gcc/ChangeLog

   1 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
   2
   3         PR target/81497
   4         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
   5         qualifier_void_pointer and qualifier_const_void_pointer.
   6         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
   7         (arm_init_builtins): Handle the above.
   8         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
   9         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
  10         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
  11         void intrinsics.
  12
  13 2018-06-05  Martin Liska  <mliska@suse.cz>
  14
  15         * auto-profile.c (read_autofdo_file): Do not use
  16         gcov_ctr_summary struct.
  17         (afdo_callsite_hot_enough_for_early_inline): Likewise.
  18         * coverage.c (struct counts_entry): Likewise.
  19         (read_counts_file): Read just single summary entry.
  20         (get_coverage_counts): Use gcov_summary struct.
  21         * coverage.h (get_coverage_counts): Likewise.
  22         * gcov-dump.c (dump_working_sets): Likewise.
  23         (tag_summary): Dump just single summary.
  24         * gcov-io.c (gcov_write_summary): Write just histogram
  25         summary.
  26         (gcov_read_summary): Read just single summary.
  27         (compute_working_sets): Use gcov_summary struct.
  28         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
  29         of GCOV_COUNTERS_SUMMABLE.
  30         (GCOV_COUNTERS_SUMMABLE): Remove.
  31         (GCOV_FIRST_VALUE_COUNTER): Replace with
  32         GCOV_COUNTER_V_INTERVAL.
  33         (struct gcov_ctr_summary): Remove.
  34         (struct gcov_summary): Directly use fields of former
  35         gcov_ctr_summary.
  36         (compute_working_sets): Use gcov_summary struct.
  37         * gcov.c (read_count_file): Do not use ctrs fields.
  38         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
  39         struct.
  40         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
  41         struct.
  42         * profile.c: Likewise.
  43         * profile.h: Likewise.
  44
  45 2018-06-05  Martin Liska  <mliska@suse.cz>
  46
  47         PR gcov-profile/84846
  48         * gcov.c (output_lines): Print working directory only
  49         in intermediate format.
  50
  51 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
  52
  53         * config/s390/s390-builtin-types.def: Add void function type.
  54         * config/s390/s390-builtins.def: Use the function type for the
  55         tbeginc builtin.
  56
  57 2018-06-04  Jim Wilson  <jimw@sifive.com>
  58
  59         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
  60         to int.
  61         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
  62         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
  63         handle EH_RETURN_DATA_REGNO registers properly.
  64         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
  65         (riscv_expand_epilogue): Update comment.  Change argument name and
  66         type.  Update code to use new name and type.  Pass new args to
  67         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
  68         EXCEPTION_RETURN.
  69         * config/riscv/riscv.md (NORMAL_RETURN): New.
  70         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
  71         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
  72         (eh_return): Call gen_eh_return_internal and emit barrier.
  73         (eh_return_internal): Call riscv_expand_epilogue.
  74
  75 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
  76
  77         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
  78         bit_insertion field and declare can_be_merged_into method.
  79         (merged_store_group::can_be_merged_into): New method.
  80         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
  81         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
  82         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
  83
  84 2018-06-04  Richard Biener  <rguenther@suse.de>
  85
  86         PR tree-optimization/85955
  87         * builtins.c (fold_builtin_sincos): Convert pointers to
  88         destination to appropriate type before dereferencing.
  89
  90 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
  91
  92         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
  93
  94 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
  95
  96         * expr.c (expand_expr_real_1): Force the operand into memory if
  97         its TYPE_MODE is BLKmode and if there is no integer mode for
  98         the number of bits being extracted.
  99
 100 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
 101
 102         PR target/85832
 103         PR target/86036
 104         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
 105         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
 106
 107 2018-06-04  Richard Biener  <rguenther@suse.de>
 108
 109         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
 110         (cleanup_tree_cfg_noloop): ... single caller.  Do
 111         start_recording_case_labels later.
 112
 113 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
 114
 115         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
 116         to _IMMINTRIN_H_INCLUDED.
 117         * config/i386/pconfigintrin.h: Ditto.
 118         * config/i386/waitpkgintrin.h: Ditto.
 119         * config/i386/immintrin.h: Add includes for sgxintrin.h,
 120         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
 121         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
 122         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
 123         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
 124         waitpkgintrin.h and cldemoteintrin.h.
 125
 126 2018-06-04  Richard Biener  <rguenther@suse.de>
 127
 128         PR tree-optimization/86038
 129         * tracer.c (find_best_successor): Check probability for
 130         being initialized, bail out if not.
 131
 132 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
 133
 134         PR target/86003
 135         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
 136         of bits to ignore when comparing architectures.
 137
 138 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
 139
 140         PR tree-optimization/69615
 141         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
 142         maximum or minimum of the type, try to merge it also as if
 143         range1 is + [-, x - 1] or + [x + 1, -].
 144
 145         PR c++/86025
 146         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
 147
 148 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
 149
 150         PR tree-optimization/86034
 151         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
 152         the unsigned bitfield type in a bit insertion sequence if it does not
 153         have a larger precision than the bitfield size.
 154         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
 155
 156 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
 157
 158         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
 159
 160 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
 161
 162         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
 163         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
 164         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
 165         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
 166
 167 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
 168
 169         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
 170         Disable -fdelete-null-pointer-checks for ELF toolchain.
 171
 172 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
 173             Kito Cheng  <kito.cheng@gmail.com>
 174
 175         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
 176         (nds32le-*-*, nds32be-*-*): Integrate checking process.
 177         (nds32*-*-*): Add glibc and uclibc conditions.
 178         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
 179         (TARGET_EXCEPT_UNWIND_INFO): Define.
 180         * config/nds32/elf.h: New file.
 181         * config/nds32/linux.h: New file.
 182         * config/nds32/nds32-elf.opt: New file.
 183         * config/nds32/nds32-linux.opt: New file.
 184         * config/nds32/nds32-fp-as-gp.c
 185         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
 186         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
 187         TARGET_LINUX_ABI.
 188         (nds32_asm_file_end): Ditto.
 189         (nds32_print_operand): Ditto.
 190         (nds32_insert_attributes): Ditto.
 191         (nds32_init_libfuncs): New function.
 192         (TARGET_HAVE_TLS): Define.
 193         (TARGET_INIT_LIBFUNCS): Define.
 194         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
 195         spec content.
 196         (TARGET_ELF): Apply different mcmodel setting.
 197         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
 198         been migrated into elf.h and linux.h files.
 199         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
 200         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
 201         (mcmodel): The content has been migrated into nds32-elf.opt and
 202         nds32-linux.opt files.
 203         * config/nds32/t-elf: New file.
 204         * config/nds32/t-linux: New file.
 205
 206 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
 207             Shiva Chen  <shiva0217@gmail.com>
 208
 209         * config/nds32/constants.md (unspec_volatile_element): Add
 210         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
 211         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
 212         optimization.
 213         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
 214         (make_pass_nds32_fp_as_gp): Declare.
 215         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
 216         optmization pass.
 217         (nds32_asm_function_end_prologue): Remove unused asm output.
 218         (nds32_asm_function_begin_epilogue): Remove unused asm output.
 219         (nds32_asm_file_start): Output necessary fp_as_gp information.
 220         (nds32_option_override): Adjust register usage.
 221         (nds32_expand_prologue): Consider fp_as_gp situation.
 222         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
 223         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
 224         (epilogue): Ditto.
 225         (return): Ditto.
 226         (simple_return): Ditto.
 227         (omit_fp_begin): Output special directive for fp_as_gp.
 228         (omit_fp_end): Output special directive for fp_as_gp.
 229         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
 230         mforbid-fp-as-gp): New options.
 231
 232 2018-06-01  Mark Wielaard  <mark@klomp.org>
 233
 234         * dwarf2out.c (dwarf2out_finish): Remove generation of
 235         DW_AT_loclists_base.
 236
 237 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
 238
 239         * gimple-ssa-store-merging.c: Include gimple-fold.h.
 240         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
 241         (struct merged_store_group): Add bit_insertion field.
 242         (dump_char_array): Use standard hexadecimal format.
 243         (merged_store_group::merged_store_group): Set bit_insertion to false.
 244         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
 245         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
 246         also print the mask in the dump file.
 247         (pass_store_merging::gate): Minor tweak.
 248         (imm_store_chain_info::coalesce_immediate): Fix wrong association
 249         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
 250         stores with INTEGER_CST stores.
 251         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
 252         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
 253         and use it throughout.  Generate bit insertion sequences if need be.
 254         (pass_store_merging::process_store): Remove redundant condition.
 255         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
 256
 257 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
 258
 259         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
 260         the 128-bit floating point types.  Fix function comment.
 261
 262 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 263
 264         * config/aarch64/aarch64-simd.md
 265         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
 266         mnemonics.
 267         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
 268         mnemonics.
 269
 270 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
 271
 272         PR tree-optimization/85989
 273         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
 274         variable.
 275         (backprop::intersect_uses): Check it when deciding whether this
 276         is a backedge reference.
 277         (backprop::process_block): Add each phi to m_visited_phis
 278         after visiting it, then clear it at the end.
 279
 280 2018-06-01  Richard Biener  <rguenther@suse.de>
 281
 282         * tree-vectorizer.h (vect_dr_stmt): New function.
 283         (vect_get_load_cost): Adjust.
 284         (vect_get_store_cost): Likewise.
 285         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
 286         Use vect_dr_stmt instead of DR_SMTT.
 287         (vect_record_base_alignments): Likewise.
 288         (vect_calculate_target_alignment): Likewise.
 289         (vect_compute_data_ref_alignment): Likewise and make static.
 290         (vect_update_misalignment_for_peel): Likewise.
 291         (vect_verify_datarefs_alignment): Likewise.
 292         (vector_alignment_reachable_p): Likewise.
 293         (vect_get_data_access_cost): Likewise.  Pass down
 294         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
 295         (vect_get_peeling_costs_all_drs): Likewise.
 296         (vect_peeling_hash_get_lowest_cost): Likewise.
 297         (vect_enhance_data_refs_alignment): Likewise.
 298         (vect_find_same_alignment_drs): Likewise.
 299         (vect_analyze_data_refs_alignment): Likewise.
 300         (vect_analyze_group_access_1): Likewise.
 301         (vect_analyze_group_access): Likewise.
 302         (vect_analyze_data_ref_access): Likewise.
 303         (vect_analyze_data_ref_accesses): Likewise.
 304         (vect_vfa_segment_size): Likewise.
 305         (vect_small_gap_p): Likewise.
 306         (vectorizable_with_step_bound_p): Likewise.
 307         (vect_prune_runtime_alias_test_list): Likewise.
 308         (vect_analyze_data_refs): Likewise.
 309         (vect_supportable_dr_alignment): Likewise.
 310         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
 311         (vect_gen_prolog_loop_niters): Likewise.
 312         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
 313         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
 314         modify DR_STMT.
 315         (vect_recog_mask_conversion_pattern): Likewise.
 316         (vect_try_gather_scatter_pattern): Likewise.
 317         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
 318         to vect_get_store_cost.
 319         (vect_get_store_cost): Get stmt_info instead of DR.
 320         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
 321         (vect_get_load_cost): Get stmt_info instead of DR.
 322
 323 2018-06-01  Richard Biener  <rguenther@suse.de>
 324
 325         PR middle-end/86017
 326         * gimple-fold.c (var_decl_component_p): Also allow offsetted
 327         vars wrapped in MEM_REFs.
 328
 329 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
 330
 331         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
 332         Fix subreg tests so that we only return a choice between
 333         GENERAL_REGS and FP_REGS if the original classes included both.
 334
 335 2018-06-01  Richard Biener  <rguenther@suse.de>
 336
 337         PR ipa/85960
 338         * tree-ssa-structalias.c (get_function_part_constraint):
 339         Handle NULL fi->decl.
 340         (find_func_aliases_for_call): Properly handle indirect
 341         fi from direct call.
 342         (find_func_clobbers): Likewise.
 343         (ipa_pta_execute): Likewise.
 344         (create_variable_info_for): For functions that are ifunc_resolver
 345         resolve to a varinfo that contains the result of the resolver
 346         call.
 347         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
 348         aliases.
 349
 350 2018-05-31  Michael Collison  <michael.collison@arm.com>
 351
 352         * config/aarch64/aarch64.md:
 353         (*fix_to_zero_extenddfdi2): New pattern.
 354         * gcc.target/aarch64/fix_extend1.c: New testcase.
 355
 356 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
 357
 358         PR middle-end/78809
 359         PR middle-end/83026
 360         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
 361         and BUILT_IN_STRNCMP_EQ.
 362         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
 363         BUILT_IN_STRNCMP_EQ.
 364         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
 365         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
 366         (gimple_fold_builtin): Likewise.
 367         * tree-ssa-strlen.c (compute_string_length): New function.
 368         (determine_min_obsize): New function.
 369         (handle_builtin_string_cmp): New function to handle calls to
 370         string compare functions.
 371         (strlen_optimize_stmt): Add handling to builtin string compare
 372         calls.
 373         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
 374         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
 375         * tree.c (build_common_builtin_nodes): Add new defines of
 376         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
 377
 378 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
 379
 380         PR target/85984
 381         * bb-reorder.c (pass_partition_blocks::gate): Return false for
 382         functions with naked attribute.
 383
 384 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
 385
 386         * config/i386/sse.md (avx_vec_concat<mode>):
 387         Substitute concat_tg_mode mode attribute with xtg_mode.
 388         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
 389         (concat_tg_mode): Remove mode attribute.
 390
 391 2018-05-31  Martin Sebor  <msebor@redhat.com>
 392
 393         PR c/82063
 394         * calls.c (alloc_max_size): Correct a logic error/typo.
 395         Treat excessive arguments as infinite.  Warn for invalid arguments.
 396         * doc/invoke.texi (-Walloc-size-larger-than): Update.
 397
 398 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
 399
 400         PR target/85829
 401         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
 402         and movx for Haswell.
 403
 404 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
 405             Cesar Philippidis  <cesar@codesourcery.com>
 406
 407         PR middle-end/85879
 408         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
 409         when emitting error on private/firstprivate reductions.
 410         * omp-low.c (lower_omp_target): Avoid reference-type processing
 411         on pointers for firstprivate clause.
 412
 413 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
 414
 415         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
 416         (st1x2): Likewise.
 417         (st1x3): Likewise.
 418         * config/aarch64/aarch64-simd.md
 419         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
 420         (aarch64_ld1_x3_<mode>): Likewise
 421         (aarch64_st1x2<VALLDIF:mode>): Likewise
 422         (aarch64_st1_x2_<mode>): Likewise
 423         (aarch64_st1x3<VALLDIF:mode>): Likewise
 424         (aarch64_st1_x3_<mode>): Likewise
 425         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
 426         (vld1_s8_x3): Likewise.
 427         (vld1_u16_x3): Likewise.
 428         (vld1_s16_x3): Likewise.
 429         (vld1_u32_x3): Likewise.
 430         (vld1_s32_x3): Likewise.
 431         (vld1_u64_x3): Likewise.
 432         (vld1_s64_x3): Likewise.
 433         (vld1_f16_x3): Likewise.
 434         (vld1_f32_x3): Likewise.
 435         (vld1_f64_x3): Likewise.
 436         (vld1_p8_x3): Likewise.
 437         (vld1_p16_x3): Likewise.
 438         (vld1_p64_x3): Likewise.
 439         (vld1q_u8_x3): Likewise.
 440         (vld1q_s8_x3): Likewise.
 441         (vld1q_u16_x3): Likewise.
 442         (vld1q_s16_x3): Likewise.
 443         (vld1q_u32_x3): Likewise.
 444         (vld1q_s32_x3): Likewise.
 445         (vld1q_u64_x3): Likewise.
 446         (vld1q_s64_x3): Likewise.
 447         (vld1q_f16_x3): Likewise.
 448         (vld1q_f32_x3): Likewise.
 449         (vld1q_f64_x3): Likewise.
 450         (vld1q_p8_x3): Likewise.
 451         (vld1q_p16_x3): Likewise.
 452         (vld1q_p64_x3): Likewise.
 453         (vst1_s64_x2): Likewise.
 454         (vst1_u64_x2): Likewise.
 455         (vst1_f64_x2): Likewise.
 456         (vst1_s8_x2): Likewise.
 457         (vst1_p8_x2): Likewise.
 458         (vst1_s16_x2): Likewise.
 459         (vst1_p16_x2): Likewise.
 460         (vst1_s32_x2): Likewise.
 461         (vst1_u8_x2): Likewise.
 462         (vst1_u16_x2): Likewise.
 463         (vst1_u32_x2): Likewise.
 464         (vst1_f16_x2): Likewise.
 465         (vst1_f32_x2): Likewise.
 466         (vst1_p64_x2): Likewise.
 467         (vst1q_s8_x2): Likewise.
 468         (vst1q_p8_x2): Likewise.
 469         (vst1q_s16_x2): Likewise.
 470         (vst1q_p16_x2): Likewise.
 471         (vst1q_s32_x2): Likewise.
 472         (vst1q_s64_x2): Likewise.
 473         (vst1q_u8_x2): Likewise.
 474         (vst1q_u16_x2): Likewise.
 475         (vst1q_u32_x2): Likewise.
 476         (vst1q_u64_x2): Likewise.
 477         (vst1q_f16_x2): Likewise.
 478         (vst1q_f32_x2): Likewise.
 479         (vst1q_f64_x2): Likewise.
 480         (vst1q_p64_x2): Likewise.
 481         (vst1_s64_x3): Likewise.
 482         (vst1_u64_x3): Likewise.
 483         (vst1_f64_x3): Likewise.
 484         (vst1_s8_x3): Likewise.
 485         (vst1_p8_x3): Likewise.
 486         (vst1_s16_x3): Likewise.
 487         (vst1_p16_x3): Likewise.
 488         (vst1_s32_x3): Likewise.
 489         (vst1_u8_x3): Likewise.
 490         (vst1_u16_x3): Likewise.
 491         (vst1_u32_x3): Likewise.
 492         (vst1_f16_x3): Likewise.
 493         (vst1_f32_x3): Likewise.
 494         (vst1_p64_x3): Likewise.
 495         (vst1q_s8_x3): Likewise.
 496         (vst1q_p8_x3): Likewise.
 497         (vst1q_s16_x3): Likewise.
 498         (vst1q_p16_x3): Likewise.
 499         (vst1q_s32_x3): Likewise.
 500         (vst1q_s64_x3): Likewise.
 501         (vst1q_u8_x3): Likewise.
 502         (vst1q_u16_x3): Likewise.
 503         (vst1q_u32_x3): Likewise.
 504         (vst1q_u64_x3): Likewise.
 505         (vst1q_f16_x3): Likewise.
 506         (vst1q_f32_x3): Likewise.
 507         (vst1q_f64_x3): Likewise.
 508         (vst1q_p64_x3): Likewise.
 509
 510 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
 511
 512         * config/msp430/msp430.c (msp430_output_labelref): Prepend
 513         user_label_prefix to name.
 514
 515         * tree-core.h: Update comment about the format of NAME string
 516         passed to handler in attribute_spec.
