASSERT_EQ_AT): Likewise.
3484         (ASSERT_KNOWN_EQ): Likewise.
3485         (ASSERT_KNOWN_EQ_AT): Likewise.
3486         (ASSERT_NE): Likewise.
3487         (ASSERT_MAYBE_NE): Likewise.
3488         (ASSERT_MAYBE_NE_AT): Likewise.
3489         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
3490         the assertion to pass.
3491         (ASSERT_STREQ_AT): Likewise.
3492
3493 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
3494
3495         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3496         interaction with -pie.
3497
3498 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3499
3500         * selftest.h: Fix alphabetization of per-source-file selftest
3501         declarations.
3502
3503 2018-04-30  Jason Merrill  <jason@redhat.com>
3504
3505         PR c++/61982 - dead stores to destroyed objects.
3506         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3507         of clobber.
3508
3509 2018-04-30  Jason Merrill  <jason@redhat.com>
3510
3511         * tree.c (build_clobber): New.
3512         * tree.h: Declare it.
3513         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3514
3515 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3516
3517         * diagnostic-show-locus.c (layout::layout): Update for
3518         location_get_source_line returning a char_span.
3519         (struct char_span): Move to input.h.
3520         (struct correction): Update for fields in char_span becoming
3521         private.
3522         (struct source_line): Update for location_get_source_line
3523         returning a char_span.
3524         (layout::print_line): Likewise.
3525         * edit-context.c (edited_file::print_content): Likewise.
3526         (edited_file::print_diff_hunk): Likewise.
3527         (edited_file::print_run_of_changed_lines): Likewise.
3528         (edited_file::get_num_lines): Likewise.
3529         (edited_line::edited_line): Likewise.
3530         * final.c (asm_show_source): Likewise.
3531         * input.c (location_get_source_line): Convert return type
3532         from const char * to char_span, losing the final "line_len"
3533         param.
3534         (dump_location_info): Update for the above.
3535         (get_substring_ranges_for_loc): Likewise.  Use a char_span
3536         when handling the literal within the line.
3537         (test_reading_source_line): Update for location_get_source_line
3538         returning a char_span.
3539         * input.h (class char_span): Move here from
3540         diagnostic-show-locus.c, converting from a struct to a class.
3541         Make data members private.
3542         (char_span::operator bool): New.
3543         (char_span::length): New.
3544         (char_span::get_buffer): New.
3545         (char_span::operator[]): New.
3546         (char_span::subspan): Make const.
3547         (char_span::xstrdup): New.
3548         (location_get_source_line): Convert return type from const char *
3549         to char_span, losing the final "line_size" param.
3550
3551 2018-04-30  Jan Hubicka  <jh@suse.cz>
3552
3553         * lto-wrapper.c (ltrans_priorities): New static var.
3554         (cmp_priority): New.
3555         (run_gcc): Read priorities and if doing parallel build order
3556         the Makefile by them.
3557
3558 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3559
3560         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3561
3562 2018-04-30  Richard Biener  <rguenther@suse.de>
3563
3564         * tree-cfg.c (verify_address): Remove base argument, add
3565         flag whether to check TREE_ADDRESSABLE and do that.
3566         (verify_expr): Remove.
3567         (verify_types_in_gimple_reference): Add pieces from verify_expr.
3568         (verify_gimple_assign_single): Likewise.
3569         (verify_gimple_switch): Likewise.
3570         (verify_expr_location_1): Dereference tp once.  Add (disabled)
3571         piece from verify_expr.
3572         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3573
3574 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3575
3576         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3577
3578 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3579
3580         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3581         (small_data_pattern): Likewise.
3582         (arc_rewrite_small_data): Likewise.
3583         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3584         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3585         (get_symbol_alignment): New function.
3586         (legitimate_small_data_address_p): Likewise.
3587         (legitimate_scaled_address): Update, call
3588         legitimate_small_data_address_p.
3589         (output_sdata): New static variable.
3590         (arc_print_operand): Update how we handle small data operands.
3591         (arc_print_operand_address): Likewise.
3592         (arc_legitimate_address_p): Update, use
3593         legitimate_small_data_address_p.
3594         (arc_rewrite_small_data_p): Remove.
3595         (arc_rewrite_small_data_1): Likewise.
3596         (arc_rewrite_small_data): Likewise.
3597         (small_data_pattern): Likewise.
3598         (compact_sda_memory_operand): Update to use
3599         legitimate_small_data_address_p and get_symbol_alignment.
3600         (prepare_move_operands): Don't rewite sdata pattern.
3601         (prepare_extend_operands): Remove.
3602         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3603         pattern.
3604         (zero_extendqisi2): Likewise.
3605         (zero_extendhisi2): Likewise.
3606         (extendqihi2): Likewise.
3607         (extendqisi2): Likewise.
3608         (extendhisi2): Likewise.
3609         (addsi3): Likewise.
3610         (subsi3): Likewise.
3611         (andsi3): Likewise.
3612         * config/arc/constraints.md (Usd): Change it to memory constraint.
3613
3614 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3615
3616         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3617         as source of std instructions.
3618         * config/arc/arc.md (movsi_insn): Update pattern predicate to
3619         allow 6-bit constants as source for store instructions.
3620         (movdi_insn): Update instruction pattern to allow 6-bit constants
3621         as source for store instructions.
3622
3623 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
3624
3625         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3626
3627 2018-04-30  Nathan Sidwell  <nathan@acm.org>
3628             Sandra Loosemore <sandra@codesourcery.com>
3629
3630         * dumpfile.c (dump_open): Allow '-' for stdout.
3631         * doc/invoke.texi (Developer Options): Document dump filename
3632         determination early.  Document stdin/stdout selection.
3633
3634 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
3635
3636         Microblaze Target: PIC data text relative
3637
3638         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3639         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3640         Add declaration.
3641         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3642         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3643         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3644         New addressing mode for data-text relative position indepenedent code.
3645         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3646         'ADDRESS_SYMBOLIC_TXT_REL'.
3647         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3648         (microblaze_legitimate_pic_operand): Exclude function calls from
3649         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3650         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3651         addresses cases.
3652         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3653         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3654         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3655         for 'address + offset'.
3656         (microblaze_expand_prologue): Add new function prologue call for
3657         'r20' assignation.
3658         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3659         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3660         table in case of TARGET_PIC_DATA_TEXT_REL.
3661         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3662         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3663         Add new macros 'UNSPEC_TEXT',
3664         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3665         + exclude function calls from 'UNSPEC_PLT' in case of data text
3666         relative mode.
3667         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3668         new target hook for generating address diff vector tables in case of
3669         flag_pic.
3670         * doc/tm.texi : Regenerate.
3671         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3672         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3673         of addr diff vector generation.
3674         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3675         target hook definition.
3676         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3677         Add default function for generate_pic_addr_diff_vec -> flag_pic.
3678         * doc/invoke.texi (Add new pic option): Add new microblaze pic
3679         option for data text relative.
3680
3681 2018-04-30  Richard Biener  <rguenther@suse.de>
3682
3683         * tree-chrec.h (evolution_function_is_constant_p): Remove
3684         redundant check.
3685         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3686
3687 2018-04-30  Richard Biener  <rguenther@suse.de>
3688
3689         PR bootstrap/85571
3690         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3691
3692 2018-04-30  Richard Biener  <rguenther@suse.de>
3693
3694         PR tree-optimization/28364
3695         PR tree-optimization/85275
3696         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3697         copying first exit test.
3698
3699 2018-04-28  Mark Wielaard  <mark@klomp.org>
3700
3701         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3702         dwarf_version >= 5.
3703         (dwarf_AT): Handle DW_AT_addr_base.
3704         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3705
3706 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
3707
3708         PR target/84431
3709         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3710         (*ashl<dwi>3_doubleword_mask_1): Ditto.
3711         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3712         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3713
3714 2018-04-28  Richard Biener  <rguenther@suse.de>
3715
3716         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3717         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3718         to reflect use.  Only add interesting stmts.
3719
3720 2018-04-27  Martin Jambor  <mjambor@suse.cz>
3721
3722         PR ipa/85549
3723         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3724         the jump function allows for passing through aggregate values.
3725
3726 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
3727
3728         * input.h (in_system_header_at): Convert from macro to inline
3729         function.
3730         (from_macro_expansion_at): Likewise.
3731         (from_macro_definition_at): Likewise.
3732
3733 2018-04-27  Jeff Law  <law@redhat.com>
3734
3735         * config.gcc: Mark tile* targets as deprecated/obsolete.
3736
3737 2018-04-27  Richard Biener  <rguenther@suse.de>
3738
3739         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3740         fix for ILP32.
3741
3742 2018-04-27  Richard Biener  <rguenther@suse.de>
3743
3744         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3745
3746 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
3747
3748         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3749         with Yd constraint. Set "preferred_for_speed" attribute from
3750         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3751         with Yd constraint.
3752         (*movdi_internal): Ditto.
3753         (movti_interunit splitters): Remove
3754         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3755         (movdi_interunit splitters): Ditto.
3756         * config/i386/constraints.md (Ye): Remove.
3757         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3758
3759 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3760
3761         PR target/85512
3762         * config/aarch64/constraints.md (Usg): Limit to 31.
3763         (Usj): Limit to 63.
3764
3765 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
3766
3767         PR tree-optimization/85529
3768         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3769         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3770         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3771         zero extension or masking of the MSB bit.
3772         (optimize_range_tests): Add FIRST_BB argument, pass it through
3773         to optimize_range_tests_var_bound.
3774         (maybe_optimize_range_tests, reassociate_bb): Adjust
3775         optimize_range_tests callers.
3776
3777 2018-04-26  Richard Biener  <rguenther@suse.de>
3778             Jakub Jelinek  <jakub@redhat.com>
3779
3780         * cgraph.h (symbol_table): Just declare debug method here.
3781         * symtab.c (symbol_table::debug): Define.
3782
3783 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
3784
3785         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3786
3787 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
3788
3789         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3790         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3791         (*movdi_internal): Substitute Yi and Yj constraint with x
3792         and Ym and Yn constraint with y constraint.  Update "isa"
3793         attribute and set "preferred_for_speed" attribute from
3794         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3795         (*movsi_internal): Ditto.
3796         (*movdf_internal): Ditto.
3797         (*movsf_internal): Ditto.
3798         (*zero_extendsidi2): Ditto.
3799         * config/i386/sse.md (vec_set<mode>_0): Ditto.
3800         (sse2_loadld): Ditto.
3801         (*vec_extract<ssevecmodelower>_0): Ditto.
3802         (*vec_extractv4si_0_zext_sse4): Ditto.
3803         (vec_concatv2di): Ditto.
3804         (*vec_dup<mode>): Ditto.
3805         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3806         * config/i386/constraints.md (Yi): Remove.
3807         (Yj): Remove.
3808         (Ym): Remove.
3809         (Yn): Remove.
3810
3811 2018-04-26  Nathan Sidwell  <nathan@acm.org>
3812
3813         * dumpfile.c (dump_open): New.
3814         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3815         (dump_finish): Detect stdio/stderr by value not name.
3816
3817 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
3818
3819         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3820
3821 2018-04-26  Tom de Vries  <tom@codesourcery.com>
3822
3823         PR target/84952
3824         * config/nvptx/nvptx.c (verify_neutering_jumps)
3825         (verify_neutering_labels): New function
3826         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3827
3828 2018-04-26  Tom de Vries  <tom@codesourcery.com>
3829
3830         PR target/84025
3831         * config/nvptx/nvptx.c (needs_neutering_p): New function.
3832         (nvptx_single): Use needs_neutering_p to skip over insns that do not
3833         need neutering.
3834
3835 2018-04-26  Richard Biener <rguenther@suse.de>
3836             Tom de Vries  <tom@codesourcery.com>
3837
3838         PR lto/85422
3839         * lto-streamer-out.c (output_function): Fixup loops if required to match
3840         discovery done in the reader.
3841
3842 2018-04-26  Richard Biener  <rguenther@suse.de>
3843
3844         PR tree-optimization/85116
3845         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3846         have a loop exit from the single latch predecessor.  Remove
3847         case of header with just condition.
3848         (ch_base::copy_headers): Exclude infinite loops from any
3849         processing.
3850         (pass_ch::execute): Record exits.
3851
3852 2018-04-26  Richard Biener  <rguenther@suse.de>
3853
3854         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3855         prologue cost vector and pass it to vect_get_load_cost.
3856         (vect_get_peeling_costs_all_drs): Likewise.
3857         (vect_peeling_hash_get_lowest_cost): Likewise.
3858         (vect_enhance_data_refs_alignment): Likewise.
3859
3860 2018-04-26  Richard Biener  <rguenther@suse.de>
3861
3862         PR middle-end/85450
3863         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3864         checking of integer<->pointer conversions.
3865         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3866         sign-/zero-extending pointer types.
3867         (expand_omp_for_static_chunk): Likewise.
3868
3869 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
3870             Jean Lee  <xiaoyur347@gmail.com>
3871
3872         * config/mips/mips.c (mips_asan_shadow_offset): New function.
3873         (TARGET_ASAN_SHADOW_OFFSET): Define.
3874         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3875         true for -fsanitize=address.
3876
3877 2018-04-25  Mark Wielaard  <mark@klomp.org>
3878
3879         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3880         shorter ones.
3881
3882 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3883
3884         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3885         than "alu", remove explicit "memory" and "imm_disp" attributes.
3886         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3887
3888         PR middle-end/85414
3889         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3890         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3891         gen_lowpart_no_emit.
3892
3893 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
3894
3895         PR target/85473
3896         * config/i386/i386.c (ix86_expand_builtin): Change memory
3897         operand to XI, extend p0 to Pmode.
3898         * config/i386/i386.md: Change unspec volatile and operand
3899         1 mode to XI, change operand 0 mode to P.
3900
3901 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3902
3903         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3904         GET_MODE_MASK before any checking.
3905         (nds32_can_use_bset_p): Likewise.
3906         (nds32_can_use_btgl_p): Likewise.
3907
3908 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3909
3910         * config/nds32/nds32-doubleword.md: New define_split pattern for
3911         illegal register number.
3912
3913 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3914
3915         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3916
3917 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3918
3919         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3920
3921 2018-04-25  Richard Biener  <rguenther@suse.de>
3922
3923         * lto-streamer.h (LTO_major_version): Bump to 8.
3924
3925 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3926
3927         * BASE-VER: Set to 9.0.0.
3928
3929 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
3930
3931         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3932         in __abskf2 and __powikf2.
3933
3934 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3935
3936         PR target/85512
3937         * config/aarch64/constraints.md (Usg, Usj): New constraints.
3938         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3939         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3940         Use the above on operand 2.  Reindent.
3941         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3942
3943 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
3944
3945         PR target/85485
3946         * common/config/i386/i386-common.c (ix86_handle_option): Don't
3947         handle OPT_mcet.
3948         * config/i386/i386.opt (mcet): Removed.
3949         * doc/install.texi: Remove -mcet documentation.
3950         * doc/invoke.texi: Likewise.
3951
3952 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
3953
3954         PR target/85485
3955         * doc/install.texi: Remove -mcet from bootstrap-cet.
3956
3957 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3958
3959         PR target/85511
3960         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3961         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3962         if TARGET_64BIT.
3963
3964         PR target/85503
3965         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3966         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3967         containing a CONST_VECTOR.
3968
3969 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
3970
3971         * doc/install.texi: Update newlib dependency for nvptx.
3972
3973 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3974
3975         PR target/85508
3976         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3977         instead of INTVAL when shifting x left.
3978
3979 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
3980
3981         PR tree-optimization/85478
3982         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3983         vect_grouped_store_supported for single element vectors.
3984
3985 2018-04-24  Richard Biener  <rguenther@suse.de>
3986
3987         PR target/85491
3988         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3989         load cost increase to the case of non-constant step.
3990
3991 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3992
3993         PR target/84828
3994         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3995         destination if any_malformed_asm.
3996
3997 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
3998
3999         PR middle-end/85496
4000         * expr.c (store_field): In the bitfield case, if the value comes from
4001         a function call and is returned in registers by means of a PARALLEL,
4002         do not change the mode of the temporary unless BLKmode and VOIDmode.
4003
4004 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
4005
4006         PR rtl-optimization/85423
4007         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4008         dependencies to debug insns when the previous insn is non-debug.
4009
4010 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
4011
4012         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4013         enums into a single definition.
4014         (fls): Fix predicates and printing.
4015         (seti): Likewise.
4016
4017 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
4018
4019         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4020         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4021         and short u6 immediate.
4022         (check_if_valid_sleep_operand): Remove.
4023         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4024
4025 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4026
4027         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4028         flag_always_save_lp condition.
4029         * config/nds32/nds32.opt (malways-save-lp): New option.
4030
4031 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
4032
4033         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4034         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4035         * config/nds32/nds32.h
4036         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4037         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4038
4039 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
4040
4041         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4042         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4043
4044 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
4045             Chung-Ju Wu  <jasonwucj@gmail.com>
4046
4047         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4048         Declare.
4049         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4050         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4051
4052 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4053
4054         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4055
4056 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4057
4058         * config/nds32/nds32-protos.h (nds32_data_alignment,
4059         nds32_local_alignment): Declare.
4060         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4061         nds32_local_alignment): New functions.
4062         (TARGET_CONSTANT_ALIGNMENT): Define.
4063         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4064
4065 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4066
4067         * config/nds32/nds32.c
4068         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4069         (TARGET_MODES_TIEABLE_P): Likewise.
4070
4071 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4072
4073         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4074         level Ofast and Og.
4075
4076 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
4077             Chung-Ju Wu  <jasonwucj@gmail.com>
4078
4079         * config/nds32/constants.md (unspec_volatile_element): Add enum values
4080         for unaligned access.
4081         * config/nds32/nds32-intrinsic.c: Implementation of expanding
4082         unaligned access.
4083         * config/nds32/nds32-intrinsic.md: Likewise.
4084         * config/nds32/nds32_intrinsic.h: Likewise.
4085         * config/nds32/nds32.h (nds32_builtins): Likewise.
4086         * config/nds32/nds32.opt (munaligned-access): New option.
4087         * config/nds32/nds32.c (nds32_asm_file_start): Display
4088         flag_unaligned_access status.
4089
4090 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
4091
4092         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4093         -mno-relax is present.
4094         * config/riscv/linux.h (LINK_SPEC): Ditto.
4095
4096 2018-04-20  Martin Sebor  <msebor@redhat.com>
4097
4098         PR c/85365
4099         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4100         for null pointers.
4101         (gimple_fold_builtin_stxcpy_chk): Same.
4102         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4103
4104 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
4105
4106         PR target/85456
4107         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4108         __powikf2 when long double is IEEE 128-bit.
4109
4110 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
4111
4112         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4113         step to make sure stack always aligned.
4114
4115 2018-04-20  Carl Love  <cel@us.ibm.com>
4116
4117         PR target/83402
4118         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4119         size check for arg0.
4120
4121 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
4122             Tom de Vries  <tom@codesourcery.com>
4123
4124         PR target/85445
4125         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4126         Emit insns for calls too.
4127         (nvptx_find_par): Always look for worker-level predecessor insn.
4128         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
4129         calls.
4130         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4131         (nvptx_process_pars): Propagate frames for calls.
4132
4133 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
4134
4135         PR target/85469
4136         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4137         Removed.
4138         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4139         (ix86_handle_option): Don't handle OPT_mibt.
4140         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4141         __SHSTK__.
4142         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4143         has_ibt and ibt.
4144         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4145         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4146         (ix86_target_macros): Define __CET__ with flag_cf_protection
4147         for -fcf-protection.
4148         * config/i386/i386.c (isa2_opts): Remove -mibt.
4149         * config/i386/i386.h (TARGET_IBT): Removed.
4150         (TARGET_IBT_P): Likewise.
4151         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4152         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4153         * config/i386/i386.opt (mcet): Update help message.
4154         (mshstk): Likewise.
4155         (mibt): Removed.
4156         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
4157         -mcet as an alias for -mshstk.
4158
4159 2018-04-20  Richard Biener <rguenther@suse.de>
4160
4161         PR middle-end/85475
4162         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4163         complexity by forcing a single use of the multiply operand.
4164
4165 2018-04-20  Martin Jambor  <mjambor@suse.cz>
4166
4167         ipa/85449
4168         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4169         recursion dependency to only apply to non-clones.
4170
4171 2018-04-20  Martin Jambor  <mjambor@suse.cz>
4172
4173         ipa/85447
4174         * ipa-cp.c (create_specialized_node): Check that clones of
4175         self-recursive edges exist during IPA-CP.
4176
4177 2018-04-19  Toon Moene  <toon@moene.org>
4178
4179         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4180         by -O3.
4181
4182 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
4183
4184         PR tree-optimization/85467
4185         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4186         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
4187         VECTOR_CST element to type.
4188
4189 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4190
4191         PR target/85397
4192         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4193         * config/i386/i386.md (builtin_setjmp_setup): Removed.
4194         (builtin_longjmp): Likewise.
4195         (save_stack_nonlocal): New pattern.
4196         (restore_stack_nonlocal): Likewise.
4197
4198 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4199
4200         PR target/85404
4201         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4202         Replace ASM_OUTPUT_LABEL with fprintf.
4203
4204 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4205
4206         PR target/85417
4207         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4208         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4209         * config/i386/i386-c.c (ix86_target_macros_internal): Also
4210         define __IBT__ and __SHSTK__ for -fcf-protection.
4211         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4212         TARGET_IBT.
4213         (ix86_trampoline_init): Likewise.
4214         (x86_output_mi_thunk): Likewise.
4215         (ix86_notrack_prefixed_insn_p): Likewise.
4216         (ix86_option_override_internal): Don't disallow -fcf-protection.
4217         * config/i386/i386.md (rdssp<mode>): Also enable for
4218         -fcf-protection.
4219         (incssp<mode>): Likewise.
4220         (nop_endbr): Likewise.
4221         * config/i386/i386.opt (mcet): Change help message to built-in
4222         functions only.
4223         (mibt): Likewise.
4224         (mshstk): Likewise.
4225         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4226         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
4227         enable CET built-in functions.
4228
4229 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
4230
4231         * common/config/i386/i386-common.c
4232         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4233         OPTION_MASK_ISA_MOVDIRI_UNSET,
4234         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4235         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4236         * config.gcc (movdirintrin.h): New header.
4237         * config/i386/cpuid.h (bit_MOVDIRI,
4238         bit_MOVDIR64B): New bits.
4239         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4240         and -mmvodir64b.
4241         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4242         (VOID, PVOID, PCVOID)): New function types.
4243         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4244         __builtin_ia32_directstoreu_u64,
4245          __builtin_ia32_movdir64b): New builtins.
4246         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4247         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4248         and -mmovdiri.
4249         (ix86_valid_target_attribute_inner_p): Ditto.
4250         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4251         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4252         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4253         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4254         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4255         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4256         (movdiri<mode>, movdir64b_<mode>): New.
4257         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4258         * config/i386/immintrin.h: Include movdirintrin.h.
4259         * config/i386/movdirintrin.h: New file.
4260         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4261
4262 2018-04-19  Richard Biener  <rguenther@suse.de>
4263
4264         PR middle-end/85455
4265         * cfg.c (clear_bb_flags): When loop state says we have
4266         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4267
4268 2018-04-19  Richard Biener  <rguenther@suse.de>
4269
4270         PR tree-optimization/84737
4271         * tree-vect-data-refs.c (vect_copy_ref_info): New function
4272         copying restrict info.
4273         (vect_setup_realignment): Use it.
4274         * tree-vectorizer.h (vect_copy_ref_info): Declare.
4275         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4276         the first DR to all generated stores.
4277         (vectorizable_load): Likewise for loads.
4278
4279 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
4280
4281         PR tree-optimization/85446
4282         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4283         the integral and pointer types to have the same precision.
4284
4285         * doc/install.texi: Document --disable-cet being the default and
4286         --enable-cet=auto.
4287
4288 2018-04-18  Martin Liska  <mliska@suse.cz>
4289
4290         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4291         style.
4292
4293 2018-04-18  Martin Liska  <mliska@suse.cz>
4294
4295         Revert
4296         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
4297
4298         PR ipa/83983
4299         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4300         arguments if they are comparable.
4301
4302 2018-04-18  Martin Liska  <mliska@suse.cz>
4303
4304         Revert
4305         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
4306
4307         PR lto/84805
4308         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4309         incomplete types.
4310
4311 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
4312
4313         PR target/85388
4314         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4315         ENDBR after calling __morestack.
4316
4317 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
4318
4319         PR jit/85384
4320         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4321         by using gcc_base_ver to generate a gcc_driver_version, and use
4322         it when generating GCC_DRIVER_NAME.
4323         * configure: Regenerate.
4324
4325 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
4326
4327         PR target/81084
4328         * config.gcc: Obsolete powerpc*-*-*spe*.
4329
4330 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4331
4332         PR debug/84637
4333         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4334         (stabstr_D): Change type of unum from unsigned int to
4335         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
4336         type.
4337
4338 2018-04-17  Jim Wilson  <jimw@sifive.com>
4339
4340         PR 84856
4341         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4342         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4343         Set arg_pointer_offset after using pretend_args_size.
4344
4345 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4346
4347         PR rtl-optimization/85431
4348         * dse.c (record_store): Ignore zero width stores.
4349
4350         PR sanitizer/85230
4351         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
4352         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4353         __builtin_stack_restore rather than after it.
4354         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4355         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4356         argument instead of virtual_dynamic_stack_rtx.
4357
4358 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4359
4360         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4361         New prototype.
4362         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4363         Add note to error message to explain internal mapping of overloaded
4364         built-in function name to non-overloaded built-in function name.
4365         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4366         function.
4367
4368 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
4369
4370         PR target/85424
4371         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4372         where the inputs overlap with the output.
4373
4374 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4375
4376         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4377         (=v, v) alternative and explicit "memory" attribute.
4378         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
4379         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4380         attributes.
4381         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4382         "sselog1" type instead of "sselog".
4383         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4384         "sselog".  Remove explicit "memory" attribute.
4385         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4386         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4387         attributes.
4388         (vec_extract_hi_v32hi): Merge all alternatives into one, use
4389         "sselog1" type instead of "sselog".  Remove explicit "memory"
4390         attribute.
4391         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4392         use "sselog1" type instead of "sselog".  Remove explicit "memory"
4393         attribute.
4394         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4395         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4396         attributes.
4397         (vec_extract_hi_v64qi): Merge all alternatives into one, use
4398         "sselog1" type instead of "sselog".  Remove explicit "memory"
4399         attribute.
4400         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4401         use "sselog1" type instead of "sselog".  Remove explicit "memory"
4402         attribute.
4403
4404         PR target/85430
4405         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4406
4407         PR middle-end/85414
4408         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4409         on a SUBREG.
4410
4411 2018-04-17  Martin Jambor  <mjambor@suse.cz>
4412
4413         PR ipa/85421
4414         * ipa-cp.c (create_specialized_node): Call
4415         expand_all_artificial_thunks if necessary.
4416
4417 2018-04-17  Martin Liska  <mliska@suse.cz>
4418
4419         PR lto/85405
4420         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4421         in message, remote space in between '_G' and '('.
4422
4423 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4424
4425         PR target/85281
4426         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4427         avx512f_vmcmp<mode>3<round_saeonly_name>,
4428         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4429         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4430         avx512f_rndscale<mode><round_saeonly_name>,
4431         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4432         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4433         Use %<iptr>2 instead of %2 for -masm=intel.
4434         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4435         avx512f_vcvttss2usi<round_saeonly_name>,
4436         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4437         -masm=intel.
4438         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4439         avx512f_vcvttsd2usi<round_saeonly_name>,
4440         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4441         Use %q1 instead of %1 for -masm=intel.
4442         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4443         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4444         of %3 for -masm=intel.
4445         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4446         -masm=intel.
4447         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4448         -masm=intel.
4449         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4450         -masm=intel.
4451         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4452         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4453         %g1.
4454         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4455         -masm=intel.
4456         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4457         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4458         %g1 and one with %0 and %1.
4459         (avx512er_vmrcp28<mode><round_saeonly_name>,
4460         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4461         %1 for -masm=intel.
4462         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4463         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4464         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4465         of %0 and %{%4%} for -masm=intel.
4466         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4467         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4468         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4469         order of %0 and %{%5%}%{z%} for -masm=intel.
4470
4471 2018-04-17  Jan Hubicka  <jh@suse.cz>
4472
4473         PR lto/85405
4474         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4475
4476 2018-04-17  Martin Liska  <mliska@suse.cz>
4477
4478         PR ipa/85329
4479         * multiple_target.c (create_dispatcher_calls): Set apostrophes
4480         for target_clone error message.  Make default implementation
4481         clone to be a local declaration.
4482         (separate_attrs): Add new argument and check for an empty
4483         string.
4484         (expand_target_clones): Handle it.
4485         (ipa_target_clone): Make redirection just for target_clones
4486         functions.
4487
4488 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
4489             Tom de Vries  <tom@codesourcery.com>
4490
4491         PR middle-end/84955
4492         * omp-expand.c (expand_oacc_for): Add dummy false branch for
4493         tiled basic blocks without omp continue statements.
4494
4495 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
4496
4497         PR target/83660
4498         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4499         vec_extract expression as having side effects to make sure it gets
4500         a cleanup point.
4501
4502 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
4503
4504         PR target/85403
4505         * config/i386/i386.c (get_builtin_code_for_version): Check
4506         error_mark_node.
4507
4508 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
4509
4510         PR target/84331
4511         * gcc/config.gcc: Support "skylake".
4512         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4513         PROCESSOR_SKYLAKE.
4514         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4515         (processor_target_table): Add "skylake".
4516         (ix86_option_override_internal): Add "skylake".
4517         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4518         PROCESSOR_CANNONLAKE.
4519         (get_builtin_code_for_version): Fix priority for
4520         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4521         PROCESSOR_SKYLAKE-AVX512.
4522         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4523         (processor_type): Add PROCESSOR_SKYLAKE.
4524
4525 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
4526             Jason Merrill  <jason@redhat.com>
4527
4528         PR c++/85112
4529         * convert.c (convert_to_integer_1): Use direct recursion for
4530         enumeral types and types with a precision less than the number
4531         of bits in their mode.
4532
4533 2018-04-16  Julia Koval  <julia.koval@intel.com>
4534
4535         PR target/84413
4536         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4537         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4538
4539 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
4540
4541         PR target/85293
4542         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4543         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4544         and -mno-direct-move.
4545
4546 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
4547
4548         PR target/83402
4549         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4550         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4551         Ensure negative shifts result in {0}.
4552
4553 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
4554
4555         PR rtl-optimization/79916
4556         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4557         regs (if any) to define how to gnerate SD moves when LRA is in
4558         progress.
4559
4560 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
4561
4562         PR rtl-optimization/85393
4563         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4564         * except.c (expand_dw2_landing_pad_for_region): Make static.
4565         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4566         a label and unconditional jump to old_bb, rather than
4567         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4568         basic block.
4569
4570         PR rtl-optimization/85376
4571         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4572         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4573         instead of a specific value.
4574
4575 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
4576             Bin Cheng  <bin.cheng@arm.com>
4577
4578         PR tree-optimization/82965
4579         PR tree-optimization/83991
4580         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4581         by_profile_only parameter.
4582         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4583         information if the loop was predicted to iterate too many times.
4584         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4585
4586 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
4587
4588         PR lto/71991
4589         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4590         always inline.
4591
4592 2018-04-13  Martin Liska  <mliska@suse.cz>
4593             Jakub Jelinek  <jakub@redhat.com>
4594
4595         PR middle-end/81657
4596         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4597         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4598         * builtins.c (expand_builtin_memory_copy_args): Use
4599         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4600         handle dest_addr == pc_rtx.
4601
4602 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
4603
4604         PR target/85291
4605         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4606         asked to not generate direct moves.
4607         (fix_trunc<mode>si2_stfiwx): Similar.
4608         (fix_trunc<mode>si2_internal): Similar.
4609
4610 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4611
4612         PR debug/83157
4613         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4614         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4615         lookup if dest in some wider mode is known to be const0_rtx and
4616         if so, record permanent equivalence for it to be ZERO_EXTEND of
4617         the narrower mode destination.
4618
4619 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
4620
4621         * lto-streamer-out.c (output_function): Revert 259346.
4622         * omp-expand.c (expand_oacc_for): Likewise.
4623
4624 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
4625
4626         PR rtl-optimization/85354
4627         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4628         * sel-sched.c (sel_global_init): ... here.
4629
4630 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
4631
4632         PR target/85238
4633         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4634         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4635         mode for PE-COFF targets.
4636         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4637         (i386_pe_asm_lto_end): Likewise.
4638         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4639         (TARGET_ASM_LTO_END): Likewise.
4640         * config/i386/winnt.c (saved_debug_info_level): New static variable.
4641         (i386_pe_asm_lto_start): New function.
4642         (i386_pe_asm_lto_end): Likewise.
4643
4644 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
4645             Richard Biener  <rguenther@suse.de>
4646
4647         PR middle-end/84955
4648         * lto-streamer-out.c (output_function): Fix CFG loop state before
4649         streaming out.
4650         * omp-expand.c (expand_oacc_for): Handle calls to internal
4651         functions like regular functions.
4652
4653 2018-04-12  Richard Biener  <rguenther@suse.de>
4654
4655         PR lto/85371
4656         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4657         for the early LTO debug to properly generate references to it
4658         during DIE emission.  Do not re-use that for the skeleton for
4659         split-dwarf.
4660         (dwarf2out_early_finish): Likewise.
4661
4662 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4663
4664         PR target/85328
4665         * config/i386/sse.md
4666         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4667         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4668         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4669         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4670         and output is a reg, avoid creating invalid lowpart subreg, but
4671         instead split into a 512-bit move.  Don't split if not AVX512VL,
4672         input is xmm16+ reg and output is a mem.
4673         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4674         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4675         xmm16+ reg and output is a mem.
4676
4677 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4678
4679         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4680         also for flag_dwarf2_cfi_asm.
4681
4682 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4683
4684         PR rtl-optimization/85342
4685         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4686         a bool scalar var inside of the loop instead.  Don't try to update
4687         recog_data.operand after failed apply_change_group.
4688
4689 2018-04-12  Tom de Vries  <tom@codesourcery.com>
4690
4691         PR target/85296
4692         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4693         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
4694         array with flexible array member as array without given dimension.
4695         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4696         argument for undefined param to true.
4697
4698 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
4699
4700         PR target/85321
4701         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4702         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4703         from PowerPC section.
4704         * config/rs6000/sysv4.opt (mcall-): Improve help text.
4705         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4706         help text that is too long.
4707         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4708         help text that is too long.
4709         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4710         help text that is too long.
4711
4712 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
4713
4714         * config/alpha/alpha.md (stack_probe_internal): Rename
4715         from "probe_stack".  Update all callers.
4716
4717 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4718
4719         PR rtl-optimization/84566
4720         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4721         sched_macro_fuse_insns.
4722
4723 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4724
4725         PR target/84301
4726         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4727         (compute_block_dependences): ... from here.
4728
4729 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4730
4731         PR tree-optimization/85331
4732         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4733         from int to HOST_WIDE_INT.
4734
4735 2018-04-11  Martin Jambor  <mjambor@suse.cz>
4736
4737         PR ipa/84149
4738         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4739         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4740         not the same as the source val.
4741         (cgraph_edge_brings_value_p): New parameter.
4742         (gather_edges_for_value): Pass destination value to
4743         cgraph_edge_brings_value_p.
4744         (perhaps_add_new_callers): Likewise.
4745         (get_info_about_necessary_edges): Likewise and exclude values brought
4746         only by self-recursive edges.
4747         (create_specialized_node): Redirect only clones of self-calling edges.
4748         (+self_recursive_pass_through_p): New function.
4749         (find_more_scalar_values_for_callers_subset): Use it.
4750         (find_aggregate_values_for_callers_subset): Likewise.
4751         (known_aggs_to_agg_replacement_list): Removed.
4752         (decide_whether_version_node): Re-calculate known constants for all
4753         remaining context clones.
4754
4755 2018-04-11  Richard Biener  <rguenther@suse.de>
4756
4757         PR lto/85339
4758         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4759         from early DWARF output.
4760         (dwarf2out_early_finish): Output line info unconditionally into
4761         early DWARF and add reference to it.
4762
4763 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4764
4765         PR target/85281
4766         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4767         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4768         other than V2DFmode using iptr mode attribute.
4769         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4770
4771 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4772
4773         PR rtl-optimization/84659
4774         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4775
4776 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4777
4778         PR debug/85302
4779         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4780         SIZEP is NULL.
4781         (output_loc_list): Pass address of a dummy size variable even in the
4782         locview handling loop.
4783         (index_location_lists): Add comment on why skip_loc_list_entry can't
4784         call size_of_locs.
4785
4786 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
4787
4788         PR target/85261
4789         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4790         into register.
4791
4792 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
4793
4794         PR target/85321
4795         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4796         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4797         and -mstring-compare-inline-limit.
4798
4799 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4800
4801         PR target/85287
4802         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4803         for stack clash protection in a register whenever we need it to be in
4804         a register.
4805
4806 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4807
4808         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4809         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4810
4811 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4812
4813         PR target/85321
4814         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4815         the help text.
4816         (mlong-double-): Ditto.
4817         * config/rs6000/sysv4.opt (msdata=): Ditto.
4818         (mtls-size=): Ditto.
4819
4820 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4821
4822         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4823         erroneous entries for
4824         "vector int vec_ldl (int, long int *)", and
4825         "vector unsigned int vec_ldl (int, unsigned long int *)".
4826         Add comments and entries for
4827         "vector bool char vec_ldl (int, bool char *)",
4828         "vector bool short vec_ldl (int, bool short *)",
4829         "vector bool int vec_ldl (int, bool int *)",
4830         "vector bool long long vec_ldl (int, bool long long *)",
4831         "vector pixel vec_ldl (int, pixel *)",
4832         "vector long long vec_ldl (int, long long *)",
4833         "vector unsigned long long vec_ldl (int, unsigned long long *)".
4834         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4835         type tree bool_long_long_type_node and correct definition of
4836         bool_V2DI_type_node to make reference to this new type tree.
4837         (rs6000_mangle_type): Replace erroneous reference to
4838         bool_long_type_node with bool_long_long_type_node.
4839         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4840         comments to emphasize sign distinctions for char and int types and
4841         replace RS6000_BTI_bool_long constant with
4842         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
4843         use of RS6000_BTI_pixel.
4844         (bool_long_type_node): Remove this macro definition.
4845         (bool_long_long_type_node): New macro definition
4846
4847 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4848
4849         PR rtl-optimization/85300
4850         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4851         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4852         simplify_unary_operation fails.
4853
4854 2018-04-10  Martin Liska  <mliska@suse.cz>
4855
4856         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4857         cgraph_edge and ipa_ref.
4858
4859 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4860
4861         PR target/85177
4862         PR target/85255
4863         * config/i386/sse.md
4864         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4865         computation of the VEC_MERGE selector from mask.
4866         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4867         Fix decoding of the VEC_MERGE selector into mask.
4868
4869 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4870
4871         PR tree-optimization/85286
4872         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4873
4874 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4875
4876         * final.c (final_1): Set insn_last_address as well as
4877         insn_current_address.
4878
4879 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4880
4881         PR target/85173
4882         * explow.c (emit_stack_probe): Call validize_mem on memory location
4883         before passing it to gen_probe_stack.  Create address operand and
4884         legitimize it for the probe_stack_address case.
4885
4886 2018-04-09  Jan Hubicka  <jh@suse.cz>
4887
4888         PR lto/85078
4889         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4890         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4891         * tree.c (free_lang_data_in_type): Fix handling of binfos;
4892         walk basetypes.
4893         (free_lang_data): Rebuild type inheritance graph.
4894
4895 2018-04-09  Martin Sebor  <msebor@redhat.com>
4896
4897         * invoke.texi (-finline-small-functions): Mention other optimization
4898         options.
4899         (-findirect-inlining, -fpartial-inlining): Same.
4900         (-finline-functions-called-once): Same.
4901         (-freorder-blocks-and-partition): Same.
4902
4903 2018-04-09  Jan Hubicka  <jh@suse.cz>
4904
4905         PR rtl/84058
4906         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4907         jumps; choose last target that matches the criteria (i.e.
4908         no partition changes for non-crossing jumps).
4909         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4910         support for redirecting crossing jumps to non-crossing.
4911
4912 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
4913
4914         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4915         also for naked functions.
4916
4917 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
4918
4919         * config/arc/arc.md (add_shift): New pattern.
4920         (add_shift2): Likewise.
4921         (sub_shift): Likewise.
4922         (sub_shift_cmp0_noout): Likewise.
4923         (compare_si_ashiftsi): Likewise.
4924         (xbfu_cmp0_noout): New combine pattern.
4925         (xbfu_cmp0"): Likewise.
4926         (movsi_set_cc_insn): Place the predicable variant first.
4927         (commutative_binary_cmp0_noout): Remove clobber.
4928         (commutative_binary_cmp0): New pattern.
4929         (noncommutative_binary_cmp0): Likewise.
4930         (noncommutative_binary_cmp0_noout): Likewise.
4931         (noncommutative_binary_comparison_result_used): Removed.
4932         (rsub_cmp0): New pattern.
4933         (rsub_cmp0_noout): Likewise.
4934         (extzvsi): Changed, keep only meaningful variants.
4935         (SQH, SEZ): New iterators.
4936         (SQH_postfix): New mode attribute.
4937         (SEZ_prefix): New code attribute.
4938         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4939         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4940         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4941         of numerical value.
4942         (noncommutative_operator): Check the availability of barrel
4943         shifter option.
4944
4945 2018-04-09  Richard Biener  <rguenther@suse.de>
4946
4947         PR tree-optimization/85284
4948         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4949         Only use the niter constraining form of simple_iv when the exit
4950         is always executed.
4951
4952 2018-04-09  Tom de Vries  <tom@codesourcery.com>
4953
4954         PR target/84041
4955         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4956         (define_expand "*memory_barrier"): New define_expand.
4957         (define_insn "memory_barrier"): New insn.
4958
4959 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4960
4961         PR rtl-optimization/80463
4962         PR rtl-optimization/83972
4963         PR rtl-optimization/83480
4964
4965         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4966         correct producer for the insn.
4967         (tidy_control_flow): Fixup seqnos in case of debug insns.
4968
4969 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4970
4971         PR rtl-optimization/83913
4972
4973         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4974         different sched-times when merging exprs.
4975
4976 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4977
4978         PR rtl-optimization/83962
4979
4980         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4981         tidy_fallthru_edge and tidy_control_flow.
4982
4983 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4984
4985         PR rtl-optimization/83530
4986
4987         * sel-sched.c (force_next_insn): New global variable.
4988         (remove_insn_for_debug): When force_next_insn is true, also leave only
4989         next insn in the ready list.
4990         (sel_sched_region): When the region wasn't scheduled, make another pass
4991         over it with force_next_insn set to 1.
4992
4993 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
4994
4995         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4996         into tm_file.
4997         * config/nds32/constants.md (unspec_volatile_element): Add enum values
4998         for interrupt control.
4999         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5000         functions for interrupt control.
5001         * config/nds32/nds32-intrinsic.md: Likewise.
5002         * config/nds32/nds32_intrinsic.h: Likewise.
5003         * config/nds32/nds32.h (nds32_builtins): Likewise.
5004
5005 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
5006
5007         * config/nds32/nds32.c (nds32_init_machine_status,
5008         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5009         strict_aligned_p field.
5010         (nds32_expand_to_rtl_hook): New function.
5011         (TARGET_EXPAND_TO_RTL_HOOK): Define.
5012         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5013
5014 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5015             Chung-Ju Wu  <jasonwucj@gmail.com>
5016
5017         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5018         * config/nds32/nds32-n7.md: New file.
5019         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5020         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5021         pipeline.
5022         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5023         * config/nds32/nds32.md (pipeline_model): Add n7.
5024         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5025         * config/nds32/pipelines.md: Include n7 settings.
5026
5027 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5028             Chung-Ju Wu  <jasonwucj@gmail.com>
5029
5030         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5031         * config/nds32/nds32-e8.md: New file.
5032         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5033         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5034         pipeline.
5035         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5036         * config/nds32/nds32.md (pipeline_model): Add e8.
5037         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5038         * config/nds32/pipelines.md: Include e8 settings.
5039
5040 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5041             Chung-Ju Wu  <jasonwucj@gmail.com>
5042
5043         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5044         * config/nds32/nds32-n8.md: New file.
5045         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5046         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5047         pipeline.
5048         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5049         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5050         * config/nds32/nds32.md (pipeline_model): Add n8.
5051         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5052         * config/nds32/pipelines.md: Include n8 settings.
5053
5054 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5055             Chung-Ju Wu  <jasonwucj@gmail.com>
5056
5057         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5058         * config/nds32/nds32-n9-2r1w.md: New file.
5059         * config/nds32/nds32-n9-3r2w.md: New file.
5060         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5061         nds32_register_ports): New or modify for cpu n9.
5062         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5063         pipeline.
5064         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5065         * config/nds32/nds32-utils.c: New file.
5066         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5067         TARGET_MUL_SLOW): Define.
5068         * config/nds32/nds32.md (pipeline_model): New attribute.
5069         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5070         New options that support cpu n9.
5071         * config/nds32/pipelines.md: Include n9 settings.
5072         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5073
5074 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
5075
5076         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5077         information if necessary.
5078         (output_cond_branch_compare_zero): Likewise.
5079         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5080         (nds32_target_alignment): Refine for alignment.
5081         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5082         (FUNCTION_BOUNDARY): Modify.
5083         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5084         align case.
5085         * config/nds32/nds32.opt (malways-align, malign-functions): New.
5086
5087 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
5088
5089         * config/nds32/constants.md (unspec_volatile_element): Add values for
5090         TLB operation and data prefetch.
5091         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5092         functions for TLB operation and data prefetch.
5093         * config/nds32/nds32-intrinsic.md: Likewise.
5094         * config/nds32/nds32_intrinsic.h: Likewise.
5095         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5096         (nds32_print_operand): Likewise.
5097         * config/nds32/nds32.h (nds32_builtins): Likewise.
5098
5099 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
5100         Andrew Pinski <pinsika@gcc.gnu.org>
5101
5102         PR middle-end/82976
5103         * match.pd: Use constant_boolean_node of correct type instead of
5104         boolean_true_node or boolean_false_node for simplifying
5105         pointer comparisons to zero.