 517
 518         * config/msp430/msp430.md: Remove erroneous subreg expression from
 519         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
 520         zero_extend{q,h}isi2.
 521
 522 2018-05-30  Borislav Petkov  <bp@suse.de>
 523
 524         * doc/extend.texi: Document some architecture specific
 525         constraints and sort entries.
 526
 527 2018-05-30  Martin Sebor  <msebor@redhat.com>
 528
 529         PR middle-end/85369
 530         * builtins.c (expand_builtin_stpcpy_1): New function.
 531         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
 532         only if the former succeeds.
 533
 534 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
 535
 536         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
 537         in saphira.
 538
 539 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 540
 541         * doc/invoke.texi (-flinker-output): Document
 542
 543 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 544
 545         * passes.c (ipa_write_summaries): Only modify statements if body
 546         is in memory.
 547         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
 548         incrementally linking.
 549         (ipa_passes): Likewise.
 550         * lto-cgraph.c (lto_output_node): When incrementally linking do not
 551         pass down resolution info.
 552         * common.opt (flag_incremental_link): Update info.
 553         * gcc.c (plugin specs): Turn flinker-output=* to
 554         -plugin-opt=-linker-output-known
 555         * toplev.c (compile_file): Also cut compilation when doing incremental
 556         link.
 557         * flag-types. (enum lto_partition_model): Add
 558         LTO_LINKER_OUTPUT_NOLTOREL.
 559         (invoke.texi): Add -flinker-output docs.
 560         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
 561         link same way as WPA; do not stream in dead initializers.
 562
 563         * dwarf2out.c (dwarf2out_die_ref_for_decl,
 564         darf2out_register_external_decl): Support incremental link.
 565
 566 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 567
 568         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
 569
 570 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 571
 572         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
 573         it down to simple_object_copy_lto_debug_sections.
 574         (run_gcc): Determine incremental LTO link time and configure
 575         lto1 into non-wpa mode, disable renaming of debug sections.
 576
 577 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 578
 579         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
 580         descriptions of various incorrectly documented functions.
 581
 582 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
 583
 584         Revert:
 585         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
 586         address check not strict.
 587
 588 2018-05-30  Richard Biener  <rguenther@suse.de>
 589
 590         PR tree-optimization/85964
 591         * tracer.c (better_p): Drop initialized count check, we only
 592         call the function with initialized counts now.
 593         (find_best_successor): Do find a best edge if one
 594         has uninitialized count.
 595         (find_best_predecessor): Likewise.  Do BB frequency check only
 596         if count is initialized.
 597
 598 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
 599
 600         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
 601         (aarch64_ldrstr_offset_compare): New.
 602         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
 603         load/store orderings.
 604         (aarch64_gen_adjusted_ldpstp): Likewise.
 605
 606 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
 607
 608         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
 609         Check for subset of GENERAL_REGS and FP_REGS.
 610         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
 611         r=w alternative.
 612
 613 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
 614
 615         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
 616         and wi::to_poly_offset.  Add the current offset and then check
 617         whether the sum fits, rather than using an unchecked addition of
 618         a checked term.  Check for a shwi rather than a uhwi.
 619         * expr.c (get_bit_range): Use tree_to_poly_uint64.
 620         (store_constructor): Use poly_int_tree_p.
 621         (expand_expr_real_1): Likewise.
 622         * function.c (assign_temp): Likewise.
 623         * fold-const.c (const_binop): Use poly_int_tree_p and
 624         wi::to_poly_offset.
 625         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
 626         division.
 627         * ipa-icf-gimple.c (func_checker::compare_operand): Use
 628         to_poly_offset for MEM offsets.
 629         * ipa-icf.c (sem_variable::equals): Likewise.
 630         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
 631         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
 632         wi::to_poly_offset for BIT_FIELD_REF offsets.
 633         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
 634         wi::to_poly_offset.
 635         * var-tracking.c (emit_note_insn_var_location): Use
 636         tree_to_poly_uint64.
 637
 638 2018-05-29  Jim Wilson  <jimw@sifive.com>
 639
 640         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
 641
 642 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
 643
 644         PR target/85950
 645         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
 646         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
 647         sequence.
 648         (sse4_1_round<mode>2): Use nonimmediate_operand
 649         for operand 1 predicate.
 650
 651 2018-05-29  Martin Sebor  <msebor@redhat.com>
 652             Richard Biener  <rguenther@suse.de>
 653
 654         PR testsuite/85888
 655         * calls.c (get_size_range): Call determine_value_range instead
 656         of get_value_range..
 657         * tree-vrp.h (determine_value_range): Declared new function.
 658         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
 659
 660 2018-05-29  Richard Biener  <rguenther@suse.de>
 661
 662         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
 663         sure to use non-pattern stmts for get_earlier_stmt arguments.
 664         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
 665         called on pattern stmts.
 666         (get_later_stmt): Likewise.
 667
 668 2018-05-29  Martin Liska  <mliska@suse.cz>
 669
 670         PR gcov-profile/85759
 671         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
 672         env variables.
 673
 674 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
 675
 676         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
 677         VEC_UNPACK_*_EXPR.
 678         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
 679         VEC_PACK_*_EXPR.
 680
 681         PR target/85918
 682         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
 683         VEC_PACK_FLOAT_EXPR): New tree codes.
 684         * tree-pretty-print.c (op_code_prio): Handle
 685         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
 686         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
 687         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
 688         * tree-inline.c (estimate_operator_cost): Likewise.
 689         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
 690         * fold-const.c (const_binop): Likewise.
 691         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
 692         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
 693         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
 694         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
 695         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
 696         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
 697         * expr.c (expand_expr_real_2): Likewise.
 698         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
 699         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
 700         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
 701         optabs.
 702         * optabs.c (expand_widen_pattern_expr): For
 703         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
 704         sign from result type rather than operand's type.
 705         (expand_binop_directly): For vec_packu_float_optab and
 706         vec_packs_float_optab allow result type to be different from operand's
 707         type.
 708         * optabs-tree.c (optab_for_tree_code): Handle
 709         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
 710         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
 711         * tree-vect-generic.c (expand_vector_operations_1):  Handle
 712         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
 713         VEC_PACK_FLOAT_EXPR.
 714         * tree-vect-stmts.c (supportable_widening_operation): Handle
 715         FIX_TRUNC_EXPR.
 716         (supportable_narrowing_operation): Handle FLOAT_EXPR.
 717         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
 718         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
 719         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
 720         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
 721         mode attributes.
 722         (vec_pack<floatprefix>_float_<mode>): New expander.
 723         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
 724         attributes.
 725         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
 726         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
 727         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
 728         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
 729         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
 730         Document.
 731         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
 732         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
 733         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
 734         VEC_PACK_FLOAT_EXPR): Document.
 735
 736 2018-05-29  Richard Biener  <rguenther@suse.de>
 737
 738         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
 739         member.
 740         (stmt_vec_info_vec): Make pointer.
 741         (init_stmt_vec_info_vec): Remove.
 742         (free_stmt_vec_info_vec): Likewise.
 743         (set_stmt_vec_info_vec): New function.
 744         (free_stmt_vec_infos): Likewise.
 745         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
 746         (set_vinfo_for_stmt): Likewise.
 747         (get_earlier_stmt): Likewise.
 748         (get_later_stmt): Likewise.
 749         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
 750         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
 751         (vec_info::~vec_info): Free stmt_vec_infos.
 752         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
 753         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
 754         (pass_slp_vectorize::execute): Likewise.
 755         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
 756         (free_stmt_vec_info_vec): Likewise.
 757         (set_stmt_vec_info_vec): New function.
 758         (free_stmt_vec_infos): Likewise.
 759         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
 760         the global stmt_vec_info_vec.
 761         * tree-parloops.c (gather_scalar_reductions): Use
 762         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
 763         vector.
 764
 765 2018-05-29  Richard Biener  <rguenther@suse.de>
 766
 767         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
 768
 769 2018-05-29  Martin Liska  <mliska@suse.cz>
 770             David Malcolm  <dmalcolm@redhat.com>
 771
 772         * vec.c (test_reverse): New.
 773         (vec_c_tests): Add new test.
 774         * vec.h (vl_ptr>::reverse): New function.
 775
 776 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
 777
 778         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
 779
 780         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
 781         and later.
 782
 783 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
 784
 785         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
 786
 787 2018-05-28  Richard Biener  <rguenther@suse.de>
 788
 789         PR tree-optimization/85933
 790         * tree-vect-data-refs.c (vect_record_base_alignments): Only
 791         look at stmts marked as vectorizable.
 792
 793 2018-05-28  Richard Biener  <rguenther@suse.de>
 794
 795         PR tree-optimization/85934
 796         * tree-vect-generic.c (expand_vector_operations_1): Hoist
 797         vector boolean check before scalar optimization.
 798
 799 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
 800
 801         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
 802         for armv5te.
 803
 804 2018-05-28  Mark Wielaard  <mark@klomp.org>
 805
 806         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
 807         if it is an expression containing a minus sign.
 808
 809 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
 810
 811         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
 812
 813 2018-05-27  Paul Koning  <ni1d@arrl.net>
 814
 815         * config/pdp11/pdp11.md (truncsihi2): Remove.
 816
 817 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
 818             Chung-Ju Wu  <jasonwucj@gmail.com>
 819
 820         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
 821         implementation.
 822         (unaligned_store_dw): Ditto.
 823         * config/nds32/nds32-memory-manipulation.c
 824         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
 825         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
 826         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
 827         (emit_setmem_word_loop): Rename to ...
 828         (emit_setmem_doubleword_loop): ... this.
 829         (nds32_gen_dup_4_byte_to_word_value): New function.
 830         (nds32_gen_dup_8_byte_to_double_word_value): New function.
 831         (nds32_expand_setmem_loop): Refine implementation.
 832         (nds32_expand_setmem_loop_v3m): Ditto.
 833         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
 834         pattern.
 835
 836 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
 837
 838         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
 839
 840 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
 841
 842         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
 843         (nds32_init_machine_status): Initialize machine->attr_naked_p and
 844         machine->attr_no_prologue_p.
 845         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
 846         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
 847         (nds32_expand_epilogue): Consider attr_naked_p.
 848         (nds32_expand_epilogue_v3pop): Likewise.
 849         (nds32_can_use_return_insn): Likewise.
 850         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
 851         attr_no_prologue_p fields.
 852         * config/nds32/nds32.opt (mret-in-naked-func): New option.
 853
 854 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
 855
 856         PR target/85918
 857         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
 858         attributes.
 859         * config/i386/sse.md
 860         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
 861         Rename to ...
 862         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
 863         ... this.
 864         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
 865         Rename to ...
 866         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
 867         ... this.
 868         (*<floatsuffix>floatv2div2sf2): Rename to ...
 869         (*float<floatunssuffix>v2div2sf2): ... this.
 870         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
 871         (float<floatunssuffix>v2div2sf2_mask): ... this.
 872         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
 873         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
 874         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
 875         to ...
 876         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
 877         ... this.
 878         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
 879         Rename to ...
 880         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
 881         ... this.
 882         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
 883         Rename to ...
 884         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
 885         ... this.
 886         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
 887         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
 888         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
 889         gen_ufix_truncv8dfv8si2.
 890         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
 891         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
 892         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
 893         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
 894         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
 895         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
 896         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
 897         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
 898
 899 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
 900
 901         PR target/85900
 902         PR target/85345
 903         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
 904
 905 2018-05-25  Jim Wilson  <jimw@sifive.com>
 906
 907         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
 908         * config/riscv/riscv.c (struct machine_function): Add
 909         interrupt_handler_p and attribute_checked_p fields.
 910         (riscv_attribute_table): Add interrupt.
 911         (riscv_interrupt_type_p): New.
 912         (riscv_save_reg_p): Save extra regs for interrupt handler.
 913         (riscv_use_save_libcall): Return false  for interrupt handler.
 914         (riscv_first_stack_step): Add forward declaration.
 915         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
 916         for interrupt handler with large frame.  Use it for saved reg list.
 917         (riscv_expand_prologue): Move flag_stack_usage_info support to
 918         eliminate duplication.
 919         (riscv_expand_epilogue): Generate mret for interrupt handler.
 920         (riscv_epilogue_uses): New.
 921         (riscv_can_use_return_insn): Return false for interrupt handler.
 922         (riscv_function_ok_for_sibcall): Likewise.
 923         (riscv_set_current_function): Add interrupt handler support.
 924         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
 925         * config/riscv/riscv.md (UNSPECV_MRET): New.
 926         (GP_REGNUM): New.
 927         (riscv_frflags, riscv_fsflags): Use tab after opcode.
 928         (riscv_mret): New.
 929         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
 930
 931 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
 932
 933         PR tree-optimization/85712
 934         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
 935         this candidate has already been replaced in-situ by a copy.
 936
 937 2018-05-25  Jason Merrill  <jason@redhat.com>
 938
 939         PR c++/80485 - inline function non-zero address.
 940         * symtab.c (nonzero_address): Check DECL_COMDAT.
 941
 942 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
 943
 944         PR target/83628
 945         * config/alpha/alpha.md (ashlsi3): New insn pattern.
 946         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
 947         extension of SImode operation.  Use const123_operand predicate.
 948         (*saddsi_1): Remove.
 949         (*saddl_se_1): Ditto.
 950         (*ssubsi_1): Ditto.
 951         (*ssubl_se_1): Ditto.
 952         * config/alpha/predicates.md (const123_operand): New predicate.
 953         * config/alpha/constraints.md (P): Use IN_RANGE.
 954
 955 2018-05-25  Richard Biener  <rguenther@suse.de>
 956
 957         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
 958         defaulted to true.
 959         (ref_maybe_used_by_stmt_p): Likewise.
 960         (stmt_may_clobber_ref_p): Likewise.
 961         (stmt_may_clobber_ref_p_1): Likewise.
 962         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
 963         and pass it along.
 964         (ref_maybe_used_by_stmt_p): Likewise.
 965         (stmt_may_clobber_ref_p): Likewise.
 966         (stmt_may_clobber_ref_p_1): Likewise.
 967         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
 968         the alias oracle to disambiguate DRs with stmts DR analysis
 969         couldn't handle.
 970         (vect_analyze_data_refs): Do not give up on not analyzable
 971         DRs for BB vectorization.  Remove code truncating the dataref
 972         vector.
 973
 974 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
 975
 976         PR target/85832
 977         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
 978         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
 979         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
 980
 981 2018-05-25  Richard Biener  <rguenther@suse.de>
 982
 983         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
 984         function, combining stmt data ref gathering and fatal analysis
 985         parts.
 986         (vect_analyze_data_refs): Remove now redudnant code and simplify.
 987         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
 988         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
 989         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
 990         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
 991
 992 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
 993
 994         PR tree-optimization/85720
 995         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
 996         SCC if all partitions are builtins.
 997         (version_loop_by_alias_check): New parameter.  Generate cancelable
 998         runtime alias check if all partitions are builtins.
 999         (distribute_loop): Update call to above function.
1000
1001 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
1002
1003         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1004         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1005         (parm_default_def_partition_arg): Ditto.
1006         (set_parm_default_def_partition): Ditto.
1007         (get_parm_default_def_partitions): Ditto and make it static.
1008         (get_undefined_value_partitions): Ditto and make it static.
1009         (remove_ssa_form): Refactor call to init_var_map here.
1010         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1011         computation for loop region.
1012         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1013         (register_default_def): Delete.
1014         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1015         (parm_default_def_partition_arg): Ditto.
1016         (set_parm_default_def_partition): Ditto.
1017         (get_parm_default_def_partitions): Ditto and make it static.
1018         (get_undefined_value_partitions): Ditto and make it static.
1019         (coalesce_with_default, coalesce_with_default): Update comment.
1020         (create_coalesce_list_for_region): New func factored out from
1021         create_outofssa_var_map.
1022         (populate_coalesce_list_for_outofssa): New func factored out from
1023         create_outofssa_var_map and coalesce_ssa_name.
1024         (create_outofssa_var_map): Delete.
1025         (coalesce_ssa_name): Refactor to support live range computation.
1026         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1027         (get_parm_default_def_partitions): Delete.
1028         (get_undefined_value_partitions): Ditto.
1029         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1030         computation for loop region.
1031         (new_tree_live_info, loe_visit_block): Ditto.
1032         (live_worklist, set_var_live_on_entry): Ditto.
1033         (calculate_live_on_exit, verify_live_on_entry): Ditto.
1034         * tree-ssa-live.h (struct _var_map): New fields.
1035         (init_var_map): Change decl.
1036         (region_contains_p): New.
1037
1038 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
1039
1040         * tree-ssa-live.h (live_merge_and_clear): Delete.
1041
1042 2018-05-25  Richard Biener  <rguenther@suse.de>
1043
1044         PR c++/85912
1045         * tree-dump.c (dequeue_and_dump): Remove access to removed
1046         operand 2 of a SWITCH_EXPR.
1047
1048 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1049
1050         * doc/sourcebuild.texi (vect_double_cond_arith): Include
1051         multiplication and division.
1052         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1053         (cond_udiv@var{m}, cond_umod@var{m}): Document.
1054         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1055         (cond_udiv_optab, cond_umod_optab): New optabs.
1056         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1057         (IFN_COND_RDIV): New internal functions.
1058         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1059         TRUNC_MOD_EXPR and RDIV_EXPR.
1060         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1061         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1062         New unspecs.
1063         (SVE_INT_BINARY): Include mult.
1064         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1065         (optab, sve_int_op): Handle mult.
1066         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1067         UNSPEC_COND_DIV.
1068         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1069         for SVE_INT_BINARY_SD.
1070
1071 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1072
1073         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1074         (optab, sve_int_op): Handle div and udiv.
1075         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1076         for SVE_INT_BINARY_SD.
1077         (*<optab><mode>3): New insn for the same.
1078
1079 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1080
1081         * tree-vect-patterns.c: Include predict.h.
1082         (vect_recog_divmod_pattern): Restrict check for division support
1083         to when optimizing for size.
1084
1085 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1086
1087         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1088         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1089         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1090         (gimple_match_op::set_op): Likewise.
1091         (gimple_resimplify4): Declare.
1092         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1093         (expr::gen_transform): Likewise.
1094         (decision_tree::gen): Generate a simplification routine for 4 operands.
1095         * gimple-match-head.c (gimple_simplify): Add an overload for
1096         4 operands.  In the top-level function, handle up to 4 call
1097         arguments and call gimple_resimplify4.
1098         (gimple_resimplify4): New function.
1099         (build_call_internal): Pass a fourth operand.
1100         (maybe_push_to_seq): Likewise.
1101         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1102         Fold VEC_COND_EXPRs of an operation and a default value into
1103         an IFN_COND_* function if possible.
1104         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1105         New unspecs.
1106         (SVE_COND_FP_BINARY): Include them.
1107         (optab, sve_fp_op): Handle them.
1108         (SVE_INT_BINARY_REV): New code iterator.
1109         (SVE_COND_FP_BINARY_REV): New int iterator.
1110         (commutative): New int attribute.
1111         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1112         Declare.
1113         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1114         function.