5106
5107 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
5108
5109         PR tree-optimization/80021
5110         * tree.c (verify_type_variant): Make error call in verify_variant_match
5111         translatable and remove final full stop.
5112
5113 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5114
5115         * config/nds32/constants.md (unspec_volatile_element): Add
5116         UNSPEC_VOLATILE_EH_RETURN.
5117         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5118         nds32_output_stack_pop): Support dwarf exception handling process.
5119         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5120         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5121         exception handling process.
5122         (nds32_compute_stack_frame): Likewise.
5123         (nds32_return_addr_rtx): Likewise.
5124         (nds32_initial_elimination_offset): Likewise.
5125         (nds32_expand_prologue): Likewise.
5126         (nds32_expand_epilogue): Likewise.
5127         (nds32_dynamic_chain_address): New function.
5128         * config/nds32/nds32.h (machine_function): Add fields for dwarf
5129         exception handling.
5130         (DYNAMIC_CHAIN_ADDRESS): Define.
5131         (EH_RETURN_DATA_REGNO): Define.
5132         (EH_RETURN_STACKADJ_RTX): Define.
5133         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5134         patterns for dwarf exception handling.
5135
5136 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5137
5138         * config/nds32/nds32.h: Clean up obsolete macros.
5139
5140 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5141
5142         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5143         Add enum values for particular instructions.
5144         * config/nds32/nds32-intrinsic.c: Implementation of expanding
5145         particular intrinsic functions.
5146         * config/nds32/nds32-intrinsic.md: Likewise.
5147         * config/nds32/nds32_intrinsic.h: Likewise.
5148         * config/nds32/nds32.h (nds32_builtins): Likewise.
5149         * config/nds32/nds32.md (type): Add pbsad and pbsada.
5150         (btst, ave): New patterns for particular instructions.
5151
5152 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5153
5154         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5155         Add enum values for atomic load/store and memory sync.
5156         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5157         and memory sync.
5158         * config/nds32/nds32-intrinsic.md: Likewise.
5159         * config/nds32/nds32_intrinsic.h: Likewise.
5160         * config/nds32/nds32.h (nds32_builtins): Likewise.
5161
5162 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
5163
5164         PR tree-optimization/85257
5165         * fold-const.c (native_encode_vector): If not all elts could fit
5166         and off is -1, return 0 rather than offset.
5167         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5168         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5169         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
5170         adjust buffer in native_interpret_expr call.
5171
5172 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5173
5174         * config/nds32/constants.md (unspec_volatile_element): Add cache
5175         control enum values.
5176         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5177         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5178         * config/nds32/nds32.c (nds32_cctl_names): New.
5179         (nds32_print_operand): Handle cache control register names.
5180         * config/nds32/nds32.h (nds32_builtins): New enum values.
5181         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5182         macros.
5183         * config/nds32/nds32.md (type): Add mmu.
5184         * config/nds32/pipelines.md (simple_insn): Add mmu.
5185
5186 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5187
5188         * config/nds32/nds32.md (type): Remove call.
5189         * config/nds32/pipelines.md (simple_insn): Likewise.
5190
5191 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5192
5193         * config/nds32/constants.md (unspec_volatile_element): Add
5194         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5195         UNSPEC_VOLATILE_FMFCFG.
5196         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5197         description for fmfcfg and fmfcsr.
5198         (bdesc_1arg): Add fmtcsr.
5199         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5200         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5201         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5202         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5203         unspec_fmfcfg): New patterns.
5204         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5205         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5206         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5207         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5208         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5209         __nds32__fmfcfg): Define.
5210
5211 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5212
5213         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5214         intrinsic register names.
5215         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5216         intrinsic register enum values and macros.
5217
5218 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5219
5220         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5221         for load/store addressing form.
5222         (nds32_print_operand_address): Likewise.
5223
5224 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
5225
5226         PR target/85196
5227         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5228         based on LABEL_REF.  Remove useless assertion.
5229         (pic_address_needs_scratch): Fix formatting.
5230         (sparc_legitimize_pic_address): Minor tweaks.
5231         (sparc_delegitimize_address): Adjust assertion accordingly.
5232         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5233         into symbolic_operand.
5234         (movsi_high_pic_label_ref): Likewise.
5235         (movsi_lo_sum_pic_label_ref): Likewise.
5236         (movdi_pic_label_ref): Likewise.
5237         (movdi_high_pic_label_ref): Likewise.
5238         (movdi_lo_sum_pic_label_ref): Likewise.
5239
5240 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
5241
5242         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5243         custom LIB_SPEC setup.
5244
5245 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
5246             Kito Cheng  <kito.cheng@gmail.com>
5247
5248         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5249         * config/riscv/freebsd.h: New.
5250
5251 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5252
5253         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5254         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5255         file.
5256
5257 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5258             Kito Cheng  <kito.cheng@gmail.com>
5259
5260         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5261         nds32_output_call, nds32_symbol_binds_local_p): New functions.
5262         * config/nds32/nds32-protos.h (nds32_output_call,
5263         nds32_output_return): Declare.
5264         * config/nds32/nds32.md: Refine all the call and return patterns.
5265
5266 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
5267
5268         PR debug/85252
5269         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5270         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5271
5272         PR rtl-optimization/84872
5273         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5274         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5275         EDGE_CROSSING edge.
5276
5277 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
5278
5279         * expr.c (copy_blkmode_to_reg): Revert 254862.
5280         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5281
5282 2018-04-06  Richard Biener  <rguenther@suse.de>
5283
5284         PR middle-end/85244
5285         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5286         after seeing a component reference with an adjacent field.  Treat
5287         refs to arrays at struct end of external decls similar to
5288         refs to unconstrained commons.
5289
5290 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
5291
5292         PR sanitizer/85213
5293         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5294         look through SAVE_EXPRs with non-side-effects argument.  Adjust
5295         recursive calls.
5296         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5297         save_p here.
5298
5299 2018-04-06  Richard Biener  <rguenther@suse.de>
5300
5301         PR middle-end/85180
5302         * alias.c (find_base_term): New wrapper around find_base_term
5303         unwinding CSELIB_VAL_PTR changes.
5304         (find_base_term): Do not restore CSELIB_VAL_PTR during the
5305         recursion.
5306
5307 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5308
5309         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5310         instructions.
5311         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5312         constant definitions.
5313         ("nop"): lr 0,0 -> nopr r0
5314         ("nop_lr0", "nop_lr1"): New insn definitions.
5315
5316 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5317
5318         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5319         NDS32_V3PUSH_AVAILABLE_P macro.
5320
5321 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
5322             Chung-Ju Wu  <jasonwucj@gmail.com>
5323
5324         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5325         (nds32*-*-*): Add float and fpu_config into supported_defaults.
5326         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5327         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5328         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5329         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5330         * config/nds32/constraints.md: New constraints and checking for hard
5331         float configuration.
5332         * config/nds32/iterators.md: New mode iterator and attribute for hard
5333         float configuration.
5334         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5335         patterns.
5336         * config/nds32/nds32-fpu.md: New file.
5337         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5338         deal with hard float code generation.
5339         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5340         ARCH_V3S.
5341         (abi_type, float_reg_number): New enum type.
5342         * config/nds32/nds32-predicates.c: New predicates for hard float.
5343         * config/nds32/nds32-protos.h: Declare functions for hard float.
5344         * config/nds32/nds32.c: Implementation for hard float configuration.
5345         * config/nds32/nds32.h: Definitions for hard float configuration.
5346         * config/nds32/nds32.md: Include hard float machine description and
5347         modify patterns for hard float configuration.
5348         * config/nds32/nds32.opt: New options for hard float configuration.
5349         * config/nds32/predicates.md: New predicates for hard float
5350         configuration.
5351
5352 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
5353
5354         * common/config/nds32/nds32-common.c
5355         (nds32_option_optimization_table): Enable -mreleax-hint by default.
5356
5357 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
5358
5359         PR middle-end/85195
5360         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5361         CONSTRUCTOR_ELT (ctor, ...)->value.
5362
5363 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
5364
5365         PR target/85193
5366         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5367
5368 2018-04-05  Tom de Vries  <tom@codesourcery.com>
5369
5370         PR target/85204
5371         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5372         cond jump.
5373
5374 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
5375             Kito Cheng  <kito.cheng@gmail.com>
5376
5377         * config/nds32/constraints.md (U33): Fine-tune checking condition.
5378         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5379         * config/nds32/nds32.h (nds32_16bit_address_type): Add
5380         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5381
5382 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
5383             Kito Cheng  <kito.cheng@gmail.com>
5384
5385         * config/nds32/constraints.md (Ufe): New memory constraint.
5386         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5387         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5388         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5389         operands.
5390         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5391         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5392
5393 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5394
5395         * config/nds32/nds32.md: Use optimize_size in the condition for
5396         alu-shift instructions.
5397
5398 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5399
5400         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5401
5402 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5403
5404         * config/nds32/nds32.md (negsi2): Refine pattern.
5405
5406 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
5407             Chung-Ju Wu  <jasonwucj@gmail.com>
5408
5409         * config/nds32/iterators.md (shift_rotate): New code iterator.
5410         (shift): New code attribute.
5411         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5412         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5413         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5414         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5415         bit-wise operations.
5416         (andsi3, *andsi3): Ditto.
5417         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5418         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5419         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5420         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5421         nds32_ior_operand, nds32_xor_operand): New predicates.
5422
5423 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5424
5425         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5426         (addsi3, subsi3): ... this.
5427
5428 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5429
5430         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5431
5432 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5433
5434         * config/nds32/nds32.md: Adjust indention.
5435
5436 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
5437
5438         * config/nds32/nds32.md (feature): New attribute.
5439
5440 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5441
5442         * config/nds32/nds32.md (subtype): New attribute.
5443
5444 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
5445
5446         PR target/85203
5447         * config/arm/arm-builtins.c (arm_expand_builtin): Change
5448         expansion to perform a bitwise AND of the argument followed by a
5449         boolean negation of the result.
5450
5451 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
5452
5453         PR rtl-optimization/84878
5454         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5455         the basic block.  Assert the use reference is not artificial and that
5456         it has an associated insn.
5457
5458 2018-04-04  Michael Matz  <matz@suse.de>
5459
5460         * builtins.c (compute_objsize): Pass correct operand
5461         to array_at_struct_end_p.
5462
5463 2018-04-04  Richard Biener  <rguenther@suse.de>
5464
5465         PR lto/85176
5466         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5467         from contexts for DINFO_LEVEL_TERSE and below.
5468
5469 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5470
5471         * config/nds32/nds32-doubleword.md (move_<mode>): Require
5472         resiter_operand condition.
5473         * config/nds32/nds32.md (*move<mode>): Ditto.
5474
5475 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5476             Monk Chiang  <sh.chiang04@gmail.com>
5477
5478         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5479
5480 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5481
5482         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5483
5484 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5485             Kito Cheng  <kito.cheng@gmail.com>
5486
5487         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5488         nds32_cond_code_str, output_cond_branch,
5489         output_cond_branch_compare_zero, nds32_expand_cbranch,
5490         nds32_expand_cstore, nds32_expand_movcc,
5491         nds32_output_cbranchsi4_equality_zero,
5492         nds32_output_cbranchsi4_equality_reg,
5493         nds32_output_cbranchsi4_equality_reg_or_const_int,
5494         nds32_output_cbranchsi4_greater_less_zero: New functions.
5495         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5496         nds32_expand_cstore, nds32_expand_movcc,
5497         nds32_output_cbranchsi4_equality_zero,
5498         nds32_output_cbranchsi4_equality_reg,
5499         nds32_output_cbranchsi4_equality_reg_or_const_int,
5500         nds32_output_cbranchsi4_greater_less_zero): Declare.
5501         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5502         nds32_rimm11s_operand): New predicates.
5503         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5504         * config/nds32/nds32.md: Rewrite all the branch and conditional move
5505         patterns.
5506
5507 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5508
5509         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5510         * config/nds32/nds32.md: Ditto.
5511         * config/nds32/pipelines.md: Ditto.
5512
5513 2018-04-04  Richard Biener  <rguenther@suse.de>
5514
5515         PR tree-optimization/85168
5516         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5517         propagating abnormals.
5518
5519 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5520
5521         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5522
5523 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5524             Kito Cheng  <kito.cheng@gmail.com>
5525
5526         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5527         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5528         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5529         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5530         * config/nds32/nds32.md (sibcall_internal): New.
5531         (sibcall_register): Remove.
5532         (sibcall_immediate): Remove.
5533         (sibcall_value_internal): New.
5534         (sibcall_value_register): Remove.
5535         (sibcall_value_immediate): Remove.
5536         * config/nds32/predicates.md (nds32_general_register_operand): New.
5537         (nds32_call_address_operand): New.
5538
5539 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
5540
5541         PR rtl-optimization/85167
5542         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5543         bb_defs if *split_p, instead preinitialize it to NULL.
5544
5545         PR tree-optimization/85156
5546         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5547         evaluating the argument multiple times.
5548
5549 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
5550
5551         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5552         than vector.
5553         (_mm_cvtpd_ps): Likewise.
5554         (_mm_cvttpd_epi32): Likewise.
5555         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5556         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5557         vector, pixel, and bool following altivec.h include.
5558
5559 2018-04-03  Martin Sebor  <msebor@redhat.com>
5560
5561         * doc/extend.texi (Common Function Attributes): Clarify.
5562         (const attribute): Likewise.
5563         (pure attribute): Likewise.
5564
5565 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
5566
5567         PR target/85169
5568         * config/i386/i386.c (ix86_expand_vector_set): Use
5569         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
5570
5571 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
5572
5573         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5574         instructions when changing rounding bits to preserve precision bits
5575         in the x87 control word.
5576
5577 2018-04-03  Martin Liska  <mliska@suse.cz>
5578
5579         PR tree-optimization/82491
5580         * rtl.h (strip_offset_and_add): Replace += suboffset with
5581         poly_uint64 () + suboffset.
5582
5583 2018-03-29  Martin Liska  <mliska@suse.cz>
5584             Martin Jambor  <mjambor@suse.cz>
5585
5586         PR ipa/84947
5587         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5588         param_type is not an integral or pointer type.
5589
5590 2018-04-03  Richard Biener  <rguenther@suse.de>
5591
5592         * sese.h (recompute_all_dominators): Remove.
5593
5594 2018-04-02  Martin Sebor  <msebor@redhat.com>
5595
5596         * doc/invoke.texi (-Wrestrict): Fix typos.
5597
5598 2018-04-02  Jim Wilson  <jimw@sifive.com>
5599
5600         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5601         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5602         (<optab>di3, <optab>si3_extend): Likewise.
5603         (<optab>si3_mask, <optab>si3_mask_1): New.
5604         (<optab>di3_mask, <optab>di3_mask_1): New.
5605         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5606         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5607         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5608
5609 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
5610
5611         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5612         example.
5613
5614 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
5615
5616         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5617         (nds32_canonicalize_comparison): New function.
5618
5619 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5620             Kito Cheng  <kito.cheng@gmail.com>
5621             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
5622
5623         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5624         * config/nds32/constants.md (unspec_volatile_element): Add
5625         UNSPEC_VOLATILE_RELAX_GROUP.
5626         * config/nds32/nds32-relax-opt.c: New file.
5627         * config/nds32/nds32-predicates.c
5628         (nds32_symbol_load_store_p): New function.
5629         * config/nds32/nds32-protos.h
5630         (nds32_symbol_load_store_p): Declare function.
5631         (make_pass_nds32_relax_opt): Declare new rtl pass function.
5632         * config/nds32/nds32.c
5633         (nds32_register_pass): New function to register pass.
5634         (nds32_register_passes): New function to register passes.
5635         * config/nds32/nds32.md (relax_group): New pattern.
5636         * config/nds32/nds32.opt (mrelax-hint): New option.
5637         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5638
5639 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
5640
5641         * config/nds32/t-nds32: Modify files dependency.
5642
5643 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5644
5645         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5646         (PROFILE_HOOK): Define its implementation.
5647
5648 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5649
5650         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5651         type and 32-bit size.
5652
5653 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
5654
5655         PR middle-end/85090
5656         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5657         (V_128_256): New mode iterator.
5658         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5659         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5660         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5661         of V.
5662         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5663         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5664
5665 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
5666
5667         PR target/83315
5668         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5669         NaN inputs correctly.
5670
5671 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
5672
5673         PR target/80546
5674         * config/rs6000/vsx.md (??r): New mode attribute.
5675         (*vsx_mov<mode>_64bit): Use it.
5676         (*vsx_mov<mode>_32bit): Likewise.
5677
5678 2018-03-30  Martin Sebor  <msebor@redhat.com>
5679
5680         PR tree-optimization/84818
5681         * builtins.c (check_access): Use warning_n.
5682
5683 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
5684
5685         PR target/83822
5686         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5687         condition.
5688         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5689         condition.
5690
5691 2018-03-30  Julia Koval  <julia.koval@intel.com>
5692
5693         PR target/84413
5694         * x86-tune.def (movx, partial_reg_dependency): Enable for
5695         m_SKYLAKE_AVX512.
5696
5697 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
5698
5699         PR inline-asm/84985
5700         * lra-constraints.c (process_alt_operands): Move setting
5701         this_alternative_matches below.
5702
5703 2018-03-29  Martin Liska  <mliska@suse.cz>
5704
5705         PR lto/84995.
5706         * doc/invoke.texi: Document how LTO works with debug info.
5707         Describe auto-load support of binutils.  Mention 'x86-64'
5708         as valid option value of -march option.
5709
5710 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
5711
5712         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5713
5714         PR c/85094
5715         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5716         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5717         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5718         checking.
5719
5720 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
5721
5722         PR target/84912
5723         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5724         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5725         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5726         for RS6000_BTM_POWERPC64.
5727         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5728         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5729         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5730         definition.
5731         (DIVDE): Use it.
5732         (DIVDEU): Likewise.
5733
5734 2018-03-28 Carl Love  <cel@us.ibm.com>
5735
5736         Revert
5737         2017-09-27  Carl Love  <cel@us.ibm.com>
5738
5739         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5740         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5741         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5742         fctiw instruction.
5743
5744 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5745
5746         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5747         instead of __vector bool.
5748         (_mm_max_pu8): Likewise.
5749         (_mm_min_pi16): Likewise.
5750
5751 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
5752
5753         PR target/84912
5754         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5755         (DIVWEUO): Likewise.
5756         (DIVDEO): Likewise.
5757         (DIVDEUO): Likewise.
5758         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5759         DIVWEUO and DIVDEUO.
5760         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5761         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5762         (div_extend): Likewise.
5763         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5764         builtin function.
5765         (__builtin_divweuo): Likewise.
5766         (__builtin_divdeo): Likewise.
5767         (__builtin_divdeuo): Likewise.
5768
5769 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
5770
5771         PR target/85095
5772         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5773         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5774
5775         PR tree-optimization/82004
5776         * gimple-match-head.c (optimize_pow_to_exp): New function.
5777         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5778         Don't fold to exp if optimize_pow_to_exp is false.
5779
5780 2018-03-28  Martin Liska  <mliska@suse.cz>
5781
5782         PR other/84819
5783         * calls.c (initialize_argument_information): Fix trailing space.
5784         * common.opt: Fix typo and provide better explanation for
5785         -fsanitize-coverage option.
5786         * config/i386/i386.opt: Fix typo.
5787
5788 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
5789             Martin Liska  <mliska@suse.cz>
5790
5791         PR sanitizer/85081
5792         * gimplify.c (asan_poison_variable): Don't do the check for
5793         gimplify_omp_ctxp here.
5794         (gimplify_decl_expr): Do it here.
5795         (gimplify_target_expr): Likewise.
5796
5797 2018-03-28  Martin Liska  <mliska@suse.cz>
5798
5799         PR target/84988
5800         * config/i386/i386.c (ix86_function_arg_advance): Do not call
5801         chkp_type_bounds_count if MPX is not enabled.
5802
5803 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
5804
5805         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5806
5807 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5808
5809         PR target/84914
5810         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5811         function to create the function decl for complex long double
5812         multiply and divide for -mabi=ieeelongdouble.
5813         (init_float128_ieee): Call it.
5814
5815 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
5816
5817         PR target/85044
5818         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5819         -fcf-protection=branch -mibt.
5820         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5821
5822 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5823
5824         PR target/81863
5825         * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5826
5827 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
5828
5829         PR target/85056
5830         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5831         extern array declarations.
5832
5833 2018-03-27  Richard Biener  <rguenther@suse.de>
5834
5835         PR middle-end/84067
5836         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5837         explicit single_use checks.
5838
5839 2018-03-27  Richard Biener  <rguenther@suse.de>
5840
5841         PR tree-optimization/85082
5842         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5843         Valueize the VUSE.
5844
5845 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5846
5847         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5848         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5849         Turn on fasynchronous-unwind-tables and funwind-tables.
5850
5851 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
5852
5853         PR target/85073
5854         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5855         (*bmi_blsr_<mode>_ccz): Ditto.
5856
5857 2018-03-26  Tom de Vries  <tom@codesourcery.com>
5858
5859         PR tree-optimization/85063
5860         * omp-general.c (offloading_function_p): New function.  Factor out
5861         of ...
5862         * omp-offload.c (pass_omp_target_link::gate): ... here.
5863         * omp-general.h (offloading_function_p): Declare.
5864         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5865         with attribute omp declare target for offloading functions.
5866
5867 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
5868
5869         PR tree-optimization/84005
5870         * tree-data-ref.h (get_base_for_alignment): Declare.
5871         * tree-data-ref.c (get_base_for_alignment_1): New function.
5872         (get_base_for_alignment): Likewise.
5873         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5874         get_base_for_alignment to find a suitable base object, instead
5875         of always using drb->base_address.
5876
5877 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5878
5879         PR inline-asm/85022
5880         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5881         known size by default.
5882
5883 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
5884
5885         PR inline-asm/85030
5886         * lra-constraints.c (process_alt_operands): Don't match BLKmode
5887         and non BLKmode operands.
5888
5889 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5890
5891         PR target/85026
5892         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5893         Clean up attributes.
5894
5895 2018-03-23  Richard Biener  <rguenther@suse.de>
5896
5897         PR debug/85020
5898         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5899         we are going to emit early debug for LTO.
5900
5901 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5902
5903         PR inline-asm/85034
5904         * function.c (match_asm_constraints_1): Don't optimize if input
5905         doesn't satisfy general_operand predicate for output's mode.
5906
5907         PR inline-asm/85022
5908         * alias.c (write_dependence_p): Don't require for x_canonicalized
5909         non-VOIDmode if x has VOIDmode.
5910
5911         PR sanitizer/85029
5912         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5913         just don't try to optimize it rather than assert it never happens.
5914
5915 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5916
5917         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5918         macro expansions for definition of ST_INTERNAL_<mode> and
5919         LD_INTERNAL_<mode> builtins.
5920         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5921         Remove prototype.
5922         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5923         function.
5924         (altivec_expand_st_builtin): Likewise.
5925         (altivec_expand_builtin): Remove calls to deleted functions.
5926         (rs6000_address_for_altivec): Delete this function.
5927         * config/rs6000/vector.md: Remove expands for
5928         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5929
5930 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
5931
5932         PR target/84826
5933         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5934         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5935         re-computing once computed.
5936         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5937         (arm_init_machine_status): Initialize
5938         machine->static_chain_stack_bytes.
5939
5940 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5941
5942         PR target/84760
5943         * doc/extend.texi: Add four new prototypes for vec_ld.
5944         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5945         definitions for more logical presentation.
5946         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5947         entries for V1TI variants of __builtin_altivec_ld builtin.
5948         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5949         handling of V1TI variant of LVX icode pattern.
5950         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5951         (rs6000_gimple_fold_builtin): Likewise.
5952         (altivec_init_builtins): Add code to define
5953         __builtin_altivec_lvx_v1ti function.
5954
5955 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
5956
5957         PR inline-asm/84941
5958         * function.c (match_asm_constraints_1): Don't do the optimization
5959         if input isn't a REG, SUBREG, MEM or constant.
5960
5961 2018-03-22  Tom de Vries  <tom@codesourcery.com>
5962
5963         PR tree-optimization/84956
5964         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5965         bb_has_abnormal_pred.
5966
5967 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
5968
5969         PR sanitizer/85018
5970         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5971         DECL_INITIAL (decl) to decl at the end.
5972         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5973         adjust the comment.
5974
5975 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
5976
5977         * doc/extend.texi (__builtin_tgmath): Document when complex
5978         integer types are treated as _Complex _Float64.
5979
5980 2018-03-21  Tom de Vries  <tom@codesourcery.com>
5981
5982         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5983
5984 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5985
5986         PR tree-optimization/84960
5987         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5988         if it is ENTRY block, move them into single succ of ENTRY in that case.
5989
5990 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
5991
5992         PR tree-optimization/84811
5993         * poly-int.h (poly_span_traits): Remove the T3 parameter and
5994         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5995         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5996         (known_subrange_p): Update accordingly.  Cast each value involved
5997         in the size comparison, rather than casting the result of the
5998         subtraction.
5999
6000 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
6001
6002         PR tree-optimization/84982
6003         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6004         by flipping the least significant bit rather than all bits from
6005         bitpos to bitpos + bitsize - 1.
6006
6007 2018-03-21  Nathan Sidwell  <nathan@acm.org>
6008
6009         * doc/extend.texi (Deprecated Features): Remove mention of
6010         long-deleted deprecations.
6011
6012 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6013
6014         PR jit/84288
6015         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6016         * configure: Regenerate.
6017
6018 2018-03-21  Tom de Vries  <tom@codesourcery.com>
6019
6020         PR tree-optimization/83126
6021         * tree-parloops.c (num_phis): New function.
6022         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6023
6024 2018-03-21  Nathan Sidwell  <nathan@acm.org>
6025
6026         * doc/extend.texi (Deprecated Features): Update deprecated flags,
6027         mention anon-struct/union members and trailing attributes.
6028
6029 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
6030
6031         PR tree-optimization/84969
6032         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6033         builtin memset partitions if they set different rhs values.
6034
6035 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
6036
6037         PR rtl-optimization/84989
6038         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6039         VEC_DUPLICATE with scalar result mode.
6040
6041 2018-03-21  Martin Liska  <mliska@suse.cz>
6042
6043         PR ipa/84963
6044         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6045         not intended return statement.
6046
6047 2018-03-21  Martin Liska  <mliska@suse.cz>
6048
6049         PR target/84988
6050         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6051         (chkp_find_bound_slots_1): Limit number of iterations.
6052
6053 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
6054
6055         PR target/84838
6056         * Minor grammar fixes for x86 options.
6057
6058 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6059
6060         PR debug/84875
6061         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6062         holding REG_CFA_RESTORE notes, instead turn them into a USE.
6063
6064 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
6065
6066         PR target/83789
6067         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6068         (altivec_lvx_<mode>_1op): Likewise.
6069         (altivec_stvx_<mode>_2op): Likewise.
6070         (altivec_stvx_<mode>_1op): Likewise.
6071         (altivec_lvx_<VM2:mode>): New define_expand.
6072         (altivec_stvx_<VM2:mode>): Likewise.
6073         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6074         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6075         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6076         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6077         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6078         (rs6000_gen_lvx): Likewise.
6079         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6080         (altivec_expand_stv_builtin): Likewise.
6081         (altivec_expand_builtin): Likewise.
6082         * config/rs6000/vector.md: Likewise.
6083
6084 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6085
6086         PR target/82518
6087         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6088         BYTES_BIG_ENDIAN.
6089
6090 2018-03-20  Richard Biener  <rguenther@suse.de>
6091
6092         PR target/84986
6093         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6094         sign-conversions as zero, fall back to standard scalar_stmt
6095         cost for the rest.
6096
6097 2018-03-20  Martin Liska  <mliska@suse.cz>
6098
6099         PR ipa/84825
6100         * predict.c (rebuild_frequencies): Handle case when we have
6101         PROFILE_ABSENT, but flag_guess_branch_prob is false.
6102
6103 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6104
6105         PR target/84990
6106         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6107         flag_section_anchors.
6108         * varasm.c (use_blocks_for_decl_p): Remove hack for
6109         dw2_force_const_mem.
6110
6111         PR target/84845
6112         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6113         to ...
6114         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
6115         be created, use lowpart_subreg of operands[0] rather than operands[0]
6116         itself.
6117         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6118         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6119         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6120         and n constraint instead of aarch64_shift_imm_di and Usd.
6121         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6122         (*aarch64_<optab>_reg_minus<mode>3): ... this.
6123
6124 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
6125
6126         PR target/82989
6127         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6128         to favor GPR over NEON registers.
6129         (<shift>di3_neon): Likewise.
6130
6131 2018-03-20  Tom de Vries  <tom@codesourcery.com>
6132
6133         PR target/84952
6134         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6135         (nvptx_process_pars): Emit bar.sync asap and alap.
6136
6137 2018-03-20  Tom de Vries  <tom@codesourcery.com>
6138
6139         PR target/84954
6140         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6141         seen_label if seen_label is already set.
6142
6143 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6144
6145         PR target/84945
6146         * config/i386/i386.c (fold_builtin_cpu): For features above 31
6147         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6148         Use 1U instead of 1.  Formatting fixes.
6149
6150         PR c/84953
6151         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6152         instead of TREE_TYPE (s1) for the return value.
6153
6154 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
6155
6156         PR tree-optimization/84946
6157         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6158         bitsize + bitsize in poly_uint64 rather than poly_int64.
6159
6160         PR sanitizer/78651
6161         * dwarf2asm.c: Include fold-const.c.
6162         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6163         of decl rather than decl itself.
6164
6165         PR rtl-optimization/84643
6166         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6167
6168 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
6169
6170         PR sanitizer/78651
6171         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6172         calling assemble_variable.
6173
6174 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
6175
6176         PR target/81647
6177         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6178         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6179
6180 2018-03-19  Jim Wilson  <jimw@sifive.com>
6181
6182         PR bootstrap/84856
6183         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6184         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6185         (riscv_first_stack_step): Likewise.
6186         (riscv_option_override): Use STACK_BOUNDARY instead of
6187         MIN_STACK_BOUNDARY.
6188         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6189         MIN_STACK_BOUNDARY.
6190         (BIGGEST_ALIGNMENT): Set to 128.
6191         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6192         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6193         STACK_BOUNDARY.
6194
6195 2018-03-19  Richard Biener  <rguenther@suse.de>
6196
6197         PR tree-optimization/84933
6198         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6199         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6200
6201 2018-03-19  Richard Biener  <rguenther@suse.de>
6202
6203         PR tree-optimization/84859
6204         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6205         (cond_if_else_store_replacement): Perform sinking operation on
6206         single-store BBs regardless of MAX_STORES_TO_SINK setting.
6207         Generalize what a BB with a single eligible store is.
6208
6209 2018-03-19  Richard Biener  <rguenther@suse.de>
6210
6211         PR tree-optimization/84929
6212         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6213         chrec_is_positive against non-chrec arg.
6214
6215 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
6216
6217         PR target/84711
6218         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6219
6220 2018-03-18  Martin Liska  <mliska@suse.cz>
6221
6222         PR rtl-optimization/84635
6223         * regrename.c (build_def_use): Use matches_mode only when
6224         matches >= 0.
6225
6226 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
6227
6228         PR tree-optimization/84913
6229         * tree-vect-loop.c (vectorizable_reduction): Don't try to
6230         vectorize chains of COND_EXPRs.
6231
6232 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6233
6234         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6235
6236 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6237
6238         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6239
6240 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6241
6242         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6243
6244 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6245             Kito Cheng  <kito.cheng@gmail.com>
6246
6247         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6248         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6249         (nds32_adjust_reg_alloc_order): New function.
6250         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6251
6252 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
6253
6254         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6255         nds32_print_operand, nds32_print_operand_address): Use
6256         HOST_WIDE_INT_PRINT_DEC instead.
6257
6258 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6259
6260         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6261
6262 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
6263
6264         PR target/84902
6265         * config/i386/i386.c (initial_ix86_tune_features,
6266         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6267         unsigned long long.
6268         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6269         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6270         rather than 1u << ix86_tune.  Formatting fix.
6271         (ix86_option_override_internal): Change ix86_arch_mask from
6272         unsigned int to unsigned HOST_WIDE_INT, initialize to
6273         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6274         (ix86_function_specific_restore): Likewise.
6275
6276 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6277
6278         PR target/84899
6279         * postreload.c (reload_combine_recognize_pattern): Perform
6280         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6281         truncate_int_for_mode the result for the destination's mode.
6282
6283         PR c/84909
6284         * hsa-gen.c (mem_type_for_type): Fix comment typo.
6285         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6286         Likewise.
6287         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6288         Likewise.
6289
6290 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
6291
6292         PR target/84876
6293         * lra-assigns.c (lra_split_hard_reg_for): Don't use
6294         regno_allocno_class_array and sorted_pseudos.
6295         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6296         insns where regno is used.
6297
6298 2018-03-16  Martin Liska  <mliska@suse.cz>
6299
6300         PR ipa/84833
6301         * multiple_target.c (create_dispatcher_calls): Redirect
6302         reference in the symbol table.
6303
6304 2018-03-16  Martin Liska  <mliska@suse.cz>
6305
6306         PR ipa/84722
6307         * multiple_target.c (create_dispatcher_calls): Redirect also
6308         an alias.
6309
6310 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6311
6312         PR c++/79937
6313         PR c++/82410
6314         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6315         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6316         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6317
6318 2018-03-16  Julia Koval  <julia.koval@intel.com>
6319
6320         * doc/invoke.texi (Skylake Server): Add CLWB.
6321         Cannonlake): Remove CLWB.
6322
6323 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6324
6325         PR tree-optimization/84841
6326         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6327         1 << 3.
6328         (FLOAT_ONE_CONST_TYPE): Define.
6329         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6330         (sort_by_operand_rank): Put entries with higher constant_type last
6331         rather than first to match comments.
6332
6333 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
6334
6335         * config/nios2/nios2.md (movsi_internal): Fix thinko in
6336         split predicate.
6337
6338 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
6339
6340         PR c++/79085
6341         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6342         check and use address of target always.
6343
6344 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
6345
6346         PR target/84574
6347         * config/i386/i386.c (indirect_thunk_needed): Update comments.
6348         (indirect_thunk_bnd_needed): Likewise.
6349         (indirect_thunks_used): Likewise.
6350         (indirect_thunks_bnd_used): Likewise.
6351         (indirect_return_needed): New.
6352         (indirect_return_bnd_needed): Likewise.
6353         (output_indirect_thunk_function): Add a bool argument for
6354         function return.
6355         (output_indirect_thunk_function): Don't generate alias for
6356         function return thunk.
6357         (ix86_code_end): Call output_indirect_thunk_function to generate
6358         function return thunks.
6359         (ix86_output_function_return): Set indirect_return_bnd_needed
6360         and indirect_return_needed instead of indirect_thunk_bnd_needed
6361         and indirect_thunk_needed.
6362
6363 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
6364
6365         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6366         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6367         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6368
6369 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
6370             Paul Hua <paul.hua.gm@gmail.com>
6371
6372         PR c/84852
6373         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6374
6375 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
6376
6377         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6378         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6379         resp. SFmode cases.
6380
6381 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
6382
6383         PR target/84711
6384         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6385         instead of GET_MODE_SIZE when comparing Units.
6386
6387 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
6388
6389         PR target/68256
6390         * varasm.c (hash_section): Return an unchangeble hash value
6391         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6392         Return !aarch64_can_use_per_function_literal_pools_p ().
6393
6394 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
6395
6396         PR target/84860
6397         * optabs.c (emit_conditional_move): Pass address of cmode's copy
6398         rather than address of cmode as last argument to prepare_cmp_insn.
6399
6400 2018-03-15  Julia Koval  <julia.koval@intel.com>
6401
6402         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6403         F_AVX512VNNI, F_AVX512BITALG): New.
6404
6405 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
6406
6407         PR target/83451
6408         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
6409         insn for floating-point loads and stores.
6410
6411 2018-03-14  Carl Love  <cel@us.ibm.com>
6412
6413         * config/rs6000/rs6000-c.c: Add macro definitions for
6414         ALTIVEC_BUILTIN_VEC_PERMXOR.
6415         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6416         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6417         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6418         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6419         UNSPEC_VPERMXOR.
6420         * config/doc/extend.texi: Add prototypes for vec_permxor.
6421
6422 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
6423
6424         PR c/84852
6425         * diagnostic-show-locus.c (class layout_point): Convert m_line
6426         from int to linenum_type.
6427         (line_span::comparator): Use linenum "compare" function when
6428         comparing line numbers.
6429         (test_line_span): New function.
6430         (layout_range::contains_point): Convert param "row" from int to
6431         linenum_type.
6432         (layout_range::intersects_line_p): Likewise.
6433         (layout::will_show_line_p): Likewise.
6434         (layout::print_source_line): Likewise.
6435         (layout::should_print_annotation_line_p): Likewise.
6436         (layout::print_annotation_line): Likewise.
6437         (layout::print_leading_fixits): Likewise.
6438         (layout::annotation_line_showed_range_p): Likewise.
6439         (struct line_corrections): Likewise for field m_row.
6440         (line_corrections::line_corrections): Likewise for param "row".
6441         (layout::print_trailing_fixits): Likewise.
6442         (layout::get_state_at_point): Likewise.
6443         (layout::get_x_bound_for_row): Likewise.
6444         (layout::print_line): Likewise.
6445         (diagnostic_show_locus): Likewise for locals "last_line" and
6446         "row".
6447         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6448         * input.c (selftest::test_linenum_comparisons): New function.
6449         (selftest::input_c_tests): Call it.
6450         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6451         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6452         * selftest.h (ASSERT_GT): New macro.
6453         (ASSERT_GT_AT): New macro.
6454         (ASSERT_LT): New macro.
6455         (ASSERT_LT_AT): New macro.
6456
6457 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
6458
6459         PR rtl-optimization/84780
6460         * combine.c (distribute_links): Don't make a link based on pc_rtx.
6461
6462 2018-03-14  Martin Liska  <mliska@suse.cz>
6463
6464         * tree.c (record_node_allocation_statistics): Use
6465         get_stats_node_kind.
6466         (get_stats_node_kind): New function extracted from
6467         record_node_allocation_statistics.
6468         (free_node): Use get_stats_node_kind.
6469
6470 2018-03-14  Richard Biener  <rguenther@suse.de>
6471
6472         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6473         that the value-set of ANTIC_IN doesn't grow.
6474
6475         Revert
6476         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6477         member.
6478         (BB_VISITED_WITH_VISITED_SUCCS): New define.
6479         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6480
6481 2018-03-14  Julia Koval  <julia.koval@intel.com>
6482
6483         * config.gcc (icelake-client, icelake-server): New.
6484         (icelake): Remove.
6485         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6486         (initial_ix86_arch_features): Ditto.
6487         (PTA_SKYLAKE): Add SGX.
6488         (PTA_ICELAKE): Remove.
6489         (PTA_ICELAKE_CLIENT): New.
6490         (PTA_ICELAKE_SERVER): New.
6491         (ix86_option_override_internal): Split up icelake on icelake client and
6492         icelake server.
6493         (get_builtin_code_for_version): Ditto.
6494         (fold_builtin_cpu): Ditto.
6495         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6496         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6497         * config/i386/i386.h (processor_type): Ditto.
6498         * doc/invoke.texi: Ditto.
6499
6500 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
6501
6502         PR sanitizer/83392
6503         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6504         INTEGER_CST offset, add it together with bitpos / 8 and
6505         sign extend based on POINTER_SIZE.
6506
6507         PR target/84844
6508         Revert
6509         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
6510
6511         PR target/78090
6512         * config/i386/constraints.md (Yc): New register constraint.
6513         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6514         Use Yc constraint for alternative 2 of operand 0.  Remove
6515         preferred_for_speed attribute.
6516
6517 2018-03-14  Richard Biener  <rguenther@suse.de>
6518
6519         PR tree-optimization/84830
6520         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6521         with the old one to avoid oscillations.
6522
6523 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
6524
6525         PR target/83712
6526         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6527         pseudos.
6528         (assign_by_spills): Return a flag of reload assignment failure.
6529         Do not process the reload assignment failures.  Do not spill other
6530         reload pseudos if they has the same reg class.  Update n if
6531         necessary.
6532         (lra_assign): Add a return arg.  Set up from the result of
6533         assign_by_spills call.
6534         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6535         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6536         usage_insns if it is not NULL.
6537         (spill_hard_reg_in_range): New function.
6538         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6539         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6540         function prototypes.
6541         (lra_assign): Change prototype.
6542         * lra.c (lra): Add code to deal with fails by splitting hard reg
6543         live ranges.
6544
6545 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
6546
6547         * config/riscv/riscv.opt (mrelax): New option.
6548         * config/riscv/riscv.c (riscv_file_start): Emit ".option
6549         "norelax" when riscv_mrelax is disabled.
6550         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6551
6552 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
6553
6554         PR target/84743
6555         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6556         reassociation for int modes.
6557
6558 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6559
6560         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6561         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6562         for big-endian.
6563         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6564         * config/aarch64/aarch64-sve.md
6565         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6566         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6567         (*extend<mode><Vwide>2): Rename to...
6568         (aarch64_sve_extend<mode><Vwide>2): ...this.
6569         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6570         renaming the old pattern to...
6571         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
6572         unsigned packs.
6573         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6574         define_expand, renaming the old pattern to...
6575         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6576         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6577         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6578         account when deciding which SVE instruction the optab should use.
6579         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6580
6581 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6582
6583         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6584         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6585         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6586         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6587         (tlsdesc_small_<mode>): Turn a define_expand and use
6588         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
6589         (tlsdesc_small_advsimd_<mode>): ...this.
6590         (tlsdesc_small_sve_<mode>): New pattern.
6591
6592 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6593
6594         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6595         (UNSPEC_UMUL_HIGHPART): New constants.
6596         (MUL_HIGHPART): New int iteraor.
6597         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6598         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6599         define_expand.
6600         (*<su>mul<mode>3_highpart): New define_insn.
6601
6602 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
6603
6604         PR lto/84805
6605         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6606         incomplete types.
6607
6608 2018-03-13  Martin Liska  <mliska@suse.cz>
6609
6610         PR ipa/84658.
6611         * (sem_item_optimizer::sem_item_optimizer): Initialize new
6612         vector.
6613         (sem_item_optimizer::~sem_item_optimizer): Release it.
6614         (sem_item_optimizer::merge_classes): Register variable aliases.
6615         (sem_item_optimizer::fixup_pt_set): New function.
6616         (sem_item_optimizer::fixup_points_to_sets): Likewise.
6617         * ipa-icf.h: Declare new variables and functions.
6618
6619 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
6620
6621         PR middle-end/84834
6622         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6623         integer_pow2p@2 and test integer_pow2p in condition.
6624         (A < 0 ? C : 0): Similarly for @1.
6625
6626         PR middle-end/84831
6627         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6628         characters starting at p contain '\0' character, don't look beyond
6629         that.
6630
6631         PR target/84827
6632         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6633         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6634
6635         PR target/84828
6636         * reg-stack.c (change_stack): Change update_end var from int to
6637         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6638         also call set_block_for_insn on the newly added insns and rescan.
6639
6640         PR target/84786
6641         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6642         on the last operand.
6643
6644         PR c++/84704
6645         * tree.c (stabilize_reference_1): Return save_expr (e) for
6646         STATEMENT_LIST even if it doesn't have side-effects.
6647
6648 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
6649
6650         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6651
6652 2018-03-12  Renlin Li  <renlin.li@arm.com>
6653
6654         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6655         aarch64_output_scalar_simd_mov_immediate.
6656
6657 2018-03-12  Martin Sebor  <msebor@redhat.com>
6658
6659         PR tree-optimization/83456
6660         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6661         for perfectly overlapping calls to memcpy.
6662         (gimple_fold_builtin_memory_chk): Same.
6663         (gimple_fold_builtin_strcpy): Handle no-warning.
6664         (gimple_fold_builtin_stxcpy_chk): Same.
6665         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6666
6667 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
6668
6669         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6670         parameter.  Use it for SFmode.
6671         (rs6000_function_arg_advance_1): Adjust.
6672         (rs6000_function_arg): Adjust.
6673         (rs6000_gimplify_va_arg): Pass false for that new parameter.
6674
6675 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
6676
6677         PR rtl-optimization/84169
6678         PR rtl-optimization/84780
6679         * combine.c (can_combine_p): Check for a 2-insn combination whether
6680         the destination register is used between the two insns, too.
6681
6682 2018-03-12  Richard Biener  <rguenther@suse.de>
6683
6684         PR tree-optimization/84803
6685         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6686         for refs DR analysis didn't process.
6687
6688 2018-03-12  Richard Biener  <rguenther@suse.de>
6689
6690         PR tree-optimization/84777
6691         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6692         force-vectorize loops ignore whether we are optimizing for size.
6693
6694 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6695
6696         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6697         (TARGET_MD_ASM_ADJUST): Define.
6698
6699 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
6700             Kito Cheng  <kito.cheng@gmail.com>
6701             Chung-Ju Wu  <jasonwucj@gmail.com>
6702
6703         * config/nds32/nds32.c (nds32_compute_stack_frame,
6704         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6705         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6706         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6707         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6708         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6709         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6710         * config/nds32/nds32.md (prologue, epilogue): Use macro
6711         NDS32_V3PUSH_AVAILABLE_P to do checking.
6712
6713 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
6714
6715         PR debug/58150
6716         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6717         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6718         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
6719         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
6720         addition of most attributes on !orig_type_die or the attribute not
6721         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6722
6723 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6724             Chung-Ju Wu  <jasonwucj@gmail.com>
6725
6726         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6727         __NDS32_VH__ macro.
6728         * config/nds32/nds32.opt (mvh): New option.
6729
6730 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6731             Chung-Ju Wu  <jasonwucj@gmail.com>
6732
6733         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6734         function.
6735         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6736         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6737         definition.
6738
6739 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6740             Chung-Ju Wu  <jasonwucj@gmail.com>
6741
6742         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6743         function.
6744         * config/nds32/nds32-multiple.md (strlensi): New pattern.
6745         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6746
6747 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
6748             Kito Cheng  <kito.cheng@gmail.com>
6749             Chung-Ju Wu  <jasonwucj@gmail.com>
6750
6751         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6752         UNSPEC_FFMISM and UNSPEC_FLMISM.
6753         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6754         for ffb, ffmism and flmism.
6755         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6756         (unspec_ffmism): Ditto.
6757         (unspec_flmism): Ditto.
6758         (nds32_expand_builtin_impl): Check if string extension is available.