1115         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1116         (*cond_<optab><mode>): New patterns for reversed operands.
1117
1118 2018-05-25  Richard Biener  <rguenther@suse.de>
1119
1120         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1121         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1122         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1123         (STMT_VINFO_GROUPED_ACCESS): Adjust.
1124         * tree-vect-data-refs.c (everywhere): Adjust users.
1125         * tree-vect-loop.c (everywhere): Likewise.
1126         * tree-vect-slp.c (everywhere): Likewise.
1127         * tree-vect-stmts.c (everywhere): Likewise.
1128         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1129
1130 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1131
1132         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1133         Rename to...
1134         (gcc_cv_as_section_exclude): ... this.
1135         Try Solaris as #exclude syntax.
1136         * configure: Regenerate.
1137         * config.in: Regenerate.
1138         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1139         SECTION_EXCLUDE.
1140         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1141         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1142
1143         * varasm.c (default_elf_asm_named_section): Don't check if
1144         HAVE_GAS_SECTION_EXCLUDE is defined.
1145
1146 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1147
1148         * doc/md.texi: Update the documentation of the cond_* optabs
1149         to mention the new final operand.  Fix GET_MODE_NUNITS call.
1150         Describe the scalar case too.
1151         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1152         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1153         instead of 2.
1154         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1155         (get_conditional_internal_fn): Update comment.
1156         * tree-vect-loop.c (vectorizable_reduction): Pass the original
1157         accumulator value as a final argument to conditional functions.
1158         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1159         a define_expand and add an "else" operand.  Assert for now that
1160         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
1161         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1162         (*cond_<optab><mode>): New patterns.
1163         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1164         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1165         (UNSPEC_COND_EOR): Delete.
1166         (optab): Remove associated mappings.
1167         (SVE_INT_BINARY): New code iterator.
1168         (sve_int_op): Remove int attribute and add "minus" to the code
1169         attribute.
1170         (SVE_COND_INT_OP): Delete.
1171         (SVE_COND_FP_OP): Rename to...
1172         (SVE_COND_FP_BINARY): ...this.
1173
1174 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1175
1176         * optabs.c (can_reuse_operands_p): New function.
1177         (maybe_legitimize_operands): Try to reuse the results for
1178         earlier operands.
1179
1180 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
1181
1182         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1183         Add {q} suffix to insn mnemonic.
1184
1185 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1186
1187         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1188         (msp430_warn_func_return): New.
1189
1190 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
1191
1192         * fold-const.c (tree_nonzero_bits): New function.
1193         * fold-const.h (tree_nonzero_bits): Likewise.
1194         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1195         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1196
1197 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
1198
1199         PR target/85900
1200         PR target/85345
1201         * varasm.c (assemble_alias): Check ifunc_resolver only on
1202         FUNCTION_DECL.
1203
1204 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
1205
1206         PR target/85903
1207         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1208         when memory input operand is handled.
1209
1210 2018-05-24  Luis Machado  <luis.machado@linaro.org>
1211
1212         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1213         global.
1214         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1215
1216 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
1217
1218         * match.pd: Delay FMA folds until after vectorization.
1219
1220 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1221
1222         PR target/83009
1223         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1224         address check not strict.
1225
1226 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
1227
1228         * gimple-match.h (gimple_match_op): New class.
1229         (mprts_hook): Replace parameters with a gimple_match_op *.
1230         (maybe_build_generic_op): Likewise.
1231         (gimple_simplified_result_is_gimple_val): Replace parameters with
1232         a const gimple_match_op *.
1233         (gimple_simplify): Replace code_helper * and tree * parameters with
1234         a gimple_match_op * parameter.
1235         (gimple_resimplify1): Replace code_helper *, tree and tree *
1236         parameters with a gimple_match_op * parameter.
1237         (gimple_resimplify2): Likewise.
1238         (gimple_resimplify3): Likewise.
1239         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1240         parameters with a gimple_match_op * parameter.
1241         * gimple-match-head.c (gimple_simplify): Change prototypes of
1242         auto-generated functions to take a gimple_match_op * instead of
1243         separate code_helper * and tree * parameters.  Make the same
1244         change in the top-level overload and update calls to the
1245         gimple_resimplify routines.  Update calls to the auto-generated
1246         functions and to maybe_push_res_to_seq in the publicly-facing
1247         operation-specific gimple_simplify overloads.
1248         (gimple_match_op::MAX_NUM_OPS): Define.
1249         (gimple_resimplify1): Replace rcode and ops with a single res_op
1250         parameter.  Update call to gimple_simplify.
1251         (gimple_resimplify2): Likewise.
1252         (gimple_resimplify3): Likewise.
1253         (mprts_hook): Replace parameters with a gimple_match_op *.
1254         (maybe_build_generic_op): Likewise.
1255         (build_call_internal): Replace type, nargs and ops with
1256         a gimple_match_op *.
1257         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1258         with a single gimple_match_op *.  Update calls to mprts_hook,
1259         build_call_internal and gimple_simplified_result_is_gimple_val.
1260         Factor out code that is common to the tree_code and combined_fn cases.
1261         * genmatch.c (expr::gen_transform): Replace tem_code and
1262         tem_ops with a gimple_match_op called tem_op.  Update calls
1263         to the gimple_resimplify functions and maybe_push_res_to_seq.
1264         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1265         res_ops.  Update call to the gimple_resimplify functions.
1266         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1267         (decision_tree::gen): Make the functions take a gimple_match_op *
1268         called res_op instead of separate res_code and res_ops parameters.
1269         Update call accordingly.
1270         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1271         and ops with a single res_op parameter.  Update calls to
1272         maybe_build_generic_op and maybe_push_res_to_seq.
1273         (fold_stmt_1): Update calls to gimple_simplify and
1274         replace_stmt_with_simplification.
1275         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1276         and gimple_simplified_result_is_gimple_val.
1277         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1278         gimple_simplify.
1279         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1280         with a gimple_match_op *.
1281         (vn_nary_build_or_lookup): Likewise.  Update call to
1282         vn_nary_build_or_lookup_1.
1283         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1284         gimple_match_op *.  Update calls to the gimple_resimplify routines
1285         and to gimple_simplified_result_is_gimple_val.
1286         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1287         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1288         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1289         (visit_nary_op): Likewise.
1290         (visit_reference_op_load): Likewise.
1291
1292 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1293
1294         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1295         modifier for printing the step amount.
1296
1297 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
1298
1299         PR target/78849
1300         * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1301         types.
1302
1303 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
1304
1305         * doc/sourcebuild.texi (Endianness): New subsubsection.
1306
1307 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1308
1309         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1310         <prefetch_dynamic_strides>: New const bool field.
1311         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1312         prefetch_dynamic_strides.
1313         (exynosm1_prefetch_tune): Likewise.
1314         (thunderxt88_prefetch_tune): Likewise.
1315         (thunderx_prefetch_tune): Likewise.
1316         (thunderx2t99_prefetch_tune): Likewise.
1317         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1318         false.
1319         (aarch64_override_options_internal): Update to set
1320         PARAM_PREFETCH_DYNAMIC_STRIDES.
1321         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1322         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1323         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1324         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1325         prefetch-dynamic-strides setting.
1326
1327 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1328
1329         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1330         <minimum_stride>: New const int field.
1331         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1332         minimum_stride field defaulting to -1.
1333         (exynosm1_prefetch_tune): Likewise.
1334         (thunderxt88_prefetch_tune): Likewise.
1335         (thunderx_prefetch_tune): Likewise.
1336         (thunderx2t99_prefetch_tune): Likewise.
1337         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1338         <default_opt_level>: Set to 3.
1339         (aarch64_override_options_internal): Update to set
1340         PARAM_PREFETCH_MINIMUM_STRIDE.
1341         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1342         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1343         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1344         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1345         stride is constant and is below the minimum stride threshold.
1346
1347 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1348
1349         * config/arm/arm-cpus.in (mode26): Delete.
1350         (armv4): Delete mode26 reference.
1351         * config/arm/arm.c (arm_configure_build_target): Delete use of
1352         isa_bit_mode26.
1353
1354 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
1355
1356         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1357         New insn pattern.
1358         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1359         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
1360         for non-SSE modes.
1361         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
1362         (floatunsdidf2): Ditto.
1363
1364 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
1365
1366         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1367         (fixuns_trunc<mode>si2_avx512f): Ditto.
1368         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1369         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1370         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1371
1372 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
1373
1374         PR rtl-optimization/79985
1375         * df-scan.c (df_insn_refs_collect): Remove special case for
1376         global registers and asm statements.
1377
1378 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
1379
1380         * extend.texi (Global Register Variables): Rewrite the bullet list.
1381         Note that the register is available for allocation. Note that access
1382         via inline asm must use constraints. Add note about async-signal
1383         handlers. Remove paragraph about automagic register selection.
1384
1385 2018-05-23  Richard Biener  <rguenther@suse.de>
1386
1387         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1388         of fixed offset from memset VN.
1389
1390 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
1391
1392         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1393         first_interp field.
1394         (alloc_cand_and_find_basis): Initialize first_interp field.
1395         (slsr_process_mul): Modify first_interp field.
1396         (slsr_process_add): Likewise.
1397         (slsr_process_cast): Modify first_interp field for each new
1398         interpretation.
1399         (slsr_process_copy): Likewise.
1400         (dump_candidate): Dump first_interp field.
1401         (replace_mult_candidate): Process all interpretations, not just
1402         subsequent ones.
1403         (replace_rhs_if_not_dup): Likewise.
1404         (replace_one_candidate): Likewise.
1405
1406 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
1407
1408         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1409         Add new boolean.
1410         (aarch64_needs_frame_chain): New function.
1411         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1412
1413 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
1414
1415         PR target/84882
1416         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1417         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1418         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1419         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1420         as true for strict-align.
1421         (aarch64_can_inline_p): Perform checks even when callee has no
1422         attributes to check for strict alignment.
1423         * doc/extend.texi (AArch64 Function Attributes): Document
1424         no-strict-align.
1425         * doc/invoke.texi: (AArch64 Options): Likewise.
1426
1427 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
1428
1429         PR tree-optimization/85853
1430         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1431         the handling of the root of the node to...
1432         (vect_slp_analyze_node_operations_1): ...this new function,
1433         and run the whole thing with the child nodes' def types
1434         set according to their SLP node's def type.
1435
1436 2018-05-23  Richard Biener  <rguenther@suse.de>
1437
1438         PR middle-end/85874
1439         * tree-data-ref.c (create_runtime_alias_checks): Defer
1440         and ignore overflow warnings.
1441
1442 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
1443
1444         PR tree-optimization/85822
1445         * tree-vrp.c (is_masked_range_test): Fix handling of negative
1446         constants.
1447
1448 2018-05-23  Richard Biener  <rguenther@suse.de>
1449
1450         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1451         memset constants via native_interpret_expr.
1452
1453 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
1454
1455         PR target/85345
1456         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1457         attribute.
1458         (cgraph_node::create_alias): Likewise.
1459         (cgraph_node::get_availability): Check ifunc_resolver instead
1460         of looking up ifunc attribute.
1461         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1462         * varasm.c (do_assemble_alias): Likewise.
1463         (assemble_alias): Likewise.
1464         (default_binds_local_p_3): Likewise.
1465         * cgraph.h (cgraph_node): Add ifunc_resolver.
1466         (cgraph_node::only_called_directly_or_aliased_p): Return false
1467         for IFUNC resolver.
1468         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1469         attribute.
1470         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1471         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1472         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1473         instead of looking up ifunc attribute.
1474
1475 2018-05-22  Luis Machado  <luis.machado@linaro.org>
1476
1477         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1478
1479 2018-05-22  Martin Sebor  <msebor@redhat.com>
1480
1481         PR middle-end/85359
1482         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1483         only when expasion succeeds.
1484         (expand_builtin_strcmp): Same.
1485         (expand_builtin_strncmp): Same.
1486
1487 2018-05-22  Martin Sebor  <msebor@redhat.com>
1488
1489         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1490
1491 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
1492             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1493
1494         * config/aarch64/aarch64-ldpstp.md: Replace uses of
1495         aarch64_mem_pair_operand with memory_operand and delete operand swapping
1496         code.
1497         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1498         Add check for legitimate_address.
1499         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1500         (aarch64_swap_ldrstr_operands): New.
1501         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1502         Define prototype.
1503
1504 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
1505             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1506
1507         * config/aarch64/aarch64.md: New patterns to generate stp
1508         and ldp.
1509         (store_pair_sw, store_pair_dw): New patterns to generate stp for
1510         single words and double words.
1511         (load_pair_sw, load_pair_dw): Likewise.
1512         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1513         Delete.
1514         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1515         Delete.
1516         * config/aarch64/aarch64-ldpstp.md: Modify peephole
1517         for different mode ldpstp and add peephole for merged zero stores.
1518         Likewise for loads.
1519         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1520         Add size check.
1521         (aarch64_gen_store_pair): Rename calls to match new patterns.
1522         (aarch64_gen_load_pair): Rename calls to match new patterns.
1523         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1524         (load_pair<DREG:mode><DREG2:mode>): ... This.
1525         (store_pair<mode>): Rename to...
1526         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1527         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1528         New mode iterators.
1529         (V_INT_EQUIV): Handle SImode.
1530         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1531         New predicate.
1532
1533 2018-05-22  Martin Sebor  <msebor@redhat.com>
1534
1535         PR c/85623
1536         * calls.c (maybe_warn_nonstring_arg): Use string length to set
1537         or ajust the presumed bound on an operation to avoid unnecessary
1538         warnings.
1539
1540 2018-05-22  Martin Sebor  <msebor@redhat.com>
1541
1542         PR tree-optimization/85826
1543         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1544         assuming that a DECL necesarily has a constant size.
1545
1546 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
1547
1548         PR middle-end/85862
1549         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1550
1551 2018-05-22  Richard Biener  <rguenther@suse.de>
1552
1553         PR tree-optimization/85834
1554         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1555         non-constant and non-zero memset arguments.
1556
1557 2018-05-22  Martin Liska  <mliska@suse.cz>
1558
1559         PR ipa/85607
1560         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1561
1562 2018-05-22  Richard Biener  <rguenther@suse.de>
1563
1564         PR tree-optimization/85863
1565         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1566         comparisons when vectype is specified.
1567         (vectorizable_condition): Do not specify vectype for
1568         vect_is_simple_cond when SLP vectorizing.
1569
1570 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
1571
1572         PR target/85657
1573         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1574         define __ibm128 as long double.
1575         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1576         as a distinct type when IEEE 128-bit support is enabled.
1577         (init_float128_ieee): Fix up conversions between IFmode and IEEE
1578         128-bit types to use the correct functions.
1579         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1580         convert between 128-bit floating point types that have different
1581         modes but the same representation, instead of using gen_lowpart to
1582         makean alias.
1583         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1584         KFmode.
1585         (IFKF_reg): New attributes to give the register constraints for
1586         IFmode and KFmode.
1587         (extend<mode>tf2_internal): New insns to mark an explicit
1588         conversion between 128-bit floating point types that have a
1589         different mode but share the same representation.
1590
1591 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
1592
1593         PR tree-optimization/85814
1594         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1595         a null return from get_strinfo when unsharing the next
1596         strinfo in the chain.
1597
1598 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
1599
1600         PR gcc/84923
1601         * varasm.c (weak_finish): Clean up weak_decls.
1602
1603 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1604
1605         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1606         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1607         UNSPEC_UADALP values.
1608         * config/aarch64/iterators.md (ABAL): New int iterator.
1609         (ABDL2): Likewise.
1610         (ADALP): Likewise.
1611         (sur): Add mappings for the above.
1612         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1613         New define_insn.
1614         (aarch64_<sur>abal<mode>_4): Likewise.
1615         (aarch64_<sur>adalp<mode>_3): Likewise.
1616         (<sur>sadv16qi): New define_expand.
1617
1618 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
1619
1620         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1621         (*movdf_internal): Ditto.
1622         (*rcpsf2_sse): Ditto.
1623         (*rsqrtsf2_sse): Ditto.
1624         (*sqrt<mode>2_sse): Ditto.
1625
1626 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
1627
1628         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1629         eor3q<mode>4.
1630         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1631         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1632         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1633         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1634         vbcaxq_s64): New.
1635         * config/aarch64/arm_neon.h: Likewise.
1636         * config/aarch64/iterators.md (VQ_I): New.
1637
1638 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
1639
1640         * config.gcc: Add arc/t-multilib-linux to tmake_file for
1641         arc*-*-linux*.
1642         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1643         MULTILIB_DIRNAMES
1644
1645 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
1646
1647         * config/nds32/constraints.md (S): New constraint.
1648         * config/nds32/nds32.md (call_internal): Use constraint S.
1649         (call_value_internal): Likewise.
1650         (sibcall_internal): Likewise.
1651         (sibcall_value_internal): Likewise.
1652
1653 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
1654             Chung-Ju Wu  <jasonwucj@gmail.com>
1655
1656         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1657         into consideration.
1658
1659 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
1660             Chung-Ju Wu  <jasonwucj@gmail.com>
1661
1662         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1663         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1664         (nds32_rtx_costs_impl): Simplify.
1665         (nds32_address_cost_impl): Simplify.
1666         (nds32_init_rtx_costs): New function.
1667         (nds32_rtx_costs_speed_prefer): Likewise.
1668         (nds32_rtx_costs_size_prefer): Likewise.
1669         (nds32_address_cost_speed_prefer): Likewise.
1670         (nds32_address_cost_speed_fwprop): Likewise.
1671         (nds32_address_cost_size_prefer): Likewise.
1672         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1673         * config/nds32/nds32.c (nds32_option_override): Use
1674         nds32_init_rtx_costs function.
1675
1676 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
1677
1678         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1679         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1680         (TARGET_PIPELINE_N8): Likewise.
1681         (TARGET_PIPELINE_N10): Likewise.
1682         (TARGET_PIPELINE_N13): Likewise.
1683         (TARGET_PIPELINE_GRAYWOLF): Likewise.
1684
1685 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
1686
1687         * config/nds32/nds32-fpu.md: Update copyright year.
1688
1689 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1690
1691         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1692
1693 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1694
1695         * config/nds32/nds32.c
1696         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1697         * config/nds32/nds32.opt (minline-asm-r15): New option.
1698
1699 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1700
1701         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1702         MASK_HW_ABS.
1703         * config/nds32/nds32.md (abssi2): New pattern.
1704
1705 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
1706
1707         * config/i386/i386.md (rex64namesuffix): New mode attribute.
1708         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1709         Merge insn pattern from sse_cvtsi2ss<round_name> and
1710         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1711         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1712         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1713         using SWI48 mode iterator.
1714         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1715         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1716         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1717         pattern from sse_cvttss2si<round_saeonly_name>
1718         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1719         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1720         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1721         using SWI48 mode iterator.
1722         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1723         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1724         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1725         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1726         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1727         using SWI48 mode iterator.
1728         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1729         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1730         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1731         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1732         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1733         SWI48 mode iterator.