6759         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6760         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6761
6762 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
6763
6764         Reverting patch:
6765         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6766
6767         PR target/83712
6768         * lra-assigns.c (assign_by_spills): Return a flag of reload
6769         assignment failure.  Do not process the reload assignment
6770         failures.  Do not spill other reload pseudos if they has the same
6771         reg class.
6772         (lra_assign): Add a return arg.  Set up from the result of
6773         assign_by_spills call.
6774         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6775         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6776         usage_insns if it is not NULL.
6777         (spill_hard_reg_in_range): New function.
6778         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6779         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6780         function prototypes.
6781         (lra_assign): Change prototype.
6782         * lra.c (lra): Add code to deal with fails by splitting hard reg
6783         live ranges.
6784
6785 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
6786
6787         PR target/84807
6788         * config/i386/i386.opt: Replace Enforcment with Enforcement.
6789
6790 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
6791
6792         PR debug/84620
6793         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6794         (dw_val_node): Add val_symbolic_view.
6795         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6796         (symview_upper_bound): New.
6797         (new_line_info_table): Initialize symviews_since_reset.
6798         (dwarf2out_source_line): Count symviews_since_reset and set
6799         symview_upper_bound.
6800         (dw_val_equal_p): Handle symview.
6801         (add_AT_symview): New.
6802         (print_dw_val): Handle symview.
6803         (attr_checksum, attr_checksum_ordered): Likewise.
6804         (same_dw_val_p, size_of_die): Likewise.
6805         (value_format, output_die): Likewise.
6806         (add_high_low_attributes): Use add_AT_symview for entry_view.
6807         (dwarf2out_finish): Reset symview_upper_bound, clear
6808         zero_view_p.
6809
6810 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
6811
6812         PR target/83969
6813         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6814         Add strict argument and use it.
6815         (rs6000_split_multireg_move): Update for new strict argument.
6816         (mem_operand_gpr): Disallow all non-offsettable addresses.
6817         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6818
6819 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
6820
6821         PR target/84772
6822         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6823         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6824         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6825
6826         PR c++/84767
6827         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6828         decl, use remap_type if we want to use the type.
6829
6830 2018-03-09  Martin Sebor  <msebor@redhat.com>
6831
6832         PR tree-optimization/84526
6833         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6834         Remove dead code.
6835         (builtin_access::generic_overlap): Be prepared to handle non-array
6836         base objects.
6837
6838 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
6839
6840         PR rtl-optimization/84682
6841         * lra-constraints.c (process_address_1): Check is_address flag
6842         for address constraints.
6843         (process_alt_operands): Likewise.
6844         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6845         preprocess_constraints.
6846         * recog.h (preprocess_constraints): Add oploc parameter.
6847         Adjust callers.
6848         * recog.c (preprocess_constraints): Test address_operand for
6849         CT_ADDRESS constraints.
6850
6851 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6852
6853         PR target/83712
6854         * lra-assigns.c (assign_by_spills): Return a flag of reload
6855         assignment failure.  Do not process the reload assignment
6856         failures.  Do not spill other reload pseudos if they has the same
6857         reg class.
6858         (lra_assign): Add a return arg.  Set up from the result of
6859         assign_by_spills call.
6860         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6861         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6862         usage_insns if it is not NULL.
6863         (spill_hard_reg_in_range): New function.
6864         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6865         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6866         function prototypes.
6867         (lra_assign): Change prototype.
6868         * lra.c (lra): Add code to deal with fails by splitting hard reg
6869         live ranges.
6870
6871 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6872
6873         PR target/83193
6874         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6875         Accept complain bool parameter.  Only emit errors if it is true.
6876         (arm_parse_cpu_option_name): Likewise.
6877         (arm_target_thumb_only): Adjust callers of the above.
6878         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6879         prototype to take a default true bool parameter.
6880         (arm_parse_arch_option_name): Likewise.
6881
6882 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
6883             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
6884
6885         PR jit/64089
6886         PR jit/84288
6887         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6888         * configure: Regenerate.
6889         * configure.ac ("linker --version-script option"): New.
6890         ("linker soname option"): New.
6891
6892 2018-03-09  Richard Biener  <rguenther@suse.de>
6893
6894         PR tree-optimization/84775
6895         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6896         immediate uses of predicate stmts and mark them modified.
6897
6898         Revert
6899         PR tree-optimization/84178
6900         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6901         to caller.
6902         (version_loop_for_if_conversion): Delay update_ssa call.
6903         (tree_if_conversion): Delay update_ssa until after predicate
6904         insertion.
6905
6906 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
6907
6908         PR target/84763
6909         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6910         when the function accesses prior frames.
6911
6912 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6913
6914         PR debug/84456
6915         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6916         gen_llsym, otherwise call maybe_gen_llsym.
6917
6918         PR inline-asm/84742
6919         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6920         has ',' character inside of it.
6921
6922 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6923
6924         PR target/84748
6925         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6926         as clobbering CC_REGNUM.
6927
6928 2018-03-08  Richard Biener  <rguenther@suse.de>
6929
6930         PR middle-end/84552
6931         * tree-scalar-evolution.c: Include tree-into-ssa.h.
6932         (follow_copies_to_constant): Do not follow SSA names registered
6933         for update.
6934
6935 2018-03-08  Richard Biener  <rguenther@suse.de>
6936
6937         PR tree-optimization/84178
6938         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6939         to caller.
6940         (version_loop_for_if_conversion): Delay update_ssa call.
6941         (tree_if_conversion): Delay update_ssa until after predicate
6942         insertion.
6943
6944 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
6945
6946         PR tree-optimization/84178
6947         * tree-if-conv.c (release_bb_predicate): Remove the
6948         the assertion that the stmts have NULL use_ops.
6949         Discard the statements, asserting that they haven't
6950         yet been added to a BB.
6951
6952 2018-03-08  Richard Biener  <rguenther@suse.de>
6953
6954         PR tree-optimization/84746
6955         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6956         (phi_translate): Pass in destination ANTIC_OUT set.
6957         (phi_translate_1): Likewise.  For a simplified result lookup
6958         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6959         (phi_translate_set): Adjust.
6960         (do_pre_regular_insertion): Likewise.
6961         (do_pre_partial_partial_insertion): Likewise.
6962
6963 2018-03-08  Martin Liska  <mliska@suse.cz>
6964
6965         PR gcov-profile/84735
6966         * doc/gcov.texi: Document usage of profile files.
6967         * gcov-io.h: Document changes in the format.
6968
6969 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
6970
6971         PR debug/84404
6972         PR debug/84408
6973         * dwarf2out.c (struct dw_line_info_table): Update comments for
6974         view == -1.
6975         (FORCE_RESET_NEXT_VIEW): New.
6976         (FORCE_RESETTING_VIEW_P): New.
6977         (RESETTING_VIEW_P): Check for -1 too.
6978         (ZERO_VIEW_P): Likewise.
6979         (new_line_info_table): Force-reset next view.
6980         (dwarf2out_begin_function): Likewise.
6981         (dwarf2out_source_line): Simplify zero_view_p initialization.
6982         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6983         view directly.  Omit view when omitting .loc at line 0.
6984
6985 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6986
6987         PR tree-optimization/84740
6988         * tree-switch-conversion.c (process_switch): Call build_constructors
6989         only if info.phi_count is non-zero.
6990
6991         PR tree-optimization/84739
6992         * tree-tailcall.c (find_tail_calls): Check call arguments against
6993         DECL_ARGUMENTS (current_function_decl) rather than
6994         DECL_ARGUMENTS (func) when checking for tail recursion.
6995
6996 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
6997
6998         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6999         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7000         Volker Reichelt's entry and add entries for people that perform
7001         GCC fuzzy testing and report numerous bugs.
7002
7003 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
7004
7005         PR target/82411
7006         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7007         readonly data in sdata, if that is disabled.
7008         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7009         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7010         -mreadonly-in-sdata option.
7011
7012 2018-03-07  Martin Sebor  <msebor@redhat.com>
7013
7014         PR tree-optimization/84468
7015         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7016         basic block when looking for nul assignment.
7017
7018 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
7019
7020         PR target/84277
7021         * except.h (output_function_exception_table): Adjust prototype.
7022         * except.c (output_function_exception_table): Remove FNNAME parameter
7023         and add SECTION parameter.  Ouput one part of the table at a time.
7024         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7025         the first part of the exception table and emit unwind directives.
7026         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7027         (i386_pe_seh_cold_init): Likewise.
7028         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7029         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7030         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7031         (ix86_output_call_insn): Emit a nop in one more case for SEH.
7032         * config/i386/winnt.c: Include except.h.
7033         (struct seh_frame_state): Add reg_offset, after_prologue and
7034         in_cold_section fields.
7035         (i386_pe_seh_end_prologue): Set seh->after_prologue.
7036         (i386_pe_seh_cold_init): New function.
7037         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7038         to seh->in_cold_section.
7039         (seh_emit_push): Record the offset of the push.
7040         (seh_emit_save): Record the offet of the save.
7041         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7042         Test seh->after_prologue to disregard the epilogue.
7043         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7044         (i386_pe_end_cold_function): New function.
7045
7046 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
7047
7048         PR fortran/84565
7049         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7050         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7051
7052         PR c++/84704
7053         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7054         on tmp_var.
7055         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7056         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7057
7058         PR middle-end/84723
7059         * multiple_target.c: Include tree-inline.h and intl.h.
7060         (expand_target_clones): Diagnose and fail if node->definition and
7061         !tree_versionable_function_p (node->decl).
7062
7063 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
7064
7065         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7066         sprint_ul.
7067         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7068         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7069         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7070
7071 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
7072
7073         PR target/84710
7074         * combine.c (try_combine): Use reg_or_subregno instead of handling
7075         just paradoxical SUBREGs and REGs.
7076
7077 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
7078
7079          * config/arc/arc.c (arc_finalize_pic): Remove function.
7080          (arc_must_save_register): We use single base PIC register, remove
7081          checks to save/restore the PIC register.
7082          (arc_expand_prologue): Likewise.
7083          * config/arc/arc-protos.h (arc_set_default_type_attributes):
7084          Remove.
7085          (arc_verify_short): Likewise.
7086          (arc_attr_type): Likewise.
7087          * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7088          (walk_stores): Likewise.
7089          (arc_address_cost): Make it static.
7090          (arc_verify_short): Likewise.
7091          (branch_dest): Likewise.
7092          (arc_attr_type): Likewise.
7093          * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7094          (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7095          (arc_final_prescan_insn): Remove inserting the nops due to
7096          hardware hazards.  It is done in reorg step.
7097          (insn_length_variant_t): Remove.
7098          (insn_length_parameters_t): Likewise.
7099          (arc_insn_length_parameters): Likewise.
7100          (arc_get_insn_variants): Likewise.
7101          * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7102
7103 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
7104
7105         PR inline-asm/84683
7106         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7107         assertion failure.
7108
7109         PR tree-optimization/84687
7110         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7111         on new_node->decl.
7112         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7113
7114 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7115
7116         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7117         Rename to ppc_speculation_barrier.
7118         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7119         __builtin_ppc_speculation_barrier.
7120
7121 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
7122
7123         PR target/84700
7124         * combine.c (combine_simplify_rtx): Don't try to simplify if
7125         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7126         are equal to x.
7127
7128 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
7129
7130         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7131         to 32 bytes when compiling for POWER9.
7132
7133 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
7134
7135         PR target/84564
7136         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7137         regparm >= 3 with no arg reg available also for calls with
7138         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
7139
7140         PR target/84524
7141         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7142         orig,vex.
7143         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
7144
7145 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
7146
7147         PR target/84264
7148         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7149
7150 2018-03-05  Richard Biener  <rguenther@suse.de>
7151
7152         PR tree-optimization/84486
7153         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7154         When inserting a __builtin_assume_aligned call set the LHS
7155         SSA name alignment info accordingly.
7156
7157 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
7158
7159         PR tree-optimization/84114
7160         * config/aarch64/aarch64.c (aarch64_reassociation_width)
7161         Avoid reassociation of FLOAT_MODE addition.
7162
7163 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
7164
7165         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7166         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7167         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7168         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7169         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7170         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7171         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7172         and -mwbnoinvd.
7173         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7174         __builtin_ia32_wbinvd): New builtins.
7175         (SPECIAL_ARGS2): New.
7176         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7177         (SPECIAL_ARGS2): New.
7178         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7179         (ix86_valid_target_attribute_inner_p): Ditto.
7180         (ix86_init_mmx_sse_builtins): Add special_args2.
7181         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7182         TARGET_WBNOINVD_P): New.
7183         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7184         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7185         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7186         * config/i386/immintrin.h (_wbinvd): New intrinsic.
7187         * config/i386/pconfigintrin.h: New file.
7188         * config/i386/wbnoinvdintrin.h: Ditto.
7189         * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7190         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7191
7192 2018-03-05  Richard Biener  <rguenther@suse.de>
7193
7194         PR tree-optimization/84670
7195         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7196         member.
7197         (BB_VISITED_WITH_VISITED_SUCCS): New define.
7198         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7199         (compute_antic_aux): Only assert the number of values in ANTIC_IN
7200         doesn't grow if all successors (recursively) were visited at least
7201         once.
7202
7203 2018-03-05  Richard Biener  <rguenther@suse.de>
7204
7205         PR tree-optimization/84650
7206         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7207         if executed in the loop pipeline.
7208
7209 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
7210
7211         * doc/configfiles.texi (Configuration Files): Move info about
7212         conditionalizing $target-protos.h to...
7213         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
7214         differs from $target-protos.h.
7215
7216 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
7217             Chung-Ju Wu  <jasonwucj@gmail.com>
7218
7219         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7220         * config/nds32/nds32-multiple.md (setmemsi): Define.
7221         * config/nds32/nds32-memory-manipulation.c
7222         (nds32_gen_dup_4_byte_to_word_value): New.
7223         (emit_setmem_word_loop): New.
7224         (emit_setmem_byte_loop): New.
7225         (nds32_expand_setmem_loop): New.
7226         (nds32_expand_setmem_loop_v3m): New.
7227         (nds32_expand_setmem_unroll): New.
7228         (nds32_expand_setmem): New.
7229
7230 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7231             Chung-Ju Wu  <jasonwucj@gmail.com>
7232
7233         * config/nds32/nds32-memory-manipulation.c
7234         (nds32_emit_load_store): New.
7235         (nds32_emit_post_inc_load_store): New.
7236         (nds32_emit_mem_move): New.
7237         (nds32_emit_mem_move_block): New.
7238         (nds32_expand_movmemsi_loop_unknown_size): New.
7239         (nds32_expand_movmemsi_loop_known_size): New.
7240         (nds32_expand_movmemsi_loop): New.
7241         (nds32_expand_movmemsi_unroll): New.
7242         (nds32_expand_movmemqi): Rename ...
7243         (nds32_expand_movmemsi): ... to this.
7244         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7245         (movmemsi): ... to this.
7246         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7247         (nds32_expand_movmemsi): ... to this.
7248
7249 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7250             Monk Chiang  <sh.chiang04@gmail.com>
7251             Chung-Ju Wu  <jasonwucj@gmail.com>
7252
7253         * config/nds32/nds32-protos.h
7254         (nds32_expand_load_multiple): New arguments.
7255         (nds32_expand_store_multiple): Ditto.
7256         (nds32_valid_multiple_load_store): Rename ...
7257         (nds32_valid_multiple_load_store_p): ... to this.
7258         * config/nds32/nds32-memory-manipulation.c
7259         (nds32_expand_load_multiple): Refine implementation.
7260         (nds32_expand_store_multiple): Ditto.
7261         * config/nds32/nds32-multiple.md
7262         (load_multiple): Update nds32_expand_load_multiple interface.
7263         (store_multiple): Update nds32_expand_store_multiple interface.
7264         * config/nds32/nds32-predicates.c
7265         (nds32_valid_multiple_load_store): Rename ...
7266         (nds32_valid_multiple_load_store_p): ... to this and refine
7267         implementation.
7268         * config/nds32/predicates.md
7269         (nds32_load_multiple_and_update_address_operation): New predicate.
7270         (nds32_store_multiple_and_update_address_operation): New predicate.
7271
7272 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7273             Chung-Ju Wu  <jasonwucj@gmail.com>
7274
7275         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7276         (combo): New attribute.
7277         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7278
7279 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
7280
7281         * config/nds32/nds32.opt: Change -mcmodel= default value.
7282
7283 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
7284             Monk Chiang  <sh.chiang04@gmail.com>
7285             Chung-Ju Wu  <jasonwucj@gmail.com>
7286
7287         * config/nds32/constants.md (unspec_element): New enum.
7288         * config/nds32/constraints.md (Umw): New constraint.
7289         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7290         * config/nds32/nds32-intrinsic.md: Likewise.
7291         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7292         (nds32_valid_smw_lwm_base_p): New.
7293         (nds32_output_smw_single_word): New.
7294         (nds32_output_lmw_single_word): New.
7295         (nds32_expand_unaligned_load): New.
7296         (nds32_expand_unaligned_store): New.
7297         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7298         (nds32_output_smw_single_word): Declare.
7299         (nds32_output_lmw_single_word): Declare.
7300         (nds32_expand_unaligned_load): Declare.
7301         (nds32_expand_unaligned_store): Declare.
7302         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7303         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7304         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7305         NDS32_BUILTIN_UASTORE_DW.
7306         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7307         predicate.
7308
7309 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
7310             Kito Cheng  <kito.cheng@gmail.com>
7311             Chung-Ju Wu  <jasonwucj@gmail.com>
7312
7313         * config/nds32/nds32-intrinsic.c
7314         (nds32_expand_builtin_null_ftype_reg): Delete.
7315         (nds32_expand_builtin_reg_ftype_imm): Ditto.
7316         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7317         (nds32_read_argument): New.
7318         (nds32_legitimize_target): Ditto.
7319         (nds32_legitimize_argument): Ditto.
7320         (nds32_check_constant_argument): Ditto.
7321         (nds32_expand_unop_builtin): Ditto.
7322         (nds32_expand_unopimm_builtin): Ditto.
7323         (nds32_expand_binop_builtin): Ditto.
7324         (nds32_builtin_decl_impl): Ditto.
7325         (builtin_description): Ditto.
7326         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7327         (nds32_init_builtins_impl): Ditto.
7328         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7329         (nds32_builtin_decl): New.
7330         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7331         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7332
7333 2018-03-02  Jeff Law  <law@redhat.com>
7334
7335         * reorg.c (stop_search_p): Handle DEBUG_INSN.
7336         (redundant_insn, fill_simple_delay_slots): Likewise.
7337         (fill_slots_from_thread): Likewise.
7338         * resource.c (mark_referenced_resources): Likewise.
7339         (mark_set_resources, find_dead_or_set_registers): Likewise.
7340
7341 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7342
7343         * substring-locations.h (format_warning_va): Formatting fix for
7344         ATTRIBUTE_GCC_DIAG.
7345         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7346         argument.
7347         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7348         * substring-locations.c: Include intl.h.
7349         (format_warning_va): Turned into small wrapper around
7350         format_warning_n_va, renamed to ...
7351         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7352         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7353         use ngettext.
7354         (format_warning_at_substring_n): New function.
7355         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7356         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
7357         format_warning_at_substring with just a shorter name instead of
7358         const function pointer.
7359         (fmtwarn_n): New function.
7360         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7361         appropriate, get rid of all the fmtstr temporaries, move conditionals
7362         with G_() wrapped string literals directly into fmtwarn arguments,
7363         cast dir.len to (int), formatting fixes.
7364
7365 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
7366
7367         * doc/invoke.texi: Remove "Cilk Plus" references.
7368
7369 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7370             Richard Biener  <rguenther@suse.de>
7371
7372         PR ipa/84628
7373         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7374         for error or warning attributes if CALL_FROM_THUNK_P is set.
7375         Formatting fixes.
7376
7377 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7378
7379         PR target/56540
7380         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7381         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7382
7383         PR target/56540
7384         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7385         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7386
7387         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7388         instead of -1U in last predictors element's probability member.
7389
7390 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
7391
7392         PR ipa/83983
7393         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7394         arguments if they are comparable.
7395
7396 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
7397
7398         PR tree-optimization/84634
7399         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7400         masks and masked_loop_p with a single loop_masks, making sure it's
7401         null for bb vectorization.
7402
7403 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
7404
7405         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7406         (vect_analyze_data_ref_access): Use loop->safe_len rather than
7407         loop->force_vectorize to check whether there is no alias.
7408
7409 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7410
7411         PR target/84614
7412         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7413         prototypes.
7414         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7415         comments.
7416         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7417         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7418         instead of a loop around prev_real_insn.
7419         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7420         prev_real_insn.
7421
7422         PR inline-asm/84625
7423         * config/i386/i386.c (ix86_print_operand): Use conditional
7424         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7425         zero vector.
7426
7427 2018-03-02  Richard Biener  <rguenther@suse.de>
7428
7429         PR tree-optimization/84427
7430         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7431         (bitmap_set_subtract_values): Rewrite to handle multiple
7432         exprs per value.
7433         (clean): Likewise.
7434         (prune_clobbered_mems): Likewise.
7435         (phi_translate): Take edge instead of pred/phiblock.
7436         (phi_translate_1): Likewise.
7437         (phi_translate_set): Likewise.  Insert all translated
7438         exprs for a value into the set, keeping possibly multiple
7439         expressions per value.
7440         (compute_antic_aux): Adjust for phi_translate changes.
7441         When intersecting union the expressions and prune those
7442         not in the final value set, keeping possibly multiple
7443         expressions per value.  Do not use value-insertion
7444         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7445         all expressions.  Add verification that the value-sets
7446         only shrink during iteration.
7447         (compute_partial_antic_aux): Adjust for the phi_translate changes.
7448         (do_pre_regular_insertion): Likewise.
7449         (do_pre_partial_partial_insertion): Likewise.
7450
7451 2018-03-02  Richard Biener  <rguenther@suse.de>
7452
7453         PR target/82005
7454         * config/darwin.c (saved_debug_info_level): New static global.
7455         (darwin_asm_lto_start): Disable debug info generation for LTO out.
7456         (darwin_asm_lto_end): Restore debug info generation settings.
7457
7458 2018-03-01  Martin Liska  <mliska@suse.cz>
7459
7460         PR sanitizer/82484
7461         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7462         volatile arguments.
7463
7464 2018-03-01  Richard Biener  <rguenther@suse.de>
7465
7466         PR debug/84645
7467         * dwarf2out.c (gen_variable_die): Properly handle late VLA
7468         type annotation with LTO when debug was disabled at compile-time.
7469
7470 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
7471
7472         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7473         XINT with INTVAL.
7474         (mips_final_postscan_insn): Likewise.
7475
7476 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
7477
7478         PR rtl-optimization/84528
7479         * alias.c (init_alias_target): Add commentary.
7480         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7481         a unique base value if the frame pointer is not eliminated
7482         to the stack pointer.
7483
7484 2018-03-01  Tom de Vries  <tom@codesourcery.com>
7485
7486         PR rtl-optimization/83327
7487         * lra-int.h (hard_regs_spilled_into): Declare.
7488         * lra.c (hard_regs_spilled_into): Define.
7489         (init_reg_info): Init hard_regs_spilled_into.
7490         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7491         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7492         (process_bb_lives): Handle hard_regs_spilled_into.
7493         (lra_create_live_ranges_1): Before doing liveness propagation, clear
7494         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7495
7496 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
7497
7498         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7499         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7500         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7501         * config/rs6000/aix72.h: New file.
7502
7503 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
7504
7505         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7506         instead of warning_at with conditional singular and plural messages
7507         where possible.
7508
7509         PR target/52991
7510         * stor-layout.c (update_alignment_for_field): For
7511         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7512         && !DECL_PACKED (field), do the alignment update, just use
7513         only desired_align instead of MAX (type_align, desired_align)
7514         as the alignment.
7515         (place_field): Don't do known_align < desired_align handling
7516         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7517         is non-NULL, instead do it after rli->prev_field handling and
7518         only if not within a bitfield word.  For DECL_PACKED (field)
7519         use type_align of BITS_PER_UNIT.
7520
7521 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
7522
7523         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7524         superfluous parentheses and trailing spaces.
7525
7526 2018-02-28  Richard Biener  <rguenther@suse.de>
7527
7528         PR tree-optimization/84584
7529         * graphite-scop-detection.c (scop_detection::add_scop): Discard
7530         SCoPs with fake exit edge.
7531
7532 2018-02-28  Martin Liska  <mliska@suse.cz>
7533
7534         PR testsuite/84597
7535         * timevar.c (timer::print): Fix format to properly print 100%
7536         values.
7537
7538 2018-02-28  Richard Biener  <rguenther@suse.de>
7539
7540         PR middle-end/84607
7541         * genmatch.c (capture_info::walk_match): Do not mark
7542         captured expressions without operands as expr_p given
7543         they act more like predicates and should be subject to
7544         "lost tail" side-effect preserving.
7545
7546 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
7547
7548         PR rtl-optimization/81611
7549         * auto-inc-dec.c (attempt_change): Move dead note from
7550         mem_insn if it's the next use of regno
7551         (find_address): Take address use of reg holding
7552         non-incremented value.  Add parm to limit search to the named
7553         reg only.
7554         (merge_in_block): Attempt to use a mem insn that is the next
7555         use of the original regno.
7556
7557 2018-02-27  Martin Sebor  <msebor@redhat.com>
7558
7559         PR c++/83871
7560         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7561         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7562
7563 2018-02-27  Martin Sebor  <msebor@redhat.com>
7564
7565         PR translation/84207
7566         * diagnostic-core.h (warning_n, error_n, inform_n): Change
7567         n argument to unsigned HOST_WIDE_INT.
7568         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7569         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
7570         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7571         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7572
7573 2018-02-27  Richard Biener  <rguenther@suse.de>
7574
7575         PR tree-optimization/84512
7576         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7577         Do not use the estimate returned from record_stmt_cost for
7578         the scalar iteration cost but sum properly using add_stmt_cost.
7579
7580 2018-02-27  Richard Biener  <rguenther@suse.de>
7581
7582         PR tree-optimization/84466
7583         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7584         Adjust last change to less strictly validate use operands.
7585
7586 2018-02-27  Martin Liska  <mliska@suse.cz>
7587
7588         PR gcov-profile/84548
7589         * gcov.c (process_file): Allow partial overlap and consider it
7590         also as group functions.
7591         (output_lines): Properly calculate range of lines for a group.
7592
7593 2018-02-27  Martin Liska  <mliska@suse.cz>
7594
7595         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7596         'ggc' suffixes.  Change first column width.
7597         (timer::print): Fix formatting of the column.
7598
7599 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
7600
7601         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7602         preserve inline entry blocks for the sake of debug inline
7603         entry point markers alone.
7604         (remove_unused_locals): Suggest in comments a better place to
7605         force the preservation of inline entry blocks that are
7606         otherwise unused, but do not preserve them.
7607
7608 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7609
7610         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7611
7612 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7613
7614         PR target/84039
7615         * config/i386/constraints.md (Bs): Replace
7616         ix86_indirect_branch_register with
7617         TARGET_INDIRECT_BRANCH_REGISTER.
7618         (Bw): Likewise.
7619         * config/i386/i386.md (indirect_jump): Likewise.
7620         (tablejump): Likewise.
7621         (*sibcall_memory): Likewise.
7622         (*sibcall_value_memory): Likewise.
7623         Peepholes of indirect call and jump via memory: Likewise.
7624         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7625         (*sibcall_value_GOT_32): Likewise.
7626         * config/i386/predicates.md (indirect_branch_operand): Likewise.
7627         (GOT_memory_operand): Likewise.
7628         (call_insn_operand): Likewise.
7629         (sibcall_insn_operand): Likewise.
7630         (GOT32_symbol_operand): Likewise.
7631         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7632
7633 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
7634
7635         PR rtl-optimization/83496
7636         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7637         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
7638         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7639         redundant insn, if any.
7640         (relax_delay_slots): Likewise.
7641         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7642
7643 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
7644
7645         PR tree-optimization/83965
7646         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7647         that grouped statements are part of a reduction chain.  Return
7648         true if the statement is not marked as a reduction itself but
7649         is part of a group.
7650         (vect_recog_dot_prod_pattern): Don't check whether the statement
7651         is part of a group here.
7652         (vect_recog_sad_pattern): Likewise.
7653         (vect_recog_widen_sum_pattern): Likewise.
7654
7655 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
7656
7657         PR debug/84545
7658         * final.c (rest_of_clean_state): Also look for calls inside sequences.
7659
7660 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7661
7662         PR target/84530
7663         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7664         the bool argument.
7665         (ix86_output_indirect_function_return): New prototype.
7666         (ix86_split_simple_return_pop_internal): Likewise.
7667         * config/i386/i386.c (indirect_return_via_cx): New.
7668         (indirect_return_via_cx_bnd): Likewise.
7669         (indirect_thunk_name): Handle return va CX_REG.
7670         (output_indirect_thunk_function): Create alias for
7671         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7672         (ix86_output_indirect_jmp): Remove the bool argument.
7673         (ix86_output_indirect_function_return): New function.
7674         (ix86_split_simple_return_pop_internal): Likewise.
7675         * config/i386/i386.md (*indirect_jump): Don't pass false
7676         to ix86_output_indirect_jmp.
7677         (*tablejump_1): Likewise.
7678         (simple_return_pop_internal): Change it to define_insn_and_split.
7679         Call ix86_split_simple_return_pop_internal to split it for
7680         -mfunction-return=.
7681         (simple_return_indirect_internal): Call
7682         ix86_output_indirect_function_return instead of
7683         ix86_output_indirect_jmp.
7684
7685 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
7686
7687         PR bootstrap/84405
7688         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7689         memset and value initialization afterwards.
7690
7691 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
7692
7693         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7694
7695 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
7696
7697         PR target/84521
7698         * common/config/aarch64/aarch64-common.c
7699         (aarch_option_optimization_table[]): Switch
7700         off fomit-frame-pointer
7701
7702 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
7703             Chung-Ju Wu  <jasonwucj@gmail.com>
7704
7705         * config/nds32/nds32-multiple.md (load_multiple): Disallow
7706         volatile memory.
7707         (store_multiple): Ditto.
7708
7709 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
7710
7711         * config.gcc: Add --with-cpu support for nds32 target.
7712         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7713         * config/nds32/nds32.opt: Add -mcpu= option.
7714
7715 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
7716
7717         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7718         isel=yes): Warn for these deprecated options.
7719
7720 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
7721
7722         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7723         ISA_2_5_MASKS_EMBEDDED.
7724
7725 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
7726
7727         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7728         p->max as pointers rather than using iterative_hash_expr.
7729
7730 2018-02-23  Carl Love  <cel@us.ibm.com>
7731
7732         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7733         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7734         BU_P8V_OVERLOAD_2.
7735         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7736         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7737         P8V_BUILTIN_VEC_VUNSIGNED2.
7738
7739 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
7740
7741         PR target/81572
7742         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7743         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7744         LRA_UNKNOWN_ALT.
7745         * lra-constraints.c (curr_insn_transform): Set up
7746         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
7747         LRA_UNKNOWN_ALT.
7748         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7749         * lra-eliminations.c (spill_pseudos): Ditto.
7750         (process_insn_for_elimination): Ditto.
7751         * lra-lives.c (reg_early_clobber_p): Use the new macros.
7752         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7753         LRA_NON_CLOBBERED_ALT.
7754
7755 2018-02-22  Martin Sebor  <msebor@redhat.com>
7756
7757         PR tree-optimization/84480
7758         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7759         to maybe_diag_stxncpy_trunc.  Call it.
7760         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7761         from gimple_fold_builtin_strcpy.  Print inlining stack.
7762         (handle_builtin_stxncpy): Print inlining stack.
7763         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7764
7765 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
7766
7767         PR target/84176
7768         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7769         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7770         and -fcheck-pointer-bounds are used together.
7771         (indirect_thunk_prefix): New enum.
7772         (indirect_thunk_need_prefix): New function.
7773         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
7774         "_nt" instead of "_bnd" for NOTRACK prefix.
7775         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7776         (output_indirect_thunk_function): Likewise.
7777         (): Likewise.
7778         (ix86_code_end): Update output_indirect_thunk_function calls.
7779         (ix86_output_indirect_branch_via_reg): Replace
7780         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7781         (ix86_output_indirect_branch_via_push): Likewise.
7782         (ix86_output_function_return): Likewise.
7783         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7784         incompatible with -fcf-protection=branch and
7785         -fcheck-pointer-bounds.
7786
7787 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
7788
7789         PR target/83335
7790         * config/aarch64/aarch64.c (aarch64_print_address_internal):
7791         Change gcc_assert call to output_operand_lossage.
7792
7793 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
7794
7795         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7796
7797 2018-02-22  DJ Delorie  <dj@redhat.com>
7798             Sebastian Perta  <sebastian.perta@renesas.com>
7799             Oleg Endo  <olegendo@gcc.gnu.org>
7800
7801         * config/rx/rx.c (rx_rtx_costs): New function.
7802         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7803
7804 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7805
7806         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7807
7808 2018-02-22  Martin Liska  <mliska@suse.cz>
7809
7810         PR driver/83193
7811         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7812         Add "native" as a possible value.
7813
7814 2018-02-22  Martin Liska  <mliska@suse.cz>
7815
7816         PR driver/83193
7817         * config/i386/i386.c (ix86_option_override_internal):
7818         Add "native" as a possible value for -march and -mtune.
7819
7820 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
7821
7822         PR target/84502
7823         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7824         to all type variants.
7825
7826         PR tree-optimization/84503
7827         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7828         width as info->bitpos + info->bitsize - start.
7829         (merged_store_group::merge_overlapping): Simplify width computation.
7830         (check_no_overlap): New function.
7831         (imm_store_chain_info::try_coalesce_bswap): Compute expected
7832         start + width and last_order of the group, fail if check_no_overlap
7833         fails.
7834         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7835         to group if check_no_overlap fails.
7836
7837 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7838
7839         * config/rs6000/altivec.md: Delete contraint arguments to
7840         define_expand, define_split, and define_peephole2, and in
7841         define_insn_and_split if always unused.
7842         * config/rs6000/darwin.md: Ditto.
7843         * config/rs6000/dfp.md: Ditto.
7844         * config/rs6000/rs6000.md: Ditto.
7845         * config/rs6000/sync.md: Ditto.
7846         * config/rs6000/vector.md: Ditto.
7847         * config/rs6000/vsx.md: Ditto.
7848
7849 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7850
7851         * config/rs6000/altivec.md: Write output control strings as braced
7852         blocks instead of double-quoted strings.
7853         * config/rs6000/darwin.md: Ditto.
7854         * config/rs6000/rs6000.md: Ditto.
7855         * config/rs6000/vector.md: Ditto.
7856         * config/rs6000/vsx.md: Ditto.
7857
7858 2018-02-21  Jason Merrill  <jason@redhat.com>
7859
7860         PR c++/84314 - ICE with templates and fastcall attribute.
7861         * attribs.c (build_type_attribute_qual_variant): Remove assert.
7862
7863 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7864
7865         * ipa-cp.c (determine_versionability): Fix comment typos.
7866
7867 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7868
7869         PR c/84229
7870         * ipa-cp.c (determine_versionability): Do not version functions caling
7871         va_arg_pack.
7872
7873 2018-02-21  Martin Liska  <mliska@suse.cz>
7874
7875         PR driver/83193
7876         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7877         Add "native" as a possible value.
7878         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
7879         the macro when native cpu detection is available.
7880
7881 2018-02-21  Martin Liska  <mliska@suse.cz>
7882
7883         PR driver/83193
7884         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7885         Add "native" as a possible value.
7886         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7887         when native cpu detection is available.
7888
7889 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
7890             Martin Sebor  <msebor@redhat.com>
7891
7892         PR tree-optimization/84478
7893         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7894         false.
7895         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7896         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7897         support which is conservatively correct, for 2 only stay conservative
7898         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
7899         argument to the 2 argument get_range_strlen, adjust 6 arg
7900         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7901         false.
7902         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7903         (gimple_fold_builtin_strlen): Pass true as last argument to
7904         get_range_strlen.
7905
7906 2018-02-20  Martin Sebor  <msebor@redhat.com>
7907
7908         PR middle-end/84095
7909         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7910         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
7911         (builtin_memref::builtin_memref): Factor out parts into
7912         set_base_and_offset and call it.
7913
7914 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7915
7916         PR middle-end/84406
7917         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7918         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7919         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
7920         search at the associated MODE_INT.
7921
7922 2018-02-20  Jeff Law  <law@redhat.com>
7923
7924         PR middle-end/82123
7925         PR tree-optimization/81592
7926         PR middle-end/79257
7927         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7928         for range data rather than using global data.
7929         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7930         range data rather than using global data.
7931         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7932         pass it to children as needed.
7933         (struct directive::fmtresult): Similarly.
7934         (struct directive::set_width): Similarly.
7935         (struct directive::set_precision): Similarly.
7936         (format_integer, format_directive, parse_directive): Similarly.
7937         (format_none): Accept unnamed vr_values parameter.
7938         (format_percent, format_floating, format_character): Similarly.
7939         (format_string, format_plain): Similarly.
7940         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7941         the EVRP range analyzer for range data rather than using global data.
7942         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7943         gimple-ssa-evrp-analyze.h
7944         (class sprintf_dom_walker): Add after_dom_children member function.
7945         Add evrp_range_analyzer member.
7946         (sprintf_dom_walker::before_dom_children): Call into the EVRP
7947         range analyzer as needed.
7948         (sprintf_dom_walker::after_dom_children): New member function.
7949         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7950         if not optimizing.
7951         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7952         (evrp_range_analyzer::pop_to_marker): Likewise.
7953
7954 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7955
7956         PR tree-optimization/84419
7957         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7958         with the required type if its current type is compatible but
7959         different.
7960
7961 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
7962
7963         PR middle-end/82004
7964         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7965         after vectorization.
7966
7967 2018-02-20  Martin Liska  <mliska@suse.cz>
7968
7969         PR driver/83193
7970         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7971         possible values if we don't have a hint.
7972
7973 2018-02-20  Martin Liska  <mliska@suse.cz>
7974
7975         PR c/84310
7976         PR target/79747
7977         * final.c (shorten_branches): Build align_tab array with one
7978         more element.
7979         * opts.c (finish_options): Add alignment option limit check.
7980         (MAX_CODE_ALIGN): Likewise.
7981         (MAX_CODE_ALIGN_VALUE): Likewise.
7982         * doc/invoke.texi: Document maximum allowed option value for
7983         all -falign-* options.
7984
7985 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
7986
7987         PR target/84146
7988         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7989         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7990         * var-tracking.c (emit_note_insn_var_location): Remove all references
7991         to NOTE_INSN_CALL_ARG_LOCATION.
7992         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7993         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7994         Use copy_rtx_if_shared.
7995         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7996         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7997         (dwarf2out_var_location): Remove handling of
7998         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7999         on call_insn.
8000         * final.c (final_scan_insn): Remove all references to
8001         NOTE_INSN_CALL_ARG_LOCATION.
8002         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
8003         before dumping final insns.
8004         * except.c (emit_note_eh_region_end): Remove all references to
8005         NOTE_INSN_CALL_ARG_LOCATION.
8006         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8007         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8008         * config/arc/arc.c (hwloop_optimize): Likewise.
8009         * config/arm/arm.c (create_fix_barrier): Likewise.
8010         * config/s390/s390.c (s390_chunkify_start): Likewise.
8011         * config/sh/sh.c (find_barrier): Likewise.
8012         * config/i386/i386.c (rest_of_insert_endbranch,
8013         ix86_seh_fixup_eh_fallthru): Likewise.
8014         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8015         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8016         * config/frv/frv.c (frv_function_prologue): Likewise.
8017         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
8018         reg note.
8019         (note_outside_basic_block_p): Remove all references to
8020         NOTE_INSN_CALL_ARG_LOCATION.
8021         * gengtype.c (adjust_field_rtx_def): Likewise.
8022         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8023         Likewise.
8024         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8025         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8026
8027         PR c++/84444
8028         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8029         is ADDR_EXPR.
8030
8031         PR tree-optimization/84452
8032         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8033         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8034         is NULL.
8035
8036 2018-02-19  Martin Liska  <mliska@suse.cz>
8037
8038         PR sanitizer/82183
8039         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8040
8041 2018-02-19  Martin Liska  <mliska@suse.cz>
8042             Richard Sandiford  <richard.sandiford@linaro.org>
8043
8044         PR tree-optimization/82491
8045         * gimple-fold.c (get_base_constructor): Make earlier bail out
8046         to prevent ubsan.
8047
8048 2018-02-19  Carl Love  <cel@us.ibm.com>
8049
8050         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8051         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8052         BU_P8V_OVERLOAD_1.
8053         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8054         P8V_BUILTIN_VEC_NEG.
8055
8056 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
8057
8058         * config/rl78/rl78.md (movdf): New define expand.
8059
8060 2018-02-19  Martin Liska  <mliska@suse.cz>
8061
8062         PR other/80589
8063         * doc/invoke.texi: Fix typo.
8064         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8065
8066 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
8067
8068         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8069         handle rs6000_single_float and rs6000_double_float specially for
8070         e500 family CPUs.
8071
8072 2018-02-16  Jeff Law  <law@redhat.com>
8073
8074         * config/rx/rx.c (add_pop_cfi_notes): New function.;
8075         (pop_regs): Use it.
8076
8077 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
8078
8079         PR ipa/84425
8080         * ipa-inline.c (inline_small_functions): Fix a typo.
8081
8082 2018-02-16  Nathan Sidwell  <nathan@acm.org>
8083
8084         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8085
8086 2018-02-16  Carl Love  <cel@us.ibm.com>
8087
8088         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8089         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8090         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8091         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8092         expansion to P8V_BUILTIN_VEC_FLOAT2.
8093
8094 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
8095
8096         PR rtl-optimization/70023
8097         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8098         src_regno into account.
8099
8100 2018-02-16  Carl Love  <cel@us.ibm.com>
8101
8102         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8103         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8104         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8105         * config/rs6000/rs6000.c: Remove case statements for
8106         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8107         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8108         and P9V_BUILTIN_VEC_VINSERT4B.
8109         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8110         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8111         * config/rs6000/vsx.md:
8112         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8113         vec_insert4b.
8114
8115 2018-02-16  Carl Love  <cel@us.ibm.com>
8116
8117         * config/rs6000/altivec.h: Add builtin names vec_extract4b
8118         vec_insert4b.
8119         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8120         definitions.
8121         * config/rs6000/rs6000-c.c: Add the definitions for
8122         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8123         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8124         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8125         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
8126         definition for insert4b and define insn *insert3b_internal.
8127         * doc/extend.texi: Add documentation for vec_extract4b.
8128
8129 2018-02-16  Nathan Sidwell  <nathan@acm.org>
8130
8131         * doc/extend.texi (Backwards Compatibility): Mention friend
8132         injection.  Note for-scope is deprecated.
8133         * doc/invoke.texi (-ffriend-injection): Deprecate.
8134
8135 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
8136
8137         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8138         that moved to I2, also allow destinations that are a paradoxical
8139         subreg (instead of a normal reg).
8140
8141 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
8142
8143         PR target/83831
8144         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8145         to QImode.
8146
8147 2018-02-16  Richard Biener  <rguenther@suse.de>
8148
8149         PR tree-optimization/84037
8150         PR tree-optimization/84016
8151         PR target/82862
8152         * config/i386/i386.c (ix86_builtin_vectorization_cost):
8153         Adjust vec_construct for the fact we need additional higher latency
8154         128bit inserts for AVX256 and AVX512 vector builds.
8155         (ix86_add_stmt_cost): Scale vector construction cost for
8156         elementwise loads.
8157
8158 2018-02-16  Richard Biener  <rguenther@suse.de>
8159
8160         PR tree-optimization/84417
8161         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8162         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8163         (non_rewritable_lvalue_p): Likewise, use poly-ints.
8164
8165 2018-02-16  Martin Liska  <mliska@suse.cz>
8166
8167         PR sanitizer/84307
8168         * internal-fn.def (ASAN_CHECK): Set proper flags.
8169         (ASAN_MARK): Likewise.
8170
8171 2018-02-16  Julia Koval  <julia.koval@intel.com>
8172
8173         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8174         from PTA_CANNONLAKE.
8175
8176 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
8177
8178         PR target/84272
8179         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8180         Use ++iter rather than iter++ for std::list iterators.
8181         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
8182         defer deleting them until all nodes in the forest are processed.  Do
8183         free even leaf nodes.  Change to_process into auto_vec.
8184
8185         PR bootstrap/84405
8186         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8187         * vec.h (vec_default_construct): Use memset instead of placement new
8188         if BROKEN_VALUE_INITIALIZATION is defined.
8189         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8190         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8191         is defined.
8192
8193         PR rtl-optimization/83723
8194         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8195         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
8196         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
8197         recursive calls.
8198         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8199         callers.
8200         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8201
8202 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
8203
8204         PR rtl-optimization/81443
8205         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8206         from inner REGs to paradoxical SUBREGs.
8207
8208 2018-02-16  Richard Biener  <rguenther@suse.de>
8209
8210         PR tree-optimization/84399
8211         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8212         For operands we can analyze at their definition make sure we can
8213         analyze them at each use as well.
8214
8215 2018-02-16  Richard Biener  <rguenther@suse.de>
8216
8217         PR tree-optimization/84190
8218         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8219         volatile accesses if the decl isn't volatile.
8220
8221 2018-02-15  Jason Merrill  <jason@redhat.com>
8222
8223         PR c++/84314 - ICE with templates and fastcall attribute.
8224         * attribs.c (build_type_attribute_qual_variant): Don't clobber
8225         TYPE_CANONICAL on an existing type.
8226
8227 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
8228
8229         PR tree-optimization/84383
8230         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8231         dstoff nor call operand_equal_p if dstbase is NULL.
8232
8233         PR tree-optimization/84334
8234         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8235         also a CONSTANT_CLASS_P, punt.
8236
8237 2018-02-14  Jim Wilson  <jimw@sifive.com>
8238
8239         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8240         first SMALL_OPERAND check.  New local min_second_step.  Move assert
8241         to where locals are set.  Add TARGET_RVC support.
8242         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8243
8244 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
8245
8246         * doc/invoke.texi: Correct -Wformat-overflow code sample.
8247
8248 2018-02-14  Martin Sebor  <msebor@redhat.com>
8249
8250         PR tree-optimization/83698
8251         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8252         arrays constrain the offset range to their bounds.
8253         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8254         (builtin_access::overlap): Avoid setting the size of overlap if it's
8255         already been set.
8256         (maybe_diag_overlap): Also consider arrays when deciding what values
8257         of offsets to include in diagnostics.
8258
8259 2018-02-14  Martin Sebor  <msebor@redhat.com>
8260
8261         PR c/84108
8262         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8263         that correspond to the kind of a declaration.
8264
8265 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
8266
8267         PR target/83984
8268         * config/pa/pa.md: Load address of PIC label using the linkage table
8269         if the label is nonlocal.
8270
8271 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8272
8273         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8274         warning message if user requests -maltivec=be.
8275         * doc/invoke.texi: Document deprecation of -maltivec=be.
8276
8277 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
8278
8279         PR target/84220
8280         * config/rs6000/rs6000-c.c: Update definitions for
8281         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8282         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8283
8284 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
8285
8286         PR target/84239
8287         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8288         add _get_ssp intrinsics. Remove argument from
8289         __builtin_ia32_rdssp[d|q].
8290         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8291         * config/i386/i386-builtin.def: Remove argument from
8292         __builtin_ia32_rdssp[d|q].
8293         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8294         ix86_expand_special_args_builtin for _rdssp[d|q].
8295         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8296         Clear register before usage.
8297         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8298         Add documentation for new _get_ssp and _inc_ssp intrinsics.
8299
8300 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
8301
8302         PR tree-optimization/84357
8303         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8304         operand 1 of an ARRAY_REF too.
8305
8306 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
8307
8308         PR target/83831
8309         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8310         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8311         declarations.
8312         (set_of_reg): New struct.
8313         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8314         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8315         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8316         functions.
8317         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8318         Split into bitclr, bitset, bitinvert patterns if appropriate.
8319         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8320         use rx_fuse_in_memory_bitop.
8321         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8322         to named insn, correct maximum insn length.
8323
8324 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
8325
8326         PR target/79242
8327         * machmode.def: Define a complex mode for PARTIAL_INT.
8328         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8329         MODE_PARTIAL_INT.
8330         * doc/rtl.texi: Document CSPImode.
8331         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8332         handling.
8333         (msp430_hard_regno_nregs_with_padding): Likewise.
8334
8335 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
8336
8337         PR target/84279
8338         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8339
8340 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
8341
8342         PR rtl-optimization/84169
8343         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
8344         we generated a parallel as new i3 and we split that to new i2 and i3
8345         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
8346         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
8347         those to i2, not i1.  Partially rewrite this scan code.
8348
8349 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
8350
8351         PR c/82210
8352         * stor-layout.c (place_field): For variable length fields, adjust
8353         offset_align afterwards not just based on the field's alignment,
8354         but also on the size.
8355
8356         PR middle-end/84309
8357         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8358         of exps and logs in the use_exp2 case.
8359
8360 2018-02-13  Jeff Law  <law@redhat.com>
8361
8362         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8363         entry for "vector".
8364
8365         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8366         ARGS as unused.
8367
8368 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
8369
8370         PR debug/84342
8371         PR debug/84319
8372         * common.opt (gas-loc-support, gas-locview-support): New.
8373         (ginline-points, ginternal-reset-location-views): New.
8374         * doc/invoke.texi: Document them.  Use @itemx where intended.
8375         (gvariable-location-views): Adjust.
8376         * target.def (reset_location_view): New.
8377         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8378         (TARGET_RESET_LOCATION_VIEW): New.
8379         * doc/tm.texi: Rebuilt.
8380         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8381         (dwarf2out_default_as_locview_support): New.
8382         (output_asm_line_debug_info): Use option variables.
8383         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8384         (output_loc_list): Likewise.
8385         (add_high_low_attributes): Check option variables.
8386         Don't output entry view attribute in strict mode.
8387         (gen_inlined_subroutine_die): Check option variables.
8388         (dwarf2out_inline_entry): Likewise.
8389         (init_sections_and_labels): Likewise.
8390         (dwarf2out_early_finish): Likewise.
8391         (maybe_reset_location_view): New, from...
8392         (dwarf2out_var_location): ... here.  Call it.
8393         * debug.h (dwarf2out_default_as_loc_support): Declare.
8394         (dwarf2out_default_as_locview_support): Declare.
8395         * hooks.c (hook_int_rtx_insn_0): New.
8396         * hooks.h (hook_int_rtx_insn_0): Declare.
8397         * toplev.c (process_options): Take -gas-loc-support and
8398         -gas-locview-support from dwarf2out.  Enable
8399         -gvariable-location-views by default only with locview
8400         assembler support.  Enable -ginternal-reset-location-views by
8401         default only if the target defines the corresponding hook.
8402         Enable -ginline-points by default if location views are
8403         enabled; force it disabled if statement frontiers are
8404         disabled.
8405         * tree-inline.c (expand_call_inline): Check option variables.
8406         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8407
8408 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
8409
8410         PR tree-optimization/84321
8411         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8412         handling.  Also check whether the anti-range contains any values
8413         that satisfy the mask; switch to a VR_RANGE if not.
8414
8415 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
8416
8417         PR sanitizer/84340
8418         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8419
8420 2018-02-13  Martin Jambor  <mjambor@suse.cz>
8421
8422         PR c++/83990
8423         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8424         of call statements, also set location of a load to a temporary.
8425
8426 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
8427
8428         * config/rl78/rl78.c (add_vector_labels): New function.
8429         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8430         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8431         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8432         which checks that no arguments are passed.
8433         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8434         * doc/extend.texi: Documentation for the new attribute.
8435
8436 2018-02-13  Andreas Schwab  <schwab@suse.de>
8437
8438         * config/riscv/linux.h (CPP_SPEC): Define.
8439
8440 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
8441
8442         PR target/84335
8443         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8444         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8445         OPTION_MASK_ISA_AES as first argument to def_builtin_const
8446         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8447         instead of OPTION_MASK_ISA_PCLMUL as first argument to
8448         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8449         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8450         temporarily for AES and PCLMUL builtins.
8451
8452         PR tree-optimization/84339
8453         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8454         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8455         Formatting fixes.
8456
8457         PR middle-end/84309
8458         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8459         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8460         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8461         inline function.
8462         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8463         inline function.
8464         * omp-simd-clone.h: New file.
8465         * omp-simd-clone.c: Include omp-simd-clone.h.
8466         (expand_simd_clones): No longer static.
8467         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8468         cgraph.h and omp-simd-clone.h.
8469         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8470         (vect_recog_widen_shift_pattern): Formatting fix.
8471         (vect_pattern_recog_1): Don't check optab for calls.
8472
8473         PR target/84336
8474         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8475         operands[2] into a REG before using gen_lowpart on it.
8476
8477 2018-02-12  Jeff Law  <law@redhat.com>
8478
8479         PR target/83760
8480         * config/sh/sh.c (find_barrier): Consider a sibling call
8481         a barrier as well.
8482
8483         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8484         successfully back substituting a reg.
8485
8486 2018-02-12  Richard Biener  <rguenther@suse.de>
8487
8488         PR tree-optimization/84037
8489         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8490         parameter, move visited init to caller.
8491         (vect_slp_analyze_operations): Separate cost from validity
8492         check, initialize visited once for all instances.
8493         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8494         for all instances.
8495         * tree-vect-stmts.c (vect_model_simple_cost): Make early
8496         out an assert.
8497         (vect_model_promotion_demotion_cost): Likewise.
8498         (vectorizable_bswap): Guard cost modeling with !slp_node
8499         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8500         SLP stmts.
8501         (vectorizable_call): Likewise.
8502         (vectorizable_conversion): Likewise.
8503         (vectorizable_assignment): Likewise.
8504         (vectorizable_shift): Likewise.
8505         (vectorizable_operation): Likewise.
8506         (vectorizable_store): Likewise.
8507         (vectorizable_load): Likewise.
8508         (vectorizable_condition): Likewise.
8509         (vectorizable_comparison): Likewise.
8510
8511 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
8512
8513         PR sanitizer/84307
8514         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8515         (ASAN_MARK): Fix fnspec to account for return value, change pointer
8516         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8517
8518 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
8519
8520         PR middle-end/83665
8521         * params.def (inline-min-speedup): Increase from 8 to 15.
8522         (max-inline-insns-auto): Decrease from 40 to 30.
8523         * ipa-split.c (consider_split): Add some buffer for function to
8524         be considered inlining candidate.
8525         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8526         default values.
8527
8528 2018-02-12  Richard Biener  <rguenther@suse.de>
8529
8530         PR tree-optimization/84037
8531         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8532         matched stmts if we cannot swap the non-matched ones.
8533
8534 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
8535
8536         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8537         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8538         _mm_maskz_scalef_round_ss): New intrinsics.
8539         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8540         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8541         __builtin_ia32_scalefss_round): Remove.
8542         (__builtin_ia32_scalefsd_mask_round,
8543         __builtin_ia32_scalefss_mask_round): New intrinsics.
8544         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8545         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8546         ((match_operand:VF_128 2 "<round_nimm_predicate>"
8547         "<round_constraint>")): Changed to ...
8548         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8549         "<round_scalar_constraint>")): ... this.
8550         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8551         %0, %1, %2<round_op3>}"): Changed to ...
8552         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8553         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8554         %2<round_scalar_mask_op3>}"): ... this.
8555         * config/i386/subst.md (round_scalar_nimm_predicate): New.
8556
8557 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
8558
8559         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8560         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8561         (_mm_maskz_sqrt_round_ss): New intrinsics.
8562         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8563         (__builtin_ia32_sqrtsd_mask_round)
8564         (__builtin_ia32_sqrtss_mask_round): New builtins.
8565         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8566         (__builtin_ia32_sqrtss_round): Remove.
8567         (__builtin_ia32_sqrtsd_mask_round)
8568         (__builtin_ia32_sqrtss_mask_round): New builtins.
8569         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8570         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8571         ((match_operand:VF_128 1 "vector_operand"
8572         "xBm,<round_constraint>")): Changed to ...
8573         ((match_operand:VF_128 1 "vector_operand"
8574         "xBm,<round_scalar_constraint>")): ... this.
8575         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8576         %0, %2, %<iptr>1<round_op3>}): Changed to ...
8577         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8578         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8579         %<iptr>1<round_scalar_mask_op3>}): ... this.
8580         ((set_attr "prefix" "<round_prefix>")): Changed to ...
8581         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8582
8583 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
8584
8585         PR target/84266
8586         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8587         Cast vec_cmpeq result to correct type.
8588         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8589         Cast vec_cmpgt result to correct type.
8590
8591 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
8592
8593         * final.c (final_scan_insn_1): Renamed from...
8594         (final_scan_insn): ... this.  New wrapper, to recover
8595         seen from the outermost call in recursive ones.
8596         * config/sparc/sparc.c (output_return): Drop seen from call.
8597         (output_sibcall): Likewise.
8598         * config/visium/visium.c (output_branch): Likewise.
8599
8600 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
8601
8602         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8603         function label.
8604
8605 2018-02-10  Alan Modra  <amodra@gmail.com>
8606
8607         PR target/84300
8608         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8609         Specify LR as an input.
8610
8611 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
8612
8613         PR sanitizer/83987
8614         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8615         remove_member_access_dummy_vars): New functions.
8616         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8617         lower_omp_1, execute_lower_omp): Use them.
8618
8619         PR rtl-optimization/84308
8620         * shrink-wrap.c (spread_components): Release todo vector.
8621
8622 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
8623
8624         PR rtl-optimization/57193
8625         * ira-color.c (struct allocno_color_data): Add member
8626         conflict_allocno_hard_prefs.
8627         (update_conflict_allocno_hard_prefs): New.
8628         (bucket_allocno_compare_func): Add a preference based on
8629         conflict_allocno_hard_prefs.
8630         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8631         (color_allocnos): Remove a dead code.  Initiate
8632         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
8633
8634 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
8635
8636         PR target/84226
8637         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8638         constraint from =wa to wa.  Avoid a subreg on the output operand,
8639         instead use a pseudo and subreg it in a move.
8640         (p9_xxbrd_<mode>): Changed to ...
8641         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8642         (p9_xxbrd_v2df): New expander.
8643         (p9_xxbrw_<mode>): Changed to ...
8644         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8645         (p9_xxbrw_v4sf): New expander.
8646
8647 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
8648
8649         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8650
8651 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
8652
8653         PR target/83926
8654         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8655         multiply in 32-bit mode.
8656         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8657         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8658         mode.
8659
8660 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
8661
8662         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8663         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8664         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8665         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8666
8667 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
8668
8669         PR lto/84213
8670         * dwarf2out.c (is_trivial_indirect_ref): New function.
8671         (dwarf2out_late_global_decl): Do not generate a location
8672         attribute for variables that have a non-trivial DECL_VALUE_EXPR
8673         and that are not defined in the current unit.
8674
8675 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
8676
8677         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8678         instead of a libcall for UNORDERED.
8679
8680 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
8681
8682         PR target/82641
8683         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8684         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8685
8686 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8687
8688         PR target/PR84295
8689         * config/s390/s390.c (s390_set_current_function): Invoke
8690         s390_indirect_branch_settings also if fndecl didn't change.
8691
8692 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
8693
8694         * config/rs6000/rs6000.md (blockage): Set length to zero.
8695
8696 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
8697
8698         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8699
8700 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
8701
8702         PR sanitizer/84285
8703         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8704         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8705         -static-lib*san.
8706
8707         PR debug/84252
8708         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8709         PARALLEL incoming that failed vt_get_decl_and_offset check.
8710
8711         PR middle-end/84237
8712         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8713         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8714         TREE_READONLY bit.
8715         (get_variable_section): For decls in named .bss* sections pass true as
8716         second argument to bss_initializer_p.
8717
8718 2018-02-09  Marek Polacek  <polacek@redhat.com>
8719             Jakub Jelinek  <jakub@redhat.com>
8720
8721         PR c++/83659
8722         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8723         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
8724         Sync some changes from cxx_fold_indirect_ref.
8725
8726 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
8727
8728         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8729         markers.
8730         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8731         (BLOCK_INLINE_ENTRY_LABEL): New.
8732         (dwarf2out_var_location): Disregard inline entry markers.
8733         (inline_entry_data): New struct.
8734         (inline_entry_data_hasher): New hashtable type.
8735         (inline_entry_data_hasher::hash): New.
8736         (inline_entry_data_hasher::equal): New.
8737         (inline_entry_data_table): New variable.
8738         (add_high_low_attributes): Add DW_AT_entry_pc and
8739         DW_AT_GNU_entry_view attributes if a pending entry is found
8740         in inline_entry_data_table.  Add old entry_pc attribute only
8741         if debug nonbinding markers are disabled.
8742         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8743         markers are enabled.
8744         (block_within_block_p, dwarf2out_inline_entry): New.
8745         (dwarf2out_finish): Check that no entries remained in
8746         inline_entry_data_table.
8747         * final.c (reemit_insn_block_notes): Handle inline entry notes.
8748         (final_scan_insn, notice_source_line): Likewise.
8749         (rest_of_clean_state): Skip inline entry markers.
8750         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8751         markers.
8752         * gimple.c (gimple_build_debug_inline_entry): New.
8753         * gimple.h (enum gimple_debug_subcode): Add
8754         GIMPLE_DEBUG_INLINE_ENTRY.
8755         (gimple_build_debug_inline_entry): Declare.
8756         (gimple_debug_inline_entry_p): New.
8757         (gimple_debug_nonbind_marker_p): Adjust.
8758         * insn-notes.def (INLINE_ENTRY): New.
8759         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8760         inline entry marker notes.
8761         (print_insn): Likewise.
8762         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8763         (INSN_DEBUG_MARKER_KIND): Likewise.
8764         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8765         * tree-inline.c (expand_call_inline): Build and insert
8766         debug_inline_entry stmt.
8767         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8768         inline entry blocks early, if nonbind markers are enabled.
8769         (dump_scope_block): Dump fragment info.
8770         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8771         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8772         (gimple_build_debug_inline_entry): New.
8773         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8774         Enable/disable inline entry points too.
8775         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8776         (DEBUG_INSN): Describe inline entry markers.
8777
8778         * common.opt (gvariable-location-views): New.
8779         (gvariable-location-views=incompat5): New.
8780         * config.in: Rebuilt.
8781         * configure: Rebuilt.
8782         * configure.ac: Test assembler for view support.
8783         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8784         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8785         * dwarf2out.c (var_loc_view): New typedef.
8786         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8787         (dwarf2out_locviews_in_attribute): New.
8788         (dwarf2out_locviews_in_loclist): New.
8789         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8790         (enum dw_line_info_opcode): Add LI_adv_address.
8791         (struct dw_line_info_table): Add view.
8792         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8793         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8794         (zero_view_p): New variable.
8795         (ZERO_VIEW_P): New macro.
8796         (output_asm_line_debug_info): New.
8797         (struct var_loc_node): Add view.
8798         (add_AT_view_list, AT_loc_list): New.
8799         (add_var_loc_to_decl): Add view param.  Test it against last.
8800         (new_loc_list): Add view params.  Record them.
8801         (AT_loc_list_ptr): Handle loc and view lists.
8802         (view_list_to_loc_list_val_node): New.
8803         (print_dw_val): Handle dw_val_class_view_list.
8804         (size_of_die): Likewise.
8805         (value_format): Likewise.
8806         (loc_list_has_views): New.
8807         (gen_llsym): Set vl_symbol too.
8808         (maybe_gen_llsym, skip_loc_list_entry): New.
8809         (dwarf2out_maybe_output_loclist_view_pair): New.
8810         (output_loc_list): Output view list or entries too.
8811         (output_view_list_offset): New.
8812         (output_die): Handle dw_val_class_view_list.
8813         (output_dwarf_version): New.
8814         (output_compilation_unit_header): Use it.
8815         (output_skeleton_debug_sections): Likewise.
8816         (output_rnglists, output_line_info): Likewise.
8817         (output_pubnames, output_aranges): Update version comments.
8818         (output_one_line_info_table): Output view numbers in asm comments.
8819         (dw_loc_list): Determine current endview, pass it to new_loc_list.
8820         Call maybe_gen_llsym.
8821         (loc_list_from_tree_1): Adjust.
8822         (add_AT_location_description): Create view list attribute if
8823         needed, check it's absent otherwise.
8824         (convert_cfa_to_fb_loc_list): Adjust.
8825         (maybe_emit_file): Call output_asm_line_debug_info for test.
8826         (dwarf2out_var_location): Reset views as needed.  Precompute
8827         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
8828         attribute.  Set view.
8829         (new_line_info_table): Reset next view.
8830         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8831         (dwarf2out_source_line): Likewise.  Output view resets and labels to
8832         the assembler, or select appropriate line info opcodes.
8833         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8834         (optimize_string_length): Catch it.  Adjust.
8835         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
8836         dw_val_class_view_list, and remove it if no longer needed.
8837         (hash_loc_list): Hash view numbers.
8838         (loc_list_hasher::equal): Compare them.
8839         (optimize_location_lists): Check whether a view list symbol is
8840         needed, and whether the locview attribute is present, and
8841         whether they match.  Remove the locview attribute if no longer
8842         needed.
8843         (index_location_lists): Call skip_loc_list_entry for test.
8844         (dwarf2out_finish): Call output_asm_line_debug_info for test.
8845         Use output_dwarf_version.
8846         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8847         (struct dw_val_node): Add val_view_list.
8848         * final.c (SEEN_NEXT_VIEW): New.
8849         (set_next_view_needed): New.
8850         (clear_next_view_needed): New.
8851         (maybe_output_next_view): New.
8852         (final_start_function): Rename to...
8853         (final_start_function_1): ... this.  Take pointer to FIRST,
8854         add SEEN parameter.  Emit param bindings in the initial view.
8855         (final_start_function): Reintroduce SEEN-less interface.
8856         (final): Rename to...
8857         (final_1): ... this.  Take SEEN parameter.  Output final pending
8858         next view at the end.
8859         (final): Reintroduce seen-less interface.
8860         (final_scan_insn): Output pending next view before switching
8861         sections or ending a block.  Mark the next view as needed when
8862         outputting variable locations.  Notify debug backend of section
8863         changes, and of location view changes.
8864         (rest_of_handle_final): Adjust.
8865         * toplev.c (process_options): Autodetect value for debug variable
8866         location views option.  Warn on incompat5 without -gdwarf-5.
8867         * doc/invoke.texi (gvariable-location-views): New.
8868         (gvariable-location-views=incompat5): New.
8869         (gno-variable-location-views): New.
8870
8871 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
8872
8873         PR tree-optimization/84136
8874         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8875         that the result of find_edge is non-NULL.
8876
8877 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8878
8879         PR target/83008
8880         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8881         storing integer register in SImode.  Fix cost of 256 and 512
8882         byte aligned SSE register store.
8883
8884 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8885
8886         * config/i386/i386.c (ix86_multiplication_cost): Fix
8887         multiplication cost for TARGET_AVX512DQ.
8888
8889 2018-02-08  Marek Polacek  <polacek@redhat.com>
8890
8891         PR tree-optimization/84238
8892         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8893         get_range_strlen.
8894
8895 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8896
8897         PR tree-optimization/84265
8898         * tree-vect-stmts.c (vectorizable_store): Don't treat
8899         VMAT_CONTIGUOUS accesses as grouped.
8900         (vectorizable_load): Likewise.
8901
8902 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8903
8904         PR tree-optimization/81635
8905         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8906         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8907         (test_round_for_mask): New functions.
8908         (wide_int_cc_tests): Call test_round_for_mask.
8909         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8910         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8911         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8912         range returned by get_range_info.
8913
8914 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
8915
8916         PR ipa/81360
8917         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8918         * symtab.c: Include builtins.h
8919         (symtab_node::output_to_lto_symbol_table_p): Move here
8920         from lto-streamer-out.c:output_symbol_p.
8921         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8922         (output_symbol_p): Move all logic to symtab.c
8923         (produce_symtab): Update.
8924
8925 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8926
8927         * config/s390/s390-opts.h (enum indirect_branch): Define.
8928         * config/s390/s390-protos.h (s390_return_addr_from_memory)
8929         (s390_indirect_branch_via_thunk)
8930         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8931         (enum s390_indirect_branch_type): Define.
8932         * config/s390/s390.c (struct s390_frame_layout, struct
8933         machine_function): Remove.
8934         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8935         (indirect_branch_table_label_no, indirect_branch_table_name):
8936         Define variables.
8937         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8938         (enum s390_indirect_branch_option): Define.
8939         (s390_return_addr_from_memory): New function.
8940         (s390_handle_string_attribute): New function.
8941         (s390_attribute_table): Add new attribute handler.
8942         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8943         (s390_indirect_branch_via_thunk): New function.
8944         (s390_indirect_branch_via_inline_thunk): New function.
8945         (s390_function_ok_for_sibcall): When jumping via thunk disallow
8946         sibling call optimization for non z10 compiles.
8947         (s390_emit_call): Force indirect branch target to be a single
8948         register.  Add r1 clobber for non-z10 compiles.
8949         (s390_emit_epilogue): Emit return jump via return_use expander.
8950         (s390_reorg): Handle JUMP_INSNs as execute targets.
8951         (s390_option_override_internal): Perform validity checks for the
8952         new command line options.
8953         (s390_indirect_branch_attrvalue): New function.
8954         (s390_indirect_branch_settings): New function.
8955         (s390_set_current_function): Invoke s390_indirect_branch_settings.
8956         (s390_output_indirect_thunk_function):  New function.
8957         (s390_code_end): Implement target hook.
8958         (s390_case_values_threshold): Implement target hook.
8959         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8960         macros.
8961         * config/s390/s390.h (struct s390_frame_layout)
8962         (struct machine_function): Move here from s390.c.
8963         (TARGET_INDIRECT_BRANCH_NOBP_RET)
8964         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8965         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8966         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8967         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8968         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8969         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8970         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8971         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8972         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8973         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8974         (mnemonic attribute): Add values which aren't recognized
8975         automatically.
8976         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8977         pattern for branch conversion.  Fix mnemonic attribute.
8978         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8979         indirect branch via thunk if requested.
8980         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8981         ("*indirect_jump"): Disable for branch conversion using out of
8982         line thunks.
8983         ("indirect_jump_via_thunk<mode>_z10")
8984         ("indirect_jump_via_thunk<mode>")
8985         ("indirect_jump_via_inlinethunk<mode>_z10")
8986         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8987         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8988         ("casesi_jump_via_inlinethunk<mode>_z10")
8989         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8990         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8991         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8992         ("*indirect2_jump"): Disable for branch conversion.
8993         ("casesi_jump"): Turn into expander and expand patterns for branch
8994         conversion.
8995         ("return_use"): New expander.
8996         ("*return"): Emit return via thunk and rename it to ...
8997         ("*return<mode>"): ... this one.
8998         * config/s390/s390.opt: Add new options and and enum for the
8999         option values.
9000
9001 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
9002
9003         * lra-constraints.c (match_reload): Unconditionally use
9004         gen_lowpart_SUBREG, rather than selecting between that
9005         and equivalent gen_rtx_SUBREG code.
9006
9007 2018-02-08  Richard Biener  <rguenther@suse.de>
9008
9009         PR tree-optimization/84233
9010         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9011         changed flag instead of boguously re-using phi_inserted.
9012
9013 2018-02-08  Martin Jambor  <mjambor@suse.cz>
9014
9015         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9016         static local variables.
9017
9018 2018-02-08  Richard Biener  <rguenther@suse.de>
9019
9020         PR tree-optimization/84278
9021         * tree-vect-stmts.c (vectorizable_store): When looking for
9022         smaller vector types to perform grouped strided loads/stores
9023         make sure the mode is supported by the target.
9024         (vectorizable_load): Likewise.
9025
9026 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
9027
9028         * config/aarch64/aarch64.c (aarch64_components_for_bb):
9029         Increase LDP/STP opportunities by adding adjacent callee-saves.
9030
9031 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
9032
9033         PR rtl-optimization/84068
9034         PR rtl-optimization/83459
9035         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9036
9037 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
9038
9039         PR tree-optimization/84224
9040         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9041         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9042         non-zero arguments.
9043
9044 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
9045
9046         PR target/84113
9047         * config/rs6000/altivec.md (*restore_world): Remove LR use.
9048         * config/rs6000/predicates.md (restore_world_operation): Adjust op
9049         count, remove one USE.
9050
9051 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
9052
9053         * doc/install.texi (Configuration): Document the
9054         --with-long-double-format={ibm,ieee} PowerPC configuration
9055         options.
9056
9057         PR target/84154
9058         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9059         Convert from define_expand to be define_insn_and_split.  Rework
9060         float/double/_Float128 conversions to QI/HI/SImode to work with
9061         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
9062         conversions to QI/HImode types did a store and then a load to
9063         truncate the value.  For conversions to VSX registers, don't split
9064         the insn, instead emit the code directly.  Use the code iterator
9065         any_fix to combine signed and unsigned conversions.
9066         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9067         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9068         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9069         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9070         (fix_<mode>di2_hw): Likewise.
9071         (fixuns_<mode>di2_hw): Likewise.
9072         (fix_<mode>si2_hw): Likewise.
9073         (fixuns_<mode>si2_hw): Likewise.
9074         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9075         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9076         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9077         fix<uns>_trunc<SFDF:mode>si2_p8.
9078         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
9079         used.
9080         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9081         (fix<uns>_<mode>_mem): Likewise.
9082         (fctiw<u>z_<mode>_mem): Likewise.
9083         (fix<uns>_<mode>_mem): Likewise.
9084         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9085         the register allocator from doing a direct move to the GPRs to do
9086         a store, and instead use the ISA 3.0 store byte/half-word from
9087         vector register instruction.  For IEEE 128-bit floating point,
9088         also optimize stores of 32-bit ints.
9089         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9090
9091 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
9092
9093         * genextract.c (push_pathstr_operand): New function to support
9094         [a-zA-Z].
9095         (walk_rtx): Call push_pathstr_operand.
9096         (print_path): Support [a-zA-Z].
9097
9098 2018-02-07  Richard Biener  <rguenther@suse.de>
9099
9100         PR tree-optimization/84037
9101         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9102         (cse_and_gimplify_to_preheader): Declare.
9103         (vect_get_place_in_interleaving_chain): Likewise.
9104         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9105         ivexpr_map.
9106         (_loop_vec_info::~_loop_vec_info): Delete it.
9107         (cse_and_gimplify_to_preheader): New function.
9108         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9109         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9110         (vectorizable_load): Likewise.  For grouped stores always base
9111         the IV on the first element.
9112         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9113         condition before gimplifying.
9114
9115 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
9116
9117         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9118         *DIV_EXPR and *MOD_EXPR.
9119
9120 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
9121
9122         PR target/84248
9123         * config/i386/i386.c (ix86_option_override_internal): Mask out
9124         the CF_SET bit when checking -fcf-protection.
9125
9126 2018-02-07  Tom de Vries  <tom@codesourcery.com>
9127
9128         PR libgomp/84217
9129         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9130         enough.
9131
9132 2018-02-07  Richard Biener  <rguenther@suse.de>
9133
9134         PR tree-optimization/84204
9135         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9136         this place.
9137
9138         PR tree-optimization/84205
9139         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9140         special-case isl_ast_op_zdiv_r.
9141
9142         PR tree-optimization/84223
9143         * graphite-scop-detection.c (gather_bbs::before_dom_children):
9144         Only add conditions from within the region.
9145         (gather_bbs::after_dom_children): Adjust.
9146
9147 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
9148
9149         PR target/84209
9150         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9151         * config/avr/avr.md: Only post-reload split REG-REG moves if
9152         either register is GENERAL_REG_P.
9153
9154 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
9155
9156         PR tree-optimization/84235
9157         * tree-ssa-scopedtables.c
9158         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9159         if the subtraction is performed in floating point type where NaNs are
9160         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9161         build 1.  Formatting fix.
9162
9163 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
9164
9165         PR target/84146
9166         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9167         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9168         and skip it regardless of bb boundaries.  Use CALL_P macro,
9169         don't test INSN_P (insn) together with CALL_P or JUMP_P check
9170         unnecessarily, formatting fix.
9171
9172 2018-02-06  Michael Collison  <michael.collison@arm.com>
9173
9174         * config/arm/thumb2.md:
9175         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9176         (*thumb_mov_notscc): Ditto.
9177
9178 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
9179
9180         PR target/84154
9181         * config/rs6000/rs6000.md (su code attribute): Use "u" for
9182         unsigned_fix, not "s".
9183
9184 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9185
9186         * configure.ac (gcc_fn_eh_frame_ro): New function.
9187         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9188         correct .eh_frame permissions.
9189         * configure: Regenerate.
9190
9191 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
9192
9193         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9194         irrelevant options.
9195
9196 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9197
9198         * config/rs6000/rs6000.c (rs6000_option_override_internal):
9199         Display warning message for -mno-speculate-indirect-jumps.
9200
9201 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
9202
9203         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9204         Undocumented.
9205         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9206
9207 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
9208
9209         PR tree-optimization/84225
9210         * tree-eh.c (find_trapping_overflow): Only call
9211         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9212
9213 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
9214
9215         PR target/84145
9216         * config/i386/i386.c: Reimplement the check of possible options
9217         -mibt/-mshstk conbination. Change error messages.
9218         * doc/invoke.texi: Fix a typo: remove extra '='.
9219
9220 2018-02-06  Marek Polacek  <polacek@redhat.com>
9221
9222         PR tree-optimization/84228
9223         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9224
9225 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
9226
9227         PR target/82641
9228         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9229         emitted arch directives.
9230         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9231         __ARM_FEATURE_COPROC before changing architectures.
9232
9233 2018-02-06  Richard Biener  <rguenther@suse.de>
9234
9235         * config/i386/i386.c (print_reg): Fix typo.
9236         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9237
9238 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
9239
9240         * configure: Regenerate.
9241
9242 2018-02-05  Martin Sebor  <msebor@redhat.com>
9243
9244         PR tree-optimization/83369
9245         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9246         inlining context.
9247
9248 2018-02-05  Martin Liska  <mliska@suse.cz>
9249
9250         * doc/invoke.texi: Cherry-pick upstream r323995.
9251
9252 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
9253
9254         * ira.c (ira_init_register_move_cost): Adjust comment.
9255
9256 2018-02-05  Martin Liska  <mliska@suse.cz>
9257
9258         PR gcov-profile/84137
9259         * doc/gcov.texi: Fix typo in documentation.
9260
9261 2018-02-05  Martin Liska  <mliska@suse.cz>
9262
9263         PR gcov-profile/83879
9264         * doc/gcov.texi: Document necessity of --dynamic-list-data when
9265         using dlopen functionality.
9266
9267 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
9268
9269         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9270         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9271         _mm_maskz_range_ss, _mm_mask_range_round_ss,
9272         _mm_maskz_range_round_ss): New intrinsics.
9273         (__builtin_ia32_rangesd128_round)
9274         (__builtin_ia32_rangess128_round): Remove.
9275         (__builtin_ia32_rangesd128_mask_round,
9276         __builtin_ia32_rangess128_mask_round): New builtins.
9277         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9278         __builtin_ia32_rangess128_round): Remove.
9279         (__builtin_ia32_rangesd128_mask_round,
9280         __builtin_ia32_rangess128_mask_round): New builtins.
9281         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9282         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9283         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9284         "<round_saeonly_constraint>")): Changed to ...
9285         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9286         "<round_saeonly_scalar_constraint>")): ... this.
9287         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9288         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9289         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9290         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9291         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9292
9293 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
9294
9295         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9296         options.
9297         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9298         Remove all values except native, 8540 and 8548.
9299
9300 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
9301
9302         * config/i386/i386.c (ix86_output_function_return): Pass
9303         INVALID_REGNUM, instead of -1, as invalid register number to
9304         indirect_thunk_name and output_indirect_thunk.
9305
9306 2018-02-02  Julia Koval  <julia.koval@intel.com>
9307
9308         * config.gcc: Add -march=icelake.
9309         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9310         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9311         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9312         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9313         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9314         (processor_target_table): Add icelake.
9315         (ix86_option_override_internal): Handle new PTAs.
9316         (get_builtin_code_for_version): Handle icelake.
9317         (M_INTEL_COREI7_ICELAKE): New.
9318         (fold_builtin_cpu): Handle icelake.
9319         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9320         * doc/invoke.texi: Add -march=icelake.
9321
9322 2018-02-02  Julia Koval  <julia.koval@intel.com>
9323
9324         * config/i386/i386.c (ix86_option_override_internal): Change flags type
9325         to wide_int_bitmask.
9326         * wide-int-bitmask.h: New.
9327
9328 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
9329
9330         PR target/84066
9331         * config/i386/i386.md: Replace Pmode with word_mode in
9332         builtin_setjmp_setup and builtin_longjmp to support x32.
9333
9334 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
9335
9336         PR target/56010
9337         PR target/83743
9338         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9339         #include "opts.h".
9340         (rs6000_supported_cpu_names): New static variable.
9341         (linux_cpu_translation_table): Likewise.
9342         (elf_platform) <cpu>: Define new static variable and use it.
9343         Translate kernel AT_PLATFORM name to canonical name if needed.
9344         Error if platform name is unknown.
9345
9346 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
9347
9348         PR target/84089
9349         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9350
9351 2018-02-01  Jeff Law  <law@redhat.com>
9352
9353         PR target/84128
9354         * config/i386/i386.c (release_scratch_register_on_entry): Add new
9355         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
9356         the scratch if RELEASE_VIA_POP is false.
9357         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9358         If we have to save a temporary register, decrement SIZE appropriately.
9359         Pass new arguments to release_scratch_register_on_entry.
9360         (ix86_adjust_stack_and_probe): Likewise.
9361         (ix86_emit_probe_stack_range): Pass new arguments to
9362         release_scratch_register_on_entry.
9363
9364 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
9365
9366         PR rtl-optimization/84157
9367         * combine.c (change_zero_ext): Use REG_P predicate in
9368         front of HARD_REGISTER_P predicate.
9369
9370 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
9371
9372         * config/avr/avr.c (avr_option_override): Move disabling of
9373         -fdelete-null-pointer-checks to...
9374         * common/config/avr/avr-common.c (avr_option_optimization_table):
9375         ...here.
9376
9377 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9378
9379         PR tree-optimization/81635
9380         * tree-data-ref.c (split_constant_offset_1): For types that
9381         wrap on overflow, try to use range info to prove that wrapping
9382         cannot occur.
9383
9384 2018-02-01  Renlin Li  <renlin.li@arm.com>
9385
9386         PR target/83370
9387         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9388         TAILCALL_ADDR_REGS.
9389         (aarch64_register_move_cost): Likewise.
9390         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9391         TAILCALL_ADDR_REGS.
9392         (REG_CLASS_NAMES): Likewise.
9393         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9394         TAILCALL_ADDR_REGS. Remove IP registers.
9395         * config/aarch64/aarch64.md (Ucs): Update register constraint.
9396
9397 2018-02-01  Richard Biener  <rguenther@suse.de>
9398
9399         * domwalk.h (dom_walker::dom_walker): Add additional constructor
9400         for specifying RPO order and allow NULL for that.
9401         * domwalk.c (dom_walker::dom_walker): Likewise.
9402         (dom_walker::walk): Handle NULL RPO order.
9403         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9404         in RPO order.
9405         (rewrite_update_dom_walker): Likewise.
9406         (mark_def_dom_walker): Likewise.
9407
9408 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9409
9410         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9411         (aarch64_maybe_expand_sve_subreg_move): Declare.
9412         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9413         * config/aarch64/predicates.md (aarch64_any_register_operand): New
9414         predicate.
9415         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9416         that are semantically a reverse operation.
9417         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9418         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9419         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9420         functions.
9421         (aarch64_can_change_mode_class): For big-endian, forbid changes
9422         between two SVE modes if they have different element sizes.
9423
9424 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9425
9426         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9427         the TImode handling for big-endian targets.
9428
9429 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9430
9431         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9432         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
9433         not just bytes.
9434         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9435         Remove BSWAP handing for big-endian targets and use the form of
9436         LD1RQ appropariate for the mode.
9437
9438 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9439
9440         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9441         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9442         duplicated element.
9443
9444 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9445
9446         PR tearget/83845
9447         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9448         check for operands that need to go through aarch64_sve_reload_be.
9449
9450 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
9451
9452         PR tree-optimization/81661
9453         PR tree-optimization/84117
9454         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9455         * tree-eh.c: Include gimplify.h.
9456         (find_trapping_overflow, replace_trapping_overflow,
9457         rewrite_to_non_trapping_overflow): New functions.
9458         * tree-vect-loop.c: Include tree-eh.h.
9459         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9460         * tree-data-ref.c: Include tree-eh.h.
9461         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9462
9463 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
9464
9465         PR rtl-optimization/84123
9466         * combine.c (change_zero_ext): Check if hard register satisfies
9467         can_change_dest_mode before calling gen_lowpart_SUBREG.
9468
9469 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
9470
9471         PR target/82444
9472         * ira.c (ira_init_register_move_cost): Remove assert.
9473
9474 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9475
9476         PR rtl-optimization/84071
9477         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9478         * doc/tm.texi: Regenerate.
9479
9480 2018-01-31  Richard Biener  <rguenther@suse.de>
9481
9482         PR tree-optimization/84132
9483         * tree-data-ref.c (analyze_miv_subscript): Properly
9484         check whether evolution_function_is_affine_multivariate_p
9485         before calling gcd_of_steps_may_divide_p.
9486
9487 2018-01-31  Julia Koval  <julia.koval@intel.com>
9488
9489         PR target/83618
9490         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9491         * config/i386/i386.md (rdpid_rex64) New.
9492         (rdpid): Make 32bit only.
9493
9494 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
9495
9496         PR lto/84105
9497         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9498         an IDENTIFIER_NODE for FUNCTION_TYPE's.
9499
9500 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9501
9502         Revert
9503         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
9504
9505         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9506
9507 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9508
9509         PR rtl-optimization/84071
9510         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9511         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9512
9513 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
9514
9515         * config/arc/arc.c (arc_handle_aux_attribute): New function.
9516         (arc_attribute_table): Add 'aux' attribute.
9517         (arc_in_small_data_p): Consider aux like variables.
9518         (arc_is_aux_reg_p): New function.
9519         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9520         (arc_get_aux_arg): New function.
9521         (prepare_move_operands): Handle aux-register access.
9522         (arc_handle_aux_attribute): New function.
9523         * doc/extend.texi (ARC Variable attributes): Add subsection.
9524
9525 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
9526
9527         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9528         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9529         (arc_attribute_table): Add 'uncached' attribute.
9530         (arc_print_operand): Print '.di' flag for uncached memory
9531         accesses.
9532         (arc_in_small_data_p): Do not consider for small data the uncached
9533         types.
9534         (arc_is_uncached_mem_p): New function.
9535         * config/arc/predicates.md (compact_store_memory_operand): Check
9536         for uncached memory accesses.
9537         (nonvol_nonimm_operand): Likewise.
9538         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9539
9540 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
9541
9542         PR c/84100
9543         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9544         falign-loops=): Add Optimization flag.
9545
9546 2018-01-30  Jeff Law  <law@redhat.com>
9547
9548         PR target/84064
9549         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9550         INT_REGISTERS_SAVED.  Check it prior to calling
9551         get_scratch_register_on_entry.
9552         (ix86_adjust_stack_and_probe): Similarly.
9553         (ix86_emit_probe_stack_range): Similarly.
9554         (ix86_expand_prologue): Corresponding changes.
9555
9556 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9557
9558         PR target/40411
9559         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9560         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9561
9562 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
9563
9564         PR target/84112
9565         * lra-constraints.c (curr_insn_transform): Process AND in the
9566         address.
9567
9568 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
9569
9570         PR rtl-optimization/83986
9571         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9572         dependence against last_pending_memory_flush in addition to
9573         pending_jump_insns.
9574
9575 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
9576
9577         PR tree-optimization/81611
9578         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9579         copies.
9580
9581 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
9582
9583         PR target/83758
9584         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9585         a reg rtx.
9586
9587 2018-01-30  Richard Biener  <rguenther@suse.de>
9588             Jakub Jelinek  <jakub@redhat.com>
9589
9590         PR tree-optimization/84111
9591         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9592         inner loops added during recursion, as they don't have up-to-date
9593         SSA form.
9594
9595 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
9596
9597         PR ipa/81360
9598         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9599         (can_inline_edge_by_limits_p): ... here.
9600         (can_early_inline_edge_p, check_callers,
9601         update_caller_keys, update_callee_keys, recursive_inlining,
9602         add_new_edges_to_heap, speculation_useful_p,
9603         inline_small_functions,
9604         inline_small_functions, flatten_function,
9605         inline_to_all_callers_1): Update.
9606
9607 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
9608
9609         * profile-count.c (profile_count::combine_with_ipa_count): Handle
9610         zeros correctly.
9611
9612 2018-01-30  Richard Biener  <rguenther@suse.de>
9613
9614         PR tree-optimization/83008
9615         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9616         invariant and constant vector uses in stmts when they need
9617         more than one stmt.
9618
9619 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9620
9621         PR bootstrap/84017
9622         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9623         * configure: Regenerate.
9624
9625 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9626
9627         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9628         pattern.
9629         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9630         Use gen_rtx_REG rather than gen_lowpart.
9631
9632 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9633
9634         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9635         rather than 0 when creating partial subregs.
9636
9637 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9638
9639         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9640         of usage.
9641
9642 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
9643
9644         PR target/81550
9645         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9646         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9647         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9648         flags.  This restores the settings used before the 2017-07-24.
9649         Turning off pre increment/decrement/modify allows IVOPTS to
9650         optimize DF/SF loops where the index is an int.
9651
9652 2018-01-29  Richard Biener  <rguenther@suse.de>
9653             Kelvin Nilsen  <kelvin@gcc.gnu.org>
9654
9655         PR bootstrap/80867
9656         * tree-vect-stmts.c (vectorizable_call): Don't call
9657         targetm.vectorize_builtin_md_vectorized_function if callee is
9658         NULL.
9659
9660 2018-01-22  Carl Love  <cel@us.ibm.com>
9661
9662         * doc/extend.tex: Fix typo in second arg in
9663         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9664
9665 2018-01-29  Richard Biener  <rguenther@suse.de>
9666
9667         PR tree-optimization/84086
9668         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9669         (flush_ssaname_freelist): When SSA names were released reset
9670         the SCEV hash table.
9671
9672 2018-01-29  Richard Biener  <rguenther@suse.de>
9673
9674         PR tree-optimization/84057
9675         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9676         removed paths when removing edges.
9677
9678 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
9679
9680         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9681         -mfunction-return=@var{choice}.
9682
9683 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9684
9685         PR diagnostic/84034
9686         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9687         Handle CR like TAB.
9688         (layout::print_source_line): Likewise.
9689         (test_get_line_width_without_trailing_whitespace): Add test cases.
9690
9691 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
9692
9693         PR middle-end/84040
9694         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9695         debug insns.
9696
9697 2018-01-26  Jim Wilson  <jimw@sifive.com>
9698
9699         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9700
9701         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9702         specified.
9703
9704 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9705
9706         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9707         and CMP + SUB-immediate -> SUBS.
9708
9709 2018-01-26  Martin Sebor  <msebor@redhat.com>
9710
9711         PR tree-optimization/83896
9712         * tree-ssa-strlen.c (get_string_len): Rename...
9713         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
9714         Avoid assuming length is constant.
9715         (handle_char_store): Use HOST_WIDE_INT for string length.
9716
9717 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
9718
9719         PR target/81763
9720         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9721         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9722
9723 2018-01-26  Richard Biener  <rguenther@suse.de>
9724
9725         PR rtl-optimization/84003
9726         * dse.c (record_store): Only record redundant stores when
9727         the earlier store aliases at least all accesses the later one does.
9728
9729 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
9730
9731         PR rtl-optimization/83985
9732         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9733         REG_CFA_RESTORE insns.
9734         (delete_unmarked_insns): Don't ignore separate shrink wrapping
9735         REG_CFA_RESTORE insns here.
9736
9737         PR c/83989
9738         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9739         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9740
9741 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9742
9743         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9744         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9745         (arc_init): Likewise.
9746         (arc_override_options): Likewise.
9747         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9748         value.
9749         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9750         support.
9751         * config/arc/arc.h (TARGET_DBNZ): Define.
9752         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9753         properly set the tune attribute.
9754         (dbnz): Use TARGET_DBNZ guard.
9755         * config/arc/arc.opt (mtune): Add core3 option.
9756
9757 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9758
9759         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9760         recognize new pic like addresses.
9761         (arc_delegitimize_address): Clean up.