1734         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1735         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1736         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1737         pattern from sse_cvttsd2si<round_saeonly_name>
1738         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1739
1740 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1741
1742         * config/nds32/nds32-md-auxiliary.c
1743         (nds32_valid_smw_lwm_base_p): Refine.
1744         (nds32_output_smw_single_word): Refine.
1745         (nds32_output_smw_double_word): New.
1746         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1747
1748 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1749
1750         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1751         (nds32_output_stack_pop): Refine.
1752         (nds32_expand_unaligned_load): Refine.
1753         (nds32_expand_unaligned_store): Refine.
1754
1755 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
1756             Chung-Ju Wu  <jasonwucj@gmail.com>
1757
1758         * config/nds32/constants.md: Add TP_REGNUM constant.
1759         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1760         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1761         UNSPEC_ADD32.
1762         * config/nds32/nds32-doubleword.md: Consider flag_pic.
1763         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1764         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1765         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1766         and PIC code generation.
1767         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1768         code generation.
1769         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1770         optimization.
1771         * config/nds32/nds32.md: Support TLS and PIC.
1772         * config/nds32/nds32.c: Support TLS and PIC.
1773         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1774         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1775         predicate.
1776
1777 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1778
1779         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1780         mode with E_ prefix.
1781
1782 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
1783             Chung-Ju Wu  <jasonwucj@gmail.com>
1784
1785         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1786         * config/nds32/nds32-md-auxiliary.c
1787         (symbolic_reference_mentioned_p): New.
1788         (nds32_legitimize_ict_address): New.
1789         (nds32_expand_ict_move): New.
1790         (nds32_indirect_call_referenced_p): New.
1791         (nds32_symbol_binds_local_p): Delete.
1792         (nds32_long_call_p): Modify.
1793         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1794         * config/nds32/nds32-protos.h
1795         (symbolic_reference_mentioned_p): Declare.
1796         (nds32_legitimize_ict_address): Declare.
1797         (nds32_expand_ict_move): Declare.
1798         (nds32_indirect_call_referenced_p): Declare.
1799         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1800         (nds32_relax_group): Use nds32_ict_const_p as condition.
1801         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1802         (nds32_asm_file_start): Output ict_model directive in asm code.
1803         (nds32_legitimate_address_p): Consider indirect call.
1804         (nds32_print_operand): Consider indirect call.
1805         (nds32_print_operand_address): Consider indirect call.
1806         (nds32_insert_attributes): Handle "indirect_call" attribute.
1807         (TARGET_LEGITIMATE_ADDRESS_P): Define.
1808         (TARGET_LEGITIMATE_CONSTANT_P): Define.
1809         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1810         (TARGET_DELEGITIMIZE_ADDRESS): Define.
1811         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1812         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1813         (TARGET_ICT_MODEL_SMALL): Define.
1814         (TARGET_ICT_MODEL_LARGE): Define.
1815         * config/nds32/nds32.md (movsi): Consider ict model.
1816         (call, call_value): Consider ict model.
1817         (sibcall, sibcall_value): Consider ict model.
1818         * config/nds32/nds32.opt (mict-model): New option.
1819         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1820         model.
1821
1822 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
1823             Monk Chiang  <sh.chiang04@gmail.com>
1824             Jim Wilson <jimw@sifive.com>
1825
1826         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1827         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
1828         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1829         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1830         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
1831         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1832         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1833         compute save_libcall_adjustment properly.
1834         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1835         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1836         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1837         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1838         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1839         (ABI_SPEC): Handle mabi=ilp32e.
1840         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1841         (RVE): Add RVE mask.
1842         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1843         <-march>: Add rv32e as an example.
1844
1845 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
1846
1847         PR c++/82899
1848         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1849         (intra_create_variable_infos): Handle C++ constructors.
1850
1851 2018-05-18  Martin Liska  <mliska@suse.cz>
1852
1853         * passes.def: Remove a redundant pass.
1854
1855 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
1856
1857         PR bootstrap/85838
1858         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1859
1860 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1861
1862         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1863         (ARMv4): Update.
1864         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1865         (ARMv6m): Update.
1866         (armv2, armv2a, armv3, armv3m): Delete architectures.
1867         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1868         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1869         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1870         Delete cpus.
1871         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1872         (*mulsidi3adddi): Likewise.
1873         (mulsidi3): Likewise.
1874         (*mulsidi3_nov6): Likewise.
1875         (umulsidi3): Likewise.
1876         (umulsidi3_nov6): Likewise.
1877         (umaddsidi4): Likewise.
1878         (*umulsidi3adddi): Likewise.
1879         (smulsi3_highpart): Likewise.
1880         (*smulsi3_highpart_nov6): Likewise.
1881         (umulsi3_highpart): Likewise.
1882         (*umulsi3_highpart_nov6): Likewise.
1883         * config/arm/arm.h (arm_arch3m): Delete.
1884         * config/arm/arm.c (arm_arch3m): Delete.
1885         (arm_option_override_internal): Update armv3-related comment.
1886         (arm_configure_build_target): Delete use of isa_bit_mode32.
1887         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1888         (arm_rtx_costs_internal): Delete check of arm_arch3m.
1889         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1890         (mulsa3): Likewise.
1891         (mulusa3): Likewise.
1892         * config/arm/arm-protos.h (arm_arch3m): Delete.
1893         * config/arm/arm-tables.opt: Regenerate.
1894         * config/arm/arm-tune.md: Likewise.
1895         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1896         deleted architectures.
1897
1898 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1899
1900         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1901         (armv5t, armv5te): New features.
1902         (ARMv5, ARMv5e): Delete fgroups.
1903         (ARMv5t, ARMv5te): Adjust for above changes.
1904         (ARMv6m): Likewise.
1905         (armv5, armv5e): Delete arches.
1906         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1907         arm_arch5.
1908         (*call_reg_arm): Likewise.
1909         (*call_value_reg_armv5): Likewise.
1910         (*call_value_reg_arm): Likewise.
1911         (*call_symbol): Likewise.
1912         (*call_value_symbol): Likewise.
1913         (*sibcall_insn): Likewise.
1914         (*sibcall_value_insn): Likewise.
1915         (clzsi2): Likewise.
1916         (prefetch): Likewise.
1917         (define_split and define_peephole2 dependent on arm_arch5):
1918         Likewise.
1919         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1920         arm_arch5e.
1921         (TARGET_ARM_QBIT): Likewise.
1922         (TARGET_DSP_MULTIPLY): Likewise.
1923         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1924         (arm_arch5, arm_arch5e): Delete.
1925         (arm_arch5t, arm_arch5te): Declare.
1926         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1927         (arm_arch5t): Declare.
1928         (arm_option_reconfigure_globals): Update for the above.
1929         (arm_options_perform_arch_sanity_checks): Update comment, replace
1930         use of arm_arch5 with arm_arch5t.
1931         (use_return_insn): Likewise.
1932         (arm_emit_call_insn): Likewise.
1933         (output_return_instruction): Likewise.
1934         (arm_final_prescan_insn): Likewise.
1935         (arm_coproc_builtin_available): Likewise.
1936         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1937         arm_arch5e with arm_arch5t and arm_arch5te.
1938         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1939         (arm_arch5t, arm_arch5te): Declare.
1940         * config/arm/arm-tables.opt: Regenerate.
1941         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1942         * config/arm/t-multilib: Likewise.
1943         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1944         instead of arm_arch5.
1945         (*call_reg_thumb1): Likewise.
1946         (*call_value_reg_thumb1_v5): Likewise.
1947         (*call_value_reg_thumb1): Likewise.
1948         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1949         unreachable path.
1950         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1951
1952 2018-05-18  Martin Liska  <mliska@suse.cz>
1953
1954         PR gcov-profile/84846
1955         * doc/gcov.texi: Document -t option of gcov tool.
1956
1957 2018-05-18  Martin Liska  <mliska@suse.cz>
1958
1959         PR gcov-profile/84846
1960         * gcov.c (print_usage): Add new -t option.
1961         (process_args): Handle the option.
1962         (generate_results): Use stdout as output when requested by
1963         the option.
1964
1965 2018-05-18  Martin Liska  <mliska@suse.cz>
1966
1967         PR gcov-profile/84846
1968         * coverage.c (coverage_init): Write PWD to .gcno file.
1969         * doc/gcov.texi: Document how working directory is printed.
1970         * gcov-dump.c (dump_gcov_file): Print PWD.
1971         * gcov.c (output_intermediate_file): Likewise.
1972         (read_graph_file): Read PWD string.
1973         (output_lines): Print PWD.
1974
1975 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1976
1977         PR middle-end/85817
1978         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1979         for retval and return false if all args to phi are zero.
1980
1981 2018-05-18  Richard Biener  <rguenther@suse.de>
1982
1983         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1984         method.
1985         (evrp_dom_walker::before_dom_children): Call it.
1986
1987 2018-05-18  Richard Biener  <rguenther@suse.de>
1988
1989         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1990         results when processing array refs with variable index.
1991
1992 2018-05-18  Toon Moene  <toon@moene.org>
1993
1994         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1995         directly after that of -floop-interchange. Indicate that both
1996         options are enabled by default when specifying -O3.
1997
1998 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1999
2000         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2001         iterator.  Delete separate integer-mode vec_set<mode> expander.
2002         (aarch64_simd_vec_setv2di): Delete.
2003         (vec_setv2di): Delete.
2004         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2005         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
2006         the "w, r" alternative.
2007
2008 2018-05-18  Martin Liska  <mliska@suse.cz>
2009
2010         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2011         * tree-pass.h (make_pass_lower_switch_O0): New function.
2012         * tree-switch-conversion.c (node_has_low_bound): Remove.
2013         (node_has_high_bound): Likewise.
2014         (node_is_bounded): Likewise.
2015         (class pass_lower_switch): Make it a template type and create
2016         two instances.
2017         (pass_lower_switch::execute): Add template argument.
2018         (make_pass_lower_switch): New function.
2019         (make_pass_lower_switch_O0): New function.
2020         (do_jump_if_equal): Remove.
2021         (emit_case_nodes): Simplify to just handle all 3 cases and leave
2022         all the hard work to tree optimization passes.
2023
2024 2018-05-18  Martin Liska  <mliska@suse.cz>
2025
2026         * dbgcnt.c (limit_low): Renamed from limit.
2027         (limit_high): New variable.
2028         (dbg_cnt_is_enabled): Check for upper limit.
2029         (dbg_cnt): Adjust dumping.
2030         (dbg_cnt_set_limit_by_index): Add new argument for high
2031         value.
2032         (dbg_cnt_set_limit_by_name): Likewise.
2033         (dbg_cnt_process_single_pair): Parse new format.
2034         (dbg_cnt_process_opt): Use strtok.
2035         (dbg_cnt_list_all_counters): Remove 'value' and add
2036         'limit_high'.
2037         * doc/invoke.texi: Document changes.
2038
2039 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
2040
2041         * doc/sourcebuild.texi (scalar_all_fma): Document.
2042         * tree.def (FMA_EXPR): Delete.
2043         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2044         * internal-fn.c (ternary_direct): New macro.
2045         (expand_ternary_optab_fn): Likewise.
2046         (direct_ternary_optab_supported_p): Likewise.
2047         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2048         * builtins.c (fold_builtin_fma): Delete.
2049         (fold_builtin_3): Don't call it.
2050         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2051         * expr.c (expand_expr_real_2): Likewise.
2052         * fold-const.c (operand_equal_p): Likewise.
2053         (fold_ternary_loc): Likewise.
2054         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2055         * gimple.c (DEFTREECODE): Likewise.
2056         * gimplify.c (gimplify_expr): Likewise.
2057         * optabs-tree.c (optab_for_tree_code): Likewise.
2058         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2059         * tree-eh.c (operation_could_trap_p): Likewise.
2060         (stmt_could_throw_1_p): Likewise.
2061         * tree-inline.c (estimate_operator_cost): Likewise.
2062         * tree-pretty-print.c (dump_generic_node): Likewise.
2063         (op_code_prio): Likewise.
2064         * tree-ssa-loop-im.c (stmt_cost): Likewise.
2065         * tree-ssa-operands.c (get_expr_operands): Likewise.
2066         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2067         * fold-const-call.h (fold_fma): Delete.
2068         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2069         CFN_FNMA and CFN_FNMS.
2070         (fold_fma): Delete.
2071         * genmatch.c (combined_fn): New enum.
2072         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2073         (commutative_op): New function.
2074         (commutate): Use it.  Handle more than 2 operands.
2075         (dt_operand::gen_gimple_expr): Use commutative_op.
2076         (parser::parse_expr): Allow :c to be used with non-binary
2077         operators if the commutative operand is known.
2078         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2079         CFN_FMS, CFN_FNMA and CFN_FNMS.
2080         (backprop::process_assign_use): Remove FMA_EXPR handling.
2081         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2082         (gen_hsa_fma): New function.
2083         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2084         IFN_FNMA and IFN_FNMS.
2085         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2086         * gimple-fold.h (follow_all_ssa_edges): Declare.
2087         * gimple-fold.c (follow_all_ssa_edges): New function.
2088         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2089         gimple_build interface and use follow_all_ssa_edges to fold the result.
2090         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2091         instead of checking for optabs directly.
2092         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2093         rather than FMA_EXPRs.
2094         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2095         call to IFN_FMA instead of an FMA_EXPR.
2096
2097 2018-05-17  Jim Wilson  <jimw@sifive.com>
2098
2099         * expr.c (do_tablejump): When converting index to Pmode, if we have a
2100         sign extended promoted subreg, and the range does not have the sign bit
2101         set, then do a sign extend.
2102
2103         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2104         test, check for sign extended subreg and/or constant operands, and
2105         do a sign extend in that case.
2106
2107 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
2108
2109         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2110         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2111         Add untyped.
2112         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2113         Change logics_shift_reg to logics_shift_imm.
2114         (thunderx2t99_fp_loadpair_basic): Delete.
2115         (thunderx2t99_fp_storepair_basic): Delete.
2116         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2117         (thunderx2t99_asimd_polynomial): Delete.
2118         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2119         and neon_fp_mul_d_scalar_q.
2120         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2121         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2122         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2123         (thunderx2t99_asimd_lut): Add missing tbl types.
2124         (thunderx2t99_asimd_ext): Delete.
2125         (thunderx2t99_asimd_load1_1_mult): Delete.
2126         (thunderx2t99_asimd_load1_2_mult): Delete.
2127         (thunderx2t99_asimd_load1_ldp): New.
2128         (thunderx2t99_asimd_load1): New.
2129         (thunderx2t99_asimd_load2): Add missing *load2* types.
2130         (thunderx2t99_asimd_load3): New.
2131         (thunderx2t99_asimd_load4): New.
2132         (thunderx2t99_asimd_store1_1_mult): Delete.
2133         (thunderx2t99_asimd_store1_2_mult): Delete.
2134         (thunderx2t99_asimd_store2_mult): Delete.
2135         (thunderx2t99_asimd_store2_onelane): Delete.
2136         (thunderx2t99_asimd_store_stp): New.
2137         (thunderx2t99_asimd_store1): New.
2138         (thunderx2t99_asimd_store2): New.
2139         (thunderx2t99_asimd_store3): New.
2140         (thunderx2t99_asimd_store4): New.
2141
2142 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
2143
2144         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2145         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
2146
2147 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
2148             Segher Boessenkool  <segher@kernel.crashing.org>
2149
2150         PR target/85698
2151         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2152         operand.
2153
2154 2018-05-17  Richard Biener  <rguenther@suse.de>
2155
2156         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2157         for pruning loop and prune defs feeding only already visited PHIs.
2158
2159 2018-05-17  Richard Biener  <rguenther@suse.de>
2160
2161         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2162
2163 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
2164             Richard Biener  <rguenther@suse.de>
2165
2166         PR tree-optimization/85793
2167         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2168         for VMAT_ELEMENTWISE.
2169
2170 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
2171
2172         * internal-fn.h (lookup_internal_fn): Declare
2173         * internal-fn.c (lookup_internal_fn): New function.
2174         * gimple.c (gimple_build_call_from_tree): Handle calls to
2175         internal functions.
2176         * gimple-pretty-print.c (dump_gimple_call): Print "." before
2177         internal function names.
2178         * tree-pretty-print.c (dump_generic_node): Likewise.
2179         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2180
2181 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
2182
2183         * gimple-fold.h (gimple_build): Make the function forms take
2184         combined_fn rather than built_in_function.
2185         (gimple_simplify): Likewise.
2186         * gimple-match-head.c (gimple_simplify): Likewise.
2187         * gimple-fold.c (gimple_build): Likewise.
2188         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2189         rather than gimple_build_call_internal.
2190         (get_initial_defs_for_reduction): Likewise.
2191         (vect_create_epilog_for_reduction): Likewise.
2192         (vectorizable_live_operation): Likewise.
2193
2194 2018-05-17  Martin Liska  <mliska@suse.cz>
2195
2196         * gimple-ssa-sprintf.c (format_directive): Do not use
2197         space in between 'G_' and '('.
2198
2199 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
2200
2201         PR target/85323
2202         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2203         even if the mask is not all ones.
2204
2205         PR target/85323
2206         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2207         vector.
2208         (ix86_gimple_fold_builtin): Likewise.
2209
2210         PR target/85323
2211         * config/i386/i386.c: Include tree-vector-builder.h.
2212         (ix86_vector_shift_count): New function.
2213         (ix86_fold_builtin): Fold shift builtins by scalar count.
2214         (ix86_gimple_fold_builtin): Likewise.
2215
2216         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2217         _mm512_setzero): New intrinsics.
2218
2219 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
2220             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2221
2222         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2223         code generation for cases where splatting a value is not useful.
2224         * simplify-rtx.c (simplify_ternary_operation): Simplify
2225         vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2226         mode to a vec_concat.
2227
2228 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
2229
2230         * config.gcc: Support "goldmont-plus".
2231         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2232         "goldmont-plus".
2233         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2234         PROCESSOR_GOLDMONT_PLUS.
2235         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2236         (processor_target_table): Add "goldmont-plus".
2237         (PTA_GOLDMONT_PLUS): Define.
2238         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2239         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2240         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2241         (fold_builtin_cpu): Add "goldmont-plus".
2242         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2243         (ix86_option_override_internal): Add "goldmont-plus".
2244         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2245         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2246         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2247         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2248
2249 2018-05-17  Richard Biener  <rguenther@suse.de>
2250
2251         PR tree-optimization/85757
2252         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2253         remove defs that only feed that PHI from further processing.
2254
2255 2018-05-16  Jim Wilson  <jimw@sifive.com>
2256
2257         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2258         asterisk to name.
2259         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2260
2261 2018-05-16  Mark Wielaard  <mark@klomp.org>
2262
2263         * dwarf2out.c (count_index_strings): New function.
2264         (output_indirect_strings): Call count_index_strings and generate
2265         header for dwarf_version >= 5.
2266
2267 2018-05-16  Mark Wielaard  <mark@klomp.org>
2268
2269         * dwarf2out.c (dwarf_FORM): New function.
2270         (set_indirect_string): Use dwarf_FORM.
2271         (reset_indirect_string): Likewise.
2272         (size_of_die): Likewise.