9762
9763 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9764
9765         * config/arc/arc-arches.def: Option mrf16 valid for all
9766         architectures.
9767         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9768         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9769         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9770         * config/arc/arc-tables.opt: Regenerate.
9771         * config/arc/arc.c (arc_conditional_register_usage): Handle
9772         reduced register file case.
9773         (arc_file_start): Set must have build attributes.
9774         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9775         mrf16 option value.
9776         * config/arc/arc.opt (mrf16): Add new option.
9777         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9778         * config/arc/genmultilib.awk: Handle new mrf16 option.
9779         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9780         * config/arc/t-multilib: Regenerate.
9781         * doc/invoke.texi (ARC Options): Document mrf16 option.
9782
9783 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9784
9785         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9786         * config/arc/arc.c (arc_handle_secure_attribute): New function.
9787         (arc_attribute_table): Add 'secure_call' attribute.
9788         (arc_print_operand): Print secure call operand.
9789         (arc_function_ok_for_sibcall): Don't optimize tail calls when
9790         secure.
9791         (arc_is_secure_call_p): New function.  * config/arc/arc.md
9792         (call_i): Add support for sjli instruction.
9793         (call_value_i): Likewise.
9794         * config/arc/constraints.md (Csc): New constraint.
9795
9796 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9797             John Eric Martin  <John.Martin@emmicro-us.com>
9798
9799         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9800         * config/arc/arc.c (_arc_jli_section): New struct.
9801         (arc_jli_section): New type.
9802         (rc_jli_sections): New static variable.
9803         (arc_handle_jli_attribute): New function.
9804         (arc_attribute_table): Add jli_always and jli_fixed attribute.
9805         (arc_file_end): New function.
9806         (TARGET_ASM_FILE_END): Define.
9807         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9808         (arc_add_jli_section): New function.
9809         (jli_call_scan): Likewise.
9810         (arc_reorg): Call jli_call_scan.
9811         (arc_output_addsi): Remove 'S' from printing asm operand.
9812         (arc_is_jli_call_p): New function.
9813         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9814         operand.
9815         (movhi_insn): Likewise.
9816         (movsi_insn): Likewise.
9817         (movsi_set_cc_insn): Likewise.
9818         (loadqi_update): Likewise.
9819         (load_zeroextendqisi_update): Likewise.
9820         (load_signextendqisi_update): Likewise.
9821         (loadhi_update): Likewise.
9822         (load_zeroextendhisi_update): Likewise.
9823         (load_signextendhisi_update): Likewise.
9824         (loadsi_update): Likewise.
9825         (loadsf_update): Likewise.
9826         (movsicc_insn): Likewise.
9827         (bset_insn): Likewise.
9828         (bxor_insn): Likewise.
9829         (bclr_insn): Likewise.
9830         (bmsk_insn): Likewise.
9831         (bicsi3_insn): Likewise.
9832         (cmpsi_cc_c_insn): Likewise.
9833         (movsi_ne): Likewise.
9834         (movsi_cond_exec): Likewise.
9835         (clrsbsi2): Likewise.
9836         (norm_f): Likewise.
9837         (normw): Likewise.
9838         (swap): Likewise.
9839         (divaw): Likewise.
9840         (flag): Likewise.
9841         (sr): Likewise.
9842         (kflag): Likewise.
9843         (ffs): Likewise.
9844         (ffs_f): Likewise.
9845         (fls): Likewise.
9846         (call_i): Remove 'S' asm letter, add jli instruction.
9847         (call_value_i): Likewise.
9848         * config/arc/arc.op (mjli-always): New option.
9849         * config/arc/constraints.md (Cji): New constraint.
9850         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9851         operand.
9852         (subsf3_fpx): Likewise.
9853         (mulsf3_fpx): Likewise.
9854         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9855         asm operand.
9856         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9857         function attrbutes.
9858         * doc/invoke.texi (ARC): Document mjli-always option.
9859
9860 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
9861
9862         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9863         avoid addition with 0 and use incw and decw where possible.
9864
9865 2018-01-26  Richard Biener  <rguenther@suse.de>
9866
9867         PR tree-optimization/81082
9868         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9869         association if it requires casting to unsigned.
9870         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9871         from fold_plusminus_mult_expr to catch important cases late when
9872         range info is available.
9873
9874 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9875
9876         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9877         * configure.ac (hidden_linkonce): New test.
9878         * configure: Regenerate.
9879         * config.in: Regenerate.
9880
9881 2018-01-26  Julia Koval  <julia.koval@intel.com>
9882
9883         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9884         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9885         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9886         _mm_mask_bitshuffle_epi64_mask): Fix type.
9887         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9888         USI_FTYPE_V4DI_V4DI_USI): Remove.
9889         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9890         __builtin_ia32_vpshufbitqmb256_mask,
9891         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9892         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9893         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9894
9895 2018-01-26  Alan Modra  <amodra@gmail.com>
9896
9897         PR target/84033
9898         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9899         UNSPEC_VBPERMQ.  Sort other unspecs.
9900
9901 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
9902
9903         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9904
9905 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
9906
9907         PR middle-end/83055
9908         * predict.c (drop_profile): Do not push/pop cfun; update also
9909         node->count.
9910         (handle_missing_profiles): Fix logic looking for zero profiles.
9911
9912 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
9913
9914         PR middle-end/83977
9915         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9916         on functions with #pragma omp declare simd or functions with simd
9917         attribute.
9918         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9919         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9920         Remove trailing \n from warning_at calls.
9921
9922 2018-01-25  Tom de Vries  <tom@codesourcery.com>
9923
9924         PR target/84028
9925         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9926         for neutered workers.
9927
9928 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
9929
9930         PR target/68467
9931         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9932         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9933
9934 2018-01-24  Jeff Law  <law@redhat.com>
9935
9936         PR target/83994
9937         * i386.c (get_probe_interval): Move to earlier point.
9938         (ix86_compute_frame_layout): If -fstack-clash-protection and
9939         the frame is larger than the probe interval, then use pushes
9940         to save registers rather than reg->mem moves.
9941         (ix86_expand_prologue): Remove conditional for int_registers_saved
9942         assertion.
9943
9944 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
9945
9946         PR target/84014
9947         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9948         min/max for never referenced object.
9949
9950 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
9951
9952         PR middle-end/83977
9953         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9954         here.
9955         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9956         attributes from DECL_ATTRIBUTES (decl) without affecting
9957         DECL_ATTRIBUTES (current_function_decl).
9958         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9959         functions with non-NULL DECL_ABSTRACT_ORIGIN.
9960
9961 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
9962
9963         PR tree-optimization/83979
9964         * fold-const.c (fold_comparison): Use constant_boolean_node
9965         instead of boolean_{true,false}_node.
9966
9967 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
9968
9969         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9970         with zero counts.
9971
9972 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9973
9974         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9975         Simplify the clause that sets the length attribute.
9976         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9977         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9978         clause that sets the length attribute.
9979         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9980
9981 2018-01-24  Tom de Vries  <tom@codesourcery.com>
9982
9983         PR target/83589
9984         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9985         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9986         Add strict parameter.
9987         (prevent_branch_around_nothing): Insert dummy insn between branch to
9988         label and label with no ptx insn inbetween.
9989         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9990
9991 2018-01-24  Tom de Vries  <tom@codesourcery.com>
9992
9993         PR target/81352
9994         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9995         for neutered threads in warp.
9996         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9997
9998 2018-01-24  Richard Biener  <rguenther@suse.de>
9999
10000         PR tree-optimization/83176
10001         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10002         operands.
10003
10004 2018-01-24  Richard Biener  <rguenther@suse.de>
10005
10006         PR tree-optimization/82819
10007         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10008         code generating pluses that are no-ops in the target precision.
10009
10010 2018-01-24  Richard Biener  <rguenther@suse.de>
10011
10012         PR middle-end/84000
10013         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10014
10015 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
10016
10017         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10018         to merge probabilities.
10019         * predict.c (probably_never_executed): Also mark as cold functions
10020         with global 0 profile and guessed local profile.
10021         * profile-count.c (profile_probability::combine_with_count): New
10022         member function.
10023         * profile-count.h (profile_probability::operator*,
10024         profile_probability::operator*=, profile_probability::operator/,
10025         profile_probability::operator/=): Reduce precision to adjusted
10026         and set value to guessed on contradictory divisions.
10027         (profile_probability::combine_with_freq): Remove.
10028         (profile_probability::combine_wiht_count): Declare.
10029         (profile_count::force_nonzero):: Set to adjusted.
10030         (profile_count::probability_in):: Set quality to adjusted.
10031         * tree-ssa-tail-merge.c (replace_block_by): Use
10032         combine_with_count.
10033
10034 2018-01-23  Andrew Waterman  <andrew@sifive.com>
10035             Jim Wilson  <jimw@sifive.com>
10036
10037         * config/riscv/riscv.c (riscv_stack_boundary): New.
10038         (riscv_option_override): Set riscv_stack_boundary.  Handle
10039         riscv_preferred_stack_boundary_arg.
10040         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10041         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10042         (STACK_BOUNDARY): Set to riscv_stack_boundary.
10043         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10044         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10045         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10046
10047 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
10048
10049         PR target/83905
10050         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10051         of struct ix86_frame.
10052         (ix86_expand_epilogue): Likewise.  Add a local variable for
10053         the reg_save_offset field in struct ix86_frame.
10054
10055 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
10056
10057         PR tree-optimization/82604
10058         * tree-loop-distribution.c (enum partition_kind): New enum item
10059         PKIND_PARTIAL_MEMSET.
10060         (partition_builtin_p): Support above new enum item.
10061         (generate_code_for_partition): Ditto.
10062         (compute_access_range): Differentiate cases that equality can be
10063         proven at all loops, the innermost loops or no loops.
10064         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10065         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
10066         (finalize_partitions, distribute_loop): Don't fuse partition of
10067         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10068         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10069         parloop is enabled.
10070
10071 2018-01-23  Martin Liska  <mliska@suse.cz>
10072
10073         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10074         order to ignore the predictor.
10075         (PRED_POLYMORPHIC_CALL): Likewise.
10076         (PRED_RECURSIVE_CALL): Likewise.
10077
10078 2018-01-23  Martin Liska  <mliska@suse.cz>
10079
10080         * tree-profile.c (tree_profiling): Print function header to
10081         aware reader which function we are working on.
10082         * value-prof.c (gimple_find_values_to_profile): Do not print
10083         not interesting value histograms.
10084
10085 2018-01-23  Martin Liska  <mliska@suse.cz>
10086
10087         * profile-count.h (enum profile_quality): Add
10088         profile_uninitialized as the first value. Do not number values
10089         as they are zero based.
10090         (profile_count::verify): Update sanity check.
10091         (profile_probability::verify): Likewise.
10092
10093 2018-01-23  Nathan Sidwell  <nathan@acm.org>
10094
10095         * doc/invoke.texi (ffor-scope): Deprecate.
10096
10097 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10098
10099         PR tree-optimization/83510
10100         * domwalk.c (set_all_edges_as_executable): New function.
10101         (dom_walker::dom_walker): Convert bool param
10102         "skip_unreachable_blocks" to enum reachability.  Move setup of
10103         edge flags to set_all_edges_as_executable and only do it when
10104         reachability is REACHABLE_BLOCKS.
10105         * domwalk.h (enum dom_walker::reachability): New enum.
10106         (dom_walker::dom_walker): Convert bool param
10107         "skip_unreachable_blocks" to enum reachability.
10108         (set_all_edges_as_executable): New decl.
10109         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
10110         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10111         "reachability".
10112         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10113         but converting true to REACHABLE_BLOCKS.
10114         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10115         * tree-vrp.c
10116         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10117         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10118         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10119         REACHABLE_BLOCKS.
10120         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10121         if check_all_array_refs will be called.
10122
10123 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10124
10125         * tree.c (selftest::test_location_wrappers): Add more test
10126         coverage.
10127
10128 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10129
10130         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10131         (selftest::test_bit_in_range): Likewise.
10132
10133 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
10134
10135         PR testsuite/83888
10136         * doc/sourcebuild.texi (vect_float): Say that the selector
10137         only describes the situation when -funsafe-math-optimizations is on.
10138         (vect_float_strict): Document.
10139
10140 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
10141
10142         PR tree-optimization/83965
10143         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10144         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10145         instead of checking only for a reduction.
10146         (vect_recog_widen_sum_pattern): Likewise.
10147
10148 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
10149
10150         * predict.c (probably_never_executed): Only use precise profile info.
10151         (compute_function_frequency): Skip after inlining hack since we now
10152         have quality checking.
10153
10154 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
10155
10156         * profile-count.h (profile_probability::very_unlikely,
10157         profile_probability::unlikely, profile_probability::even): Set
10158         precision to guessed.
10159
10160 2018-01-23  Richard Biener  <rguenther@suse.de>
10161
10162         PR tree-optimization/83963
10163         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10164         Properly terminate dominator walk when crossing the exit edge not
10165         when visiting its source block.
10166
10167 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
10168
10169         PR c++/83918
10170         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10171         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10172
10173 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
10174
10175         PR tree-optimization/83957
10176         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
10177         semicolon after for body surrounded by braces.
10178
10179         PR tree-optimization/83081
10180         * profile-count.h (profile_probability::split): New method.
10181         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10182         Use profile_probability::split.
10183         (do_compare_rtx_and_jump): Fix adjustment of probabilities
10184         when splitting a single conditional jump into 2.
10185
10186 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
10187
10188         PR tree-optimization/69452
10189         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10190         decl.
10191
10192 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10193
10194         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10195         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10196         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10197
10198 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10199
10200         * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10201         * config/rl78/rl78.md (movdi): New define_expand.
10202         * config/rl78/rl78.c (rl78_split_movdi): New function.
10203
10204 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
10205
10206         PR target/83862
10207         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10208         no longer used.
10209         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10210         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10211         128-bit to produce an UNSPEC move to get the double word with the
10212         signbit and then a shift directly to do signbit.
10213         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10214         implementation with a new version that just does either a direct
10215         move or a regular move.  Move memory interface to separate insns.
10216         Move insns so they are next to the expander.
10217         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10218         with signbit move.  Split big and little endian case.
10219         (signbit<mode>2_dm_mem_le): Likewise.
10220         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10221         (signbit<mode>2_dm2): Likewise.
10222
10223 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10224
10225         * config/rl78/rl78.md (anddi3): New define_expand.
10226
10227 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10228
10229         * config/rl78/rl78.md (umindi3): New define_expand.
10230
10231 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10232
10233         * config/rl78/rl78.md (smindi3): New define_expand.
10234
10235 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10236
10237         * config/rl78/rl78.md (smaxdi3): New define_expand.
10238
10239 2018-01-22  Carl Love  <cel@us.ibm.com>
10240
10241         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10242         LVX_V1TI): Add macro expansion.
10243         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10244         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10245         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10246         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10247         Change check to determine if the instruction is a byte reversing
10248         entry.  Fix typo in comment.
10249         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10250         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10251         Add def_builtin calls for new builtins.
10252         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10253         Add define_insn expansion.
10254
10255 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10256
10257         * config/rl78/rl78.md (umaxdi3): New define_expand.
10258
10259 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10260
10261         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10262         for non-QImode registers.
10263
10264 2018-01-22  Richard Biener  <rguenther@suse.de>
10265
10266         PR tree-optimization/83963
10267         * graphite-scop-detection.c (scop_detection::get_sese): Delay
10268         including the loop exit block.
10269         (scop_detection::merge_sese): Likewise.
10270         (scop_detection::add_scop): Do it here instead.
10271
10272 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10273
10274         * doc/sourcebuild.texi (arm_softfloat): Document.
10275
10276 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
10277
10278         PR gcc/77734
10279         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10280         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10281         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10282
10283 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10284             David Edelsohn  <dje.gcc@gmail.com>
10285
10286         PR target/83946
10287         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10288         Change "crset eq" to "crset 2".
10289         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10290         (*call_indirect_aix<mode>_nospec): Likewise.
10291         (*call_value_indirect_aix<mode>_nospec): Likewise.
10292         (*call_indirect_elfv2<mode>_nospec): Likewise.
10293         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10294         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10295         change assembly output from . to $.
10296         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10297         (indirect_jump<mode>_nospec): Change assembly output from . to $.
10298         (*tablejump<mode>_internal1_nospec): Likewise.
10299
10300 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
10301
10302         PR target/80870
10303         * config/sh/sh_optimize_sett_clrt.cc:
10304         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10305
10306 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
10307
10308         PR tree-optimization/83940
10309         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10310         offset_dt to vect_constant_def rather than vect_unknown_def_type.
10311         (vect_check_load_store_mask): Add a mask_dt_out parameter and
10312         use it to pass back the definition type.
10313         (vect_check_store_rhs): Likewise rhs_dt_out.
10314         (vect_build_gather_load_calls): Add a mask_dt argument and use
10315         it instead of a call to vect_is_simple_use.
10316         (vectorizable_store): Update calls to vect_check_load_store_mask
10317         and vect_check_store_rhs.  Use the dt returned by the latter instead
10318         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
10319         instead of calls to vect_is_simple_use.  Pass the scalar rather
10320         than the vector operand to vect_is_simple_use when handling
10321         second and subsequent copies of an rhs value.
10322         (vectorizable_load): Update calls to vect_check_load_store_mask
10323         and vect_build_gather_load_calls.  Use the cached mask_dt and
10324         gs_info.offset_dt instead of calls to vect_is_simple_use.
10325
10326 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
10327
10328         PR middle-end/83945
10329         * tree-emutls.c: Include gimplify.h.
10330         (lower_emutls_2): New function.
10331         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10332         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10333         it before further processing.
10334
10335         PR target/83930
10336         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10337         UINTVAL (trueop1) instead of INTVAL (op1).
10338
10339 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
10340
10341         PR debug/81570
10342         PR debug/83728
10343         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10344         INCOMING_FRAME_SP_OFFSET if not defined.
10345         (scan_trace): Add ENTRY argument.  If true and
10346         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10347         emit a note to adjust the CFA offset.
10348         (create_cfi_notes): Adjust scan_trace callers.
10349         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10350         INCOMING_FRAME_SP_OFFSET in the CIE.
10351         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10352         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10353         Likewise.
10354         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10355         * doc/tm.texi: Regenerated.
10356
10357 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
10358
10359         PR rtl-optimization/83147
10360         * lra-constraints.c (remove_inheritance_pseudos): Use
10361         lra_substitute_pseudo_within_insn.
10362
10363 2018-01-19  Tom de Vries  <tom@codesourcery.com>
10364             Cesar Philippidis  <cesar@codesourcery.com>
10365
10366         PR target/83920
10367         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10368
10369 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
10370
10371         PR target/83790
10372         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10373         spaces for function labels.
10374
10375 2018-01-19  Martin Liska  <mliska@suse.cz>
10376
10377         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10378         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10379         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10380         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10381         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10382         (PRED_CONST_RETURN): Change from 69 to 65.
10383         (PRED_NULL_RETURN): Change from 91 to 71.
10384         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10385         (PRED_LOOP_GUARD): Change from 66 to 73.
10386
10387 2018-01-19  Martin Liska  <mliska@suse.cz>
10388
10389         * predict.c (predict_insn_def): Add new assert.
10390         (struct branch_predictor): Change type to signed integer.
10391         (test_prediction_value_range): Amend test to cover
10392         PROB_UNINITIALIZED.
10393         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10394         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10395         (PRED_LOOP_ITERATIONS_MAX): Likewise.
10396         (PRED_LOOP_IV_COMPARE): Likewise.
10397         * predict.h (PROB_UNINITIALIZED): Define new constant.
10398
10399 2018-01-19  Martin Liska  <mliska@suse.cz>
10400
10401         * predict.c (dump_prediction): Add new format for
10402         analyze_brprob.py script which is enabled with -details
10403         suboption.
10404         * profile-count.h (precise_p): New function.
10405
10406 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
10407
10408         PR tree-optimization/83922
10409         * tree-vect-loop.c (vect_verify_full_masking): Return false if
10410         there are no statements that need masking.
10411         (vect_active_double_reduction_p): New function.
10412         (vect_analyze_loop_operations): Use it when handling phis that
10413         are not in the loop header.
10414
10415 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
10416
10417         PR tree-optimization/83914
10418         * tree-vect-loop.c (vectorizable_induction): Don't convert
10419         init_expr or apply the peeling adjustment for inductions
10420         that are nested within the vectorized loop.
10421
10422 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10423
10424         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10425         instead of NEG.
10426
10427 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
10428
10429         PR sanitizer/81715
10430         PR testsuite/83882
10431         * function.h (gimplify_parameters): Add gimple_seq * argument.
10432         * function.c: Include gimple.h and options.h.
10433         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10434         for the added local temporaries if needed.
10435         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10436         if there are any parameter cleanups, wrap whole body into a
10437         try/finally with the cleanups.
10438
10439 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
10440
10441         PR target/82964
10442         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10443         Use GET_MODE_CLASS for scalar floating point.
10444
10445 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
10446
10447         PR ipa/82256
10448         patch by PaX Team
10449         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10450         Fix call of call_cgraph_insertion_hooks.
10451
10452 2018-01-18  Martin Sebor  <msebor@redhat.com>
10453
10454         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10455
10456 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
10457
10458         PR ipa/83619
10459         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10460         frequencies.
10461
10462 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
10463
10464         PR other/70268
10465         * common.opt: (-ffile-prefix-map): New option.
10466         * opts.c (common_handle_option): Defer it.
10467         * opts-global.c (handle_common_deferred_options): Handle it.
10468         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10469         * file-prefix-map.h: New file.
10470         (remap_debug_filename, add_debug_prefix_map): ...here.
10471         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10472         * final.c (debug_prefix_map, add_debug_prefix_map
10473         remap_debug_filename): Move to...
10474         * file-prefix-map.c: New file.
10475         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10476         generalize, get rid of alloca(), use strrchr() instead of strchr().
10477         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10478         Implement in terms of add_prefix_map().
10479         (remap_macro_filename, remap_debug_filename): Implement in term of
10480         remap_filename().
10481         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10482         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10483         * dbxout.c: Include file-prefix-map.h.
10484         * varasm.c: Likewise.
10485         * vmsdbgout.c: Likewise.
10486         * xcoffout.c: Likewise.
10487         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10488         * doc/cppopts.texi (-fmacro-prefix-map): Document.
10489         * doc/invoke.texi (-ffile-prefix-map): Document.
10490         (-fdebug-prefix-map): Update description.
10491
10492 2018-01-18  Martin Liska  <mliska@suse.cz>
10493
10494         * config/i386/i386.c (indirect_thunk_name): Document that also
10495         lfence is emitted.
10496         (output_indirect_thunk): Document why both instructions
10497         (pause and lfence) are generated.
10498
10499 2018-01-18  Richard Biener  <rguenther@suse.de>
10500
10501         PR tree-optimization/83887
10502         * graphite-scop-detection.c
10503         (scop_detection::get_nearest_dom_with_single_entry): Remove.
10504         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10505         (scop_detection::merge_sese): Re-implement with a flood-fill
10506         algorithm that properly finds a SESE region if it exists.
10507
10508 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
10509
10510         PR c/61240
10511         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10512         pointer_diff optimizations use view_convert instead of convert.
10513
10514 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10515
10516         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10517         Generate different code for -mno-speculate-indirect-jumps.
10518         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10519         (*call_indirect_aix<mode>): Disable for
10520         -mno-speculate-indirect-jumps.
10521         (*call_indirect_aix<mode>_nospec): New define_insn.
10522         (*call_value_indirect_aix<mode>): Disable for
10523         -mno-speculate-indirect-jumps.
10524         (*call_value_indirect_aix<mode>_nospec): New define_insn.
10525         (*sibcall_nonlocal_sysv<mode>): Generate different code for
10526         -mno-speculate-indirect-jumps.
10527         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10528
10529 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
10530
10531         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10532         long double type, set the flags for noting the default long double
10533         type, even if we don't pass or return a long double type.
10534
10535 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
10536
10537         PR ipa/83051
10538         * ipa-inline.c (flatten_function): Do not overwrite final inlining
10539         failure.
10540
10541 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
10542
10543         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10544         support for merge[hl].
10545         (fold_mergehl_helper): New helper function.
10546         (tree-vector-builder.h): New #include for tree_vector_builder usage.
10547         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10548         (altivec_vmrglw_direct): Add xxmrglw insn.
10549
10550 2018-01-17  Andrew Waterman  <andrew@sifive.com>
10551
10552         * config/riscv/riscv.c (riscv_conditional_register_usage): If
10553         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10554
10555 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
10556
10557         PR lto/83121
10558         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10559         call the lto_location_cache before reading the
10560         DECL_SOURCE_LOCATION of the types.
10561
10562 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
10563             Richard Sandiford  <richard.sandiford@linaro.org>
10564
10565         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10566         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10567         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10568         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10569         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10570         Add declaration.
10571         * config/aarch64/constraints.md (aarch64_movti_operand):
10572         Limit immediates.
10573         * config/aarch64/predicates.md (Uti): Add new constraint.
10574
10575 2018-01-17 Carl Love  <cel@us.ibm.com>
10576
10577         * config/rs6000/vsx.md (define_expand xl_len_r,
10578         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10579         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10580         lxvll.
10581         (define_expand, define_insn): Move the shift left from  the
10582         define_insn to the define_expand for lxvl and stxvl instructions.
10583         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10584         and XL_LEN_R definitions to PURE.
10585
10586 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
10587
10588         * config/i386/i386.c (indirect_thunk_name): Declare regno
10589         as unsigned int.  Compare regno with INVALID_REGNUM.
10590         (output_indirect_thunk): Ditto.
10591         (output_indirect_thunk_function): Ditto.
10592         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
10593         in the call to output_indirect_thunk_function.
10594
10595 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
10596
10597         PR middle-end/83884
10598         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10599         rather than the size of inner_type to determine the stack slot size
10600         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10601
10602 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
10603
10604         PR target/83546
10605         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10606         to PTA_SILVERMONT.
10607
10608 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
10609
10610         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10611         endian Linux systems to optionally enable multilibs for selecting
10612         the long double type if the user configured an explicit type.
10613         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10614         have no long double multilibs if not defined.
10615         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10616         warn if the user used -mabi={ieee,ibm}longdouble and we built
10617         multilibs for long double.
10618         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10619         appropriate multilib option.
10620         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10621         multilib options.
10622         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10623         for building long double multilibs.
10624         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10625
10626 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
10627
10628         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10629         copies.
10630
10631         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10632         64 bits.
10633         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10634         128 bits.
10635
10636         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10637         variables.
10638
10639         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10640         return value.
10641
10642 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
10643
10644         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10645         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10646
10647 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10648
10649         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10650         different rtl trees depending on TARGET_64BIT.
10651         (rs6000_gen_lvx): Likewise.
10652
10653 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
10654
10655         * config/visium/visium.md (nop): Tweak comment.
10656         (hazard_nop): Likewise.
10657
10658 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10659
10660         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10661         -mspeculate-indirect-jumps.
10662         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10663         for -mno-speculate-indirect-jumps.
10664         (*call_indirect_elfv2<mode>_nospec): New define_insn.
10665         (*call_value_indirect_elfv2<mode>): Disable for
10666         -mno-speculate-indirect-jumps.
10667         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10668         (indirect_jump): Emit different RTL for
10669         -mno-speculate-indirect-jumps.
10670         (*indirect_jump<mode>): Disable for
10671         -mno-speculate-indirect-jumps.
10672         (*indirect_jump<mode>_nospec): New define_insn.
10673         (tablejump): Emit different RTL for
10674         -mno-speculate-indirect-jumps.
10675         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10676         (tablejumpsi_nospec): New define_expand.
10677         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10678         (tablejumpdi_nospec): New define_expand.
10679         (*tablejump<mode>_internal1): Disable for
10680         -mno-speculate-indirect-jumps.
10681         (*tablejump<mode>_internal1_nospec): New define_insn.
10682         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10683         option.
10684
10685 2018-01-16  Artyom Skrobov tyomitch@gmail.com
10686
10687         * caller-save.c (insert_save): Drop unnecessary parameter.  All
10688         callers updated.
10689
10690 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10691             Richard Biener  <rguenth@suse.de>
10692
10693         PR libgomp/83590
10694         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10695         return early, inline manually is_gimple_sizepos.  Make sure if we
10696         call gimplify_expr we don't end up with a gimple constant.
10697         * tree.c (variably_modified_type_p): Don't return true for
10698         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
10699         * gimplify.h (is_gimple_sizepos): Remove.
10700
10701 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10702
10703         PR tree-optimization/83857
10704         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10705         vectorizable_live_operation for pure SLP statements.
10706         (vectorizable_live_operation): Handle PHIs.
10707
10708 2018-01-16  Richard Biener  <rguenther@suse.de>
10709
10710         PR tree-optimization/83867
10711         * tree-vect-stmts.c (vect_transform_stmt): Precompute
10712         nested_in_vect_loop_p since the scalar stmt may get invalidated.
10713
10714 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10715
10716         PR c/83844
10717         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10718         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10719         If off is not INTEGER_CST, issue a may not be aligned warning
10720         rather than isn't aligned.  Use isn%'t rather than isn't.
10721         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10722         into MULT_EXPR.
10723         <case MULT_EXPR>: Improve the case when bottom and one of the
10724         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10725         operand, in that case check if the other operand is multiple of
10726         bottom divided by the INTEGER_CST operand.
10727
10728 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10729
10730         PR target/83858
10731         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10732         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10733         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10734         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10735         * config/pa/pa.c (pa_function_arg_advance): Likewise.
10736         (pa_function_arg, pa_arg_partial_bytes): Likewise.
10737         (pa_function_arg_size): New function.
10738
10739 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10740
10741         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10742         in a separate statement.
10743
10744 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10745
10746         PR tree-optimization/83847
10747         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10748         group gathers and scatters.
10749
10750 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10751
10752         PR rtl-optimization/86620
10753         * params.def (max-sched-ready-insns): Bump minimum value to 1.
10754
10755         PR rtl-optimization/83213
10756         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10757         to last if both are JUMP_INSNs.
10758
10759         PR tree-optimization/83843
10760         * gimple-ssa-store-merging.c
10761         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10762         store_immediate_info for bswap/nop orig_stores.
10763
10764 2018-01-15  Andrew Waterman  <andrew@sifive.com>
10765
10766         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10767         !TARGET_MUL.
10768         <UDIV>: Increase cost if !TARGET_DIV.
10769
10770 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
10771
10772         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10773         (define_attr "cr_logical_3op"): New.
10774         (cceq_ior_compare): Adjust.
10775         (cceq_ior_compare_complement): Adjust.
10776         (*cceq_rev_compare): Adjust.
10777         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10778         (is_cracked_insn): Adjust.
10779         (insn_must_be_first_in_group): Adjust.
10780         * config/rs6000/40x.md: Adjust.
10781         * config/rs6000/440.md: Adjust.
10782         * config/rs6000/476.md: Adjust.
10783         * config/rs6000/601.md: Adjust.
10784         * config/rs6000/603.md: Adjust.
10785         * config/rs6000/6xx.md: Adjust.
10786         * config/rs6000/7450.md: Adjust.
10787         * config/rs6000/7xx.md: Adjust.
10788         * config/rs6000/8540.md: Adjust.
10789         * config/rs6000/cell.md: Adjust.
10790         * config/rs6000/e300c2c3.md: Adjust.
10791         * config/rs6000/e500mc.md: Adjust.
10792         * config/rs6000/e500mc64.md: Adjust.
10793         * config/rs6000/e5500.md: Adjust.
10794         * config/rs6000/e6500.md: Adjust.
10795         * config/rs6000/mpc.md: Adjust.
10796         * config/rs6000/power4.md: Adjust.
10797         * config/rs6000/power5.md: Adjust.
10798         * config/rs6000/power6.md: Adjust.
10799         * config/rs6000/power7.md: Adjust.
10800         * config/rs6000/power8.md: Adjust.
10801         * config/rs6000/power9.md: Adjust.
10802         * config/rs6000/rs64.md: Adjust.
10803         * config/rs6000/titan.md: Adjust.
10804
10805 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10806
10807         * config/i386/predicates.md (indirect_branch_operand): Rewrite
10808         ix86_indirect_branch_register logic.
10809
10810 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10811
10812         * config/i386/constraints.md (Bs): Update
10813         ix86_indirect_branch_register check.  Don't check
10814         ix86_indirect_branch_register with GOT_memory_operand.
10815         (Bw): Likewise.
10816         * config/i386/predicates.md (GOT_memory_operand): Don't check
10817         ix86_indirect_branch_register here.
10818         (GOT32_symbol_operand): Likewise.
10819
10820 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10821
10822         * config/i386/predicates.md (constant_call_address_operand):
10823         Rewrite ix86_indirect_branch_register logic.
10824         (sibcall_insn_operand): Likewise.
10825
10826 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10827
10828         * config/i386/constraints.md (Bs): Replace
10829         ix86_indirect_branch_thunk_register with
10830         ix86_indirect_branch_register.
10831         (Bw): Likewise.
10832         * config/i386/i386.md (indirect_jump): Likewise.
10833         (tablejump): Likewise.
10834         (*sibcall_memory): Likewise.
10835         (*sibcall_value_memory): Likewise.
10836         Peepholes of indirect call and jump via memory: Likewise.
10837         * config/i386/i386.opt: Likewise.
10838         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10839         (GOT_memory_operand): Likewise.
10840         (call_insn_operand): Likewise.
10841         (sibcall_insn_operand): Likewise.
10842         (GOT32_symbol_operand): Likewise.
10843
10844 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10845
10846         PR middle-end/83837
10847         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10848         type rather than type addr's type points to.
10849         (expand_omp_atomic_mutex): Likewise.
10850         (expand_omp_atomic): Likewise.
10851
10852 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10853
10854         PR target/83839
10855         * config/i386/i386.c (output_indirect_thunk_function): Use
10856         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10857         for  __x86_return_thunk.
10858
10859 2018-01-15  Richard Biener  <rguenther@suse.de>
10860
10861         PR middle-end/83850
10862         * expmed.c (extract_bit_field_1): Fix typo.
10863
10864 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10865
10866         PR target/83687
10867         * config/arm/iterators.md (VF): New mode iterator.
10868         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10869         Remove integer-related logic from pattern.
10870         (neon_vabd<mode>_3): Likewise.
10871
10872 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10873
10874         PR middle-end/82694
10875         * common.opt (fstrict-overflow): No longer an alias.
10876         (fwrapv-pointer): New option.
10877         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10878         also for pointer types based on flag_wrapv_pointer.
10879         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10880         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10881         opts->x_flag_wrapv got set.
10882         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10883         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10884         POINTER_TYPE_OVERFLOW_UNDEFINED.
10885         * match.pd: Likewise in address comparison pattern.
10886         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10887
10888 2018-01-15  Richard Biener  <rguenther@suse.de>
10889
10890         PR lto/83804
10891         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10892         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
10893         Reset type names to their identifier if their TYPE_DECL doesn't
10894         have linkage (and thus is used for ODR and devirt).
10895         (save_debug_info_for_decl): Remove.
10896         (save_debug_info_for_type): Likewise.
10897         (add_tree_to_fld_list): Adjust.
10898         * tree-pretty-print.c (dump_generic_node): Make dumping of
10899         type names more robust.
10900
10901 2018-01-15  Richard Biener  <rguenther@suse.de>
10902
10903         * BASE-VER: Bump to 8.0.1.
10904
10905 2018-01-14  Martin Sebor  <msebor@redhat.com>
10906
10907         PR other/83508
10908         * builtins.c (check_access): Avoid warning when the no-warning bit
10909         is set.
10910
10911 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
10912
10913         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10914         * ira-color (allocno_hard_regs_compare): Likewise.
10915
10916 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
10917
10918         PR target/83013
10919         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10920         Use .pushsection/.popsection.
10921
10922 2018-01-14  Martin Sebor  <msebor@redhat.com>
10923
10924         PR c++/81327
10925         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10926
10927 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
10928
10929         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10930         entry from extra_headers.
10931         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10932         extra_headers, make the list bitwise identical to the i?86-*-* one.
10933
10934 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10935
10936         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10937         -mcmodel=large with -mindirect-branch=thunk,
10938         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10939         -mfunction-return=thunk-extern.
10940         * doc/invoke.texi: Document -mcmodel=large is incompatible with
10941         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10942         -mfunction-return=thunk and -mfunction-return=thunk-extern.
10943
10944 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10945
10946         * config/i386/i386.c (print_reg): Print the name of the full
10947         integer register without '%'.
10948         (ix86_print_operand): Handle 'V'.
10949          * doc/extend.texi: Document 'V' modifier.
10950
10951 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10952
10953         * config/i386/constraints.md (Bs): Disallow memory operand for
10954         -mindirect-branch-register.
10955         (Bw): Likewise.
10956         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10957         (GOT_memory_operand): Likewise.
10958         (call_insn_operand): Likewise.
10959         (sibcall_insn_operand): Likewise.
10960         (GOT32_symbol_operand): Likewise.
10961         * config/i386/i386.md (indirect_jump): Call convert_memory_address
10962         for -mindirect-branch-register.
10963         (tablejump): Likewise.
10964         (*sibcall_memory): Likewise.
10965         (*sibcall_value_memory): Likewise.
10966         Disallow peepholes of indirect call and jump via memory for
10967         -mindirect-branch-register.
10968         (*call_pop): Replace m with Bw.
10969         (*call_value_pop): Likewise.
10970         (*sibcall_pop_memory): Replace m with Bs.
10971         * config/i386/i386.opt (mindirect-branch-register): New option.
10972         * doc/invoke.texi: Document -mindirect-branch-register option.
10973
10974 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10975
10976         * config/i386/i386-protos.h (ix86_output_function_return): New.
10977         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10978         set function_return_type.
10979         (indirect_thunk_name): Add ret_p to indicate thunk for function
10980         return.
10981         (output_indirect_thunk_function): Pass false to
10982         indirect_thunk_name.
10983         (ix86_output_indirect_branch_via_reg): Likewise.
10984         (ix86_output_indirect_branch_via_push): Likewise.
10985         (output_indirect_thunk_function): Create alias for function
10986         return thunk if regno < 0.
10987         (ix86_output_function_return): New function.
10988         (ix86_handle_fndecl_attribute): Handle function_return.
10989         (ix86_attribute_table): Add function_return.
10990         * config/i386/i386.h (machine_function): Add
10991         function_return_type.
10992         * config/i386/i386.md (simple_return_internal): Use
10993         ix86_output_function_return.
10994         (simple_return_internal_long): Likewise.
10995         * config/i386/i386.opt (mfunction-return=): New option.
10996         (indirect_branch): Mention -mfunction-return=.
10997         * doc/extend.texi: Document function_return function attribute.
10998         * doc/invoke.texi: Document -mfunction-return= option.
10999
11000 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
11001
11002         * config/i386/i386-opts.h (indirect_branch): New.
11003         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11004         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11005         with local indirect jump when converting indirect call and jump.
11006         (ix86_set_indirect_branch_type): New.
11007         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11008         (indirectlabelno): New.
11009         (indirect_thunk_needed): Likewise.
11010         (indirect_thunk_bnd_needed): Likewise.
11011         (indirect_thunks_used): Likewise.
11012         (indirect_thunks_bnd_used): Likewise.
11013         (INDIRECT_LABEL): Likewise.
11014         (indirect_thunk_name): Likewise.
11015         (output_indirect_thunk): Likewise.
11016         (output_indirect_thunk_function): Likewise.
11017         (ix86_output_indirect_branch_via_reg): Likewise.
11018         (ix86_output_indirect_branch_via_push): Likewise.
11019         (ix86_output_indirect_branch): Likewise.
11020         (ix86_output_indirect_jmp): Likewise.
11021         (ix86_code_end): Call output_indirect_thunk_function if needed.
11022         (ix86_output_call_insn): Call ix86_output_indirect_branch if
11023         needed.
11024         (ix86_handle_fndecl_attribute): Handle indirect_branch.
11025         (ix86_attribute_table): Add indirect_branch.
11026         * config/i386/i386.h (machine_function): Add indirect_branch_type
11027         and has_local_indirect_jump.
11028         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11029         to true.
11030         (tablejump): Likewise.
11031         (*indirect_jump): Use ix86_output_indirect_jmp.
11032         (*tablejump_1): Likewise.
11033         (simple_return_indirect_internal): Likewise.
11034         * config/i386/i386.opt (mindirect-branch=): New option.
11035         (indirect_branch): New.
11036         (keep): Likewise.
11037         (thunk): Likewise.
11038         (thunk-inline): Likewise.
11039         (thunk-extern): Likewise.
11040         * doc/extend.texi: Document indirect_branch function attribute.
11041         * doc/invoke.texi: Document -mindirect-branch= option.
11042
11043 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
11044
11045         PR ipa/83051
11046         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11047
11048 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
11049
11050         * ipa-inline.c (want_inline_small_function_p): Return false if
11051         inlining has already failed with CIF_FINAL_ERROR.
11052         (update_caller_keys): Call want_inline_small_function_p before
11053         can_inline_edge_p.
11054         (update_callee_keys): Likewise.
11055
11056 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11057
11058         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11059         New function.
11060         (rs6000_quadword_masked_address_p): Likewise.
11061         (quad_aligned_load_p): Likewise.
11062         (quad_aligned_store_p): Likewise.
11063         (const_load_sequence_p): Add comment to describe the outer-most loop.
11064         (mimic_memory_attributes_and_flags): New function.
11065         (rs6000_gen_stvx): Likewise.
11066         (replace_swapped_aligned_store): Likewise.
11067         (rs6000_gen_lvx): Likewise.
11068         (replace_swapped_aligned_load): Likewise.
11069         (replace_swapped_load_constant): Capitalize argument name in
11070         comment describing this function.
11071         (rs6000_analyze_swaps): Add a third pass to search for vector loads
11072         and stores that access quad-word aligned addresses and replace
11073         with stvx or lvx instructions when appropriate.
11074         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11075         New function prototype.
11076         (rs6000_quadword_masked_address_p): Likewise.
11077         (rs6000_gen_lvx): Likewise.
11078         (rs6000_gen_stvx): Likewise.
11079         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11080         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11081         when memory address is aligned.
11082         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11083         this split to select lvx instruction when memory address is aligned.
11084         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11085         instruction when memory address is aligned.
11086         (*vsx_le_perm_load_v16qi): Likewise.
11087         (four unnamed splitters): Modify to select the stvx instruction
11088         when memory is aligned.
11089
11090 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
11091
11092         * predict.c (determine_unlikely_bbs): Handle correctly BBs
11093         which appears in the queue multiple times.
11094
11095 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11096             Alan Hayward  <alan.hayward@arm.com>
11097             David Sherwood  <david.sherwood@arm.com>
11098
11099         * tree-vectorizer.h (vec_lower_bound): New structure.
11100         (_loop_vec_info): Add check_nonzero and lower_bounds.
11101         (LOOP_VINFO_CHECK_NONZERO): New macro.
11102         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11103         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11104         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11105         fields.  Make seg_len the distance travelled, not including the
11106         access size.
11107         (dr_direction_indicator): Declare.
11108         (dr_zero_step_indicator): Likewise.
11109         (dr_known_forward_stride_p): Likewise.
11110         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11111         tree-ssanames.h.
11112         (runtime_alias_check_p): Allow runtime alias checks with
11113         variable strides.
11114         (operator ==): Compare access_size and align.
11115         (prune_runtime_alias_test_list): Rework for new distinction between
11116         the access_size and seg_len.
11117         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
11118         segment lengths.
11119         (get_segment_min_max): New function.
11120         (create_intersect_range_checks): Use it.
11121         (dr_step_indicator): New function.
11122         (dr_direction_indicator): Likewise.
11123         (dr_zero_step_indicator): Likewise.
11124         (dr_known_forward_stride_p): Likewise.
11125         * tree-loop-distribution.c (data_ref_segment_size): Return
11126         DR_STEP * (niters - 1).
11127         (compute_alias_check_pairs): Update call to the dr_with_seg_len
11128         constructor.
11129         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11130         (vect_preserves_scalar_order_p): New function, split out from...
11131         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
11132         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11133         (vect_vfa_access_size): New function.
11134         (vect_vfa_align): Likewise.
11135         (vect_compile_time_alias): Take access_size_a and access_b arguments.
11136         (dump_lower_bound): New function.
11137         (vect_check_lower_bound): Likewise.
11138         (vect_small_gap_p): Likewise.
11139         (vectorizable_with_step_bound_p): Likewise.
11140         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11141         depencies if the vectorization factor is 1.  Convert the checks
11142         for nonzero steps into checks on the bounds of DR_STEP.  Try using
11143         a bunds check for variable steps if the minimum required step is
11144         relatively small. Update calls to the dr_with_seg_len
11145         constructor and to vect_compile_time_alias.
11146         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11147         function.
11148         (vect_loop_versioning): Call it.
11149         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11150         when retrying.
11151         (vect_estimate_min_profitable_iters): Account for any bounds checks.
11152
11153 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11154             Alan Hayward  <alan.hayward@arm.com>
11155             David Sherwood  <david.sherwood@arm.com>
11156
11157         * doc/sourcebuild.texi (vect_scatter_store): Document.
11158         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11159         optabs.
11160         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11161         Document.
11162         * genopinit.c (main): Add supports_vec_scatter_store and
11163         supports_vec_scatter_store_cached to target_optabs.
11164         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11165         IFN_MASK_SCATTER_STORE.
11166         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11167         functions.
11168         * internal-fn.h (internal_store_fn_p): Declare.
11169         (internal_fn_stored_value_index): Likewise.
11170         * internal-fn.c (scatter_store_direct): New macro.
11171         (expand_scatter_store_optab_fn): New function.
11172         (direct_scatter_store_optab_supported_p): New macro.
11173         (internal_store_fn_p): New function.
11174         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11175         IFN_MASK_SCATTER_STORE.
11176         (internal_fn_mask_index): Likewise.
11177         (internal_fn_stored_value_index): New function.
11178         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11179         for scatter stores.
11180         * optabs-query.h (supports_vec_scatter_store_p): Declare.
11181         * optabs-query.c (supports_vec_scatter_store_p): New function.
11182         * tree-vectorizer.h (vect_get_store_rhs): Declare.
11183         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11184         true for scatter stores.
11185         (vect_gather_scatter_fn_p): Handle scatter stores too.
11186         (vect_check_gather_scatter): Consider using scatter stores if
11187         supports_vec_scatter_store_p.
11188         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11189         scatter stores too.
11190         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11191         internal_fn_stored_value_index.
11192         (check_load_store_masking): Handle scatter stores too.
11193         (vect_get_store_rhs): Make public.
11194         (vectorizable_call): Use internal_store_fn_p.
11195         (vectorizable_store): Handle scatter store internal functions.
11196         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11197         when deciding whether the end of the group has been reached.
11198         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11199         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11200         (mask_scatter_store<mode>): New insns.
11201
11202 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11203             Alan Hayward  <alan.hayward@arm.com>
11204             David Sherwood  <david.sherwood@arm.com>
11205
11206         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11207         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11208         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11209         function.
11210         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11211         Use vect_truncate_gather_scatter_offset if we can't treat the
11212         operation as a normal gather load or scatter store.
11213         (get_group_load_store_type): Take the gather_scatter_info
11214         as argument.  Try using a gather load or scatter store for
11215         single-element groups.
11216         (get_load_store_type): Update calls to get_group_load_store_type
11217         and vect_use_strided_gather_scatters_p.
11218
11219 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11220             Alan Hayward  <alan.hayward@arm.com>
11221             David Sherwood  <david.sherwood@arm.com>
11222
11223         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11224         optional tree argument.
11225         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11226         null target hooks.
11227         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11228         but continue to use the current value as a fallback.
11229         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11230         to compare the updates.
11231         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11232         (get_load_store_type): Use it when handling a strided access.
11233         (vect_get_strided_load_store_ops): New function.
11234         (vect_get_data_ptr_increment): Likewise.
11235         (vectorizable_load): Handle strided gather loads.  Always pass
11236         a step to vect_create_data_ref_ptr and bump_vector_ptr.
11237
11238 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11239             Alan Hayward  <alan.hayward@arm.com>
11240             David Sherwood  <david.sherwood@arm.com>
11241
11242         * doc/md.texi (gather_load@var{m}): Document.
11243         (mask_gather_load@var{m}): Likewise.
11244         * genopinit.c (main): Add supports_vec_gather_load and
11245         supports_vec_gather_load_cached to target_optabs.
11246         * optabs-tree.c (init_tree_optimization_optabs): Use
11247         ggc_cleared_alloc to allocate target_optabs.
11248         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11249         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11250         functions.
11251         * internal-fn.h (internal_load_fn_p): Declare.
11252         (internal_gather_scatter_fn_p): Likewise.
11253         (internal_fn_mask_index): Likewise.
11254         (internal_gather_scatter_fn_supported_p): Likewise.
11255         * internal-fn.c (gather_load_direct): New macro.
11256         (expand_gather_load_optab_fn): New function.
11257         (direct_gather_load_optab_supported_p): New macro.
11258         (direct_internal_fn_optab): New function.
11259         (internal_load_fn_p): Likewise.
11260         (internal_gather_scatter_fn_p): Likewise.
11261         (internal_fn_mask_index): Likewise.
11262         (internal_gather_scatter_fn_supported_p): Likewise.
11263         * optabs-query.c (supports_at_least_one_mode_p): New function.
11264         (supports_vec_gather_load_p): Likewise.
11265         * optabs-query.h (supports_vec_gather_load_p): Declare.
11266         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11267         and memory_type field.
11268         (NUM_PATTERNS): Bump to 15.
11269         * tree-vect-data-refs.c: Include internal-fn.h.
11270         (vect_gather_scatter_fn_p): New function.
11271         (vect_describe_gather_scatter_call): Likewise.
11272         (vect_check_gather_scatter): Try using internal functions for
11273         gather loads.  Recognize existing calls to a gather load function.
11274         (vect_analyze_data_refs): Consider using gather loads if
11275         supports_vec_gather_load_p.
11276         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11277         (vect_get_gather_scatter_offset_type): Likewise.
11278         (vect_convert_mask_for_vectype): Likewise.
11279         (vect_add_conversion_to_patterm): Likewise.
11280         (vect_try_gather_scatter_pattern): Likewise.
11281         (vect_recog_gather_scatter_pattern): New pattern recognizer.
11282         (vect_vect_recog_func_ptrs): Add it.
11283         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11284         internal_fn_mask_index and internal_gather_scatter_fn_p.
11285         (check_load_store_masking): Take the gather_scatter_info as an
11286         argument and handle gather loads.
11287         (vect_get_gather_scatter_ops): New function.
11288         (vectorizable_call): Check internal_load_fn_p.
11289         (vectorizable_load): Likewise.  Handle gather load internal
11290         functions.
11291         (vectorizable_store): Update call to check_load_store_masking.
11292         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11293         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11294         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11295         (aarch64_gather_scale_operand_d): New predicates.
11296         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11297         (mask_gather_load<mode>): New insns.
11298
11299 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11300             Alan Hayward  <alan.hayward@arm.com>
11301             David Sherwood  <david.sherwood@arm.com>
11302
11303         * optabs.def (fold_left_plus_optab): New optab.
11304         * doc/md.texi (fold_left_plus_@var{m}): Document.
11305         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11306         * internal-fn.c (fold_left_direct): Define.
11307         (expand_fold_left_optab_fn): Likewise.
11308         (direct_fold_left_optab_supported_p): Likewise.
11309         * fold-const-call.c (fold_const_fold_left): New function.
11310         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11311         * tree-parloops.c (valid_reduction_p): New function.
11312         (gather_scalar_reductions): Use it.
11313         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11314         (vect_finish_replace_stmt): Declare.
11315         * tree-vect-loop.c (fold_left_reduction_fn): New function.
11316         (needs_fold_left_reduction_p): New function, split out from...
11317         (vect_is_simple_reduction): ...here.  Accept reductions that
11318         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11319         (vect_force_simple_reduction): Also store the reduction type in
11320         the assignment's STMT_VINFO_REDUC_TYPE.
11321         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11322         (merge_with_identity): New function.
11323         (vect_expand_fold_left): Likewise.
11324         (vectorize_fold_left_reduction): Likewise.
11325         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
11326         scalar phi in place for it.  Check for target support and reject
11327         cases that would reassociate the operation.  Defer the transform
11328         phase to vectorize_fold_left_reduction.
11329         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11330         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11331         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11332
11333 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11334
11335         * tree-if-conv.c (predicate_mem_writes): Remove redundant
11336         call to ifc_temp_var.
11337
11338 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11339             Alan Hayward  <alan.hayward@arm.com>
11340             David Sherwood  <david.sherwood@arm.com>
11341
11342         * target.def (legitimize_address_displacement): Take the original
11343         offset as a poly_int.
11344         * targhooks.h (default_legitimize_address_displacement): Update
11345         accordingly.
11346         * targhooks.c (default_legitimize_address_displacement): Likewise.
11347         * doc/tm.texi: Regenerate.
11348         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11349         as an argument, moving assert of ad->disp == ad->disp_term to...
11350         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
11351         Try calling targetm.legitimize_address_displacement before expanding
11352         the address rather than afterwards, and adjust for the new interface.
11353         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11354         Match the new hook interface.  Handle SVE addresses.
11355         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11356         new hook interface.
11357
11358 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11359
11360         * Makefile.in (OBJS): Add early-remat.o.
11361         * target.def (select_early_remat_modes): New hook.
11362         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11363         * doc/tm.texi: Regenerate.
11364         * targhooks.h (default_select_early_remat_modes): Declare.
11365         * targhooks.c (default_select_early_remat_modes): New function.
11366         * timevar.def (TV_EARLY_REMAT): New timevar.
11367         * passes.def (pass_early_remat): New pass.
11368         * tree-pass.h (make_pass_early_remat): Declare.
11369         * early-remat.c: New file.
11370         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11371         function.
11372         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11373
11374 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11375             Alan Hayward  <alan.hayward@arm.com>
11376             David Sherwood  <david.sherwood@arm.com>
11377
11378         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11379         vfm1 with a bound_epilog parameter.
11380         (vect_do_peeling): Update calls accordingly, and move the prologue
11381         call earlier in the function.  Treat the base bound_epilog as 0 for
11382         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
11383         this base when peeling for gaps.
11384         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11385         with fully-masked loops.
11386         (vect_estimate_min_profitable_iters): Handle the single peeled
11387         iteration in that case.
11388
11389 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11390             Alan Hayward  <alan.hayward@arm.com>
11391             David Sherwood  <david.sherwood@arm.com>
11392
11393         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11394         single-element interleaving even if the size is not a power of 2.
11395         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11396         accesses for single-element interleaving if the group size is
11397         not a power of 2.
11398
11399 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11400             Alan Hayward  <alan.hayward@arm.com>
11401             David Sherwood  <david.sherwood@arm.com>
11402
11403         * doc/md.texi (fold_extract_last_@var{m}): Document.
11404         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11405         * optabs.def (fold_extract_last_optab): New optab.
11406         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11407         * internal-fn.c (fold_extract_direct): New macro.
11408         (expand_fold_extract_optab_fn): Likewise.
11409         (direct_fold_extract_optab_supported_p): Likewise.
11410         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11411         * tree-vect-loop.c (vect_model_reduction_cost): Handle
11412         EXTRACT_LAST_REDUCTION.
11413         (get_initial_def_for_reduction): Do not create an initial vector
11414         for EXTRACT_LAST_REDUCTION reductions.
11415         (vectorizable_reduction): Leave the scalar phi in place for
11416         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
11417         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
11418         epilogue code for EXTRACT_LAST_REDUCTION and defer the
11419         transform phase to vectorizable_condition.
11420         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11421         split out from...
11422         (vect_finish_stmt_generation): ...here.
11423         (vect_finish_replace_stmt): New function.
11424         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11425         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11426         pattern.
11427         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11428
11429 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11430             Alan Hayward  <alan.hayward@arm.com>
11431             David Sherwood  <david.sherwood@arm.com>
11432
11433         * doc/md.texi (extract_last_@var{m}): Document.
11434         * optabs.def (extract_last_optab): New optab.
11435         * internal-fn.def (EXTRACT_LAST): New internal function.
11436         * internal-fn.c (cond_unary_direct): New macro.
11437         (expand_cond_unary_optab_fn): Likewise.
11438         (direct_cond_unary_optab_supported_p): Likewise.
11439         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11440         loops using EXTRACT_LAST.
11441         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11442         (extract_last_<mode>): ...this optab.
11443         (vec_extract<mode><Vel>): Update accordingly.
11444
11445 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11446             Alan Hayward  <alan.hayward@arm.com>
11447             David Sherwood  <david.sherwood@arm.com>
11448
11449         * target.def (empty_mask_is_expensive): New hook.
11450         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11451         * doc/tm.texi: Regenerate.
11452         * targhooks.h (default_empty_mask_is_expensive): Declare.
11453         * targhooks.c (default_empty_mask_is_expensive): New function.
11454         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11455         if the target says that empty masks are expensive.
11456         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11457         New function.
11458         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11459
11460 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11461             Alan Hayward  <alan.hayward@arm.com>
11462             David Sherwood  <david.sherwood@arm.com>
11463
11464         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11465         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11466         (vect_use_loop_mask_for_alignment_p): New function.
11467         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11468         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11469         niters_skip argument.  Make sure that the first niters_skip elements
11470         of the first iteration are inactive.
11471         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11472         Update call to vect_set_loop_masks_directly.
11473         (get_misalign_in_elems): New function, split out from...
11474         (vect_gen_prolog_loop_niters): ...here.
11475         (vect_update_init_of_dr): Take a code argument that specifies whether
11476         the adjustment should be added or subtracted.
11477         (vect_update_init_of_drs): Likewise.
11478         (vect_prepare_for_masked_peels): New function.
11479         (vect_do_peeling): Skip prologue peeling if we're using a mask
11480         instead.  Update call to vect_update_inits_of_drs.
11481         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11482         mask_skip_niters.
11483         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11484         alignment.  Do not include the number of peeled iterations in
11485         the minimum threshold in that case.
11486         (vectorizable_induction): Adjust the start value down by
11487         LOOP_VINFO_MASK_SKIP_NITERS iterations.
11488         (vect_transform_loop): Call vect_prepare_for_masked_peels.
11489         Take the number of skipped iterations into account when calculating
11490         the loop bounds.
11491         * tree-vect-stmts.c (vect_gen_while_not): New function.
11492
11493 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11494             Alan Hayward  <alan.hayward@arm.com>
11495             David Sherwood  <david.sherwood@arm.com>
11496
11497         * doc/sourcebuild.texi (vect_fully_masked): Document.
11498         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11499         default value to 0.
11500         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11501         split out from...
11502         (vect_analyze_loop_2): ...here. Don't check the vectorization
11503         factor against the number of loop iterations if the loop is
11504         fully-masked.
11505
11506 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11507             Alan Hayward  <alan.hayward@arm.com>
11508             David Sherwood  <david.sherwood@arm.com>
11509
11510         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11511         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11512         (dump_groups): Update accordingly.
11513         (iv_use::mem_type): New member variable.
11514         (address_p): New function.
11515         (record_use): Add a mem_type argument and initialize the new
11516         mem_type field.
11517         (record_group_use): Add a mem_type argument.  Use address_p.
11518         Remove obsolete null checks of base_object.  Update call to record_use.
11519         (find_interesting_uses_op): Update call to record_group_use.
11520         (find_interesting_uses_cond): Likewise.
11521         (find_interesting_uses_address): Likewise.
11522         (get_mem_type_for_internal_fn): New function.
11523         (find_address_like_use): Likewise.
11524         (find_interesting_uses_stmt): Try find_address_like_use before
11525         calling find_interesting_uses_op.
11526         (addr_offset_valid_p): Use the iv mem_type field as the type
11527         of the addressed memory.
11528         (add_autoinc_candidates): Likewise.
11529         (get_address_cost): Likewise.
11530         (split_small_address_groups_p): Use address_p.
11531         (split_address_groups): Likewise.
11532         (add_iv_candidate_for_use): Likewise.
11533         (autoinc_possible_for_pair): Likewise.
11534         (rewrite_groups): Likewise.
11535         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11536         (determine_group_iv_cost): Update after split of USE_ADDRESS.
11537         (get_alias_ptr_type_for_ptr_address): New function.
11538         (rewrite_use_address): Rewrite address uses in calls that were
11539         identified by find_address_like_use.
11540
11541 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11542             Alan Hayward  <alan.hayward@arm.com>
11543             David Sherwood  <david.sherwood@arm.com>
11544
11545         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11546         TARGET_MEM_REFs.
11547         * gimple-expr.h (is_gimple_addressable: Likewise.
11548         * gimple-expr.c (is_gimple_address): Likewise.
11549         * internal-fn.c (expand_call_mem_ref): New function.
11550         (expand_mask_load_optab_fn): Use it.
11551         (expand_mask_store_optab_fn): Likewise.
11552
11553 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11554             Alan Hayward  <alan.hayward@arm.com>
11555             David Sherwood  <david.sherwood@arm.com>
11556
11557         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11558         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11559         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11560         (cond_umax@var{mode}): Document.
11561         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11562         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11563         (cond_umin_optab, cond_umax_optab): New optabs.
11564         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11565         (COND_IOR, COND_XOR): New internal functions.
11566         * internal-fn.h (get_conditional_internal_fn): Declare.
11567         * internal-fn.c (cond_binary_direct): New macro.
11568         (expand_cond_binary_optab_fn): Likewise.
11569         (direct_cond_binary_optab_supported_p): Likewise.
11570         (get_conditional_internal_fn): New function.
11571         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11572         Cope with reduction statements that are vectorized as calls rather
11573         than assignments.
11574         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11575         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11576         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11577         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11578         (UNSPEC_COND_EOR): New unspecs.
11579         (optab): Add mappings for them.
11580         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11581         (sve_int_op, sve_fp_op): New int attributes.
11582
11583 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11584             Alan Hayward  <alan.hayward@arm.com>
11585             David Sherwood  <david.sherwood@arm.com>
11586
11587         * optabs.def (while_ult_optab): New optab.
11588         * doc/md.texi (while_ult@var{m}@var{n}): Document.
11589         * internal-fn.def (WHILE_ULT): New internal function.
11590         * internal-fn.h (direct_internal_fn_supported_p): New override
11591         that takes two types as argument.
11592         * internal-fn.c (while_direct): New macro.
11593         (expand_while_optab_fn): New function.
11594         (convert_optab_supported_p): Likewise.
11595         (direct_while_optab_supported_p): New macro.
11596         * wide-int.h (wi::udiv_ceil): New function.
11597         * tree-vectorizer.h (rgroup_masks): New structure.
11598         (vec_loop_masks): New typedef.
11599         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11600         and fully_masked_p.
11601         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11602         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11603         (vect_max_vf): New function.
11604         (slpeel_make_loop_iterate_ntimes): Delete.
11605         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11606         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11607         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11608         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11609         internal-fn.h, stor-layout.h and optabs-query.h.
11610         (vect_set_loop_mask): New function.
11611         (add_preheader_seq): Likewise.
11612         (add_header_seq): Likewise.
11613         (interleave_supported_p): Likewise.
11614         (vect_maybe_permute_loop_masks): Likewise.
11615         (vect_set_loop_masks_directly): Likewise.
11616         (vect_set_loop_condition_masked): Likewise.
11617         (vect_set_loop_condition_unmasked): New function, split out from
11618         slpeel_make_loop_iterate_ntimes.
11619         (slpeel_make_loop_iterate_ntimes): Rename to..
11620         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
11621         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11622         (vect_do_peeling): Update call accordingly.
11623         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11624         loops.
11625         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11626         mask_compare_type, can_fully_mask_p and fully_masked_p.
11627         (release_vec_loop_masks): New function.
11628         (_loop_vec_info): Use it to free the loop masks.
11629         (can_produce_all_loop_masks_p): New function.
11630         (vect_get_max_nscalars_per_iter): Likewise.
11631         (vect_verify_full_masking): Likewise.
11632         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11633         retries, and free the mask rgroups before retrying.  Check loop-wide
11634         reasons for disallowing fully-masked loops.  Make the final decision
11635         about whether use a fully-masked loop or not.
11636         (vect_estimate_min_profitable_iters): Do not assume that peeling
11637         for the number of iterations will be needed for fully-masked loops.
11638         (vectorizable_reduction): Disable fully-masked loops.
11639         (vectorizable_live_operation): Likewise.
11640         (vect_halve_mask_nunits): New function.
11641         (vect_double_mask_nunits): Likewise.
11642         (vect_record_loop_mask): Likewise.
11643         (vect_get_loop_mask): Likewise.
11644         (vect_transform_loop): Handle the case in which the final loop
11645         iteration might handle a partial vector.  Call vect_set_loop_condition
11646         instead of slpeel_make_loop_iterate_ntimes.
11647         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11648         (check_load_store_masking): New function.
11649         (prepare_load_store_mask): Likewise.
11650         (vectorizable_store): Handle fully-masked loops.
11651         (vectorizable_load): Likewise.
11652         (supportable_widening_operation): Use vect_halve_mask_nunits for
11653         booleans.
11654         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11655         (vect_gen_while): New function.
11656         * config/aarch64/aarch64.md (umax<mode>3): New expander.
11657         (aarch64_uqdec<mode>): New insn.
11658
11659 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11660             Alan Hayward  <alan.hayward@arm.com>
11661             David Sherwood  <david.sherwood@arm.com>
11662
11663         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11664         (reduc_xor_scal_optab): New optabs.
11665         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11666         (reduc_xor_scal_@var{m}): Document.
11667         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11668         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11669         internal functions.
11670         * fold-const-call.c (fold_const_call): Handle them.
11671         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11672         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11673         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11674         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11675         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11676         (UNSPEC_XORV): New unspecs.
11677         (optab): Add entries for them.
11678         (BITWISEV): New int iterator.
11679         (bit_reduc_op): New int attributes.
11680
11681 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11682             Alan Hayward  <alan.hayward@arm.com>
11683             David Sherwood  <david.sherwood@arm.com>
11684
11685         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11686         * internal-fn.def (VEC_SHL_INSERT): New internal function.
11687         * optabs.def (vec_shl_insert_optab): New optab.
11688         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11689         (duplicate_and_interleave): Likewise.
11690         * tree-vect-loop.c: Include internal-fn.h.
11691         (neutral_op_for_slp_reduction): New function, split out from
11692         get_initial_defs_for_reduction.
11693         (get_initial_def_for_reduction): Handle option 2 for variable-length
11694         vectors by loading the neutral value into a vector and then shifting
11695         the initial value into element 0.
11696         (get_initial_defs_for_reduction): Replace the code argument with
11697         the neutral value calculated by neutral_op_for_slp_reduction.
11698         Use gimple_build_vector for constant-length vectors.
11699         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11700         but the first group_size elements have a neutral value.
11701         Use duplicate_and_interleave otherwise.
11702         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11703         Update call to get_initial_defs_for_reduction.  Handle SLP
11704         reductions for variable-length vectors by creating one vector
11705         result for each scalar result, with the elements associated
11706         with other scalar results stubbed out with the neutral value.
11707         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11708         Require IFN_VEC_SHL_INSERT for double reductions on
11709         variable-length vectors, or SLP reductions that have
11710         a neutral value.  Require can_duplicate_and_interleave_p
11711         support for variable-length unchained SLP reductions if there
11712         is no neutral value, such as for MIN/MAX reductions.  Also require
11713         the number of vector elements to be a multiple of the number of
11714         SLP statements when doing variable-length unchained SLP reductions.
11715         Update call to vect_create_epilog_for_reduction.
11716         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11717         and remove initial values.
11718         (duplicate_and_interleave): Make public.
11719         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11720         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11721
11722 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11723             Alan Hayward  <alan.hayward@arm.com>
11724             David Sherwood  <david.sherwood@arm.com>
11725
11726         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11727         (can_duplicate_and_interleave_p): New function.
11728         (vect_get_and_check_slp_defs): Take the vector of statements
11729         rather than just the current one.  Remove excess parentheses.
11730         Restriction rejectinon of vect_constant_def and vect_external_def
11731         for variable-length vectors to boolean types, or types for which
11732         can_duplicate_and_interleave_p is false.
11733         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11734         (duplicate_and_interleave): New function.
11735         (vect_get_constant_vectors): Use gimple_build_vector for
11736         constant-length vectors and suitable variable-length constant
11737         vectors.  Use duplicate_and_interleave for other variable-length
11738         vectors.  Don't defer the update when inserting new statements.
11739
11740 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11741             Alan Hayward  <alan.hayward@arm.com>
11742             David Sherwood  <david.sherwood@arm.com>
11743
11744         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11745         min_profitable_iters doesn't go negative.
11746
11747 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11748             Alan Hayward  <alan.hayward@arm.com>
11749             David Sherwood  <david.sherwood@arm.com>
11750
11751         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11752         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11753         * optabs.def (vec_mask_load_lanes_optab): New optab.
11754         (vec_mask_store_lanes_optab): Likewise.
11755         * internal-fn.def (MASK_LOAD_LANES): New internal function.
11756         (MASK_STORE_LANES): Likewise.
11757         * internal-fn.c (mask_load_lanes_direct): New macro.
11758         (mask_store_lanes_direct): Likewise.
11759         (expand_mask_load_optab_fn): Handle masked operations.
11760         (expand_mask_load_lanes_optab_fn): New macro.
11761         (expand_mask_store_optab_fn): Handle masked operations.
11762         (expand_mask_store_lanes_optab_fn): New macro.
11763         (direct_mask_load_lanes_optab_supported_p): Likewise.
11764         (direct_mask_store_lanes_optab_supported_p): Likewise.
11765         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11766         parameter.
11767         (vect_load_lanes_supported): Likewise.
11768         * tree-vect-data-refs.c (strip_conversion): New function.
11769         (can_group_stmts_p): Likewise.
11770         (vect_analyze_data_ref_accesses): Use it instead of checking
11771         for a pair of assignments.
11772         (vect_store_lanes_supported): Take a masked_p parameter.
11773         (vect_load_lanes_supported): Likewise.
11774         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11775         vect_store_lanes_supported and vect_load_lanes_supported.
11776         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11777         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11778         parameter.  Don't allow gaps for masked accesses.
11779         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
11780         and vect_load_lanes_supported.
11781         (get_load_store_type): Take a masked_p parameter and update
11782         call to get_group_load_store_type.
11783         (vectorizable_store): Update call to get_load_store_type.
11784         Handle IFN_MASK_STORE_LANES.
11785         (vectorizable_load): Update call to get_load_store_type.
11786         Handle IFN_MASK_LOAD_LANES.
11787
11788 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11789             Alan Hayward  <alan.hayward@arm.com>
11790             David Sherwood  <david.sherwood@arm.com>
11791
11792         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11793         modes for SVE.
11794         * config/aarch64/aarch64-protos.h
11795         (aarch64_sve_struct_memory_operand_p): Declare.
11796         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11797         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11798         (VPRED, vpred): Handle SVE structure modes.
11799         * config/aarch64/constraints.md (Utx): New constraint.
11800         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11801         (aarch64_sve_struct_nonimmediate_operand): New predicates.
11802         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11803         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11804         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11805         structure modes.  Split into pieces after RA.
11806         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11807         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11808         New patterns.
11809         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11810         SVE structure modes.
11811         (aarch64_classify_address): Likewise.
11812         (sizetochar): Move earlier in file.
11813         (aarch64_print_operand): Handle SVE register lists.
11814         (aarch64_array_mode): New function.
11815         (aarch64_sve_struct_memory_operand_p): Likewise.
11816         (TARGET_ARRAY_MODE): Redefine.
11817
11818 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11819             Alan Hayward  <alan.hayward@arm.com>
11820             David Sherwood  <david.sherwood@arm.com>
11821
11822         * target.def (array_mode): New target hook.
11823         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11824         * doc/tm.texi: Regenerate.
11825         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11826         * hooks.c (hook_optmode_mode_uhwi_none): New function.
11827         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11828         targetm.array_mode.
11829         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
11830         type sizes.
11831
11832 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11833             Alan Hayward  <alan.hayward@arm.com>
11834             David Sherwood  <david.sherwood@arm.com>
11835
11836         * fold-const.c (fold_binary_loc): Check the argument types
11837         rather than the result type when testing for a vector operation.
11838
11839 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11840
11841         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11842         * doc/tm.texi: Regenerate.
11843
11844 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11845             Alan Hayward  <alan.hayward@arm.com>
11846             David Sherwood  <david.sherwood@arm.com>
11847
11848         * doc/invoke.texi (-msve-vector-bits=): Document new option.
11849         (sve): Document new AArch64 extension.
11850         * doc/md.texi (w): Extend the description of the AArch64
11851         constraint to include SVE vectors.
11852         (Upl, Upa): Document new AArch64 predicate constraints.
11853         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11854         enum.
11855         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11856         (msve-vector-bits=): New option.
11857         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11858         SVE when these are disabled.
11859         (sve): New extension.
11860         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11861         modes.  Adjust their number of units based on aarch64_sve_vg.
11862         (MAX_BITSIZE_MODE_ANY_MODE): Define.
11863         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11864         aarch64_addr_query_type.
11865         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11866         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11867         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11868         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11869         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11870         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11871         (aarch64_simd_imm_zero_p): Delete.
11872         (aarch64_check_zero_based_sve_index_immediate): Declare.
11873         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11874         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11875         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11876         (aarch64_sve_float_mul_immediate_p): Likewise.
11877         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11878         rather than an rtx.
11879         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11880         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11881         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11882         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11883         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11884         (aarch64_regmode_natural_size): Likewise.
11885         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11886         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11887         left one place.
11888         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11889         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11890         for VG and the SVE predicate registers.
11891         (V_ALIASES): Add a "z"-prefixed alias.
11892         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11893         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11894         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11895         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11896         (REG_CLASS_NAMES): Add entries for them.
11897         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
11898         and the predicate registers.
11899         (aarch64_sve_vg): Declare.
11900         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11901         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11902         (REGMODE_NATURAL_SIZE): Define.
11903         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11904         SVE macros.
11905         * config/aarch64/aarch64.c: Include cfgrtl.h.
11906         (simd_immediate_info): Add a constructor for series vectors,
11907         and an associated step field.
11908         (aarch64_sve_vg): New variable.
11909         (aarch64_dbx_register_number): Handle VG and the predicate registers.
11910         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11911         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11912         (VEC_ANY_DATA, VEC_STRUCT): New constants.
11913         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11914         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11915         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11916         (aarch64_get_mask_mode): New functions.
11917         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11918         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11919         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
11920         predicate modes and predicate registers.  Explicitly restrict
11921         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
11922         to store a vector mode if it is recognized by
11923         aarch64_classify_vector_mode.
11924         (aarch64_regmode_natural_size): New function.
11925         (aarch64_hard_regno_caller_save_mode): Return the original mode
11926         for predicates.
11927         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11928         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11929         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11930         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11931         functions.
11932         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
11933         does not overlap dest if the function is frame-related.  Handle
11934         SVE constants.
11935         (aarch64_split_add_offset): New function.
11936         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11937         them aarch64_add_offset.
11938         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11939         and update call to aarch64_sub_sp.
11940         (aarch64_add_cfa_expression): New function.
11941         (aarch64_expand_prologue): Pass extra temporary registers to the
11942         functions above.  Handle the case in which we need to emit new
11943         DW_CFA_expressions for registers that were originally saved
11944         relative to the stack pointer, but now have to be expressed
11945         relative to the frame pointer.
11946         (aarch64_output_mi_thunk): Pass extra temporary registers to the
11947         functions above.
11948         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
11949         IP0 and IP1 values for SVE frames.
11950         (aarch64_expand_vec_series): New function.
11951         (aarch64_expand_sve_widened_duplicate): Likewise.
11952         (aarch64_expand_sve_const_vector): Likewise.
11953         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11954         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
11955         into the register, rather than emitting a SET directly.
11956         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11957         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11958         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11959         (offset_9bit_signed_scaled_p): New functions.
11960         (aarch64_replicate_bitmask_imm): New function.
11961         (aarch64_bitmask_imm): Use it.
11962         (aarch64_cannot_force_const_mem): Reject expressions involving
11963         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
11964         (aarch64_classify_index): Handle SVE indices, by requiring
11965         a plain register index with a scale that matches the element size.
11966         (aarch64_classify_address): Handle SVE addresses.  Assert that
11967         the mode of the address is VOIDmode or an integer mode.
11968         Update call to aarch64_classify_symbol.
11969         (aarch64_classify_symbolic_expression): Update call to
11970         aarch64_classify_symbol.
11971         (aarch64_const_vec_all_in_range_p): New function.
11972         (aarch64_print_vector_float_operand): Likewise.
11973         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
11974         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
11975         and the FP immediates 1.0 and 0.5.
11976         (aarch64_print_address_internal): Handle SVE addresses.
11977         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11978         (aarch64_regno_regclass): Handle predicate registers.
11979         (aarch64_secondary_reload): Handle big-endian reloads of SVE
11980         data modes.
11981         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11982         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11983         (aarch64_convert_sve_vector_bits): New function.
11984         (aarch64_override_options): Use it to handle -msve-vector-bits=.
11985         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11986         rather than an rtx.
11987         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11988         Handle SVE vector and predicate modes.  Accept VL-based constants
11989         that need only one temporary register, and VL offsets that require
11990         no temporary registers.
11991         (aarch64_conditional_register_usage): Mark the predicate registers
11992         as fixed if SVE isn't available.
11993         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11994         Return true for SVE vector and predicate modes.
11995         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11996         rather than an unsigned int.  Handle SVE modes.
11997         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
11998         SVE modes.
11999         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12000         if SVE is enabled.
12001         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12002         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12003         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12004         (aarch64_sve_float_mul_immediate_p): New functions.
12005         (aarch64_sve_valid_immediate): New function.
12006         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12007         Explicitly reject structure modes.  Check for INDEX constants.
12008         Handle PTRUE and PFALSE constants.
12009         (aarch64_check_zero_based_sve_index_immediate): New function.
12010         (aarch64_simd_imm_zero_p): Delete.
12011         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12012         vector modes.  Accept constants in the range of CNT[BHWD].
12013         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12014         ask for an Advanced SIMD mode.
12015         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12016         (aarch64_simd_vector_alignment): Handle SVE predicates.
12017         (aarch64_vectorize_preferred_vector_alignment): New function.
12018         (aarch64_simd_vector_alignment_reachable): Use it instead of
12019         the vector size.
12020         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12021         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12022         functions.
12023         (MAX_VECT_LEN): Delete.
12024         (expand_vec_perm_d): Add a vec_flags field.
12025         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12026         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12027         (aarch64_evpc_ext): Don't apply a big-endian lane correction
12028         for SVE modes.
12029         (aarch64_evpc_rev): Rename to...
12030         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
12031         (aarch64_evpc_rev_global): New function.
12032         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12033         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12034         MAX_VECT_LEN.
12035         (aarch64_evpc_sve_tbl): New function.
12036         (aarch64_expand_vec_perm_const_1): Update after rename of
12037         aarch64_evpc_rev.  Handle SVE permutes too, trying
12038         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12039         than aarch64_evpc_tbl.
12040         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12041         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12042         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12043         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12044         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12045         (aarch64_expand_sve_vcond): New functions.
12046         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12047         of aarch64_vector_mode_p.
12048         (aarch64_dwarf_poly_indeterminate_value): New function.
12049         (aarch64_compute_pressure_classes): Likewise.
12050         (aarch64_can_change_mode_class): Likewise.
12051         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12052         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12053         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12054         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12055         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12056         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12057         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12058         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12059         constraints.
12060         (Dn, Dl, Dr): Accept const as well as const_vector.
12061         (Dz): Likewise.  Compare against CONST0_RTX.
12062         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12063         of "vector" where appropriate.
12064         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12065         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12066         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12067         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12068         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12069         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12070         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12071         (v_int_equiv): Extend to SVE modes.
12072         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12073         mode attributes.
12074         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12075         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12076         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12077         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12078         (SVE_COND_FP_CMP): New int iterators.
12079         (perm_hilo): Handle the new unpack unspecs.
12080         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12081         attributes.
12082         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12083         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12084         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12085         (aarch64_equality_operator, aarch64_constant_vector_operand)
12086         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12087         (aarch64_sve_nonimmediate_operand): Likewise.
12088         (aarch64_sve_general_operand): Likewise.
12089         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12090         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12091         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12092         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12093         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12094         (aarch64_sve_float_arith_immediate): Likewise.
12095         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12096         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12097         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12098         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12099         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12100         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12101         (aarch64_sve_float_arith_operand): Likewise.
12102         (aarch64_sve_float_arith_with_sub_operand): Likewise.
12103         (aarch64_sve_float_mul_operand): Likewise.
12104         (aarch64_sve_vec_perm_operand): Likewise.
12105         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12106         (aarch64_mov_operand): Accept const_poly_int and const_vector.
12107         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12108         as well as const_vector.
12109         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12110         in file.  Use CONST0_RTX and CONSTM1_RTX.
12111         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
12112         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12113         Use aarch64_simd_imm_zero.
12114         * config/aarch64/aarch64-sve.md: New file.
12115         * config/aarch64/aarch64.md: Include it.
12116         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12117         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12118         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12119         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12120         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12121         (sve): New attribute.
12122         (enabled): Disable instructions with the sve attribute unless
12123         TARGET_SVE.
12124         (movqi, movhi): Pass CONST_POLY_INT operaneds through
12125         aarch64_expand_mov_immediate.
12126         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12127         CNT[BHSD] immediates.
12128         (movti): Split CONST_POLY_INT moves into two halves.
12129         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12130         Split additions that need a temporary here if the destination
12131         is the stack pointer.
12132         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12133         (*add<mode>3_poly_1): New instruction.
12134         (set_clobber_cc): New expander.
12135
12136 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12137
12138         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12139         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
12140         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12141         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12142         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
12143         Change innermode from fixed_mode_size to machine_mode.
12144         (simplify_subreg): Update call accordingly.  Handle a constant-sized
12145         subreg of a variable-length CONST_VECTOR.
12146
12147 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12148             Alan Hayward  <alan.hayward@arm.com>
12149             David Sherwood  <david.sherwood@arm.com>
12150
12151         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12152         (add_offset_to_base): New function, split out from...
12153         (create_mem_ref): ...here.  When handling a scale other than 1,
12154         check first whether the address is valid without the offset.
12155         Add it into the base if so, leaving the index and scale as-is.
12156
12157 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
12158
12159         PR c++/83778
12160         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12161         fold_for_warn before checking if arg2 is INTEGER_CST.
12162
12163 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
12164
12165         * config/rs6000/predicates.md (load_multiple_operation): Delete.
12166         (store_multiple_operation): Delete.
12167         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12168         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12169         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12170         guarded by TARGET_STRING.
12171         (rs6000_output_load_multiple): Delete.
12172         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12173         OPTION_MASK_STRING / TARGET_STRING handling.
12174         (print_operand) <'N', 'O'>: Add comment that these are unused now.
12175         (const rs6000_opt_masks) <"string">: Change mask to 0.
12176         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12177         (MASK_STRING): Delete.
12178         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12179         parts.  Simplify.
12180         (load_multiple): Delete.
12181         (*ldmsi8): Delete.
12182         (*ldmsi7): Delete.
12183         (*ldmsi6): Delete.
12184         (*ldmsi5): Delete.
12185         (*ldmsi4): Delete.
12186         (*ldmsi3): Delete.
12187         (store_multiple): Delete.
12188         (*stmsi8): Delete.
12189         (*stmsi7): Delete.
12190         (*stmsi6): Delete.
12191         (*stmsi5): Delete.
12192         (*stmsi4): Delete.
12193         (*stmsi3): Delete.
12194         (movmemsi_8reg): Delete.
12195         (corresponding unnamed define_insn): Delete.
12196         (movmemsi_6reg): Delete.
12197         (corresponding unnamed define_insn): Delete.
12198         (movmemsi_4reg): Delete.
12199         (corresponding unnamed define_insn): Delete.
12200         (movmemsi_2reg): Delete.
12201         (corresponding unnamed define_insn): Delete.
12202         (movmemsi_1reg): Delete.
12203         (corresponding unnamed define_insn): Delete.
12204         * config/rs6000/rs6000.opt (mno-string): New.
12205         (mstring): Replace by deprecation warning stub.
12206         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12207
12208 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
12209
12210         * regrename.c (regrename_do_replace): If replacing the same
12211         reg multiple times, try to reuse last created gen_raw_REG.
12212
12213         PR debug/81155
12214         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12215         main to workaround a bug in GDB.
12216
12217 2018-01-12  Tom de Vries  <tom@codesourcery.com>
12218
12219         PR target/83737
12220         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12221
12222 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
12223
12224         PR rtl-optimization/80481
12225         * ira-color.c (get_cap_member): New function.
12226         (allocnos_conflict_by_live_ranges_p): Use it.
12227         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12228         (setup_slot_coalesced_allocno_live_ranges): Ditto.
12229
12230 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
12231
12232         PR target/83628
12233         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12234         (*saddl_se_1): Ditto.
12235         (*ssubsi_1): Ditto.
12236         (*ssubl_se_1): Ditto.
12237
12238 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12239
12240         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12241         rather than wi::to_widest for DR_INITs.
12242         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12243         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12244         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12245         INTEGER_CSTs.
12246         (vect_analyze_group_access_1): Note that here.
12247
12248 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12249
12250         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12251         polynomial type sizes.
12252
12253 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12254
12255         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12256         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12257         (gimple_add_tmp_var): Likewise.
12258
12259 2018-01-12  Martin Liska  <mliska@suse.cz>
12260
12261         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12262         (gimple_alloc_sizes): Likewise.
12263         (dump_gimple_statistics): Use PRIu64 in printf format.
12264         * gimple.h: Change uint64_t to int.
12265
12266 2018-01-12  Martin Liska  <mliska@suse.cz>
12267
12268         * tree-core.h: Use uint64_t instead of int.
12269         * tree.c (tree_node_counts): Likewise.
12270         (tree_node_sizes): Likewise.
12271         (dump_tree_statistics): Use PRIu64 in printf format.
12272
12273 2018-01-12  Martin Liska  <mliska@suse.cz>
12274
12275         * Makefile.in: As qsort_chk is implemented in vec.c, add
12276         vec.o to linkage of gencfn-macros.
12277         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12278         passing the info to record_node_allocation_statistics.
12279         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12280         and pass the info.
12281         * ggc-common.c (struct ggc_usage): Add operator== and use
12282         it in operator< and compare function.
12283         * mem-stats.h (struct mem_usage): Likewise.
12284         * vec.c (struct vec_usage): Remove operator< and compare
12285         function. Can be simply inherited.
12286
12287 2018-01-12  Martin Jambor  <mjambor@suse.cz>
12288
12289         PR target/81616
12290         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12291         * tree-ssa-math-opts.c: Include domwalk.h.
12292         (convert_mult_to_fma_1): New function.
12293         (fma_transformation_info): New type.
12294         (fma_deferring_state): Likewise.
12295         (cancel_fma_deferring): New function.
12296         (result_of_phi): Likewise.
12297         (last_fma_candidate_feeds_initial_phi): Likewise.
12298         (convert_mult_to_fma): Added deferring logic, split actual
12299         transformation to convert_mult_to_fma_1.
12300         (math_opts_dom_walker): New type.
12301         (math_opts_dom_walker::after_dom_children): New method, body moved
12302         here from pass_optimize_widening_mul::execute, added deferring logic
12303         bits.
12304         (pass_optimize_widening_mul::execute): Moved most of code to
12305         math_opts_dom_walker::after_dom_children.
12306         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12307         * config/i386/i386.c (ix86_option_override_internal): Added
12308         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12309
12310 2018-01-12  Richard Biener  <rguenther@suse.de>
12311
12312         PR debug/83157
12313         * dwarf2out.c (gen_variable_die): Do not reset old_die for
12314         inline instance vars.
12315
12316 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
12317
12318         PR target/81819
12319         * config/rx/rx.c (rx_is_restricted_memory_address):
12320         Handle SUBREG case.
12321
12322 2018-01-12  Richard Biener  <rguenther@suse.de>
12323
12324         PR tree-optimization/80846
12325         * target.def (split_reduction): New target hook.
12326         * targhooks.c (default_split_reduction): New function.
12327         * targhooks.h (default_split_reduction): Declare.
12328         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12329         target requests first reduce vectors by combining low and high
12330         parts.
12331         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12332         (get_vectype_for_scalar_type_and_size): Export.
12333         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12334         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12335         * doc/tm.texi: Regenerate.
12336         * config/i386/i386.c (ix86_split_reduction): Implement
12337         TARGET_VECTORIZE_SPLIT_REDUCTION.
12338
12339 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
12340
12341         PR target/83368
12342         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12343         in PIC mode except for TARGET_VXWORKS_RTP.
12344         * config/sparc/sparc.c: Include cfgrtl.h.
12345         (TARGET_INIT_PIC_REG): Define.
12346         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12347         (sparc_pic_register_p): New predicate.
12348         (sparc_legitimate_address_p): Use it.
12349         (sparc_legitimize_pic_address): Likewise.
12350         (sparc_delegitimize_address): Likewise.
12351         (sparc_mode_dependent_address_p): Likewise.
12352         (gen_load_pcrel_sym): Remove 4th parameter.
12353         (load_got_register): Adjust call to above.  Remove obsolete stuff.
12354         (sparc_expand_prologue): Do not call load_got_register here.
12355         (sparc_flat_expand_prologue): Likewise.
12356         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12357         (sparc_use_pseudo_pic_reg): New function.
12358         (sparc_init_pic_reg): Likewise.
12359         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12360         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12361
12362 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
12363
12364         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12365         Add item for branch_cost.
12366
12367 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
12368
12369         PR rtl-optimization/83565
12370         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12371         not extend the result to a larger mode for rotate operations.
12372         (num_sign_bit_copies1): Likewise.
12373
12374 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12375
12376         PR target/40411
12377         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12378         -symbolic.
12379         Use values-Xc.o for -pedantic.
12380         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12381
12382 2018-01-12  Martin Liska  <mliska@suse.cz>
12383
12384         PR ipa/83054
12385         * ipa-devirt.c (final_warning_record::grow_type_warnings):
12386         New function.
12387         (possible_polymorphic_call_targets): Use it.
12388         (ipa_devirt): Likewise.
12389
12390 2018-01-12  Martin Liska  <mliska@suse.cz>
12391
12392         * profile-count.h (enum profile_quality): Use 0 as invalid
12393         enum value of profile_quality.
12394
12395 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
12396
12397         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12398         -mext-string options.
12399
12400 2018-01-12  Richard Biener  <rguenther@suse.de>
12401
12402         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12403         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12404         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12405         Likewise.
12406         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12407
12408 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
12409
12410         * configure.ac (--with-long-double-format): Add support for the
12411         configuration option to change the default long double format on
12412         PowerPC systems.
12413         * config.gcc (powerpc*-linux*-*): Likewise.
12414         * configure: Regenerate.
12415         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12416         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12417         used without modification.
12418
12419 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
12420
12421         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12422         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12423         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12424         MISC_BUILTIN_SPEC_BARRIER.
12425         (rs6000_init_builtins): Likewise.
12426         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12427         enum value.
12428         (speculation_barrier): New define_insn.
12429         * doc/extend.texi: Document __builtin_speculation_barrier.
12430
12431 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
12432
12433         PR target/83203
12434         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12435         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12436         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12437         iterators.
12438         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
12439         integral modes instead of "ss" and "sd".
12440         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12441         vectors with 32-bit and 64-bit elements.
12442         (vecdupssescalarmodesuffix): New mode attribute.
12443         (vec_dup<mode>): Use it.
12444
12445 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
12446
12447         PR target/83330
12448         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12449         frame if argument is passed on stack.
12450
12451 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
12452
12453         PR target/82682
12454         * ree.c (combine_reaching_defs): Optimize also
12455         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12456         reg2=any_extend(exp); reg1=reg2;, formatting fix.
12457
12458 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
12459
12460         PR middle-end/83189
12461         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12462
12463 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
12464
12465         PR middle-end/83718
12466         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12467         after they are computed.
12468
12469 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
12470
12471         PR tree-optimization/83695
12472         * gimple-loop-linterchange.cc
12473         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12474         reset cached scev information after interchange.
12475         (pass_linterchange::execute): Remove call to scev_reset_htab.
12476
12477 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12478
12479         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12480         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12481         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12482         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12483         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12484         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12485         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12486         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12487         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12488         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12489         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12490         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12491         (V_lane_reg): Likewise.
12492         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12493         New define_expand.
12494         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12495         (vfmal_lane_low<mode>_intrinsic,
12496         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12497         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12498         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12499         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12500         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12501         vfmsl_lane_high<mode>_intrinsic): New define_insns.
12502
12503 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12504
12505         * config/arm/arm-cpus.in (fp16fml): New feature.