2273         (value_format): Likewise.
2274         (output_die): Likewise.
2275         (add_skeleton_AT_string): Likewise.
2276         (output_macinfo_op): Likewise.
2277         (index_string): Likewise.
2278         (output_index_string_offset): Likewise.
2279         (output_index_string): Likewise.
2280         (count_index_strings): Likewise.
2281
2282 2018-05-16  Carl Love  <cel@us.ibm.com>
2283
2284         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2285         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2286
2287 2018-05-16  Martin Jambor  <mjambor@suse.cz>
2288
2289         * ipa-prop.c (ipa_free_all_edge_args): Remove.
2290         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2291
2292 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
2293
2294         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2295         (fnma<mode>4): Likewise.
2296         (fms<mode>4): Likewise.
2297         (fnms<mode>4): Likewise.
2298         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2299         (aarch64_fnma<mode>4): Likewise.
2300         (aarch64_fms<mode>4): Likewise.
2301         (aarch64_fnms<mode>4): Likewise.
2302         (aarch64_fnmadd<mode>4): Likewise.
2303
2304 2018-05-16  Jason Merrill  <jason@redhat.com>
2305
2306         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
2307
2308 2018-05-16  Richard Biener  <rguenther@suse.de>
2309
2310         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2311         (dump_stmt_cost): Declare.
2312         (add_stmt_cost): Dump cost we add.
2313         (add_stmt_costs): New function.
2314         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2315         No longer exported.
2316         (vect_analyze_stmt): Adjust prototype.
2317         (vectorizable_condition): Likewise.
2318         (vectorizable_live_operation): Likewise.
2319         (vectorizable_reduction): Likewise.
2320         (vectorizable_induction): Likewise.
2321         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2322         cost vector to pass to vectorizable_ and record afterwards.
2323         (vect_model_reduction_cost): Take cost vector argument and adjust.
2324         (vect_model_induction_cost): Likewise.
2325         (vectorizable_reduction): Likewise.
2326         (vectorizable_induction): Likewise.
2327         (vectorizable_live_operation): Likewise.
2328         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2329         SLP_TREE_NUMBER_OF_VEC_STMTS.
2330         (vect_analyze_slp_cost_1): Remove.
2331         (vect_analyze_slp_cost): Likewise.
2332         (vect_slp_analyze_node_operations): Take visited args and
2333         a target cost vector.  Avoid processing already visited stmt sets.
2334         (vect_slp_analyze_operations): Use a local cost vector to gather
2335         costs and register those of non-discarded instances.
2336         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2337         (vect_schedule_slp_instance): Remove copying of
2338         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
2339         zero.
2340         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2341         adding cost.  Record cost entry location.
2342         (vect_prologue_cost_for_slp_op): Function to compute cost of
2343         a constant or invariant generated for SLP vect in the prologue,
2344         split out from vect_analyze_slp_cost_1.
2345         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
2346         (vect_model_promotion_demotion_cost): Likewise.
2347         (vect_model_store_cost): Likewise, make static.
2348         (vect_model_load_cost): Likewise.
2349         (vectorizable_bswap): Add cost vector arg and adjust.
2350         (vectorizable_call): Likewise.
2351         (vectorizable_simd_clone_call): Likewise.
2352         (vectorizable_conversion): Likewise.
2353         (vectorizable_assignment): Likewise.
2354         (vectorizable_shift): Likewise.
2355         (vectorizable_operation): Likewise.
2356         (vectorizable_store): Likewise.
2357         (vectorizable_load): Likewise.
2358         (vectorizable_condition): Likewise.
2359         (vectorizable_comparison): Likewise.
2360         (can_vectorize_live_stmts): Likewise.
2361         (vect_analyze_stmt): Likewise.
2362         (vect_transform_stmt): Adjust calls to vectorizable_*.
2363         * tree-vectorizer.c: Include gimple-pretty-print.h.
2364         (dump_stmt_cost): New function.
2365
2366 2018-05-16  Richard Biener  <rguenther@suse.de>
2367
2368         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2369         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2370         * tree-ssa-dse.c: Include tree-ssa-loop.h.
2371         (check_name): New callback.
2372         (dse_classify_store): Track cycles via a visited bitmap of PHI
2373         defs and simplify handling of in-loop and across loop dead stores
2374         and properly fail for loop-variant refs.  Handle byte-tracking with
2375         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2376         limiting the walk.
2377
2378 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
2379
2380         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2381         (vect_get_mask_type_for_stmt): Likewise.
2382         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2383         split out from...
2384         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
2385         to determine the statement's vector type and the vector type that
2386         should be used for calculating nunits.  Deal with cases in which
2387         the type has to be deferred.
2388         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2389         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2390         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2391         (vect_determine_vf_for_stmt): New functions, split out from...
2392         (vect_determine_vectorization_factor): ...here.
2393         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2394         (vect_get_mask_type_for_stmt): New functions, split out from
2395         vect_determine_vectorization_factor.
2396
2397 2018-05-16  Richard Biener  <rguenther@suse.de>
2398
2399         * tree-cfg.c (verify_gimple_assign_ternary): Properly
2400         verify the [VEC_]COND_EXPR embedded comparison.
2401
2402 2018-05-15  Martin Sebor  <msebor@redhat.com>
2403
2404         PR tree-optimization/85753
2405         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2406         RECORD_TYPE in addition to ARRAY_TYPE.
2407
2408 2018-05-15  Martin Sebor  <msebor@redhat.com>
2409
2410         PR middle-end/85643
2411         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2412
2413 2018-05-15  Richard Biener  <rguenther@suse.de>
2414
2415         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2416         add by_clobber_p one.  Change algorithm to collect all defs
2417         representing uses we need to walk and try reducing them to
2418         a single one before failing.
2419         (dse_dom_walker::dse_optimize_stmt): Adjust.
2420
2421 2018-05-13  Mark Wielaard  <mark@klomp.org>
2422
2423         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2424         (size_of_loc_descr): Likewise.
2425         (output_loc_operands): Likewise.
2426         (output_loc_operands_raw): Likewise.
2427         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2428         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2429         (hash_loc_operands): Likewise.
2430         (compare_loc_operands): Likewise.
2431
2432 2018-05-14  Mark Wielaard  <mark@klomp.org>
2433
2434         * dwarf2out.c (count_index_addrs): New function.
2435         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2436
2437 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2438
2439         PR tree-optimization/83648
2440         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2441         return value as malloc candidate.
2442
2443 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2444
2445         PR ipa/85734
2446         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2447         as true in call to suggest_attribute.
2448
2449 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
2450
2451         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2452         -mreadonly-in-sdata.
2453
2454 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2455
2456         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2457         New pattern.
2458         (aarch64_crypto_aesd_fused): Likewise.
2459
2460 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
2461
2462         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2463         (movsi_aarch64): Likewise.
2464         (load_pairsi): Likewise.
2465         (load_pairdi): Likewise.
2466         (store_pairsi): Likewise.
2467         (store_pairdi): Likewise.
2468         (load_pairsf): Likewise.
2469         (load_pairdf): Likewise.
2470         (store_pairsf): Likewise.
2471         (store_pairdf): Likewise.
2472         (zero_extend): Likewise.
2473         (trunc): Swap alternatives.
2474         (fcvt_target): Add '?' to prefer w over r.
2475
2476 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
2477
2478         PR target/85756
2479         * config/i386/i386.md: Disallow non-commutative arithmetics in
2480         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2481         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
2482         in the peephole2 before it.
2483
2484 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
2485
2486         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2487         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2488         (ix86_handle_option): Handle -mcldemote.
2489         * config.gcc: New header.
2490         * config/i386/cldemoteintrin.h: New file.
2491         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2492         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2493         -mcldemote.
2494         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2495         OPTION_MASK_ISA_CLDEMOTE.
2496         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2497         (ix86_valid_target_attribute_inner_p): Ditto.
2498         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2499         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2500         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2501         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2502         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2503         (cldemote): New.
2504         * config/i386/i386.opt: Add -mcldemote.
2505         * config/i386/x86intrin.h: New header.
2506         * doc/invoke.texi: Add -mcldemote.
2507
2508 2018-05-14  Richard Biener  <rguenther@suse.de>
2509
2510         * doc/match-and-simplify.texi: Adjust :s documentation.
2511
2512 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
2513
2514         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2515         intended memcpy size.
2516         (REORDER_45): Likewise.
2517
2518 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
2519
2520         * sort.cc: New file.
2521         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2522         * vec.c (qsort_chk): Use gcc_qsort.
2523         * Makefile.in (OBJS-libcommon): Add sort.o.
2524         (build/sort.o): New target.  Use it...
2525         (BUILD_RTL): ... here, and...
2526         (build/gencfn-macros): ... here, and...
2527         (build/genmatch): ... here.
2528
2529 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2530             Chung-Ju Wu  <jasonwucj@gmail.com>
2531
2532         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2533         * config/nds32/nds32-graywolf.md: New file.
2534         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2535         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2536         pipeline.
2537         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2538         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2539         * config/nds32/nds32.md (pipeline_model): Add graywolf.
2540         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2541         * config/nds32/pipelines.md: Include n15 settings.
2542
2543 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2544             Chung-Ju Wu  <jasonwucj@gmail.com>
2545
2546         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2547         * config/nds32/nds32-n13.md: New file.
2548         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2549         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2550         pipeline.
2551         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2552         * config/nds32/nds32.md (pipeline_model): Add n13.
2553         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2554         * config/nds32/pipelines.md: Include n13 settings.
2555
2556 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2557             Chung-Ju Wu  <jasonwucj@gmail.com>
2558
2559         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2560         * config/nds32/nds32-n10.md: New file.
2561         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2562         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2563         pipeline.
2564         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2565         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2566         * config/nds32/nds32.md (pipeline_model): Add n10.
2567         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2568         * config/nds32/pipelines.md: Include n10 settings.
2569
2570 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
2571             Kito Cheng  <kito.cheng@gmail.com>
2572             Chung-Ju Wu  <jasonwucj@gmail.com>
2573
2574         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2575         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2576         Add enum values for DSP extension instructions.
2577         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2578         New constraints.
2579         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2580         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2581         New code iterators.
2582         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2583         * config/nds32/nds32-dspext.md: New file for DSP implementation.
2584         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2585         * config/nds32/nds32-intrinsic.md: Likewise.
2586         * config/nds32/nds32_intrinsic.h: Likewise.
2587         * config/nds32/nds32-md-auxiliary.c: Likewise.
2588         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2589         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2590         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2591         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2592         * config/nds32/nds32-protos.h: New declarations for DSP extension.
2593         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2594         TYPE_DMAC in switch statement.
2595         * config/nds32/nds32.c: New checking and implementation for DSP
2596         extension instructions.
2597         * config/nds32/nds32.h: Likewise.
2598         * config/nds32/nds32.md: Likewise.
2599         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2600         * config/nds32/predicates.md: Implement new predicates for DSP
2601         extension.
2602
2603 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2604
2605         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2606         Reformat alternatives and attributes so it is easier to identify
2607         which constraints/attributes go with which instruction.
2608         (mov<mode>_hardfloat32, FMOVE64): Likewise.
2609         (mov<mode>_softfloat32, FMOVE64): Likewise.
2610         (mov<mode>_hardfloat64, FMOVE64): Likewise.
2611         (mov<mode>_softfloat64, FMOVE64): Likewise.
2612
2613 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2614
2615         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2616         subsection.
2617         (Basic PowerPC Built-in Functions): The new name of the
2618         subsection previously known as "PowerPC Built-in Functions".
2619         (Basic PowerPC Built-in Functions Available on all Configurations):
2620         New subsubsection.
2621         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2622         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2623         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2624         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2625
2626 2018-05-11  Martin Jambor  <mjambor@suse.cz>
2627
2628         PR ipa/85655
2629         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2630         single const.
2631
2632 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
2633
2634         PR target/85733
2635         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2636
2637 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
2638
2639         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2640         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2641         (ix86_handle_option): Handle -mwaitpkg.
2642         * config.gcc: New header.
2643         * config/i386/cpuid.h (bit_WAITPKG): New bit.
2644         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2645         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2646         function type.
2647         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2648         OPTION_MASK_ISA_WAITPKG.
2649         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2650         (ix86_option_override_internal): Add PTA_WAITPKG.
2651         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2652         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2653         IX86_BUILTIN_TPAUSE.
2654         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2655         __builtin_ia32_umwait and __builtin_ia32_tpause.
2656         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2657         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2658         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2659         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2660         UNSPECV_TPAUSE): New.
2661         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2662         * config/i386/i386.opt: Add -mwaitpkg.
2663         * config/i386/waitpkgintrin.h: New file.
2664         * config/i386/x86intrin.h: New header.
2665         * doc/invoke.texi: Add -mwaitpkg.
2666
2667 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
2668
2669         PR target/85606
2670         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2671         equivalent.
2672         (cortex-m0): Use armv6s-m isa.
2673         (cortex-m0plus): Likewise.
2674         (cortex-m1): Likewise.
2675         (cortex-m0.small-multiply): Likewise.
2676         (cortex-m0plus.small-multiply): Likewise.
2677         (cortex-m1.small-multiply): Likewise.
2678
2679 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
2680             Jakub Jelinek  <jakub@redhat.com>
2681
2682         PR tree-optimization/85692
2683         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2684         source permute as well.
2685
2686 2018-05-11  Martin Liska  <mliska@suse.cz>
2687
2688         PR sanitizer/85556
2689         * doc/extend.texi: Document LLVM style format for no_sanitize
2690         attribute.
2691
2692 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
2693
2694         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2695         mode_supports_vsx_dform_quad to mode_supports_dq_form.
2696         (mode_supports_vsx_dform_quad): Likewise.
2697         (mode_supports_vmx_dform): Move these functions to be next to the
2698         other mode_supports functions.
2699         (mode_supports_dq_form): Likewise.
2700         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2701         mode_supports_dq_form.
2702         (reg_offset_addressing_ok_p): Likewise.
2703         (offsettable_ok_by_alignment): Likewise.
2704         (rs6000_legitimate_offset_address_p): Likewise.
2705         (legitimate_lo_sum_address_p): Likewise.
2706         (rs6000_legitimize_address): Likewise.
2707         (rs6000_legitimize_reload_address): Likewise.
2708         (rs6000_secondary_reload_inner): Likewise.
2709         (rs6000_preferred_reload_class): Likewise.
2710         (rs6000_output_move_128bit): Likewise.
2711
2712 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
2713
2714         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2715         Generate SImode target register for null target.
2716         <case IX86_BUILTIN_XGETBV>: Ditto.
2717         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2718         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2719
2720 2018-05-10  Carl Love  <cel@us.ibm.com>
2721
2722         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2723         dcbtt and dcbtstt if operands[2] is 0.
2724
2725 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
2726
2727         PR target/85693
2728         * config/i386/sse.md (usadv64qi): New expander.
2729
2730 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
2731
2732         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2733         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2734         -maltivec=be support.
2735         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2736         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2737         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2738         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2739         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2740         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2741         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2742         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2743         altivec_vsumsws): Adjust.
2744         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2745         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2746         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2747         support.
2748         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2749         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2750         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2751         (altivec_lve<VI_char>x): Delete expand.
2752         (*altivec_lve<VI_char>x_internal): Rename to...
2753         (altivec_lve<VI_char>x): ... this.
2754         (altivec_lvxl_<mode>): Delete expand.
2755         (*altivec_lvxl_<mode>_internal): Rename to ...
2756         (altivec_lvxl_<mode>): ... this.
2757         (altivec_stvxl_<mode>): Delete expand.
2758         (*altivec_stvxl_<mode>_internal): Rename to ...
2759         (altivec_stvxl_<mode>): ... this.
2760         (altivec_stve<VI_char>x): Delete expand.
2761         (*altivec_stve<VI_char>x_internal): Rename to ...
2762         (altivec_stve<VI_char>x): ... this.
2763         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2764         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2765         reduc_plus_scal_<mode>): Adjust.
2766         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2767         comment.
2768         (rs6000_cpu_cpp_builtins): Adjust.
2769         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2770         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2771         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2772         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2773         -maltivec=be support.
2774         (rs6000_split_vec_extract_var): Adjust.
2775         (rs6000_split_v4si_init): Adjust.
2776         (swap_selector_for_mode): Delete.
2777         (altivec_expand_lvx_be, altivec_expand_stvx_be,
2778         altivec_expand_stvex_be): Delete.
2779         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2780         -maltivec=be support.
2781         (rs6000_gimple_fold_builtin): Ditto.
2782         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2783         Adjust.
2784         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2785         (TARGET_DIRECT_MOVE_64BIT): Adjust.
2786         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2787         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2788         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2789         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2790         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2791         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2792         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2793         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2794         anonymous split): Adjust.
2795         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2796         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2797
2798 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
2799
2800         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2801         when --with-gxx-include-dir is also specified.
2802         * configure: Regenerate.
2803
2804 2018-05-09  Jim Wilson  <jimw@sifive.com>
2805
2806         PR target/84797
2807         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2808         * config/riscv/t-withmultilib: New.
2809         * config/riscv/withmultilib.h: New.
2810         * doc/install.texi: Document RISC-V --with-multilib-list support.
2811
2812 2018-05-09  Richard Biener  <rguenther@suse.de>
2813
2814         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2815         vector.
2816         (vect_bb_vectorization_profitable_p): Adjust.  Compute
2817         actual scalar cost using the cost vector and the add_stmt_cost
2818         machinery.
2819
2820 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2821
2822         PR rtl-optimization/85645
2823         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2824         in the REG_CFA_REGISTER note for LR, don't leave it empty.
2825
2826 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2827
2828         PR rtl-optimization/85645
2829         * shrink-wrap.c (spread_components): Return a boolean saying if
2830         anything was changed.
2831         (try_shrink_wrapping_separate): Iterate spread_components until
2832         nothing changes anymore.
2833
2834 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2835
2836         PR rtl-optimization/85645
2837         * regrename.c (build_def_use): Also kill the chains that include the
2838         destination of a REG_CFA_REGISTER note.
2839
2840 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2841
2842         PR rtl-optimization/85645
2843         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2844         insn that has a REG_CFA_REGISTER note.
2845
2846 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
2847
2848         * cfgexpand.c (expand_clobber): New function.
2849         (expand_gimple_stmt_1): Use it.
2850         * tree-vect-stmts.c (vect_clobber_variable): New function,
2851         split out from...
2852         (vectorizable_simd_clone_call): ...here.
2853         (vectorizable_store): Emit a clobber either side of an
2854         IFN_STORE_LANES sequence.
2855         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2856
2857 2018-05-09  Tom de Vries  <tom@codesourcery.com>
2858
2859         PR target/85626
2860         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2861         (define_insn "trap_if_false"): Add exit after trap.
2862
2863 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
2864
2865         PR rtl-optimization/85638
2866         * bb-reorder.c: Include common/common-target.h.
2867         (create_forwarder_block): New function extracted from...
2868         (fix_up_crossing_landing_pad): ...here.  Rename into...
2869         (dw2_fix_up_crossing_landing_pad): ...this.