12506         (ALL_SIMD): Add fp16fml.
12507         (armv8.2-a): Add fp16fml as an option.
12508         (armv8.3-a): Likewise.
12509         (armv8.4-a): Add fp16fml as part of fp16.
12510         * config/arm/arm.h (TARGET_FP16FML): Define.
12511         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12512         when appropriate.
12513         * config/arm/arm-modes.def (V2HF): Define.
12514         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12515         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12516         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12517         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12518         vfmsl_low, vfmsl_high): New set of builtins.
12519         * config/arm/iterators.md (PLUSMINUS): New code iterator.
12520         (vfml_op): New code attribute.
12521         (VFMLHALVES): New int iterator.
12522         (VFML, VFMLSEL): New mode attributes.
12523         (V_reg): Define mapping for V2HF.
12524         (V_hi, V_lo): New mode attributes.
12525         (VF_constraint): Likewise.
12526         (vfml_half, vfml_half_selector): New int attributes.
12527         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12528         define_expand.
12529         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12530         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12531         New define_insn.
12532         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12533         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12534         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12535         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
12536         documentation.
12537         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12538         Document new effective target and option set.
12539
12540 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12541
12542         * config/arm/arm-cpus.in (armv8_4): New feature.
12543         (ARMv8_4a): New fgroup.
12544         (armv8.4-a): New arch.
12545         * config/arm/arm-tables.opt: Regenerate.
12546         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12547         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12548         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12549         Add matching rules for -march=armv8.4-a and extensions.
12550         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12551
12552 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
12553
12554         PR target/81821
12555         * config/rx/rx.md (BW): New mode attribute.
12556         (sync_lock_test_and_setsi): Add mode suffix to insn output.
12557
12558 2018-01-11  Richard Biener  <rguenther@suse.de>
12559
12560         PR tree-optimization/83435
12561         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12562         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12563         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12564
12565 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12566             Alan Hayward  <alan.hayward@arm.com>
12567             David Sherwood  <david.sherwood@arm.com>
12568
12569         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12570         field.
12571         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
12572         (aarch64_print_address_internal): Use it to check for a zero offset.
12573
12574 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12575             Alan Hayward  <alan.hayward@arm.com>
12576             David Sherwood  <david.sherwood@arm.com>
12577
12578         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12579         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12580         Return a poly_int64 rather than a HOST_WIDE_INT.
12581         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12582         rather than a HOST_WIDE_INT.
12583         * config/aarch64/aarch64.h (aarch64_frame): Protect with
12584         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
12585         hard_fp_offset, frame_size, initial_adjust, callee_offset and
12586         final_offset from HOST_WIDE_INT to poly_int64.
12587         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12588         to_constant when getting the number of units in an Advanced SIMD
12589         mode.
12590         (aarch64_builtin_vectorized_function): Check for a constant number
12591         of units.
12592         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12593         GET_MODE_SIZE.
12594         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12595         attribute instead of GET_MODE_NUNITS.
12596         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12597         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12598         GET_MODE_SIZE for fixed-size registers.
12599         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12600         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12601         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12602         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12603         (aarch64_print_operand, aarch64_print_address_internal)
12604         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12605         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12606         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12607         Handle polynomial GET_MODE_SIZE.
12608         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
12609         wider than SImode without modification.
12610         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12611         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12612         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12613         passing and returning SVE modes.
12614         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12615         rather than GEN_INT.
12616         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12617         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12618         (aarch64_allocate_and_probe_stack_space): Likewise.
12619         (aarch64_layout_frame): Cope with polynomial offsets.
12620         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12621         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
12622         polynomial offsets.
12623         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12624         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12625         poly_int64 rather than a HOST_WIDE_INT.
12626         (aarch64_get_separate_components, aarch64_process_components)
12627         (aarch64_expand_prologue, aarch64_expand_epilogue)
12628         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12629         (aarch64_anchor_offset): New function, split out from...
12630         (aarch64_legitimize_address): ...here.
12631         (aarch64_builtin_vectorization_cost): Handle polynomial
12632         TYPE_VECTOR_SUBPARTS.
12633         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12634         GET_MODE_NUNITS.
12635         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12636         number of elements from the PARALLEL rather than the mode.
12637         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12638         rather than GET_MODE_BITSIZE.
12639         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12640         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12641         (aarch64_expand_vec_perm_const_1): Handle polynomial
12642         d->perm.length () and d->perm elements.
12643         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
12644         Apply to_constant to d->perm elements.
12645         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12646         polynomial CONST_VECTOR_NUNITS.
12647         (aarch64_move_pointer): Take amount as a poly_int64 rather
12648         than an int.
12649         (aarch64_progress_pointer): Avoid temporary variable.
12650         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12651         the mode attribute instead of GET_MODE.
12652
12653 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12654             Alan Hayward  <alan.hayward@arm.com>
12655             David Sherwood  <david.sherwood@arm.com>
12656
12657         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12658         x exists before using it.
12659         (aarch64_add_constant_internal): Rename to...
12660         (aarch64_add_offset_1): ...this.  Replace regnum with separate
12661         src and dest rtxes.  Handle the case in which they're different,
12662         including when the offset is zero.  Replace scratchreg with an rtx.
12663         Use 2 additions if there is no spare register into which we can
12664         move a 16-bit constant.
12665         (aarch64_add_constant): Delete.
12666         (aarch64_add_offset): Replace reg with separate src and dest
12667         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
12668         Use aarch64_add_offset_1.
12669         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12670         an rtx rather than an int.  Take the delta as a poly_int64
12671         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
12672         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12673         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12674         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12675         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12676         and aarch64_add_sp.
12677         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12678         aarch64_add_constant.
12679
12680 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12681
12682         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12683         Use scalar_float_mode.
12684
12685 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12686
12687         * config/aarch64/aarch64-simd.md
12688         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12689         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12690         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12691         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12692         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12693         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12694         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12695         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12696         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12697         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12698
12699 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12700
12701         PR target/83514
12702         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12703         targ_options->x_arm_arch_string is non NULL.
12704
12705 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
12706
12707         * config/aarch64/aarch64.h
12708         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
12709
12710 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
12711
12712         PR target/82096
12713         * expmed.c (emit_store_flag_force): Swap if const op0
12714         and change VOIDmode to mode of op0.
12715
12716 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12717
12718         PR rtl-optimization/83761
12719         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12720         than bytes to mode_for_size.
12721
12722 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12723
12724         PR middle-end/83189
12725         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12726         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12727         profile.
12728
12729 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12730
12731         PR middle-end/83575
12732         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12733         when in layout mode.
12734         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12735         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12736         partition fixup.
12737
12738 2018-01-10  Michael Collison  <michael.collison@arm.com>
12739
12740         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12741         * config/aarch64/aarch64-option-extension.def: Add
12742         AARCH64_OPT_EXTENSION of 'fp16fml'.
12743         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12744         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12745         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12746         * config/aarch64/constraints.md (Ui7): New constraint.
12747         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12748         (VFMLA_SEL_W): Ditto.
12749         (f16quad): Ditto.
12750         (f16mac1): Ditto.
12751         (VFMLA16_LOW): New int iterator.
12752         (VFMLA16_HIGH): Ditto.
12753         (UNSPEC_FMLAL): New unspec.
12754         (UNSPEC_FMLSL): Ditto.
12755         (UNSPEC_FMLAL2): Ditto.
12756         (UNSPEC_FMLSL2): Ditto.
12757         (f16mac): New code attribute.
12758         * config/aarch64/aarch64-simd-builtins.def
12759         (aarch64_fmlal_lowv2sf): Ditto.
12760         (aarch64_fmlsl_lowv2sf): Ditto.
12761         (aarch64_fmlalq_lowv4sf): Ditto.
12762         (aarch64_fmlslq_lowv4sf): Ditto.
12763         (aarch64_fmlal_highv2sf): Ditto.
12764         (aarch64_fmlsl_highv2sf): Ditto.
12765         (aarch64_fmlalq_highv4sf): Ditto.
12766         (aarch64_fmlslq_highv4sf): Ditto.
12767         (aarch64_fmlal_lane_lowv2sf): Ditto.
12768         (aarch64_fmlsl_lane_lowv2sf): Ditto.
12769         (aarch64_fmlal_laneq_lowv2sf): Ditto.
12770         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12771         (aarch64_fmlalq_lane_lowv4sf): Ditto.
12772         (aarch64_fmlsl_lane_lowv4sf): Ditto.
12773         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12774         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12775         (aarch64_fmlal_lane_highv2sf): Ditto.
12776         (aarch64_fmlsl_lane_highv2sf): Ditto.
12777         (aarch64_fmlal_laneq_highv2sf): Ditto.
12778         (aarch64_fmlsl_laneq_highv2sf): Ditto.
12779         (aarch64_fmlalq_lane_highv4sf): Ditto.
12780         (aarch64_fmlsl_lane_highv4sf): Ditto.
12781         (aarch64_fmlalq_laneq_highv4sf): Ditto.
12782         (aarch64_fmlsl_laneq_highv4sf): Ditto.
12783         * config/aarch64/aarch64-simd.md:
12784         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12785         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12786         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12787         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12788         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12789         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12790         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12791         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12792         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12793         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12794         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12795         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12796         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12797         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12798         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12799         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12800         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12801         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12802         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12803         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12804         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12805         (vfmlsl_low_u32): Ditto.
12806         (vfmlalq_low_u32): Ditto.
12807         (vfmlslq_low_u32): Ditto.
12808         (vfmlal_high_u32): Ditto.
12809         (vfmlsl_high_u32): Ditto.
12810         (vfmlalq_high_u32): Ditto.
12811         (vfmlslq_high_u32): Ditto.
12812         (vfmlal_lane_low_u32): Ditto.
12813         (vfmlsl_lane_low_u32): Ditto.
12814         (vfmlal_laneq_low_u32): Ditto.
12815         (vfmlsl_laneq_low_u32): Ditto.
12816         (vfmlalq_lane_low_u32): Ditto.
12817         (vfmlslq_lane_low_u32): Ditto.
12818         (vfmlalq_laneq_low_u32): Ditto.
12819         (vfmlslq_laneq_low_u32): Ditto.
12820         (vfmlal_lane_high_u32): Ditto.
12821         (vfmlsl_lane_high_u32): Ditto.
12822         (vfmlal_laneq_high_u32): Ditto.
12823         (vfmlsl_laneq_high_u32): Ditto.
12824         (vfmlalq_lane_high_u32): Ditto.
12825         (vfmlslq_lane_high_u32): Ditto.
12826         (vfmlalq_laneq_high_u32): Ditto.
12827         (vfmlslq_laneq_high_u32): Ditto.
12828         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12829         (AARCH64_FL_FOR_ARCH8_4): New.
12830         (AARCH64_ISA_F16FML): New ISA flag.
12831         (TARGET_F16FML): New feature flag for fp16fml.
12832         (doc/invoke.texi): Document new fp16fml option.
12833
12834 2018-01-10  Michael Collison  <michael.collison@arm.com>
12835
12836         * config/aarch64/aarch64-builtins.c:
12837         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12838         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12839         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12840         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12841         (AARCH64_ISA_SHA3): New ISA flag.
12842         (TARGET_SHA3): New feature flag for sha3.
12843         * config/aarch64/iterators.md (sha512_op): New int attribute.
12844         (CRYPTO_SHA512): New int iterator.
12845         (UNSPEC_SHA512H): New unspec.
12846         (UNSPEC_SHA512H2): Ditto.
12847         (UNSPEC_SHA512SU0): Ditto.
12848         (UNSPEC_SHA512SU1): Ditto.
12849         * config/aarch64/aarch64-simd-builtins.def
12850         (aarch64_crypto_sha512hqv2di): New builtin.
12851         (aarch64_crypto_sha512h2qv2di): Ditto.
12852         (aarch64_crypto_sha512su0qv2di): Ditto.
12853         (aarch64_crypto_sha512su1qv2di): Ditto.
12854         (aarch64_eor3qv8hi): Ditto.
12855         (aarch64_rax1qv2di): Ditto.
12856         (aarch64_xarqv2di): Ditto.
12857         (aarch64_bcaxqv8hi): Ditto.
12858         * config/aarch64/aarch64-simd.md:
12859         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12860         (aarch64_crypto_sha512su0qv2di): Ditto.
12861         (aarch64_crypto_sha512su1qv2di): Ditto.
12862         (aarch64_eor3qv8hi): Ditto.
12863         (aarch64_rax1qv2di): Ditto.
12864         (aarch64_xarqv2di): Ditto.
12865         (aarch64_bcaxqv8hi): Ditto.
12866         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12867         (vsha512h2q_u64): Ditto.
12868         (vsha512su0q_u64): Ditto.
12869         (vsha512su1q_u64): Ditto.
12870         (veor3q_u16): Ditto.
12871         (vrax1q_u64): Ditto.
12872         (vxarq_u64): Ditto.
12873         (vbcaxq_u16): Ditto.
12874         * config/arm/types.md (crypto_sha512): New type attribute.
12875         (crypto_sha3): Ditto.
12876         (doc/invoke.texi): Document new sha3 option.
12877
12878 2018-01-10  Michael Collison  <michael.collison@arm.com>
12879
12880         * config/aarch64/aarch64-builtins.c:
12881         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12882         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12883         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12884         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12885         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12886         (AARCH64_ISA_SM4): New ISA flag.
12887         (TARGET_SM4): New feature flag for sm4.
12888         * config/aarch64/aarch64-simd-builtins.def
12889         (aarch64_sm3ss1qv4si): Ditto.
12890         (aarch64_sm3tt1aq4si): Ditto.
12891         (aarch64_sm3tt1bq4si): Ditto.
12892         (aarch64_sm3tt2aq4si): Ditto.
12893         (aarch64_sm3tt2bq4si): Ditto.
12894         (aarch64_sm3partw1qv4si): Ditto.
12895         (aarch64_sm3partw2qv4si): Ditto.
12896         (aarch64_sm4eqv4si): Ditto.
12897         (aarch64_sm4ekeyqv4si): Ditto.
12898         * config/aarch64/aarch64-simd.md:
12899         (aarch64_sm3ss1qv4si): Ditto.
12900         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12901         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12902         (aarch64_sm4eqv4si): Ditto.
12903         (aarch64_sm4ekeyqv4si): Ditto.
12904         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12905         (sm3part_op): Ditto.
12906         (CRYPTO_SM3TT): Ditto.
12907         (CRYPTO_SM3PART): Ditto.
12908         (UNSPEC_SM3SS1): New unspec.
12909         (UNSPEC_SM3TT1A): Ditto.
12910         (UNSPEC_SM3TT1B): Ditto.
12911         (UNSPEC_SM3TT2A): Ditto.
12912         (UNSPEC_SM3TT2B): Ditto.
12913         (UNSPEC_SM3PARTW1): Ditto.
12914         (UNSPEC_SM3PARTW2): Ditto.
12915         (UNSPEC_SM4E): Ditto.
12916         (UNSPEC_SM4EKEY): Ditto.
12917         * config/aarch64/constraints.md (Ui2): New constraint.
12918         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12919         * config/arm/types.md (crypto_sm3): New type attribute.
12920         (crypto_sm4): Ditto.
12921         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12922         (vsm3tt1aq_u32): Ditto.
12923         (vsm3tt1bq_u32): Ditto.
12924         (vsm3tt2aq_u32): Ditto.
12925         (vsm3tt2bq_u32): Ditto.
12926         (vsm3partw1q_u32): Ditto.
12927         (vsm3partw2q_u32): Ditto.
12928         (vsm4eq_u32): Ditto.
12929         (vsm4ekeyq_u32): Ditto.
12930         (doc/invoke.texi): Document new sm4 option.
12931
12932 2018-01-10  Michael Collison  <michael.collison@arm.com>
12933
12934         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12935         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12936         (AARCH64_FL_FOR_ARCH8_4): New.
12937         (AARCH64_FL_V8_4): New flag.
12938         (doc/invoke.texi): Document new armv8.4-a option.
12939
12940 2018-01-10  Michael Collison  <michael.collison@arm.com>
12941
12942         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12943         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12944         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12945         * config/aarch64/aarch64-option-extension.def: Add
12946         AARCH64_OPT_EXTENSION of 'sha2'.
12947         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12948         (crypto): Disable sha2 and aes if crypto disabled.
12949         (crypto): Enable aes and sha2 if enabled.
12950         (simd): Disable sha2 and aes if simd disabled.
12951         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12952         New flags.
12953         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12954         (TARGET_SHA2): New feature flag for sha2.
12955         (TARGET_AES): New feature flag for aes.
12956         * config/aarch64/aarch64-simd.md:
12957         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12958         conditional on TARGET_AES.
12959         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12960         (aarch64_crypto_sha1hsi): Make pattern conditional
12961         on TARGET_SHA2.
12962         (aarch64_crypto_sha1hv4si): Ditto.
12963         (aarch64_be_crypto_sha1hv4si): Ditto.
12964         (aarch64_crypto_sha1su1v4si): Ditto.
12965         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12966         (aarch64_crypto_sha1su0v4si): Ditto.
12967         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12968         (aarch64_crypto_sha256su0v4si): Ditto.
12969         (aarch64_crypto_sha256su1v4si): Ditto.
12970         (doc/invoke.texi): Document new aes and sha2 options.
12971
12972 2018-01-10  Martin Sebor  <msebor@redhat.com>
12973
12974         PR tree-optimization/83781
12975         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12976         as string arrays.
12977
12978 2018-01-11  Martin Sebor  <msebor@gmail.com>
12979             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12980
12981         PR tree-optimization/83501
12982         PR tree-optimization/81703
12983
12984         * tree-ssa-strlen.c (get_string_cst): Rename...
12985         (get_string_len): ...to this.  Handle global constants.
12986         (handle_char_store): Adjust.
12987
12988 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
12989             Jim Wilson  <jimw@sifive.com>
12990
12991         * config/riscv/riscv-protos.h (riscv_output_return): New.
12992         * config/riscv/riscv.c (struct machine_function): New naked_p field.
12993         (riscv_attribute_table, riscv_output_return),
12994         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12995         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12996         (riscv_compute_frame_info): Only compute frame->mask if not a naked
12997         function.
12998         (riscv_expand_prologue): Add early return for naked function.
12999         (riscv_expand_epilogue): Likewise.
13000         (riscv_function_ok_for_sibcall): Return false for naked function.
13001         (riscv_set_current_function): New.
13002         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13003         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13004         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13005         * doc/extend.texi (RISC-V Function Attributes): New.
13006
13007 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
13008
13009         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13010         check for 128-bit long double before checking TCmode.
13011         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13012         128-bit long doubles before checking TFmode or TCmode.
13013         (FLOAT128_IBM_P): Likewise.
13014
13015 2018-01-10  Martin Sebor  <msebor@redhat.com>
13016
13017         PR tree-optimization/83671
13018         * builtins.c (c_strlen): Unconditionally return zero for the empty
13019         string.
13020         Use -Warray-bounds for warnings.
13021         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13022         for non-constant array indices with COMPONENT_REF, arrays of
13023         arrays, and pointers to arrays.
13024         (gimple_fold_builtin_strlen): Determine and set length range for
13025         non-constant character arrays.
13026
13027 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
13028
13029         PR middle-end/81897
13030         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13031         empty blocks.
13032
13033 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
13034
13035         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13036
13037 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
13038
13039         PR target/83399
13040         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13041         VECTOR_MEM_ALTIVEC_OR_VSX_P.
13042         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13043         indexed_or_indirect_operand predicate.
13044         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13045         (*vsx_le_perm_load_v8hi): Likewise.
13046         (*vsx_le_perm_load_v16qi): Likewise.
13047         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13048         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13049         (*vsx_le_perm_store_v8hi): Likewise.
13050         (*vsx_le_perm_store_v16qi): Likewise.
13051         (eight unnamed splitters): Likewise.
13052
13053 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
13054
13055         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13056         * config/rs6000/emmintrin.h: Likewise.
13057         * config/rs6000/mmintrin.h: Likewise.
13058         * config/rs6000/xmmintrin.h: Likewise.
13059
13060 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
13061
13062         PR c++/43486
13063         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13064         "public_flag".
13065         * tree.c (tree_nop_conversion): Return true for location wrapper
13066         nodes.
13067         (maybe_wrap_with_location): New function.
13068         (selftest::check_strip_nops): New function.
13069         (selftest::test_location_wrappers): New function.
13070         (selftest::tree_c_tests): Call it.
13071         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13072         (maybe_wrap_with_location): New decl.
13073         (EXPR_LOCATION_WRAPPER_P): New macro.
13074         (location_wrapper_p): New inline function.
13075         (tree_strip_any_location_wrapper): New inline function.
13076
13077 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
13078
13079         PR target/83735
13080         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13081         stack_realign_offset for the largest alignment of stack slot
13082         actually used.
13083         (ix86_find_max_used_stack_alignment): New function.
13084         (ix86_finalize_stack_frame_flags): Use it.  Set
13085         max_used_stack_alignment if we don't realign stack.
13086         * config/i386/i386.h (machine_function): Add
13087         max_used_stack_alignment.
13088
13089 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
13090
13091         * config/arm/arm.opt (-mbranch-cost): New option.
13092         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13093         account.
13094
13095 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
13096
13097         PR target/83629
13098         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13099         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13100
13101 2018-01-10  Richard Biener  <rguenther@suse.de>
13102
13103         PR debug/83765
13104         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13105         early out so it also covers the case where we have a non-NULL
13106         origin.
13107
13108 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
13109
13110         PR tree-optimization/83753
13111         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13112         for non-strided grouped accesses if the number of elements is 1.
13113
13114 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
13115
13116         PR target/81616
13117         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13118         * i386.h (TARGET_USE_GATHER): Define.
13119         * x86-tune.def (X86_TUNE_USE_GATHER): New.
13120
13121 2018-01-10  Martin Liska  <mliska@suse.cz>
13122
13123         PR bootstrap/82831
13124         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13125         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13126         partitioning.
13127         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13128         CLEANUP_NO_PARTITIONING is not set.
13129
13130 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
13131
13132         * doc/rtl.texi: Remove documentation of (const ...) wrappers
13133         for vectors, as a partial revert of r254296.
13134         * rtl.h (const_vec_p): Delete.
13135         (const_vec_duplicate_p): Don't test for vector CONSTs.
13136         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13137         * expmed.c (make_tree): Likewise.
13138
13139         Revert:
13140         * common.md (E, F): Use CONSTANT_P instead of checking for
13141         CONST_VECTOR.
13142         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13143         checking for CONST_VECTOR.
13144
13145 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13146
13147         PR middle-end/83575
13148         * predict.c (force_edge_cold): Handle in more sane way edges
13149         with no prediction.
13150
13151 2018-01-09  Carl Love  <cel@us.ibm.com>
13152
13153         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13154         V4SI, V4SF types.
13155         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13156         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13157         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13158         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
13159         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13160         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
13161         * config/rs6000/rs6000-protos.h: Add extern defition for
13162         rs6000_generate_float2_double_code.
13163         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13164         function.
13165         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13166         (float2_v2df): Add define_expand.
13167
13168 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
13169
13170         PR target/83628
13171         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13172         op_mode in the force_to_mode call.
13173
13174 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13175
13176         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13177         instead of checking each element individually.
13178         (aarch64_evpc_uzp): Likewise.
13179         (aarch64_evpc_zip): Likewise.
13180         (aarch64_evpc_ext): Likewise.
13181         (aarch64_evpc_rev): Likewise.
13182         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13183         instead of checking each element individually.  Return true without
13184         generating rtl if
13185         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13186         whether all selected elements come from the same input, instead of
13187         checking each element individually.  Remove calls to gen_rtx_REG,
13188         start_sequence and end_sequence and instead assert that no rtl is
13189         generated.
13190
13191 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13192
13193         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13194         order of HIGH and CONST checks.
13195
13196 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13197
13198         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13199         if the destination isn't an SSA_NAME.
13200
13201 2018-01-09  Richard Biener  <rguenther@suse.de>
13202
13203         PR tree-optimization/83668
13204         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13205         move prologue...
13206         (canonicalize_loop_form): ... here, renamed from ...
13207         (canonicalize_loop_closed_ssa_form): ... this and amended to
13208         swap successor edges for loop exit blocks to make us use
13209         the RPO order we need for initial schedule generation.
13210
13211 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
13212
13213         PR tree-optimization/64811
13214         * match.pd: When optimizing comparisons with Inf, avoid
13215         introducing or losing exceptions from comparisons with NaN.
13216
13217 2018-01-09  Martin Liska  <mliska@suse.cz>
13218
13219         PR sanitizer/82517
13220         * asan.c (shadow_mem_size): Add gcc_assert.
13221
13222 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
13223
13224         Don't save registers in main().
13225
13226         PR target/83738
13227         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13228         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13229         * config/avr/avr.c (avr_set_current_function): Don't error if
13230         naked, OS_task or OS_main are specified at the same time.
13231         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13232         OS_main.
13233         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13234         attribute.
13235         * common/config/avr/avr-common.c (avr_option_optimization_table):
13236         Switch on -mmain-is-OS_task for optimizing compilations.
13237
13238 2018-01-09  Richard Biener  <rguenther@suse.de>
13239
13240         PR tree-optimization/83572
13241         * graphite.c: Include cfganal.h.
13242         (graphite_transform_loops): Connect infinite loops to exit
13243         and remove fake edges at the end.
13244
13245 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13246
13247         * ipa-inline.c (edge_badness): Revert accidental checkin.
13248
13249 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13250
13251         PR ipa/80763
13252         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13253         symbols; not inline clones.
13254
13255 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
13256
13257         PR target/83507
13258         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13259         hard registers.  Formatting fixes.
13260
13261         PR preprocessor/83722
13262         * gcc.c (try_generate_repro): Pass
13263         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13264         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13265         do_report_bug.
13266
13267 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
13268             Kito Cheng  <kito.cheng@gmail.com>
13269
13270         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13271         (riscv_leaf_function_p): Delete.
13272         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13273
13274 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13275
13276         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13277         function.
13278         (do_ifelse): New function.
13279         (do_isel): New function.
13280         (do_sub3): New function.
13281         (do_add3): New function.
13282         (do_load_mask_compare): New function.
13283         (do_overlap_load_compare): New function.
13284         (expand_compare_loop): New function.
13285         (expand_block_compare): Call expand_compare_loop() when appropriate.
13286         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13287         option description.
13288         (-mblock-compare-inline-loop-limit): New option.
13289
13290 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13291
13292         PR target/83677
13293         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13294         Reverse order of second and third operands in first alternative.
13295         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13296         of first and second elements in UNSPEC_VPERMR vector.
13297         (altivec_expand_vec_perm_le): Likewise.
13298
13299 2018-01-08  Jeff Law  <law@redhat.com>
13300
13301         PR rtl-optimizatin/81308
13302         * tree-switch-conversion.c (cfg_altered): New file scoped static.
13303         (process_switch): If group_case_labels makes a change, then set
13304         cfg_altered.
13305         (pass_convert_switch::execute): If a switch is converted, then
13306         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
13307
13308         PR rtl-optimization/81308
13309         * recog.c (split_all_insns): Conditionally cleanup the CFG after
13310         splitting insns.
13311
13312 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
13313
13314         PR target/83663 - Revert r255946
13315         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13316         generation for cases where splatting a value is not useful.
13317         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13318         across a vec_duplicate and a paradoxical subreg forming a vector
13319         mode to a vec_concat.
13320
13321 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13322
13323         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13324         -march=armv8.3-a variants.
13325         * config/arm/t-multilib: Likewise.
13326         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
13327
13328 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13329
13330         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13331         to generate rtl.
13332         (cceq_ior_compare_complement): Give it a name so I can use it, and
13333         change boolean_or_operator predicate to boolean_operator so it can
13334         be used to generate a crand.
13335         (eqne): New code iterator.
13336         (bd/bd_neg): New code_attrs.
13337         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13338         a single define_insn.
13339         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13340         decrement (bdnzt/bdnzf/bdzt/bdzf).
13341         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13342         with the new names of the branch decrement patterns, and added the
13343         names of the branch decrement conditional patterns.
13344
13345 2018-01-08  Richard Biener  <rguenther@suse.de>
13346
13347         PR tree-optimization/83563
13348         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13349         cache.
13350
13351 2018-01-08  Richard Biener  <rguenther@suse.de>
13352
13353         PR middle-end/83713
13354         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13355
13356 2018-01-08  Richard Biener  <rguenther@suse.de>
13357
13358         PR tree-optimization/83685
13359         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13360         references to abnormals.
13361
13362 2018-01-08  Richard Biener  <rguenther@suse.de>
13363
13364         PR lto/83719
13365         * dwarf2out.c (output_indirect_strings): Handle empty
13366         skeleton_debug_str_hash.
13367         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13368
13369 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
13370
13371         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13372         (emit_store_direct): Likewise.
13373         (arc_trampoline_adjust_address): Likewise.
13374         (arc_asm_trampoline_template): New function.
13375         (arc_initialize_trampoline): Use asm_trampoline_template.
13376         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13377         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13378         * config/arc/arc.md (flush_icache): Delete pattern.
13379
13380 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
13381
13382         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13383         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13384         munaligned-access.
13385
13386 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13387
13388         PR target/83681
13389         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13390         by not USED_FOR_TARGET.
13391         (make_pass_resolve_sw_modes): Likewise.
13392
13393 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13394
13395         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13396         USED_FOR_TARGET.
13397
13398 2018-01-08  Richard Biener  <rguenther@suse.de>
13399
13400         PR middle-end/83580
13401         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13402
13403 2018-01-08  Richard Biener  <rguenther@suse.de>
13404
13405         PR middle-end/83517
13406         * match.pd ((t * 2) / 2) -> t): Add missing :c.
13407
13408 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
13409
13410         PR middle-end/81897
13411         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13412         basic blocks with a small number of successors.
13413         (convert_control_dep_chain_into_preds): Improve handling of
13414         forwarder blocks.
13415         (dump_predicates): Split apart into...
13416         (dump_pred_chain): ...here...
13417         (dump_pred_info): ...and here.
13418         (can_one_predicate_be_invalidated_p): Add debugging printfs.
13419         (can_chain_union_be_invalidated_p): Improve check for invalidation
13420         of paths.
13421         (uninit_uses_cannot_happen): Avoid unnecessary if
13422         convert_control_dep_chain_into_preds yielded nothing.
13423
13424 2018-01-06  Martin Sebor  <msebor@redhat.com>
13425
13426         PR tree-optimization/83640
13427         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13428         subtracting negative offset from size.
13429         (builtin_access::overlap): Adjust offset bounds of the access to fall
13430         within the size of the object if possible.
13431
13432 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
13433
13434         PR rtl-optimization/83699
13435         * expmed.c (extract_bit_field_1): Restrict the vector usage of
13436         extract_bit_field_as_subreg to cases in which the extracted
13437         value is also a vector.
13438
13439         * lra-constraints.c (process_alt_operands): Test for the equivalence
13440         substitutions when detecting a possible reload cycle.
13441
13442 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
13443
13444         PR debug/83480
13445         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13446         by default if flag_selective_schedling{,2}.  Formatting fixes.
13447
13448         PR rtl-optimization/83682
13449         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13450         if it has non-VECTOR_MODE element mode.
13451         (vec_duplicate_p): Likewise.
13452
13453         PR middle-end/83694
13454         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13455         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13456
13457 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
13458
13459         PR target/83604
13460         * config/i386/i386-builtin.def
13461         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13462         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13463         Require also OPTION_MASK_ISA_AVX512F in addition to
13464         OPTION_MASK_ISA_GFNI.
13465         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13466         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13467         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13468         to OPTION_MASK_ISA_GFNI.
13469         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13470         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13471         OPTION_MASK_ISA_AVX512BW.
13472         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13473         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13474         addition to OPTION_MASK_ISA_GFNI.
13475         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13476         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13477         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13478         to OPTION_MASK_ISA_GFNI.
13479         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13480         a requirement for all ISAs rather than any of them with a few
13481         exceptions.
13482         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13483         processing.
13484         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13485         bitmasks to be enabled with 3 exceptions, instead of requiring any
13486         enabled ISA with lots of exceptions.
13487         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13488         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13489         Change avx512bw in isa attribute to avx512f.
13490         * config/i386/sgxintrin.h: Add license boilerplate.
13491         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
13492         to __AVX512F__ and __AVX512VL to __AVX512VL__.
13493         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13494         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13495         defined.
13496         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13497         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13498         temporarily sse2 rather than sse if not enabled already.
13499
13500         PR target/83604
13501         * config/i386/sse.md (VI248_VLBW): Rename to ...
13502         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
13503         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13504         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13505         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13506         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13507         mode iterator instead of VI248_VLBW.
13508
13509 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
13510
13511         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13512         (record_modified): Skip clobbers; add debug output.
13513         (param_change_prob): Use sreal frequencies.
13514
13515 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
13516
13517         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13518         punt for user-aligned variables.
13519
13520 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
13521
13522         * tree-chrec.c (chrec_contains_symbols): Return true for
13523         POLY_INT_CST.
13524
13525 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
13526
13527         PR target/82439
13528         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13529         of (x|y) == x for BICS pattern.
13530
13531 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
13532
13533         PR tree-optimization/83605
13534         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13535         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13536         can throw.
13537
13538 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13539
13540         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13541         * config/epiphany/rtems.h: New file.
13542
13543 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13544             Uros Bizjak  <ubizjak@gmail.com>
13545
13546         PR target/83554
13547         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13548         QIreg_operand instead of register_operand predicate.
13549         * config/i386/i386.c (ix86_rop_should_change_byte_p,
13550         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13551         comments instead of -fmitigate[-_]rop.
13552
13553 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13554
13555         PR bootstrap/81926
13556         * cgraphunit.c (symbol_table::compile): Switch to text_section
13557         before calling assembly_start debug hook.
13558         * run-rtl-passes.c (run_rtl_passes): Likewise.
13559         Include output.h.
13560
13561 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13562
13563         * tree-vrp.c (extract_range_from_binary_expr_1): Check
13564         range_int_cst_p rather than !symbolic_range_p before calling
13565         extract_range_from_multiplicative_op_1.
13566
13567 2018-01-04  Jeff Law  <law@redhat.com>
13568
13569         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13570         redundant test in assertion.
13571
13572 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13573
13574         * doc/rtl.texi: Document machine_mode wrapper classes.
13575
13576 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13577
13578         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13579         using tree_to_uhwi.
13580
13581 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13582
13583         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13584         the VEC_PERM_EXPR fold to fail.
13585
13586 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13587
13588         PR debug/83585
13589         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13590         to switched_sections.
13591
13592 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13593
13594         PR target/83680
13595         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13596         test for d.testing.
13597
13598 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
13599
13600         PR target/83387
13601         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13602         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13603
13604 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13605
13606         PR debug/83666
13607         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13608         is BLKmode and bitpos not zero or mode change is needed.
13609
13610 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13611
13612         PR target/83675
13613         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13614         TARGET_VIS2.
13615
13616 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
13617
13618         PR target/83628
13619         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13620         instead of MULT rtx.  Update all corresponding splitters.
13621         (*saddl_se): Ditto.
13622         (*ssub<modesuffix>): Ditto.
13623         (*ssubl_se): Ditto.
13624         (*cmp_sadd_di): Update split patterns.
13625         (*cmp_sadd_si): Ditto.
13626         (*cmp_sadd_sidi): Ditto.
13627         (*cmp_ssub_di): Ditto.
13628         (*cmp_ssub_si): Ditto.
13629         (*cmp_ssub_sidi): Ditto.
13630         * config/alpha/predicates.md (const23_operand): New predicate.
13631         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13632         Look for ASHIFT, not MULT inner operand.
13633         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13634
13635 2018-01-04  Martin Liska  <mliska@suse.cz>
13636
13637         PR gcov-profile/83669
13638         * gcov.c (output_intermediate_file): Add version to intermediate
13639         gcov file.
13640         * doc/gcov.texi: Document new field 'version' in intermediate
13641         file format. Fix location of '-k' option of gcov command.
13642
13643 2018-01-04  Martin Liska  <mliska@suse.cz>
13644
13645         PR ipa/82352
13646         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13647
13648 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13649
13650         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13651
13652 2018-01-03  Martin Sebor  <msebor@redhat.com>
13653
13654         PR tree-optimization/83655
13655         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13656         checking calls with invalid arguments.
13657
13658 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13659
13660         * tree-vect-stmts.c (vect_get_store_rhs): New function.
13661         (vectorizable_mask_load_store): Delete.
13662         (vectorizable_call): Return false for masked loads and stores.
13663         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
13664         instead of gimple_assign_rhs1.
13665         (vectorizable_load): Handle IFN_MASK_LOAD.
13666         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13667
13668 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13669
13670         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13671         split out from..,
13672         (vectorizable_mask_load_store): ...here.
13673         (vectorizable_load): ...and here.
13674
13675 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13676
13677         * tree-vect-stmts.c (vect_build_all_ones_mask)
13678         (vect_build_zero_merge_argument): New functions, split out from...
13679         (vectorizable_load): ...here.
13680
13681 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13682
13683         * tree-vect-stmts.c (vect_check_store_rhs): New function,
13684         split out from...
13685         (vectorizable_mask_load_store): ...here.
13686         (vectorizable_store): ...and here.
13687
13688 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13689
13690         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13691         split out from...
13692         (vectorizable_mask_load_store): ...here.
13693
13694 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13695
13696         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13697         (vect_model_store_cost): Take a vec_load_store_type instead of a
13698         vect_def_type.
13699         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13700         (vect_model_store_cost): Take a vec_load_store_type instead of a
13701         vect_def_type.
13702         (vectorizable_mask_load_store): Update accordingly.
13703         (vectorizable_store): Likewise.
13704         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13705
13706 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13707
13708         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13709         IFN_MASK_LOAD calls here rather than...
13710         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13711
13712 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13713             Alan Hayward  <alan.hayward@arm.com>
13714             David Sherwood  <david.sherwood@arm.com>
13715
13716         * expmed.c (extract_bit_field_1): For vector extracts,
13717         fall back to extract_bit_field_as_subreg if vec_extract
13718         isn't available.
13719
13720 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13721             Alan Hayward  <alan.hayward@arm.com>
13722             David Sherwood  <david.sherwood@arm.com>
13723
13724         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13725         they are variable or constant sized.
13726         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13727         slots for constant-sized data.
13728
13729 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13730             Alan Hayward  <alan.hayward@arm.com>
13731             David Sherwood  <david.sherwood@arm.com>
13732
13733         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13734         handling COND_EXPRs with boolean comparisons, try to find a better
13735         basis for the mask type than the boolean itself.
13736
13737 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13738
13739         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13740         is calculated and how it can be overridden.
13741         * genmodes.c (max_bitsize_mode_any_mode): New variable.
13742         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13743         if defined.
13744         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13745         if nonzero.
13746
13747 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13748             Alan Hayward  <alan.hayward@arm.com>
13749             David Sherwood  <david.sherwood@arm.com>
13750
13751         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13752         Remove the mode argument.
13753         (aarch64_simd_valid_immediate): Remove the mode and inverse
13754         arguments.
13755         * config/aarch64/iterators.md (bitsize): New iterator.
13756         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13757         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13758         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13759         aarch64_simd_valid_immediate.
13760         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13761         (aarch64_reg_or_bic_imm): Likewise.
13762         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13763         with an insn_type enum and msl with a modifier_type enum.
13764         Replace element_width with a scalar_mode.  Change the shift
13765         to unsigned int.  Add constructors for scalar_float_mode and
13766         scalar_int_mode elements.
13767         (aarch64_vect_float_const_representable_p): Delete.
13768         (aarch64_can_const_movi_rtx_p)
13769         (aarch64_simd_scalar_immediate_valid_for_move)
13770         (aarch64_simd_make_constant): Update call to
13771         aarch64_simd_valid_immediate.
13772         (aarch64_advsimd_valid_immediate_hs): New function.
13773         (aarch64_advsimd_valid_immediate): Likewise.
13774         (aarch64_simd_valid_immediate): Remove mode and inverse
13775         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
13776         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13777         and aarch64_float_const_representable_p on the result.
13778         (aarch64_output_simd_mov_immediate): Remove mode argument.
13779         Update call to aarch64_simd_valid_immediate and use of
13780         simd_immediate_info.
13781         (aarch64_output_scalar_simd_mov_immediate): Update call
13782         accordingly.
13783
13784 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13785             Alan Hayward  <alan.hayward@arm.com>
13786             David Sherwood  <david.sherwood@arm.com>
13787
13788         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13789         (mode_nunits): Likewise CONST_MODE_NUNITS.
13790         * machmode.def (ADJUST_NUNITS): Document.
13791         * genmodes.c (mode_data::need_nunits_adj): New field.
13792         (blank_mode): Update accordingly.
13793         (adj_nunits): New variable.
13794         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13795         parameter.
13796         (emit_mode_size_inline): Set need_bytesize_adj for all modes
13797         listed in adj_nunits.
13798         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13799         listed in adj_nunits.  Don't emit case statements for such modes.
13800         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13801         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
13802         nothing if adj_nunits is nonnull.
13803         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13804         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13805         (emit_mode_fbit): Update use of print_maybe_const_decl.
13806         (emit_move_size): Likewise.  Treat the array as non-const
13807         if adj_nunits.
13808         (emit_mode_adjustments): Handle adj_nunits.
13809
13810 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13811
13812         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13813         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13814         (VECTOR_MODES): Use it.
13815         (make_vector_modes): Take the prefix as an argument.
13816
13817 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13818             Alan Hayward  <alan.hayward@arm.com>
13819             David Sherwood  <david.sherwood@arm.com>
13820
13821         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13822         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13823         for MODE_VECTOR_BOOL.
13824         * machmode.def (VECTOR_BOOL_MODE): Document.
13825         * genmodes.c (VECTOR_BOOL_MODE): New macro.
13826         (make_vector_bool_mode): New function.
13827         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13828         MODE_VECTOR_BOOL.
13829         * lto-streamer-in.c (lto_input_mode_table): Likewise.
13830         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13831         Likewise.
13832         * stor-layout.c (int_mode_for_mode): Likewise.
13833         * tree.c (build_vector_type_for_mode): Likewise.
13834         * varasm.c (output_constant_pool_2): Likewise.
13835         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13836         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
13837         for MODE_VECTOR_BOOL.
13838         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13839         of mode class checks.
13840         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13841         instead of a list of mode class checks.
13842         (expand_vector_scalar_condition): Likewise.
13843         (type_for_widest_vector_mode): Handle BImode as an inner mode.
13844
13845 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13846             Alan Hayward  <alan.hayward@arm.com>
13847             David Sherwood  <david.sherwood@arm.com>
13848
13849         * machmode.h (mode_size): Change from unsigned short to
13850         poly_uint16_pod.
13851         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13852         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13853         or if measurement_type is not polynomial.
13854         (fixed_size_mode::includes_p): Check for constant-sized modes.
13855         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13856         return a poly_uint16 rather than an unsigned short.
13857         (emit_mode_size): Change the type of mode_size from unsigned short
13858         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
13859         (emit_mode_adjustments): Cope with polynomial vector sizes.
13860         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13861         for GET_MODE_SIZE.
13862         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13863         for GET_MODE_SIZE.
13864         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13865         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13866         * caller-save.c (setup_save_areas): Likewise.
13867         (replace_reg_with_saved_mem): Likewise.
13868         * calls.c (emit_library_call_value_1): Likewise.
13869         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13870         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13871         (gen_lowpart_for_combine): Likewise.
13872         * convert.c (convert_to_integer_1): Likewise.
13873         * cse.c (equiv_constant, cse_insn): Likewise.
13874         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13875         (cselib_subst_to_values): Likewise.
13876         * dce.c (word_dce_process_block): Likewise.
13877         * df-problems.c (df_word_lr_mark_ref): Likewise.
13878         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13879         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13880         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13881         (rtl_for_decl_location): Likewise.
13882         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13883         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13884         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13885         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13886         (expand_expr_real_1): Likewise.
13887         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13888         (pad_below): Likewise.
13889         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13890         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13891         * ira.c (get_subreg_tracking_sizes): Likewise.
13892         * ira-build.c (ira_create_allocno_objects): Likewise.
13893         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13894         (ira_sort_regnos_for_alter_reg): Likewise.
13895         * ira-costs.c (record_operand_costs): Likewise.
13896         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13897         (resolve_simple_move): Likewise.
13898         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13899         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13900         (lra_constraints): Likewise.
13901         (CONST_POOL_OK_P): Reject variable-sized modes.
13902         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13903         (add_pseudo_to_slot, lra_spill): Likewise.
13904         * omp-low.c (omp_clause_aligned_alignment): Likewise.
13905         * optabs-query.c (get_best_extraction_insn): Likewise.
13906         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13907         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13908         (expand_mult_highpart, valid_multiword_target_p): Likewise.
13909         * recog.c (offsettable_address_addr_space_p): Likewise.
13910         * regcprop.c (maybe_mode_change): Likewise.
13911         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13912         * regrename.c (build_def_use): Likewise.
13913         * regstat.c (dump_reg_info): Likewise.
13914         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13915         (find_reloads, find_reloads_subreg_address): Likewise.
13916         * reload1.c (eliminate_regs_1): Likewise.
13917         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13918         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13919         (simplify_binary_operation_1, simplify_subreg): Likewise.
13920         * targhooks.c (default_function_arg_padding): Likewise.
13921         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13922         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13923         (verify_gimple_assign_ternary): Likewise.
13924         * tree-inline.c (estimate_move_cost): Likewise.
13925         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13926         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13927         (get_address_cost_ainc): Likewise.
13928         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13929         (vect_supportable_dr_alignment): Likewise.
13930         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13931         (vectorizable_reduction): Likewise.
13932         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13933         (vectorizable_operation, vectorizable_load): Likewise.
13934         * tree.c (build_same_sized_truth_vector_type): Likewise.
13935         * valtrack.c (cleanup_auto_inc_dec): Likewise.
13936         * var-tracking.c (emit_note_insn_var_location): Likewise.
13937         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13938         (ADDR_VEC_ALIGN): Likewise.
13939
13940 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13941             Alan Hayward  <alan.hayward@arm.com>
13942             David Sherwood  <david.sherwood@arm.com>
13943
13944         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13945         unsigned short.
13946         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13947         or if measurement_type is polynomial.
13948         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13949         * combine.c (make_extraction): Likewise.
13950         * dse.c (find_shift_sequence): Likewise.
13951         * dwarf2out.c (mem_loc_descriptor): Likewise.
13952         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13953         (extract_bit_field, extract_low_bits): Likewise.
13954         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13955         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13956         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13957         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13958         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13959         * reload.c (find_reloads): Likewise.
13960         * reload1.c (alter_reg): Likewise.
13961         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13962         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13963         * tree-if-conv.c (predicate_mem_writes): Likewise.
13964         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13965         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13966         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13967         * valtrack.c (dead_debug_insert_temp): Likewise.
13968         * varasm.c (mergeable_constant_section): Likewise.
13969         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13970
13971 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13972             Alan Hayward  <alan.hayward@arm.com>
13973             David Sherwood  <david.sherwood@arm.com>
13974
13975         * expr.c (expand_assignment): Cope with polynomial mode sizes
13976         when assigning to a CONCAT.
13977
13978 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13979             Alan Hayward  <alan.hayward@arm.com>
13980             David Sherwood  <david.sherwood@arm.com>
13981
13982         * machmode.h (mode_precision): Change from unsigned short to
13983         poly_uint16_pod.
13984         (mode_to_precision): Return a poly_uint16 rather than an unsigned
13985         short.
13986         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13987         or if measurement_type is not polynomial.
13988         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
13989         in which the mode is already known to be a scalar_int_mode.
13990         * genmodes.c (emit_mode_precision): Change the type of mode_precision
13991         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
13992         initializer.
13993         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13994         for GET_MODE_PRECISION.
13995         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13996         for GET_MODE_PRECISION.
13997         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13998         as polynomial.
13999         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14000         (expand_field_assignment, make_extraction): Likewise.
14001         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14002         (get_last_value): Likewise.
14003         * convert.c (convert_to_integer_1): Likewise.
14004         * cse.c (cse_insn): Likewise.
14005         * expr.c (expand_expr_real_1): Likewise.
14006         * lra-constraints.c (simplify_operand_subreg): Likewise.
14007         * optabs-query.c (can_atomic_load_p): Likewise.
14008         * optabs.c (expand_atomic_load): Likewise.
14009         (expand_atomic_store): Likewise.
14010         * ree.c (combine_reaching_defs): Likewise.
14011         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14012         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14013         * tree.h (type_has_mode_precision_p): Likewise.
14014         * ubsan.c (instrument_si_overflow): Likewise.
14015
14016 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14017             Alan Hayward  <alan.hayward@arm.com>
14018             David Sherwood  <david.sherwood@arm.com>
14019
14020         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14021         polynomial numbers of units.
14022         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14023         (valid_vector_subparts_p): New function.
14024         (build_vector_type): Remove temporary shim and take the number
14025         of units as a poly_uint64 rather than an int.
14026         (build_opaque_vector_type): Take the number of units as a
14027         poly_uint64 rather than an int.
14028         * tree.c (build_vector_from_ctor): Handle polynomial
14029         TYPE_VECTOR_SUBPARTS.
14030         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14031         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14032         (build_vector_from_val): If the number of units is variable,
14033         use build_vec_duplicate_cst for constant operands and
14034         VEC_DUPLICATE_EXPR otherwise.
14035         (make_vector_type): Remove temporary is_constant ().
14036         (build_vector_type, build_opaque_vector_type): Take the number of
14037         units as a poly_uint64 rather than an int.
14038         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14039         VECTOR_CST_NELTS.
14040         * cfgexpand.c (expand_debug_expr): Likewise.
14041         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14042         (store_constructor, expand_expr_real_1): Likewise.
14043         (const_scalar_mask_from_tree): Likewise.
14044         * fold-const-call.c (fold_const_reduction): Likewise.
14045         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14046         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14047         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14048         (fold_relational_const): Likewise.
14049         (native_interpret_vector): Likewise.  Change the size from an
14050         int to an unsigned int.
14051         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14052         TYPE_VECTOR_SUBPARTS.
14053         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14054         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14055         duplicating a non-constant operand into a variable-length vector.
14056         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14057         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14058         * ipa-icf.c (sem_variable::equals): Likewise.
14059         * match.pd: Likewise.
14060         * omp-simd-clone.c (simd_clone_subparts): Likewise.
14061         * print-tree.c (print_node): Likewise.
14062         * stor-layout.c (layout_type): Likewise.
14063         * targhooks.c (default_builtin_vectorization_cost): Likewise.
14064         * tree-cfg.c (verify_gimple_comparison): Likewise.
14065         (verify_gimple_assign_binary): Likewise.
14066         (verify_gimple_assign_ternary): Likewise.
14067         (verify_gimple_assign_single): Likewise.
14068         * tree-pretty-print.c (dump_generic_node): Likewise.
14069         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14070         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14071         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14072         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14073         (vect_shift_permute_load_chain): Likewise.
14074         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14075         (expand_vector_condition, optimize_vector_constructor): Likewise.
14076         (lower_vec_perm, get_compute_type): Likewise.
14077         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14078         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14079         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14080         (vect_recog_mask_conversion_pattern): Likewise.
14081         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14082         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14083         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14084         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14085         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14086         (vectorizable_shift, vectorizable_operation, vectorizable_store)
14087         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14088         (supportable_widening_operation): Likewise.
14089         (supportable_narrowing_operation): Likewise.
14090         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14091         Likewise.
14092         * varasm.c (output_constant): Likewise.
14093
14094 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14095             Alan Hayward  <alan.hayward@arm.com>
14096             David Sherwood  <david.sherwood@arm.com>
14097
14098         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14099         so that both the length == 3 and length != 3 cases set up their
14100         own permute vectors.  Add comments explaining why we know the
14101         number of elements is constant.
14102         (vect_permute_load_chain): Likewise.
14103
14104 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14105             Alan Hayward  <alan.hayward@arm.com>
14106             David Sherwood  <david.sherwood@arm.com>
14107
14108         * machmode.h (mode_nunits): Change from unsigned char to
14109         poly_uint16_pod.
14110         (ONLY_FIXED_SIZE_MODES): New macro.
14111         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14112         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14113         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14114         New typedefs.
14115         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14116         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14117         or if measurement_type is not polynomial.
14118         * genmodes.c (ZERO_COEFFS): New macro.
14119         (emit_mode_nunits_inline): Make mode_nunits_inline return a
14120         poly_uint16.
14121         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14122         Use ZERO_COEFFS when emitting initializers.
14123         * data-streamer.h (bp_pack_poly_value): New function.
14124         (bp_unpack_poly_value): Likewise.
14125         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14126         for GET_MODE_NUNITS.
14127         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14128         for GET_MODE_NUNITS.
14129         * tree.c (make_vector_type): Remove temporary shim and make
14130         the real function take the number of units as a poly_uint64
14131         rather than an int.
14132         (build_vector_type_for_mode): Handle polynomial nunits.
14133         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14134         * emit-rtl.c (const_vec_series_p_1): Likewise.
14135         (gen_rtx_CONST_VECTOR): Likewise.
14136         * fold-const.c (test_vec_duplicate_folding): Likewise.
14137         * genrecog.c (validate_pattern): Likewise.
14138         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14139         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14140         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14141         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14142         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14143         * rtlanal.c (subreg_get_info): Likewise.
14144         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14145         (vect_grouped_load_supported): Likewise.
14146         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14147         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14148         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14149         (simplify_const_unary_operation, simplify_binary_operation_1)
14150         (simplify_const_binary_operation, simplify_ternary_operation)
14151         (test_vector_ops_duplicate, test_vector_ops): Likewise.
14152         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14153         instead of CONST_VECTOR_NUNITS.
14154         * varasm.c (output_constant_pool_2): Likewise.
14155         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14156         explicit-encoded elements in the XVEC for variable-length vectors.
14157
14158 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14159
14160         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14161
14162 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14163             Alan Hayward  <alan.hayward@arm.com>
14164             David Sherwood  <david.sherwood@arm.com>
14165
14166         * coretypes.h (fixed_size_mode): Declare.
14167         (fixed_size_mode_pod): New typedef.
14168         * builtins.h (target_builtins::x_apply_args_mode)
14169         (target_builtins::x_apply_result_mode): Change type to
14170         fixed_size_mode_pod.
14171         * builtins.c (apply_args_size, apply_result_size, result_vector)
14172         (expand_builtin_apply_args_1, expand_builtin_apply)
14173         (expand_builtin_return): Update accordingly.
14174
14175 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14176
14177         * cse.c (hash_rtx_cb): Hash only the encoded elements.
14178         * cselib.c (cselib_hash_rtx): Likewise.
14179         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14180         CONST_VECTOR encoding.
14181
14182 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14183             Jeff Law  <law@redhat.com>
14184
14185         PR target/83641
14186         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14187         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14188         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14189         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14190
14191         PR target/83641
14192         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14193         explicitly probe *sp in a noreturn function if there were any callee
14194         register saves or frame pointer is needed.
14195
14196 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14197
14198         PR debug/83621
14199         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14200         BLKmode for ternary, binary or unary expressions.
14201
14202         PR debug/83645
14203         * var-tracking.c (delete_vta_debug_insn): New inline function.
14204         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14205         insns from get_insns () to NULL instead of each bb separately.
14206         Use delete_vta_debug_insn.  No longer static.
14207         (vt_debug_insns_local, variable_tracking_main_1): Adjust
14208         delete_vta_debug_insns callers.
14209         * rtl.h (delete_vta_debug_insns): Declare.
14210         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14211         instead of variable_tracking_main.
14212
14213 2018-01-03  Martin Sebor  <msebor@redhat.com>
14214
14215         PR tree-optimization/83603
14216         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14217         arguments past the endof the argument list in functions declared
14218         without a prototype.
14219         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14220         Avoid checking when arguments are null.
14221
14222 2018-01-03  Martin Sebor  <msebor@redhat.com>
14223
14224         PR c/83559
14225         * doc/extend.texi (attribute const): Fix a typo.
14226         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14227         issuing -Wsuggest-attribute for void functions.
14228
14229 2018-01-03  Martin Sebor  <msebor@redhat.com>
14230
14231         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14232         offset_int::from instead of wide_int::to_shwi.
14233         (maybe_diag_overlap): Remove assertion.
14234         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14235         * gimple-ssa-sprintf.c (format_directive): Same.
14236         (parse_directive): Same.
14237         (sprintf_dom_walker::compute_format_length): Same.
14238         (try_substitute_return_value): Same.
14239
14240 2018-01-03  Jeff Law  <law@redhat.com>
14241
14242         PR middle-end/83654
14243         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14244         non-constant residual for zero at runtime and avoid probing in
14245         that case.  Reorganize code for trailing problem to mirror handling
14246         of the residual.
14247
14248 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14249
14250         PR tree-optimization/83501
14251         * tree-ssa-strlen.c (get_string_cst): New.
14252         (handle_char_store): Call get_string_cst.
14253
14254 2018-01-03  Martin Liska  <mliska@suse.cz>
14255
14256         PR tree-optimization/83593
14257         * tree-ssa-strlen.c: Include tree-cfg.h.
14258         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14259         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14260         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14261         to false.
14262         (strlen_dom_walker::before_dom_children): Call
14263         gimple_purge_dead_eh_edges. Dump tranformation with details
14264         dump flags.
14265         (strlen_dom_walker::before_dom_children): Update call by adding
14266         new argument cleanup_eh.
14267         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14268
14269 2018-01-03  Martin Liska  <mliska@suse.cz>
14270
14271         PR ipa/83549
14272         * cif-code.def (VARIADIC_THUNK): New enum value.
14273         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14274         thunks.
14275
14276 2018-01-03  Jan Beulich  <jbeulich@suse.com>
14277
14278         * sse.md (mov<mode>_internal): Tighten condition for when to use
14279         vmovdqu<ssescalarsize> for TI and OI modes.
14280
14281 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14282
14283         Update copyright years.
14284
14285 2018-01-03  Martin Liska  <mliska@suse.cz>
14286
14287         PR ipa/83594
14288         * ipa-visibility.c (function_and_variable_visibility): Skip
14289         functions with noipa attribure.
14290
14291 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14292
14293         * gcc.c (process_command): Update copyright notice dates.
14294         * gcov-dump.c (print_version): Ditto.
14295         * gcov.c (print_version): Ditto.
14296         * gcov-tool.c (print_version): Ditto.
14297         * gengtype.c (create_file): Ditto.
14298         * doc/cpp.texi: Bump @copying's copyright year.
14299         * doc/cppinternals.texi: Ditto.
14300         * doc/gcc.texi: Ditto.
14301         * doc/gccint.texi: Ditto.
14302         * doc/gcov.texi: Ditto.
14303         * doc/install.texi: Ditto.
14304         * doc/invoke.texi: Ditto.
14305
14306 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14307
14308         * vector-builder.h (vector_builder::m_full_nelts): Change from
14309         unsigned int to poly_uint64.
14310         (vector_builder::full_nelts): Update prototype accordingly.
14311         (vector_builder::new_vector): Likewise.
14312         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14313         (vector_builder::operator ==): Likewise.
14314         (vector_builder::finalize): Likewise.
14315         * int-vector-builder.h (int_vector_builder::int_vector_builder):
14316         Take the number of elements as a poly_uint64 rather than an
14317         unsigned int.
14318         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14319         from unsigned int to poly_uint64.
14320         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14321         (vec_perm_indices::new_vector): Likewise.
14322         (vec_perm_indices::length): Likewise.
14323         (vec_perm_indices::nelts_per_input): Likewise.
14324         (vec_perm_indices::input_nelts): Likewise.
14325         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14326         number of elements per input as a poly_uint64 rather than an
14327         unsigned int.  Use the original encoding for variable-length
14328         vectors, rather than clamping each individual element.
14329         For the second and subsequent elements in each pattern,
14330         clamp the step and base before clamping their sum.
14331         (vec_perm_indices::series_p): Handle polynomial element counts.
14332         (vec_perm_indices::all_in_range_p): Likewise.
14333         (vec_perm_indices_to_tree): Likewise.
14334         (vec_perm_indices_to_rtx): Likewise.
14335         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14336         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14337         (tree_vector_builder::new_binary_operation): Handle polynomial
14338         element counts.  Return false if we need to know the number
14339         of elements at compile time.
14340         * fold-const.c (fold_vec_perm): Punt if the number of elements
14341         isn't known at compile time.
14342
14343 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14344
14345         * vec-perm-indices.h (vec_perm_builder): Change element type
14346         from HOST_WIDE_INT to poly_int64.
14347         (vec_perm_indices::element_type): Update accordingly.
14348         (vec_perm_indices::clamp): Handle polynomial element_types.
14349         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14350         (vec_perm_indices::all_in_range_p): Likewise.
14351         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14352         than shwi trees.
14353         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14354         polynomial vec_perm_indices element types.
14355         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14356         * fold-const.c (fold_vec_perm): Likewise.
14357         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14358         * tree-vect-generic.c (lower_vec_perm): Likewise.
14359         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14360         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14361         element type to HOST_WIDE_INT.
14362
14363 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14364             Alan Hayward  <alan.hayward@arm.com>
14365             David Sherwood  <david.sherwood@arm.com>
14366
14367         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14368         rather than an int.  Use plus_constant.
14369         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14370         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14371
14372 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14373             Alan Hayward  <alan.hayward@arm.com>
14374             David Sherwood  <david.sherwood@arm.com>
14375
14376         * calls.c (emit_call_1, expand_call): Change struct_value_size from
14377         a HOST_WIDE_INT to a poly_int64.
14378
14379 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14380             Alan Hayward  <alan.hayward@arm.com>
14381             David Sherwood  <david.sherwood@arm.com>
14382
14383         * calls.c (load_register_parameters): Cope with polynomial
14384         mode sizes.  Require a constant size for BLKmode parameters
14385         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
14386         forces a parameter to be padded at the lsb end in order to
14387         fill a complete number of words, require the parameter size
14388         to be ordered wrt UNITS_PER_WORD.
14389
14390 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14391             Alan Hayward  <alan.hayward@arm.com>
14392             David Sherwood  <david.sherwood@arm.com>
14393
14394         * reload1.c (spill_stack_slot_width): Change element type
14395         from unsigned int to poly_uint64_pod.
14396         (alter_reg): Treat mode sizes as polynomial.
14397
14398 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14399             Alan Hayward  <alan.hayward@arm.com>
14400             David Sherwood  <david.sherwood@arm.com>
14401
14402         * reload.c (complex_word_subreg_p): New function.
14403         (reload_inner_reg_of_subreg, push_reload): Use it.
14404
14405 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14406             Alan Hayward  <alan.hayward@arm.com>
14407             David Sherwood  <david.sherwood@arm.com>
14408
14409         * lra-constraints.c (process_alt_operands): Reject matched
14410         operands whose sizes aren't ordered.
14411         (match_reload): Refer to this check here.
14412
14413 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14414             Alan Hayward  <alan.hayward@arm.com>
14415             David Sherwood  <david.sherwood@arm.com>
14416
14417         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14418         that the mode size is in the set {1, 2, 4, 8, 16}.
14419
14420 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14421             Alan Hayward  <alan.hayward@arm.com>
14422             David Sherwood  <david.sherwood@arm.com>
14423
14424         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14425         Use plus_constant instead of gen_rtx_PLUS.
14426
14427 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14428             Alan Hayward  <alan.hayward@arm.com>
14429             David Sherwood  <david.sherwood@arm.com>
14430
14431         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14432         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14433         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14434         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14435         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14436         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14437         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14438         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14439         * config/i386/i386.c (ix86_push_rounding): ...this new function.
14440         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14441         a poly_int64.
14442         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14443         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14444         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14445         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14446         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14447         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14448         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14449         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14450         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14451         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14452         function.
14453         * expr.c (emit_move_resolve_push): Treat the input and result
14454         of PUSH_ROUNDING as a poly_int64.
14455         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14456         (emit_push_insn): Likewise.
14457         * lra-eliminations.c (mark_not_eliminable): Likewise.
14458         * recog.c (push_operand): Likewise.
14459         * reload1.c (elimination_effects): Likewise.
14460         * rtlanal.c (nonzero_bits1): Likewise.
14461         * calls.c (store_one_arg): Likewise.  Require the padding to be
14462         known at compile time.
14463
14464 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14465             Alan Hayward  <alan.hayward@arm.com>
14466             David Sherwood  <david.sherwood@arm.com>
14467
14468         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14469         Use plus_constant instead of gen_rtx_PLUS.
14470
14471 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14472             Alan Hayward  <alan.hayward@arm.com>
14473             David Sherwood  <david.sherwood@arm.com>
14474
14475         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14476         rather than an int.
14477
14478 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14479             Alan Hayward  <alan.hayward@arm.com>
14480             David Sherwood  <david.sherwood@arm.com>
14481
14482         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14483         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14484         via stack temporaries.  Treat the mode size as polynomial too.
14485
14486 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14487             Alan Hayward  <alan.hayward@arm.com>
14488             David Sherwood  <david.sherwood@arm.com>
14489
14490         * expr.c (expand_expr_real_2): When handling conversions involving
14491         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14492         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
14493         as a poly_uint64 too.
14494
14495 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14496             Alan Hayward  <alan.hayward@arm.com>
14497             David Sherwood  <david.sherwood@arm.com>
14498
14499         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14500
14501 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14502             Alan Hayward  <alan.hayward@arm.com>
14503             David Sherwood  <david.sherwood@arm.com>
14504
14505         * combine.c (can_change_dest_mode): Handle polynomial
14506         REGMODE_NATURAL_SIZE.
14507         * expmed.c (store_bit_field_1): Likewise.
14508         * expr.c (store_constructor): Likewise.
14509         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14510         and polynomial REGMODE_NATURAL_SIZE.
14511         (gen_lowpart_common): Likewise.
14512         * reginfo.c (record_subregs_of_mode): Likewise.
14513         * rtlanal.c (read_modify_subreg_p): Likewise.
14514
14515 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14516             Alan Hayward  <alan.hayward@arm.com>
14517             David Sherwood  <david.sherwood@arm.com>
14518
14519         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14520         numbers of elements.
14521
14522 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14523             Alan Hayward  <alan.hayward@arm.com>
14524             David Sherwood  <david.sherwood@arm.com>
14525
14526         * match.pd: Cope with polynomial numbers of vector elements.
14527
14528 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14529             Alan Hayward  <alan.hayward@arm.com>
14530             David Sherwood  <david.sherwood@arm.com>
14531
14532         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14533         in a POINTER_PLUS_EXPR.
14534
14535 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14536             Alan Hayward  <alan.hayward@arm.com>
14537             David Sherwood  <david.sherwood@arm.com>
14538
14539         * omp-simd-clone.c (simd_clone_subparts): New function.
14540         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14541         (ipa_simd_modify_function_body): Likewise.
14542
14543 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14544             Alan Hayward  <alan.hayward@arm.com>
14545             David Sherwood  <david.sherwood@arm.com>
14546
14547         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14548         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14549         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14550         (expand_vector_condition, vector_element): Likewise.
14551         (subparts_gt): New function.
14552         (get_compute_type): Use subparts_gt.
14553         (count_type_subparts): Delete.
14554         (expand_vector_operations_1): Use subparts_gt instead of
14555         count_type_subparts.
14556
14557 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14558             Alan Hayward  <alan.hayward@arm.com>
14559             David Sherwood  <david.sherwood@arm.com>
14560
14561         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14562         (vect_compile_time_alias): ...this new function.  Do the calculation
14563         on poly_ints rather than trees.
14564         (vect_prune_runtime_alias_test_list): Update call accordingly.
14565
14566 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14567             Alan Hayward  <alan.hayward@arm.com>
14568             David Sherwood  <david.sherwood@arm.com>
14569
14570         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14571         numbers of units.
14572         (vect_schedule_slp_instance): Likewise.
14573
14574 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14575             Alan Hayward  <alan.hayward@arm.com>
14576             David Sherwood  <david.sherwood@arm.com>
14577
14578         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14579         constant and extern definitions for variable-length vectors.
14580         (vect_get_constant_vectors): Note that the number of units
14581         is known to be constant.
14582
14583 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14584             Alan Hayward  <alan.hayward@arm.com>
14585             David Sherwood  <david.sherwood@arm.com>
14586
14587         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14588         of units as polynomial.  Choose between WIDE and NARROW based
14589         on multiple_p.
14590
14591 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14592             Alan Hayward  <alan.hayward@arm.com>
14593             David Sherwood  <david.sherwood@arm.com>
14594
14595         * tree-vect-stmts.c (simd_clone_subparts): New function.
14596         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14597
14598 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14599             Alan Hayward  <alan.hayward@arm.com>
14600             David Sherwood  <david.sherwood@arm.com>
14601
14602         * tree-vect-stmts.c (vectorizable_call): Treat the number of
14603         vectors as polynomial.  Use build_index_vector for
14604         IFN_GOMP_SIMD_LANE.
14605
14606 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14607             Alan Hayward  <alan.hayward@arm.com>
14608             David Sherwood  <david.sherwood@arm.com>
14609
14610         * tree-vect-stmts.c (get_load_store_type): Treat the number of
14611         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14612         for variable-length vectors.
14613         (vectorizable_mask_load_store): Treat the number of units as
14614         polynomial, asserting that it is constant if the condition has
14615         already been enforced.
14616         (vectorizable_store, vectorizable_load): Likewise.
14617
14618 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14619             Alan Hayward  <alan.hayward@arm.com>
14620             David Sherwood  <david.sherwood@arm.com>
14621
14622         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14623         of units as polynomial.  Punt if we can't tell at compile time
14624         which vector contains the final result.
14625
14626 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14627             Alan Hayward  <alan.hayward@arm.com>
14628             David Sherwood  <david.sherwood@arm.com>
14629
14630         * tree-vect-loop.c (vectorizable_induction): Treat the number
14631         of units as polynomial.  Punt on SLP inductions.  Use an integer
14632         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
14633         cast of such a series for variable-length floating-point
14634         reductions.
14635
14636 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14637             Alan Hayward  <alan.hayward@arm.com>
14638             David Sherwood  <david.sherwood@arm.com>
14639
14640         * tree.h (build_index_vector): Declare.
14641         * tree.c (build_index_vector): New function.
14642         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14643         of units as polynomial, forcibly converting it to a constant if
14644         vectorizable_reduction has already enforced the condition.
14645         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
14646         to create a {1,2,3,...} vector.
14647         (vectorizable_reduction): Treat the number of units as polynomial.
14648         Choose vectype_in based on the largest scalar element size rather
14649         than the smallest number of units.  Enforce the restrictions
14650         relied on above.
14651
14652 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14653             Alan Hayward  <alan.hayward@arm.com>
14654             David Sherwood  <david.sherwood@arm.com>
14655
14656         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14657         number of units as polynomial.
14658
14659 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14660             Alan Hayward  <alan.hayward@arm.com>
14661             David Sherwood  <david.sherwood@arm.com>
14662
14663         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14664         * target.def (autovectorize_vector_sizes): Return the vector sizes
14665         by pointer, using vector_sizes rather than a bitmask.
14666         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14667         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14668         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14669         Likewise.
14670         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14671         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14672         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14673         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14674         * omp-general.c (omp_max_vf): Likewise.
14675         * omp-low.c (omp_clause_aligned_alignment): Likewise.
14676         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14677         * tree-vect-loop.c (vect_analyze_loop): Likewise.
14678         * tree-vect-slp.c (vect_slp_bb): Likewise.
14679         * doc/tm.texi: Regenerate.
14680         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14681         to a poly_uint64.
14682         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14683         the vector size as a poly_uint64 rather than an unsigned int.
14684         (current_vector_size): Change from an unsigned int to a poly_uint64.
14685         (get_vectype_for_scalar_type): Update accordingly.
14686         * tree.h (build_truth_vector_type): Take the size and number of
14687         units as a poly_uint64 rather than an unsigned int.
14688         (build_vector_type): Add a temporary overload that takes
14689         the number of units as a poly_uint64 rather than an unsigned int.
14690         * tree.c (make_vector_type): Likewise.
14691         (build_truth_vector_type): Take the number of units as a poly_uint64
14692         rather than an unsigned int.
14693
14694 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14695             Alan Hayward  <alan.hayward@arm.com>
14696             David Sherwood  <david.sherwood@arm.com>
14697
14698         * target.def (get_mask_mode): Take the number of units and length
14699         as poly_uint64s rather than unsigned ints.
14700         * targhooks.h (default_get_mask_mode): Update accordingly.
14701         * targhooks.c (default_get_mask_mode): Likewise.
14702         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14703         * doc/tm.texi: Regenerate.
14704
14705 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14706             Alan Hayward  <alan.hayward@arm.com>
14707             David Sherwood  <david.sherwood@arm.com>
14708
14709         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14710         * omp-general.c (omp_max_vf): Likewise.
14711         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14712         (expand_omp_simd): Handle polynomial safelen.
14713         * omp-low.c (omplow_simd_context): Add a default constructor.
14714         (omplow_simd_context::max_vf): Change from int to poly_uint64.
14715         (lower_rec_simd_input_clauses): Update accordingly.
14716         (lower_rec_input_clauses): Likewise.
14717
14718 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14719             Alan Hayward  <alan.hayward@arm.com>
14720             David Sherwood  <david.sherwood@arm.com>
14721
14722         * tree-vectorizer.h (vect_nunits_for_cost): New function.
14723         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14724         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14725         (vect_analyze_slp_cost): Likewise.
14726         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14727         (vect_model_load_cost): Likewise.
14728
14729 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14730             Alan Hayward  <alan.hayward@arm.com>
14731             David Sherwood  <david.sherwood@arm.com>
14732
14733         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14734         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14735         from an unsigned int * to a poly_uint64_pod *.
14736         (calculate_unrolling_factor): New function.
14737         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
14738
14739 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14740             Alan Hayward  <alan.hayward@arm.com>
14741             David Sherwood  <david.sherwood@arm.com>
14742
14743         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14744         from an unsigned int to a poly_uint64.
14745         (_loop_vec_info::slp_unrolling_factor): Likewise.
14746         (_loop_vec_info::vectorization_factor): Change from an int
14747         to a poly_uint64.
14748         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14749         (vect_get_num_vectors): New function.
14750         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14751         (vect_get_num_copies): Use vect_get_num_vectors.
14752         (vect_analyze_data_ref_dependences): Change max_vf from an int *
14753         to an unsigned int *.
14754         (vect_analyze_data_refs): Change min_vf from an int * to a
14755         poly_uint64 *.
14756         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14757         than an unsigned HOST_WIDE_INT.
14758         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14759         (vect_analyze_data_ref_dependence): Change max_vf from an int *
14760         to an unsigned int *.
14761         (vect_analyze_data_ref_dependences): Likewise.
14762         (vect_compute_data_ref_alignment): Handle polynomial vf.
14763         (vect_enhance_data_refs_alignment): Likewise.
14764         (vect_prune_runtime_alias_test_list): Likewise.
14765         (vect_shift_permute_load_chain): Likewise.
14766         (vect_supportable_dr_alignment): Likewise.
14767         (dependence_distance_ge_vf): Take the vectorization factor as a
14768         poly_uint64 rather than an unsigned HOST_WIDE_INT.
14769         (vect_analyze_data_refs): Change min_vf from an int * to a
14770         poly_uint64 *.
14771         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14772         vfm1 as a poly_uint64 rather than an int.  Make the same change
14773         for the returned bound_scalar.
14774         (vect_gen_vector_loop_niters): Handle polynomial vf.
14775         (vect_do_peeling): Likewise.  Update call to
14776         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14777         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14778         be constant.
14779         * tree-vect-loop.c (vect_determine_vectorization_factor)
14780         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14781         (vect_get_known_peeling_cost): Likewise.
14782         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14783         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14784         (vect_transform_loop): Likewise.  Use the lowest possible VF when
14785         updating the upper bounds of the loop.
14786         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
14787         rather than an int.
14788         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14789         polynomial unroll factors.
14790         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14791         (vect_make_slp_decision): Likewise.
14792         (vect_supported_load_permutation_p): Likewise, and polynomial
14793         vf too.
14794         (vect_analyze_slp_cost): Handle polynomial vf.
14795         (vect_slp_analyze_node_operations): Likewise.
14796         (vect_slp_analyze_bb_1): Likewise.
14797         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14798         than an unsigned HOST_WIDE_INT.
14799         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14800         (vectorizable_load): Handle polynomial vf.
14801         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14802         a poly_uint64.
14803         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14804
14805 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14806             Alan Hayward  <alan.hayward@arm.com>
14807             David Sherwood  <david.sherwood@arm.com>
14808
14809         * match.pd: Handle bit operations involving three constants
14810         and try to fold one pair.
14811
14812 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14813
14814         * tree-vect-loop-manip.c: Include gimple-fold.h.
14815         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14816         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
14817         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14818         Add a path that uses a step of VF instead of 1, but disable it
14819         for now.
14820         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14821         and niters_no_overflow parameters.  Update calls to
14822         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14823         Create a new SSA name if the latter choses to use a ste other
14824         than zero, and return it via niters_vector_mult_vf_var.
14825         * tree-vect-loop.c (vect_transform_loop): Update calls to
14826         vect_do_peeling, vect_gen_vector_loop_niters and
14827         slpeel_make_loop_iterate_ntimes.
14828         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14829         (vect_gen_vector_loop_niters): Update declarations after above changes.
14830
14831 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
14832
14833         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14834         128-bit round to integer instructions.
14835         (ceil<mode>2): Likewise.
14836         (btrunc<mode>2): Likewise.
14837         (round<mode>2): Likewise.
14838
14839 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14840
14841         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14842         unaligned VSX load/store on P8/P9.
14843         (expand_block_clear): Allow the use of unaligned VSX
14844         load/store on P8/P9.
14845
14846 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14847
14848         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14849         New function.
14850         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14851         swap associated with both a load and a store.
14852
14853 2018-01-02  Andrew Waterman  <andrew@sifive.com>
14854
14855         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14856         * config/riscv/riscv.md (clear_cache): Use it.
14857
14858 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
14859
14860         * web.c: Remove out-of-date comment.
14861
14862 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14863
14864         * expr.c (fixup_args_size_notes): Check that any existing
14865         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14866         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14867         (emit_single_push_insn): ...here.
14868
14869 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14870
14871         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14872         (const_vector_encoded_nelts): New function.
14873         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14874         (const_vector_int_elt, const_vector_elt): Declare.
14875         * emit-rtl.c (const_vector_int_elt_1): New function.
14876         (const_vector_elt): Likewise.
14877         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14878         of CONST_VECTOR_ELT.
14879
14880 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14881
14882         * expr.c: Include rtx-vector-builder.h.
14883         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14884         directly on the tree encoding.
14885         (const_vector_from_tree): Likewise.
14886         * optabs.c: Include rtx-vector-builder.h.
14887         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14888         sequence of "u" values.
14889         * vec-perm-indices.c: Include rtx-vector-builder.h.
14890         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14891         directly on the vec_perm_indices encoding.
14892
14893 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14894
14895         * doc/rtl.texi (const_vector): Describe new encoding scheme.
14896         * Makefile.in (OBJS): Add rtx-vector-builder.o.
14897         * rtx-vector-builder.h: New file.
14898         * rtx-vector-builder.c: Likewise.
14899         * rtl.h (rtx_def::u2): Add a const_vector field.
14900         (CONST_VECTOR_NPATTERNS): New macro.
14901         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14902         (CONST_VECTOR_DUPLICATE_P): Likewise.
14903         (CONST_VECTOR_STEPPED_P): Likewise.
14904         (CONST_VECTOR_ENCODED_ELT): Likewise.
14905         (const_vec_duplicate_p): Check for a duplicated vector encoding.
14906         (unwrap_const_vec_duplicate): Likewise.
14907         (const_vec_series_p): Check for a non-duplicated vector encoding.
14908         Say that the function only returns true for integer vectors.
14909         * emit-rtl.c: Include rtx-vector-builder.h.
14910         (gen_const_vec_duplicate_1): Delete.
14911         (gen_const_vector): Call gen_const_vec_duplicate instead of
14912         gen_const_vec_duplicate_1.
14913         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14914         (gen_const_vec_duplicate): Use rtx_vector_builder.
14915         (gen_const_vec_series): Likewise.
14916         (gen_rtx_CONST_VECTOR): Likewise.
14917         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14918         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14919         Build a new vector rather than modifying a CONST_VECTOR in-place.
14920         (handle_special_swappables): Update call accordingly.
14921         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14922         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14923         Build a new vector rather than modifying a CONST_VECTOR in-place.
14924         (handle_special_swappables): Update call accordingly.
14925
14926 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14927
14928         * simplify-rtx.c (simplify_const_binary_operation): Use
14929         CONST_VECTOR_ELT instead of XVECEXP.
14930
14931 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14932
14933         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14934         the selector elements to be different from the data elements
14935         if the selector is a VECTOR_CST.
14936         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14937         ssizetype for the selector.
14938
14939 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14940
14941         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14942         before testing each element individually.
14943         * tree-vect-generic.c (lower_vec_perm): Likewise.
14944
14945 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14946
14947         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14948         * selftest-run-tests.c (selftest::run_tests): Call it.
14949         * vector-builder.h (vector_builder::operator ==): New function.
14950         (vector_builder::operator !=): Likewise.
14951         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14952         (vec_perm_indices::all_from_input_p): New function.
14953         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14954         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14955         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14956         instead of reading the VECTOR_CST directly.  Detect whether both
14957         vector inputs are the same before constructing the vec_perm_indices,
14958         and update the number of inputs argument accordingly.  Use the
14959         utility functions added above.  Only construct sel2 if we need to.
14960
14961 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14962
14963         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14964         the broadcast of the low byte.
14965         (expand_mult_highpart): Use an explicit encoding for the permutes.
14966         * optabs-query.c (can_mult_highpart_p): Likewise.
14967         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14968         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14969         (vectorizable_bswap): Likewise.
14970         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14971         explicit encoding for the power-of-2 permutes.
14972         (vect_permute_store_chain): Likewise.
14973         (vect_grouped_load_supported): Likewise.
14974         (vect_permute_load_chain): Likewise.
14975
14976 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14977
14978         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14979         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14980         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14981         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14982         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14983         (vect_gen_perm_mask_any): Likewise.
14984
14985 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14986
14987         * int-vector-builder.h: New file.
14988         * vec-perm-indices.h: Include int-vector-builder.h.
14989         (vec_perm_indices): Redefine as an int_vector_builder.
14990         (auto_vec_perm_indices): Delete.
14991         (vec_perm_builder): Redefine as a stand-alone class.
14992         (vec_perm_indices::vec_perm_indices): New function.
14993         (vec_perm_indices::clamp): Likewise.
14994         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14995         (vec_perm_indices::new_vector): New function.
14996         (vec_perm_indices::new_expanded_vector): Update for new
14997         vec_perm_indices class.
14998         (vec_perm_indices::rotate_inputs): New function.
14999         (vec_perm_indices::all_in_range_p): Operate directly on the
15000         encoded form, without computing elided elements.
15001         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15002         encoding.  Update for new vec_perm_indices class.
15003         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15004         the given vec_perm_builder.
15005         (expand_vec_perm_var): Update vec_perm_builder constructor.
15006         (expand_mult_highpart): Use vec_perm_builder instead of
15007         auto_vec_perm_indices.
15008         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15009         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
15010         or double series encoding as appropriate.
15011         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15012         vec_perm_indices instead of auto_vec_perm_indices.
15013         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15014         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15015         (vect_permute_store_chain): Likewise.
15016         (vect_grouped_load_supported): Likewise.
15017         (vect_permute_load_chain): Likewise.
15018         (vect_shift_permute_load_chain): Likewise.
15019         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15020         (vect_transform_slp_perm_load): Likewise.
15021         (vect_schedule_slp_instance): Likewise.
15022         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15023         (vectorizable_mask_load_store): Likewise.
15024         (vectorizable_bswap): Likewise.
15025         (vectorizable_store): Likewise.
15026         (vectorizable_load): Likewise.
15027         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15028         vec_perm_indices instead of auto_vec_perm_indices.  Use
15029         tree_to_vec_perm_builder to read the vector from a tree.
15030         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15031         vec_perm_builder instead of a vec_perm_indices.
15032         (have_whole_vector_shift): Use vec_perm_builder and
15033         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
15034         truncation to calc_vec_perm_mask_for_shift.
15035         (vect_create_epilog_for_reduction): Likewise.
15036         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15037         from auto_vec_perm_indices to vec_perm_indices.
15038         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15039         instead of changing individual elements.
15040         (aarch64_vectorize_vec_perm_const): Use new_vector to install
15041         the vector in d.perm.
15042         * config/arm/arm.c (expand_vec_perm_d::perm): Change
15043         from auto_vec_perm_indices to vec_perm_indices.
15044         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15045         instead of changing individual elements.
15046         (arm_vectorize_vec_perm_const): Use new_vector to install
15047         the vector in d.perm.
15048         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15049         Update vec_perm_builder constructor.
15050         (rs6000_expand_interleave): Likewise.
15051         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15052         (rs6000_expand_interleave): Likewise.
15053
15054 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15055
15056         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15057         to qimode could truncate the indices.
15058         * optabs.c (expand_vec_perm_var): Likewise.
15059
15060 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15061
15062         * Makefile.in (OBJS): Add vec-perm-indices.o.
15063         * vec-perm-indices.h: New file.
15064         * vec-perm-indices.c: Likewise.
15065         * target.h (vec_perm_indices): Replace with a forward class
15066         declaration.
15067         (auto_vec_perm_indices): Move to vec-perm-indices.h.
15068         * optabs.h: Include vec-perm-indices.h.
15069         (expand_vec_perm): Delete.
15070         (selector_fits_mode_p, expand_vec_perm_var): Declare.
15071         (expand_vec_perm_const): Declare.
15072         * target.def (vec_perm_const_ok): Replace with...
15073         (vec_perm_const): ...this new hook.
15074         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15075         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15076         * doc/tm.texi: Regenerate.
15077         * optabs.def (vec_perm_const): Delete.
15078         * doc/md.texi (vec_perm_const): Likewise.
15079         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15080         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15081         expand_vec_perm for constant permutation vectors.  Assert that
15082         the mode of variable permutation vectors is the integer equivalent
15083         of the mode that is being permuted.
15084         * optabs-query.h (selector_fits_mode_p): Declare.
15085         * optabs-query.c: Include vec-perm-indices.h.
15086         (selector_fits_mode_p): New function.
15087         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15088         is defined, instead of checking whether the vec_perm_const_optab
15089         exists.  Use targetm.vectorize.vec_perm_const instead of
15090         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
15091         fit in the vector mode before using a variable permute.
15092         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15093         vec_perm_indices instead of an rtx.
15094         (expand_vec_perm): Replace with...
15095         (expand_vec_perm_const): ...this new function.  Take the selector
15096         as a vec_perm_indices rather than an rtx.  Also take the mode of
15097         the selector.  Update call to shift_amt_for_vec_perm_mask.
15098         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15099         Use vec_perm_indices::new_expanded_vector to expand the original
15100         selector into bytes.  Check whether the indices fit in the vector
15101         mode before using a variable permute.
15102         (expand_vec_perm_var): Make global.
15103         (expand_mult_highpart): Use expand_vec_perm_const.
15104         * fold-const.c: Includes vec-perm-indices.h.
15105         * tree-ssa-forwprop.c: Likewise.
15106         * tree-vect-data-refs.c: Likewise.
15107         * tree-vect-generic.c: Likewise.
15108         * tree-vect-loop.c: Likewise.
15109         * tree-vect-slp.c: Likewise.
15110         * tree-vect-stmts.c: Likewise.
15111         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15112         Delete.
15113         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15114         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15115         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15116         (aarch64_vectorize_vec_perm_const): ...this new function.
15117         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15118         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15119         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15120         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15121         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15122         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15123         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15124         into...
15125         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
15126         check for NEON modes.
15127         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15128         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15129         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15130         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15131         into...
15132         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
15133         the old VEC_PERM_CONST conditions.
15134         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15135         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15136         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15137         (ia64_vectorize_vec_perm_const_ok): Merge into...
15138         (ia64_vectorize_vec_perm_const): ...this new function.
15139         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15140         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15141         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15142         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15143         * config/mips/mips.c (mips_expand_vec_perm_const)
15144         (mips_vectorize_vec_perm_const_ok): Merge into...
15145         (mips_vectorize_vec_perm_const): ...this new function.
15146         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15147         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15148         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15149         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15150         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15151         (rs6000_expand_vec_perm_const): Delete.
15152         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15153         Delete.
15154         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15155         (altivec_expand_vec_perm_const_le): Take each operand individually.
15156         Operate on constant selectors rather than rtxes.
15157         (altivec_expand_vec_perm_const): Likewise.  Update call to
15158         altivec_expand_vec_perm_const_le.
15159         (rs6000_expand_vec_perm_const): Delete.
15160         (rs6000_vectorize_vec_perm_const_ok): Delete.
15161         (rs6000_vectorize_vec_perm_const): New function.
15162         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15163         an element count and rtx array.
15164         (rs6000_expand_extract_even): Update call accordingly.
15165         (rs6000_expand_interleave): Likewise.
15166         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15167         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15168         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15169         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15170         (rs6000_expand_vec_perm_const): Delete.
15171         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15172         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15173         (altivec_expand_vec_perm_const_le): Take each operand individually.
15174         Operate on constant selectors rather than rtxes.
15175         (altivec_expand_vec_perm_const): Likewise.  Update call to
15176         altivec_expand_vec_perm_const_le.
15177         (rs6000_expand_vec_perm_const): Delete.
15178         (rs6000_vectorize_vec_perm_const_ok): Delete.
15179         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
15180         reference to the SPE evmerge intructions.
15181         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15182         an element count and rtx array.
15183         (rs6000_expand_extract_even): Update call accordingly.
15184         (rs6000_expand_interleave): Likewise.
15185         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15186         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15187         new function.
15188         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15189
15190 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15191
15192         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15193         vector mode and that that mode matches the mode of the data
15194         being permuted.
15195         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15196         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
15197         directly using expand_vec_perm_1 when forcing selectors into
15198         registers.
15199         (expand_vec_perm_var): New function, split out from expand_vec_perm.
15200
15201 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15202
15203         * optabs-query.h (can_vec_perm_p): Delete.
15204         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15205         * optabs-query.c (can_vec_perm_p): Split into...
15206         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15207         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15208         particular selector is valid.
15209         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15210         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15211         (vect_grouped_load_supported): Likewise.
15212         (vect_shift_permute_load_chain): Likewise.
15213         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15214         (vect_transform_slp_perm_load): Likewise.
15215         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15216         (vectorizable_bswap): Likewise.
15217         (vect_gen_perm_mask_checked): Likewise.
15218         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
15219         implementations of variable permutation vectors into account
15220         when deciding which selector to use.
15221         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15222         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15223         with a false third argument.
15224         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15225         to test whether the constant selector is valid and can_vec_perm_var_p
15226         to test whether a variable selector is valid.
15227
15228 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15229
15230         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15231         * optabs-query.c (can_vec_perm_p): Likewise.
15232         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15233         instead of vec_perm_indices.
15234         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15235         (vect_gen_perm_mask_checked): Likewise,
15236         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15237         (vect_gen_perm_mask_checked): Likewise,
15238
15239 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15240
15241         * optabs-query.h (qimode_for_vec_perm): Declare.
15242         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15243         (qimode_for_vec_perm): ...this new function.
15244         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15245
15246 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
15247
15248         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15249         does not have a conditional at the top.
15250
15251 2018-01-02  Richard Biener  <rguenther@suse.de>
15252
15253         * ipa-inline.c (big_speedup_p): Fix expression.
15254
15255 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
15256
15257         PR target/81616
15258         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15259         for generic 4->6.
15260
15261 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
15262
15263         PR target/81616
15264         Generic tuning.
15265         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15266         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15267         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15268         cond_taken_branch_cost 3->4.
15269
15270 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
15271
15272         PR tree-optimization/83581
15273         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15274         TODO_cleanup_cfg if any changes have been made.
15275
15276         PR middle-end/83608
15277         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15278         convert_modes if target mode has the right side, but different mode
15279         class.
15280
15281         PR middle-end/83609
15282         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15283         last argument when extracting from CONCAT.  If either from_real or
15284         from_imag is NULL, use expansion through memory.  If result is not
15285         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15286         the parts directly to inner mode, if even that fails, use expansion
15287         through memory.
15288
15289         PR middle-end/83623
15290         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15291         check for bswap in mode rather than HImode and use that in expand_unop
15292         too.
15293 \f
15294 Copyright (C) 2018 Free Software Foundation, Inc.
15295
15296 Copying and distribution of this file, with or without modification,
15297 are permitted in any medium without royalty provided the copyright
15298 notice and this notice are preserved.