2870         (sjlj_fix_up_crossing_landing_pad): New function.
2871         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2872         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2873         from both partitions and exit the loop after one iteration.
2874
2875 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2876
2877         Revert:
2878         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2879         subsection.
2880         (Basic PowerPC Built-in Functions): The new name of the
2881         subsection previously known as "PowerPC Built-in Functions".
2882         (Basic PowerPC Built-in Functions Available on all Configurations):
2883         New subsubsection.
2884         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2885         subsubsection.
2886         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2887         subsubsection.
2888         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2889         subsubsection.
2890         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2891         subsubsection.
2892
2893 2018-05-08  Jim Wilson  <jimw@sifive.com>
2894
2895         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2896         (LD_EMUL_SUFFIX): New.
2897         (LINK_SPEC): Use it.
2898
2899 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2900
2901         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2902         subsection.
2903         (Basic PowerPC Built-in Functions): The new name of the
2904         subsection previously known as "PowerPC Built-in Functions".
2905         (Basic PowerPC Built-in Functions Available on all Configurations):
2906         New subsubsection.
2907         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2908         subsubsection.
2909         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2910         subsubsection.
2911         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2912         subsubsection.
2913         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2914         subsubsection.
2915
2916 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
2917
2918         PR target/85683
2919         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2920         after cmpelim optimization.
2921
2922 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
2923
2924         * config.gcc: Support "goldmont".
2925         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2926         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2927         PROCESSOR_GOLDMONT.
2928         * config/i386/i386.c (m_GOLDMONT): Define.
2929         (processor_target_table): Add "goldmont".
2930         (PTA_GOLDMONT): Define.
2931         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2932         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2933         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2934         (fold_builtin_cpu): Add "goldmont".
2935         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2936         (ix86_option_override_internal): Add "goldmont".
2937         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2938         (processor_type): Add PROCESSOR_GOLDMONT.
2939         * config/i386/i386.md: Add CPU "glm".
2940         * config/i386/glm.md: New file.
2941         * config/i386/x86-tune.def: Add m_GOLDMONT.
2942         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2943
2944 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
2945
2946         PR target/85572
2947         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2948         E_V4DImode.
2949         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2950         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
2951         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
2952
2953         PR target/85317
2954         * config/i386/i386.c (ix86_fold_builtin): Handle
2955         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2956
2957         PR target/85480
2958         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2959         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2960
2961 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
2962
2963         PR target/85658
2964         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2965         (check_arch): Likewise.
2966         (check_fpu): Return the result rather than printing it.
2967         (end arch): Fix operator precedence.
2968         (end cpu): Likewise.
2969         (END): Print the result from check_fpu.
2970
2971 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
2972             Alan Hayward  <alan.hayward@arm.com>
2973             David Sherwood  <david.sherwood@arm.com>
2974
2975         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2976         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2977         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2978         (*fcmuo<mode>_and): New patterns.
2979
2980 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
2981
2982         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2983         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2984         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2985         (cmp_op, sve_imm_con): New code attributes.
2986         (SVE_COND_INT_CMP, imm_con): Delete.
2987         (cmp_op): Remove above unspecs from int attribute.
2988         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2989         to...
2990         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
2991         comparison-specific unspecs.
2992         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2993         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2994         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2995         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2996         (*vec_fcm<cmp_op><mode>): Rename to...
2997         (*fcm<cmp_op><mode>): ...this and adjust likewise.
2998         (*vec_fcmuo<mode>): Rename to...
2999         (*fcmuo<mode>): ...this and adjust likewise.
3000         (*pred_fcm<cmp_op><mode>): New pattern.
3001         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3002         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3003         functions.
3004         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3005         and UNORDERED.
3006         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3007         (aarch64_emit_sve_predicated_cond): New function.
3008         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3009         (aarch64_emit_unspec_cond_or): Replace with...
3010         (aarch64_emit_sve_or_conds): ...this new function.  Use
3011         aarch64_emit_sve_ptrue_op for the individual comparisons and
3012         aarch64_emit_binop to OR them together.
3013         (aarch64_emit_inverted_unspec_cond): Replace with...
3014         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
3015         aarch64_emit_sve_ptrue_op for the comparison and
3016         aarch64_emit_unop to invert the result.
3017         (aarch64_expand_sve_vec_cmp_float): Update after the above
3018         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
3019
3020 2018-05-07  Nathan Sidwell  <nathan@acm.org>
3021
3022         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3023         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3024         (Backwards Compatibility): Likewise.
3025
3026 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3027
3028         PR bootstrap/85681
3029         Revert:
3030         2018-05-07  Luis Machado  <luis.machado@linaro.org>
3031
3032         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3033         <prefetch_dynamic_strides>: New const bool field.
3034         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3035         prefetch_dynamic_strides.
3036         (exynosm1_prefetch_tune): Likewise.
3037         (thunderxt88_prefetch_tune): Likewise.
3038         (thunderx_prefetch_tune): Likewise.
3039         (thunderx2t99_prefetch_tune): Likewise.
3040         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3041         (aarch64_override_options_internal): Update to set
3042         PARAM_PREFETCH_DYNAMIC_STRIDES.
3043         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3044         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3045         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3046         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3047         prefetch-dynamic-strides setting.
3048
3049         2018-05-07  Luis Machado  <luis.machado@linaro.org>
3050
3051         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3052         <minimum_stride>: New const int field.
3053         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3054         minimum_stride field.
3055         (exynosm1_prefetch_tune): Likewise.
3056         (thunderxt88_prefetch_tune): Likewise.
3057         (thunderx_prefetch_tune): Likewise.
3058         (thunderx2t99_prefetch_tune): Likewise.
3059         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3060         (aarch64_override_options_internal): Update to set
3061         PARAM_PREFETCH_MINIMUM_STRIDE.
3062         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3063         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3064         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3065         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3066         stride is constant and is below the minimum stride threshold.
3067
3068 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3069
3070         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3071         to 512.
3072
3073 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3074
3075         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3076         <prefetch_dynamic_strides>: New const bool field.
3077         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3078         prefetch_dynamic_strides.
3079         (exynosm1_prefetch_tune): Likewise.
3080         (thunderxt88_prefetch_tune): Likewise.
3081         (thunderx_prefetch_tune): Likewise.
3082         (thunderx2t99_prefetch_tune): Likewise.
3083         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3084         (aarch64_override_options_internal): Update to set
3085         PARAM_PREFETCH_DYNAMIC_STRIDES.
3086         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3087         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3088         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3089         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3090         prefetch-dynamic-strides setting.
3091
3092 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3093
3094         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3095         <minimum_stride>: New const int field.
3096         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3097         minimum_stride field.
3098         (exynosm1_prefetch_tune): Likewise.
3099         (thunderxt88_prefetch_tune): Likewise.
3100         (thunderx_prefetch_tune): Likewise.
3101         (thunderx2t99_prefetch_tune): Likewise.
3102         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3103         (aarch64_override_options_internal): Update to set
3104         PARAM_PREFETCH_MINIMUM_STRIDE.
3105         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3106         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3107         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3108         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3109         stride is constant and is below the minimum stride threshold.
3110
3111 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
3112
3113         PR c++/85659
3114         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3115         the type is addressable.  Don't force op into register if it has
3116         BLKmode.
3117
3118 2018-05-05  Roland McGrath  <mcgrathr@google.com>
3119
3120         PR other/77609
3121         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3122         any section for which we don't know a specific type it should have,
3123         regardless of name.  Previously this was done only for the exact
3124         names ".init_array", ".fini_array", and ".preinit_array".
3125         (default_elf_asm_named_section): Add comment about
3126         relationship with default_section_type_flags and SECTION_NOTYPE.
3127         (get_section): Don't consider it a type conflict if one side has
3128         SECTION_NOTYPE and the other doesn't, as long as neither has the
3129         SECTION_BSS et al used in the default_section_type_flags logic.
3130
3131 2018-05-05  Tom de Vries  <tom@codesourcery.com>
3132
3133         PR target/85653
3134         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3135         (workaround_barsyncs): New function.
3136         (nvptx_reorg): Use workaround_barsyncs.
3137         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3138         (define_expand "nvptx_membar_cta"): New define_expand.
3139         (define_insn "*nvptx_membar_cta"): New insn.
3140
3141 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
3142
3143         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3144         To improve optimization opportunities.
3145         * builtin-types.def: The new needed builtin types for the above.
3146
3147 2018-05-04  Richard Biener  <rguenther@suse.de>
3148
3149         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3150         * gimple-ssa-store-merging.c
3151         (imm_store_chain_info::output_merged_store): Remove redundant create,
3152         release split_store vector contents on failure.
3153         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3154         scalar stmt vector on cache hit.
3155
3156 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
3157
3158         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3159         Xilinx FP support.
3160         * config.gcc (powerpc-xilinx-eabi*): Remove.
3161         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3162         support.
3163         (fusion_addis_mem_combo_load): Ditto.
3164         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3165         FP support.
3166         (rs6000_cpu_cpp_builtins): Ditto.
3167         * config/rs6000/rs6000-linux.c
3168         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3169         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3170         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3171         support.
3172         (rs6000_setup_reg_addr_masks): Ditto.
3173         (rs6000_init_hard_regno_mode_ok): Ditto.
3174         (rs6000_option_override_internal): Ditto.
3175         (legitimate_lo_sum_address_p): Ditto.
3176         (rs6000_legitimize_address): Ditto.
3177         (rs6000_legitimize_reload_address): Ditto.
3178         (rs6000_legitimate_address_p): Ditto.
3179         (abi_v4_pass_in_fpr): Ditto.
3180         (setup_incoming_varargs): Ditto.
3181         (rs6000_gimplify_va_arg): Ditto.
3182         (rs6000_split_multireg_move): Ditto.
3183         (rs6000_savres_strategy): Ditto.
3184         (rs6000_emit_prologue_components): Ditto.
3185         (rs6000_emit_epilogue_components): Ditto.
3186         (rs6000_emit_prologue): Ditto.
3187         (rs6000_emit_epilogue): Ditto.
3188         (rs6000_elf_file_end): Ditto.
3189         (rs6000_function_value): Ditto.
3190         (rs6000_libcall_value): Ditto.
3191         * config/rs6000/rs6000.h: Ditto.
3192         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3193         (TARGET_MINMAX): ... this.  New.
3194         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3195         * config/rs6000/rs6000.md: Remove Xilinx FP support.
3196         (*movsi_internal1_single): Delete.
3197         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3198         mfpu=, mxilinx-fpu): Delete.
3199         * config/rs6000/singlefp.h: Delete.
3200         * config/rs6000/sysv4.h: Remove Xilinx FP support.
3201         * config/rs6000/t-rs6000: Ditto.
3202         * config/rs6000/t-xilinx: Delete.
3203         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3204         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3205         (VStype_simple): Delete.
3206         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3207         * config/rs6000/xfpu.h: Delete.
3208         * config/rs6000/xfpu.md: Delete.
3209         * config/rs6000/xilinx.h: Delete.
3210         * config/rs6000/xilinx.opt: Delete.
3211         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3212         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3213
3214 2018-05-04  Tom de Vries  <tom@codesourcery.com>
3215
3216         PR libgomp/85639
3217         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3218         if ignore == 0.
3219
3220 2018-05-04  Richard Biener  <rguenther@suse.de>
3221
3222         PR middle-end/85627
3223         * tree-complex.c (update_complex_assignment): We are always in SSA form.
3224         (expand_complex_div_wide): Likewise.
3225         (expand_complex_operations_1): Likewise.
3226         (expand_complex_libcall): Preserve EH info of the original stmt.
3227         (tree_lower_complex): Handle removed blocks.
3228         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3229         on complex multiplication and division libcall builtins.
3230
3231 2018-05-04  Richard Biener  <rguenther@suse.de>
3232
3233         PR middle-end/85574
3234         * fold-const.c (negate_expr_p): Restrict negation of operand
3235         zero of a division to when we know that can happen without
3236         overflow.
3237         (fold_negate_expr_1): Likewise.
3238
3239 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
3240
3241         PR libstdc++/85466
3242         * real.h (real_nextafter): Declare.
3243         * real.c (real_nextafter): New function.
3244         * fold-const-call.c (fold_const_nextafter): New function.
3245         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3246         CASE_CFN_NEXTTOWARD.
3247         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3248         even when arg1_mode is different from arg0_mode.
3249
3250 2018-05-03  Nathan Sidwell  <nathan@acm.org>
3251
3252         * doc/extend.texi (Deprecated Features): Remove
3253         -ffriend-injection.
3254         (Backwards Compatibility): Likewise.
3255         * doc/invoke.texi (C++ Language Options): Likewise.
3256         (C++ Dialect Options): Likewise.
3257
3258 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
3259
3260         PR target/85530
3261         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3262         _mm512_mask_mullox_epi64): New intrinsics.
3263
3264 2018-05-03  Tom de Vries  <tom@codesourcery.com>
3265
3266         PR testsuite/85106
3267         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3268         dump files): Add offload-tree.
3269
3270 2018-05-03  Richard Biener  <rguenther@suse.de>
3271
3272         PR tree-optimization/85615
3273         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3274         to loops not nested in BBs loop father to avoid creating multi-entry
3275         loops.
3276
3277 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3278
3279         PR tree-optimization/70291
3280         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3281         arguments.  Change return type to tree.  Emit libcall as a new
3282         statement rather than replacing existing one when inplace_p is true.
3283         (expand_complex_multiplication_components): New function.
3284         (expand_complex_multiplication): Expand floating-point complex
3285         multiplication using the above.
3286         (expand_complex_division): Rename inner_type parameter to type.
3287         Update expand_complex_libcall call-site.
3288         (expand_complex_operations_1): Update expand_complex_multiplication
3289         and expand_complex_division call-sites.
3290
3291 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
3292
3293         PR target/85582
3294         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3295         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3296         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3297         the highest significant bit of the shift count mask is clear.  In
3298         check whether and[sq]i3 is needed verify that all significant bits
3299         of the shift count other than the highest are set.
3300
3301 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3302
3303         PR libgomp/82428
3304         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3305         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3306         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3307         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3308         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3309         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3310         __builtin_goacc_parlevel_size.
3311
3312 2018-05-02  Richard Biener  <rguenther@suse.de>
3313
3314         PR tree-optimization/85597
3315         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3316         do not use split vect_get_vec_defs call but call vect_get_slp_defs
3317         directly.
3318
3319 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3320
3321         PR testsuite/85106
3322         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3323         dump files): Add ltrans-tree.
3324
3325 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3326
3327         PR testsuite/85106
3328         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3329         dump files): Add wpa-ipa.
3330
3331 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
3332
3333         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
3334         powerpc*-*-linux*paired* target.
3335         * config/rs6000/750cl.h: Delete.
3336         * config/rs6000/paired.h: Delete.
3337         * config/rs6000/paired.md: Delete.
3338         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3339         float support.
3340         * config/rs6000/rs6000-builtin.def: Remove paired float support.
3341         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3342         comment.  Remove paired float support.
3343         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3344         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3345         VECTOR_PAIRED.
3346         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3347         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3348         declarations.
3349         * config/rs6000/rs6000.c: Remove paired float support.
3350         (paired_expand_vector_init, paired_expand_vector_move,
3351         paired_emit_vector_compare, paired_emit_vector_cond_expr,
3352         (paired_expand_lv_builtin, paired_expand_stv_builtin,
3353         paired_expand_builtin, paired_expand_predicate_builtin,
3354         paired_init_builtins): Delete.
3355         * config/rs6000/rs6000.h: Remove paired float support.
3356         * config/rs6000/rs6000.md: Remove paired float support.
3357         (move_from_CR_ov_bit): Delete.
3358         * config/rs6000/rs6000.opt (mpaired): Delete.
3359         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3360         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3361
3362 2018-05-02  Richard Biener  <rguenther@suse.de>
3363
3364         PR middle-end/85567
3365         * gimplify.c (gimplify_save_expr): When in SSA form allow
3366         SAVE_EXPRs to compute to SSA vars.
3367
3368 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
3369
3370         PR target/85582
3371         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3372         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3373         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3374         clobber operands[2], instead use a new pseudo.  Formatting fixes.
3375
3376 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
3377
3378         PR tree-optimization/85586
3379         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3380         exit early for statements in the same group if the accesses are
3381         not strided.
3382
3383 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3384
3385         PR lto/85451
3386         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3387         error message.
3388
3389 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
3390
3391         PR tree-optimization/85143
3392         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3393
3394 2018-05-01  Tom de Vries  <tom@codesourcery.com>
3395
3396         PR lto/85451
3397         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3398         not found" error message.
3399
3400 2018-05-01  Tom de Vries  <tom@codesourcery.com>
3401
3402         PR other/83786
3403         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3404         * vec.c (test_ordered_remove_if): New function.
3405         (vec_c_tests): Call test_ordered_remove_if.
3406         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3407         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3408         * tree-vect-patterns.c (vect_pattern_recog_1): Use
3409         VEC_ORDERED_REMOVE_IF.
3410
3411 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3412
3413         PR tree-optimization/82665
3414         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3415         pointer subtraction where arguments come from a memchr call.
3416
3417 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
3418
3419         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3420         --push-state --as-needed and --pop-state instead of --as-needed and
3421         --no-as-needed if ld supports it.
3422         * configure: Regenerated.
3423
3424         PR web/85578
3425         * doc/install.texi2html: Replace _002d with - and _002a with * in
3426         generated html files using sed.
3427
3428 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3429
3430         PR c++/85523
3431         * gcc-rich-location.c (blank_line_before_p): New function.
3432         (use_new_line): New function.
3433         (gcc_rich_location::add_fixit_insert_formatted): New function.
3434         * gcc-rich-location.h
3435         (gcc_rich_location::add_fixit_insert_formatted): New function.
3436
3437 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3438
3439         * selftest.c (assert_streq): Rename "expected" and "actual" to
3440         "val1" and "val2".  Extend NULL-handling to cover both inputs
3441         symmetrically, while still requiring both to be non-NULL for a pass.
3442         * selftest.h (assert_streq): Rename "expected" and "actual" to
3443         "val1" and "val2".
3444         (ASSERT_EQ): Likewise.
3445         (ASSERT_EQ_AT): Likewise.
3446         (ASSERT_KNOWN_EQ): Likewise.
3447         (ASSERT_KNOWN_EQ_AT): Likewise.
3448         (ASSERT_NE): Likewise.
3449         (ASSERT_MAYBE_NE): Likewise.
3450         (ASSERT_MAYBE_NE_AT): Likewise.
3451         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
3452         the assertion to pass.
3453         (ASSERT_STREQ_AT): Likewise.
3454
3455 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
3456
3457         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3458         interaction with -pie.
3459
3460 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3461
3462         * selftest.h: Fix alphabetization of per-source-file selftest
3463         declarations.
3464
3465 2018-04-30  Jason Merrill  <jason@redhat.com>
3466
3467         PR c++/61982 - dead stores to destroyed objects.
3468         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3469         of clobber.
3470
3471 2018-04-30  Jason Merrill  <jason@redhat.com>
3472
3473         * tree.c (build_clobber): New.
3474         * tree.h: Declare it.
3475         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3476
3477 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3478
3479         * diagnostic-show-locus.c (layout::layout): Update for
3480         location_get_source_line returning a char_span.
3481         (struct char_span): Move to input.h.
3482         (struct correction): Update for fields in char_span becoming
3483         private.
3484         (struct source_line): Update for location_get_source_line
3485         returning a char_span.
3486         (layout::print_line): Likewise.
3487         * edit-context.c (edited_file::print_content): Likewise.
3488         (edited_file::print_diff_hunk): Likewise.
3489         (edited_file::print_run_of_changed_lines): Likewise.
3490         (edited_file::get_num_lines): Likewise.
3491         (edited_line::edited_line): Likewise.
3492         * final.c (asm_show_source): Likewise.
3493         * input.c (location_get_source_line): Convert return type
3494         from const char * to char_span, losing the final "line_len"
3495         param.
3496         (dump_location_info): Update for the above.
3497         (get_substring_ranges_for_loc): Likewise.  Use a char_span
3498         when handling the literal within the line.
3499         (test_reading_source_line): Update for location_get_source_line
3500         returning a char_span.
3501         * input.h (class char_span): Move here from
3502         diagnostic-show-locus.c, converting from a struct to a class.
3503         Make data members private.
3504         (char_span::operator bool): New.
3505         (char_span::length): New.
3506         (char_span::get_buffer): New.
3507         (char_span::operator[]): New.
3508         (char_span::subspan): Make const.
3509         (char_span::xstrdup): New.
3510         (location_get_source_line): Convert return type from const char *
3511         to char_span, losing the final "line_size" param.
3512
3513 2018-04-30  Jan Hubicka  <jh@suse.cz>
3514
3515         * lto-wrapper.c (ltrans_priorities): New static var.
3516         (cmp_priority): New.
3517         (run_gcc): Read priorities and if doing parallel build order
3518         the Makefile by them.
3519
3520 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3521
3522         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3523
3524 2018-04-30  Richard Biener  <rguenther@suse.de>
3525
3526         * tree-cfg.c (verify_address): Remove base argument, add
3527         flag whether to check TREE_ADDRESSABLE and do that.
3528         (verify_expr): Remove.
3529         (verify_types_in_gimple_reference): Add pieces from verify_expr.
3530         (verify_gimple_assign_single): Likewise.
3531         (verify_gimple_switch): Likewise.
3532         (verify_expr_location_1): Dereference tp once.  Add (disabled)
3533         piece from verify_expr.
3534         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3535
3536 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3537
3538         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3539
3540 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3541
3542         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3543         (small_data_pattern): Likewise.
3544         (arc_rewrite_small_data): Likewise.
3545         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3546         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3547         (get_symbol_alignment): New function.
3548         (legitimate_small_data_address_p): Likewise.
3549         (legitimate_scaled_address): Update, call
3550         legitimate_small_data_address_p.
3551         (output_sdata): New static variable.
3552         (arc_print_operand): Update how we handle small data operands.
3553         (arc_print_operand_address): Likewise.
3554         (arc_legitimate_address_p): Update, use
3555         legitimate_small_data_address_p.
3556         (arc_rewrite_small_data_p): Remove.
3557         (arc_rewrite_small_data_1): Likewise.
3558         (arc_rewrite_small_data): Likewise.
3559         (small_data_pattern): Likewise.
3560         (compact_sda_memory_operand): Update to use
3561         legitimate_small_data_address_p and get_symbol_alignment.
3562         (prepare_move_operands): Don't rewite sdata pattern.
3563         (prepare_extend_operands): Remove.
3564         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3565         pattern.
3566         (zero_extendqisi2): Likewise.
3567         (zero_extendhisi2): Likewise.
3568         (extendqihi2): Likewise.
3569         (extendqisi2): Likewise.
3570         (extendhisi2): Likewise.
3571         (addsi3): Likewise.
3572         (subsi3): Likewise.
3573         (andsi3): Likewise.
3574         * config/arc/constraints.md (Usd): Change it to memory constraint.
3575
3576 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3577
3578         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3579         as source of std instructions.
3580         * config/arc/arc.md (movsi_insn): Update pattern predicate to
3581         allow 6-bit constants as source for store instructions.
3582         (movdi_insn): Update instruction pattern to allow 6-bit constants
3583         as source for store instructions.
3584
3585 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
3586
3587         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3588
3589 2018-04-30  Nathan Sidwell  <nathan@acm.org>
3590             Sandra Loosemore <sandra@codesourcery.com>
3591
3592         * dumpfile.c (dump_open): Allow '-' for stdout.
3593         * doc/invoke.texi (Developer Options): Document dump filename
3594         determination early.  Document stdin/stdout selection.
3595
3596 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
3597
3598         Microblaze Target: PIC data text relative
3599
3600         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3601         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3602         Add declaration.
3603         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3604         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3605         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3606         New addressing mode for data-text relative position indepenedent code.
3607         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3608         'ADDRESS_SYMBOLIC_TXT_REL'.
3609         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3610         (microblaze_legitimate_pic_operand): Exclude function calls from
3611         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3612         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3613         addresses cases.
3614         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3615         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3616         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3617         for 'address + offset'.
3618         (microblaze_expand_prologue): Add new function prologue call for
3619         'r20' assignation.
3620         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3621         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3622         table in case of TARGET_PIC_DATA_TEXT_REL.
3623         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3624         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3625         Add new macros 'UNSPEC_TEXT',
3626         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3627         + exclude function calls from 'UNSPEC_PLT' in case of data text
3628         relative mode.
3629         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3630         new target hook for generating address diff vector tables in case of
3631         flag_pic.
3632         * doc/tm.texi : Regenerate.
3633         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3634         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3635         of addr diff vector generation.
3636         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3637         target hook definition.
3638         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3639         Add default function for generate_pic_addr_diff_vec -> flag_pic.
3640         * doc/invoke.texi (Add new pic option): Add new microblaze pic
3641         option for data text relative.
3642
3643 2018-04-30  Richard Biener  <rguenther@suse.de>
3644
3645         * tree-chrec.h (evolution_function_is_constant_p): Remove
3646         redundant check.
3647         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3648
3649 2018-04-30  Richard Biener  <rguenther@suse.de>
3650
3651         PR bootstrap/85571
3652         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3653
3654 2018-04-30  Richard Biener  <rguenther@suse.de>
3655
3656         PR tree-optimization/28364
3657         PR tree-optimization/85275
3658         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3659         copying first exit test.
3660
3661 2018-04-28  Mark Wielaard  <mark@klomp.org>
3662
3663         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3664         dwarf_version >= 5.
3665         (dwarf_AT): Handle DW_AT_addr_base.
3666         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3667
3668 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
3669
3670         PR target/84431
3671         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3672         (*ashl<dwi>3_doubleword_mask_1): Ditto.
3673         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3674         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3675
3676 2018-04-28  Richard Biener  <rguenther@suse.de>
3677
3678         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3679         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3680         to reflect use.  Only add interesting stmts.
3681
3682 2018-04-27  Martin Jambor  <mjambor@suse.cz>
3683
3684         PR ipa/85549
3685         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3686         the jump function allows for passing through aggregate values.
3687
3688 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
3689
3690         * input.h (in_system_header_at): Convert from macro to inline
3691         function.
3692         (from_macro_expansion_at): Likewise.
3693         (from_macro_definition_at): Likewise.
3694
3695 2018-04-27  Jeff Law  <law@redhat.com>
3696
3697         * config.gcc: Mark tile* targets as deprecated/obsolete.
3698
3699 2018-04-27  Richard Biener  <rguenther@suse.de>
3700
3701         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3702         fix for ILP32.
3703
3704 2018-04-27  Richard Biener  <rguenther@suse.de>
3705
3706         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3707
3708 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
3709
3710         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3711         with Yd constraint. Set "preferred_for_speed" attribute from
3712         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3713         with Yd constraint.
3714         (*movdi_internal): Ditto.
3715         (movti_interunit splitters): Remove
3716         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3717         (movdi_interunit splitters): Ditto.
3718         * config/i386/constraints.md (Ye): Remove.
3719         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3720
3721 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3722
3723         PR target/85512
3724         * config/aarch64/constraints.md (Usg): Limit to 31.
3725         (Usj): Limit to 63.
3726
3727 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
3728
3729         PR tree-optimization/85529
3730         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3731         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3732         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3733         zero extension or masking of the MSB bit.
3734         (optimize_range_tests): Add FIRST_BB argument, pass it through
3735         to optimize_range_tests_var_bound.
3736         (maybe_optimize_range_tests, reassociate_bb): Adjust
3737         optimize_range_tests callers.
3738
3739 2018-04-26  Richard Biener  <rguenther@suse.de>
3740             Jakub Jelinek  <jakub@redhat.com>
3741
3742         * cgraph.h (symbol_table): Just declare debug method here.
3743         * symtab.c (symbol_table::debug): Define.
3744
3745 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
3746
3747         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3748
3749 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
3750
3751         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3752         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3753         (*movdi_internal): Substitute Yi and Yj constraint with x
3754         and Ym and Yn constraint with y constraint.  Update "isa"
3755         attribute and set "preferred_for_speed" attribute from
3756         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3757         (*movsi_internal): Ditto.
3758         (*movdf_internal): Ditto.
3759         (*movsf_internal): Ditto.
3760         (*zero_extendsidi2): Ditto.
3761         * config/i386/sse.md (vec_set<mode>_0): Ditto.
3762         (sse2_loadld): Ditto.
3763         (*vec_extract<ssevecmodelower>_0): Ditto.
3764         (*vec_extractv4si_0_zext_sse4): Ditto.
3765         (vec_concatv2di): Ditto.
3766         (*vec_dup<mode>): Ditto.
3767         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3768         * config/i386/constraints.md (Yi): Remove.
3769         (Yj): Remove.
3770         (Ym): Remove.
3771         (Yn): Remove.
3772
3773 2018-04-26  Nathan Sidwell  <nathan@acm.org>
3774
3775         * dumpfile.c (dump_open): New.
3776         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3777         (dump_finish): Detect stdio/stderr by value not name.
3778
3779 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
3780
3781         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3782
3783 2018-04-26  Tom de Vries  <tom@codesourcery.com>
3784
3785         PR target/84952
3786         * config/nvptx/nvptx.c (verify_neutering_jumps)
3787         (verify_neutering_labels): New function
3788         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3789
3790 2018-04-26  Tom de Vries  <tom@codesourcery.com>
3791
3792         PR target/84025
3793         * config/nvptx/nvptx.c (needs_neutering_p): New function.
3794         (nvptx_single): Use needs_neutering_p to skip over insns that do not
3795         need neutering.
3796
3797 2018-04-26  Richard Biener <rguenther@suse.de>
3798             Tom de Vries  <tom@codesourcery.com>
3799
3800         PR lto/85422
3801         * lto-streamer-out.c (output_function): Fixup loops if required to match
3802         discovery done in the reader.
3803
3804 2018-04-26  Richard Biener  <rguenther@suse.de>
3805
3806         PR tree-optimization/85116
3807         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3808         have a loop exit from the single latch predecessor.  Remove
3809         case of header with just condition.
3810         (ch_base::copy_headers): Exclude infinite loops from any
3811         processing.
3812         (pass_ch::execute): Record exits.
3813
3814 2018-04-26  Richard Biener  <rguenther@suse.de>
3815
3816         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3817         prologue cost vector and pass it to vect_get_load_cost.
3818         (vect_get_peeling_costs_all_drs): Likewise.
3819         (vect_peeling_hash_get_lowest_cost): Likewise.
3820         (vect_enhance_data_refs_alignment): Likewise.
3821
3822 2018-04-26  Richard Biener  <rguenther@suse.de>
3823
3824         PR middle-end/85450
3825         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3826         checking of integer<->pointer conversions.
3827         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3828         sign-/zero-extending pointer types.
3829         (expand_omp_for_static_chunk): Likewise.
3830
3831 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
3832             Jean Lee  <xiaoyur347@gmail.com>
3833
3834         * config/mips/mips.c (mips_asan_shadow_offset): New function.
3835         (TARGET_ASAN_SHADOW_OFFSET): Define.
3836         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3837         true for -fsanitize=address.
3838
3839 2018-04-25  Mark Wielaard  <mark@klomp.org>
3840
3841         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3842         shorter ones.
3843
3844 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3845
3846         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3847         than "alu", remove explicit "memory" and "imm_disp" attributes.
3848         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3849
3850         PR middle-end/85414
3851         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3852         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3853         gen_lowpart_no_emit.
3854
3855 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
3856
3857         PR target/85473
3858         * config/i386/i386.c (ix86_expand_builtin): Change memory
3859         operand to XI, extend p0 to Pmode.
3860         * config/i386/i386.md: Change unspec volatile and operand
3861         1 mode to XI, change operand 0 mode to P.
3862
3863 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3864
3865         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3866         GET_MODE_MASK before any checking.
3867         (nds32_can_use_bset_p): Likewise.
3868         (nds32_can_use_btgl_p): Likewise.
3869
3870 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3871
3872         * config/nds32/nds32-doubleword.md: New define_split pattern for
3873         illegal register number.
3874
3875 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3876
3877         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3878
3879 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3880
3881         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3882
3883 2018-04-25  Richard Biener  <rguenther@suse.de>
3884
3885         * lto-streamer.h (LTO_major_version): Bump to 8.
3886
3887 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3888
3889         * BASE-VER: Set to 9.0.0.
3890
3891 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
3892
3893         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3894         in __abskf2 and __powikf2.
3895
3896 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3897
3898         PR target/85512
3899         * config/aarch64/constraints.md (Usg, Usj): New constraints.
3900         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3901         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3902         Use the above on operand 2.  Reindent.
3903         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3904
3905 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
3906
3907         PR target/85485
3908         * common/config/i386/i386-common.c (ix86_handle_option): Don't
3909         handle OPT_mcet.
3910         * config/i386/i386.opt (mcet): Removed.
3911         * doc/install.texi: Remove -mcet documentation.
3912         * doc/invoke.texi: Likewise.
3913
3914 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
3915
3916         PR target/85485
3917         * doc/install.texi: Remove -mcet from bootstrap-cet.
3918
3919 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3920
3921         PR target/85511
3922         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3923         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3924         if TARGET_64BIT.
3925
3926         PR target/85503
3927         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3928         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3929         containing a CONST_VECTOR.
3930
3931 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
3932
3933         * doc/install.texi: Update newlib dependency for nvptx.
3934
3935 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3936
3937         PR target/85508
3938         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3939         instead of INTVAL when shifting x left.
3940
3941 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
3942
3943         PR tree-optimization/85478
3944         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3945         vect_grouped_store_supported for single element vectors.
3946
3947 2018-04-24  Richard Biener  <rguenther@suse.de>
3948
3949         PR target/85491
3950         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3951         load cost increase to the case of non-constant step.
3952
3953 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3954
3955         PR target/84828
3956         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3957         destination if any_malformed_asm.
3958
3959 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
3960
3961         PR middle-end/85496
3962         * expr.c (store_field): In the bitfield case, if the value comes from
3963         a function call and is returned in registers by means of a PARALLEL,
3964         do not change the mode of the temporary unless BLKmode and VOIDmode.
3965
3966 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
3967
3968         PR rtl-optimization/85423
3969         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3970         dependencies to debug insns when the previous insn is non-debug.
3971
3972 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
3973
3974         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3975         enums into a single definition.
3976         (fls): Fix predicates and printing.
3977         (seti): Likewise.
3978
3979 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
3980
3981         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3982         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3983         and short u6 immediate.
3984         (check_if_valid_sleep_operand): Remove.
3985         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3986
3987 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3988
3989         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3990         flag_always_save_lp condition.
3991         * config/nds32/nds32.opt (malways-save-lp): New option.
3992
3993 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
3994
3995         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3996         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3997         * config/nds32/nds32.h
3998         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3999         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4000
4001 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
4002
4003         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4004         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4005
4006 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
4007             Chung-Ju Wu  <jasonwucj@gmail.com>
4008
4009         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4010         Declare.
4011         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4012         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4013
4014 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4015
4016         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4017
4018 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4019
4020         * config/nds32/nds32-protos.h (nds32_data_alignment,
4021         nds32_local_alignment): Declare.
4022         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4023         nds32_local_alignment): New functions.
4024         (TARGET_CONSTANT_ALIGNMENT): Define.
4025         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4026
4027 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4028
4029         * config/nds32/nds32.c
4030         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4031         (TARGET_MODES_TIEABLE_P): Likewise.
4032
4033 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4034
4035         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4036         level Ofast and Og.
4037
4038 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
4039             Chung-Ju Wu  <jasonwucj@gmail.com>
4040
4041         * config/nds32/constants.md (unspec_volatile_element): Add enum values
4042         for unaligned access.
4043         * config/nds32/nds32-intrinsic.c: Implementation of expanding
4044         unaligned access.
4045         * config/nds32/nds32-intrinsic.md: Likewise.
4046         * config/nds32/nds32_intrinsic.h: Likewise.
4047         * config/nds32/nds32.h (nds32_builtins): Likewise.
4048         * config/nds32/nds32.opt (munaligned-access): New option.
4049         * config/nds32/nds32.c (nds32_asm_file_start): Display
4050         flag_unaligned_access status.
4051
4052 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
4053
4054         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4055         -mno-relax is present.
4056         * config/riscv/linux.h (LINK_SPEC): Ditto.
4057
4058 2018-04-20  Martin Sebor  <msebor@redhat.com>
4059
4060         PR c/85365
4061         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4062         for null pointers.
4063         (gimple_fold_builtin_stxcpy_chk): Same.
4064         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4065
4066 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
4067
4068         PR target/85456
4069         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4070         __powikf2 when long double is IEEE 128-bit.
4071
4072 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
4073
4074         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4075         step to make sure stack always aligned.
4076
4077 2018-04-20  Carl Love  <cel@us.ibm.com>
4078
4079         PR target/83402
4080         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4081         size check for arg0.
4082
4083 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
4084             Tom de Vries  <tom@codesourcery.com>
4085
4086         PR target/85445
4087         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4088         Emit insns for calls too.
4089         (nvptx_find_par): Always look for worker-level predecessor insn.
4090         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
4091         calls.
4092         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4093         (nvptx_process_pars): Propagate frames for calls.
4094
4095 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
4096
4097         PR target/85469
4098         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4099         Removed.
4100         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4101         (ix86_handle_option): Don't handle OPT_mibt.
4102         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4103         __SHSTK__.
4104         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4105         has_ibt and ibt.
4106         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4107         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4108         (ix86_target_macros): Define __CET__ with flag_cf_protection
4109         for -fcf-protection.
4110         * config/i386/i386.c (isa2_opts): Remove -mibt.
4111         * config/i386/i386.h (TARGET_IBT): Removed.
4112         (TARGET_IBT_P): Likewise.
4113         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4114         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4115         * config/i386/i386.opt (mcet): Update help message.
4116         (mshstk): Likewise.
4117         (mibt): Removed.
4118         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
4119         -mcet as an alias for -mshstk.
4120
4121 2018-04-20  Richard Biener <rguenther@suse.de>
4122
4123         PR middle-end/85475
4124         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4125         complexity by forcing a single use of the multiply operand.
4126
4127 2018-04-20  Martin Jambor  <mjambor@suse.cz>
4128
4129         ipa/85449
4130         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4131         recursion dependency to only apply to non-clones.
4132
4133 2018-04-20  Martin Jambor  <mjambor@suse.cz>
4134
4135         ipa/85447
4136         * ipa-cp.c (create_specialized_node): Check that clones of
4137         self-recursive edges exist during IPA-CP.
4138
4139 2018-04-19  Toon Moene  <toon@moene.org>
4140
4141         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4142         by -O3.
4143
4144 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
4145
4146         PR tree-optimization/85467
4147         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4148         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
4149         VECTOR_CST element to type.
4150
4151 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4152
4153         PR target/85397
4154         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4155         * config/i386/i386.md (builtin_setjmp_setup): Removed.
4156         (builtin_longjmp): Likewise.
4157         (save_stack_nonlocal): New pattern.
4158         (restore_stack_nonlocal): Likewise.
4159
4160 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4161
4162         PR target/85404
4163         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4164         Replace ASM_OUTPUT_LABEL with fprintf.
4165
4166 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4167
4168         PR target/85417
4169         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4170         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4171         * config/i386/i386-c.c (ix86_target_macros_internal): Also
4172         define __IBT__ and __SHSTK__ for -fcf-protection.
4173         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4174         TARGET_IBT.
4175         (ix86_trampoline_init): Likewise.
4176         (x86_output_mi_thunk): Likewise.
4177         (ix86_notrack_prefixed_insn_p): Likewise.
4178         (ix86_option_override_internal): Don't disallow -fcf-protection.
4179         * config/i386/i386.md (rdssp<mode>): Also enable for
4180         -fcf-protection.
4181         (incssp<mode>): Likewise.
4182         (nop_endbr): Likewise.
4183         * config/i386/i386.opt (mcet): Change help message to built-in
4184         functions only.
4185         (mibt): Likewise.
4186         (mshstk): Likewise.
4187         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4188         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
4189         enable CET built-in functions.
4190
4191 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
4192
4193         * common/config/i386/i386-common.c
4194         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4195         OPTION_MASK_ISA_MOVDIRI_UNSET,
4196         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4197         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4198         * config.gcc (movdirintrin.h): New header.
4199         * config/i386/cpuid.h (bit_MOVDIRI,
4200         bit_MOVDIR64B): New bits.
4201         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4202         and -mmvodir64b.
4203         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4204         (VOID, PVOID, PCVOID)): New function types.
4205         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4206         __builtin_ia32_directstoreu_u64,
4207          __builtin_ia32_movdir64b): New builtins.
4208         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4209         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4210         and -mmovdiri.
4211         (ix86_valid_target_attribute_inner_p): Ditto.
4212         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4213         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4214         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4215         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4216         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4217         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4218         (movdiri<mode>, movdir64b_<mode>): New.
4219         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4220         * config/i386/immintrin.h: Include movdirintrin.h.
4221         * config/i386/movdirintrin.h: New file.
4222         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4223
4224 2018-04-19  Richard Biener  <rguenther@suse.de>
4225
4226         PR middle-end/85455
4227         * cfg.c (clear_bb_flags): When loop state says we have
4228         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4229
4230 2018-04-19  Richard Biener  <rguenther@suse.de>
4231
4232         PR tree-optimization/84737
4233         * tree-vect-data-refs.c (vect_copy_ref_info): New function
4234         copying restrict info.
4235         (vect_setup_realignment): Use it.
4236         * tree-vectorizer.h (vect_copy_ref_info): Declare.
4237         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4238         the first DR to all generated stores.
4239         (vectorizable_load): Likewise for loads.
4240
4241 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
4242
4243         PR tree-optimization/85446
4244         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4245         the integral and pointer types to have the same precision.
4246
4247         * doc/install.texi: Document --disable-cet being the default and
4248         --enable-cet=auto.
4249
4250 2018-04-18  Martin Liska  <mliska@suse.cz>
4251
4252         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4253         style.
4254
4255 2018-04-18  Martin Liska  <mliska@suse.cz>
4256
4257         Revert
4258         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
4259
4260         PR ipa/83983
4261         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4262         arguments if they are comparable.
4263
4264 2018-04-18  Martin Liska  <mliska@suse.cz>
4265
4266         Revert
4267         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
4268
4269         PR lto/84805
4270         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4271         incomplete types.
4272
4273 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
4274
4275         PR target/85388
4276         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4277         ENDBR after calling __morestack.
4278
4279 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
4280
4281         PR jit/85384
4282         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4283         by using gcc_base_ver to generate a gcc_driver_version, and use
4284         it when generating GCC_DRIVER_NAME.
4285         * configure: Regenerate.
4286
4287 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
4288
4289         PR target/81084
4290         * config.gcc: Obsolete powerpc*-*-*spe*.
4291
4292 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4293
4294         PR debug/84637
4295         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4296         (stabstr_D): Change type of unum from unsigned int to
4297         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
4298         type.
4299
4300 2018-04-17  Jim Wilson  <jimw@sifive.com>
4301
4302         PR 84856
4303         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4304         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4305         Set arg_pointer_offset after using pretend_args_size.
4306
4307 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4308
4309         PR rtl-optimization/85431
4310         * dse.c (record_store): Ignore zero width stores.
4311
4312         PR sanitizer/85230
4313         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
4314         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4315         __builtin_stack_restore rather than after it.
4316         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4317         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4318         argument instead of virtual_dynamic_stack_rtx.
4319
4320 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4321
4322         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4323         New prototype.
4324         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4325         Add note to error message to explain internal mapping of overloaded
4326         built-in function name to non-overloaded built-in function name.
4327         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4328         function.
4329
4330 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
4331
4332         PR target/85424
4333         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4334         where the inputs overlap with the output.
4335
4336 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4337
4338         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4339         (=v, v) alternative and explicit "memory" attribute.
4340         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
4341         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4342         attributes.
4343         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4344         "sselog1" type instead of "sselog".
4345         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4346         "sselog".  Remove explicit "memory" attribute.
4347         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4348         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4349         attributes.
4350         (vec_extract_hi_v32hi): Merge all alternatives into one, use
4351         "sselog1" type instead of "sselog".  Remove explicit "memory"
4352         attribute.
4353         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4354         use "sselog1" type instead of "sselog".  Remove explicit "memory"
4355         attribute.
4356         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4357         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4358         attributes.
4359         (vec_extract_hi_v64qi): Merge all alternatives into one, use
4360         "sselog1" type instead of "sselog".  Remove explicit "memory"
4361         attribute.
4362         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4363         use "sselog1" type instead of "sselog".  Remove explicit "memory"
4364         attribute.
4365
4366         PR target/85430
4367         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4368
4369         PR middle-end/85414
4370         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4371         on a SUBREG.
4372
4373 2018-04-17  Martin Jambor  <mjambor@suse.cz>
4374
4375         PR ipa/85421
4376         * ipa-cp.c (create_specialized_node): Call
4377         expand_all_artificial_thunks if necessary.
4378
4379 2018-04-17  Martin Liska  <mliska@suse.cz>
4380
4381         PR lto/85405
4382         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4383         in message, remote space in between '_G' and '('.
4384
4385 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4386
4387         PR target/85281
4388         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4389         avx512f_vmcmp<mode>3<round_saeonly_name>,
4390         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4391         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4392         avx512f_rndscale<mode><round_saeonly_name>,
4393         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4394         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4395         Use %<iptr>2 instead of %2 for -masm=intel.
4396         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4397         avx512f_vcvttss2usi<round_saeonly_name>,
4398         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4399         -masm=intel.
4400         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4401         avx512f_vcvttsd2usi<round_saeonly_name>,
4402         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4403         Use %q1 instead of %1 for -masm=intel.
4404         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4405         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4406         of %3 for -masm=intel.
4407         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4408         -masm=intel.
4409         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4410         -masm=intel.
4411         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4412         -masm=intel.
4413         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4414         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4415         %g1.
4416         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4417         -masm=intel.
4418         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4419         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4420         %g1 and one with %0 and %1.
4421         (avx512er_vmrcp28<mode><round_saeonly_name>,
4422         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4423         %1 for -masm=intel.
4424         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4425         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4426         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4427         of %0 and %{%4%} for -masm=intel.
4428         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4429         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4430         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4431         order of %0 and %{%5%}%{z%} for -masm=intel.
4432
4433 2018-04-17  Jan Hubicka  <jh@suse.cz>
4434
4435         PR lto/85405
4436         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4437
4438 2018-04-17  Martin Liska  <mliska@suse.cz>
4439
4440         PR ipa/85329
4441         * multiple_target.c (create_dispatcher_calls): Set apostrophes
4442         for target_clone error message.  Make default implementation
4443         clone to be a local declaration.
4444         (separate_attrs): Add new argument and check for an empty
4445         string.
4446         (expand_target_clones): Handle it.
4447         (ipa_target_clone): Make redirection just for target_clones
4448         functions.
4449
4450 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
4451             Tom de Vries  <tom@codesourcery.com>
4452
4453         PR middle-end/84955
4454         * omp-expand.c (expand_oacc_for): Add dummy false branch for
4455         tiled basic blocks without omp continue statements.
4456
4457 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
4458
4459         PR target/83660
4460         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4461         vec_extract expression as having side effects to make sure it gets
4462         a cleanup point.
4463
4464 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
4465
4466         PR target/85403
4467         * config/i386/i386.c (get_builtin_code_for_version): Check
4468         error_mark_node.
4469
4470 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
4471
4472         PR target/84331
4473         * gcc/config.gcc: Support "skylake".
4474         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4475         PROCESSOR_SKYLAKE.
4476         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4477         (processor_target_table): Add "skylake".
4478         (ix86_option_override_internal): Add "skylake".
4479         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4480         PROCESSOR_CANNONLAKE.
4481         (get_builtin_code_for_version): Fix priority for
4482         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4483         PROCESSOR_SKYLAKE-AVX512.
4484         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4485         (processor_type): Add PROCESSOR_SKYLAKE.
4486
4487 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
4488             Jason Merrill  <jason@redhat.com>
4489
4490         PR c++/85112
4491         * convert.c (convert_to_integer_1): Use direct recursion for
4492         enumeral types and types with a precision less than the number
4493         of bits in their mode.
4494
4495 2018-04-16  Julia Koval  <julia.koval@intel.com>
4496
4497         PR target/84413
4498         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4499         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4500
4501 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
4502
4503         PR target/85293
4504         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4505         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4506         and -mno-direct-move.
4507
4508 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
4509
4510         PR target/83402
4511         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4512         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4513         Ensure negative shifts result in {0}.
4514
4515 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
4516
4517         PR rtl-optimization/79916
4518         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4519         regs (if any) to define how to gnerate SD moves when LRA is in
4520         progress.
4521
4522 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
4523
4524         PR rtl-optimization/85393
4525         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4526         * except.c (expand_dw2_landing_pad_for_region): Make static.
4527         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4528         a label and unconditional jump to old_bb, rather than
4529         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4530         basic block.
4531
4532         PR rtl-optimization/85376
4533         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4534         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4535         instead of a specific value.
4536
4537 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
4538             Bin Cheng  <bin.cheng@arm.com>
4539
4540         PR tree-optimization/82965
4541         PR tree-optimization/83991
4542         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4543         by_profile_only parameter.
4544         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4545         information if the loop was predicted to iterate too many times.
4546         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4547
4548 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
4549
4550         PR lto/71991
4551         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4552         always inline.
4553
4554 2018-04-13  Martin Liska  <mliska@suse.cz>
4555             Jakub Jelinek  <jakub@redhat.com>
4556
4557         PR middle-end/81657
4558         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4559         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4560         * builtins.c (expand_builtin_memory_copy_args): Use
4561         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4562         handle dest_addr == pc_rtx.
4563
4564 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
4565
4566         PR target/85291
4567         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4568         asked to not generate direct moves.
4569         (fix_trunc<mode>si2_stfiwx): Similar.
4570         (fix_trunc<mode>si2_internal): Similar.
4571
4572 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4573
4574         PR debug/83157
4575         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4576         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4577         lookup if dest in some wider mode is known to be const0_rtx and
4578         if so, record permanent equivalence for it to be ZERO_EXTEND of
4579         the narrower mode destination.
4580
4581 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
4582
4583         * lto-streamer-out.c (output_function): Revert 259346.
4584         * omp-expand.c (expand_oacc_for): Likewise.
4585
4586 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
4587
4588         PR rtl-optimization/85354
4589         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4590         * sel-sched.c (sel_global_init): ... here.
4591
4592 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
4593
4594         PR target/85238
4595         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4596         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4597         mode for PE-COFF targets.
4598         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4599         (i386_pe_asm_lto_end): Likewise.
4600         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4601         (TARGET_ASM_LTO_END): Likewise.
4602         * config/i386/winnt.c (saved_debug_info_level): New static variable.
4603         (i386_pe_asm_lto_start): New function.
4604         (i386_pe_asm_lto_end): Likewise.
4605
4606 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
4607             Richard Biener  <rguenther@suse.de>
4608
4609         PR middle-end/84955
4610         * lto-streamer-out.c (output_function): Fix CFG loop state before
4611         streaming out.
4612         * omp-expand.c (expand_oacc_for): Handle calls to internal
4613         functions like regular functions.
4614
4615 2018-04-12  Richard Biener  <rguenther@suse.de>
4616
4617         PR lto/85371
4618         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4619         for the early LTO debug to properly generate references to it
4620         during DIE emission.  Do not re-use that for the skeleton for
4621         split-dwarf.
4622         (dwarf2out_early_finish): Likewise.
4623
4624 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4625
4626         PR target/85328
4627         * config/i386/sse.md
4628         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4629         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4630         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4631         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4632         and output is a reg, avoid creating invalid lowpart subreg, but
4633         instead split into a 512-bit move.  Don't split if not AVX512VL,
4634         input is xmm16+ reg and output is a mem.
4635         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4636         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4637         xmm16+ reg and output is a mem.
4638
4639 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4640
4641         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4642         also for flag_dwarf2_cfi_asm.
4643
4644 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4645
4646         PR rtl-optimization/85342
4647         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4648         a bool scalar var inside of the loop instead.  Don't try to update
4649         recog_data.operand after failed apply_change_group.
4650
4651 2018-04-12  Tom de Vries  <tom@codesourcery.com>
4652
4653         PR target/85296
4654         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4655         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
4656         array with flexible array member as array without given dimension.
4657         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4658         argument for undefined param to true.
4659
4660 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
4661
4662         PR target/85321
4663         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4664         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4665         from PowerPC section.
4666         * config/rs6000/sysv4.opt (mcall-): Improve help text.
4667         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4668         help text that is too long.
4669         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4670         help text that is too long.
4671         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4672         help text that is too long.
4673
4674 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
4675
4676         * config/alpha/alpha.md (stack_probe_internal): Rename
4677         from "probe_stack".  Update all callers.
4678
4679 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4680
4681         PR rtl-optimization/84566
4682         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4683         sched_macro_fuse_insns.
4684
4685 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4686
4687         PR target/84301
4688         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4689         (compute_block_dependences): ... from here.
4690
4691 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4692
4693         PR tree-optimization/85331
4694         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4695         from int to HOST_WIDE_INT.
4696
4697 2018-04-11  Martin Jambor  <mjambor@suse.cz>
4698
4699         PR ipa/84149
4700         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4701         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4702         not the same as the source val.
4703         (cgraph_edge_brings_value_p): New parameter.
4704         (gather_edges_for_value): Pass destination value to
4705         cgraph_edge_brings_value_p.
4706         (perhaps_add_new_callers): Likewise.
4707         (get_info_about_necessary_edges): Likewise and exclude values brought
4708         only by self-recursive edges.
4709         (create_specialized_node): Redirect only clones of self-calling edges.
4710         (+self_recursive_pass_through_p): New function.
4711         (find_more_scalar_values_for_callers_subset): Use it.
4712         (find_aggregate_values_for_callers_subset): Likewise.
4713         (known_aggs_to_agg_replacement_list): Removed.
4714         (decide_whether_version_node): Re-calculate known constants for all
4715         remaining context clones.
4716
4717 2018-04-11  Richard Biener  <rguenther@suse.de>
4718
4719         PR lto/85339
4720         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4721         from early DWARF output.
4722         (dwarf2out_early_finish): Output line info unconditionally into
4723         early DWARF and add reference to it.
4724
4725 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4726
4727         PR target/85281
4728         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4729         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4730         other than V2DFmode using iptr mode attribute.
4731         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4732
4733 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4734
4735         PR rtl-optimization/84659
4736         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4737
4738 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4739
4740         PR debug/85302
4741         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4742         SIZEP is NULL.
4743         (output_loc_list): Pass address of a dummy size variable even in the
4744         locview handling loop.
4745         (index_location_lists): Add comment on why skip_loc_list_entry can't
4746         call size_of_locs.
4747
4748 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
4749
4750         PR target/85261
4751         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4752         into register.
4753
4754 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
4755
4756         PR target/85321
4757         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4758         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4759         and -mstring-compare-inline-limit.
4760
4761 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4762
4763         PR target/85287
4764         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4765         for stack clash protection in a register whenever we need it to be in
4766         a register.
4767
4768 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4769
4770         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4771         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4772
4773 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4774
4775         PR target/85321
4776         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4777         the help text.
4778         (mlong-double-): Ditto.
4779         * config/rs6000/sysv4.opt (msdata=): Ditto.
4780         (mtls-size=): Ditto.
4781
4782 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4783
4784         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4785         erroneous entries for
4786         "vector int vec_ldl (int, long int *)", and
4787         "vector unsigned int vec_ldl (int, unsigned long int *)".
4788         Add comments and entries for
4789         "vector bool char vec_ldl (int, bool char *)",
4790         "vector bool short vec_ldl (int, bool short *)",
4791         "vector bool int vec_ldl (int, bool int *)",
4792         "vector bool long long vec_ldl (int, bool long long *)",
4793         "vector pixel vec_ldl (int, pixel *)",
4794         "vector long long vec_ldl (int, long long *)",
4795         "vector unsigned long long vec_ldl (int, unsigned long long *)".
4796         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4797         type tree bool_long_long_type_node and correct definition of
4798         bool_V2DI_type_node to make reference to this new type tree.
4799         (rs6000_mangle_type): Replace erroneous reference to
4800         bool_long_type_node with bool_long_long_type_node.
4801         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4802         comments to emphasize sign distinctions for char and int types and
4803         replace RS6000_BTI_bool_long constant with
4804         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
4805         use of RS6000_BTI_pixel.
4806         (bool_long_type_node): Remove this macro definition.
4807         (bool_long_long_type_node): New macro definition
4808
4809 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4810
4811         PR rtl-optimization/85300
4812         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4813         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4814         simplify_unary_operation fails.
4815
4816 2018-04-10  Martin Liska  <mliska@suse.cz>
4817
4818         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4819         cgraph_edge and ipa_ref.
4820
4821 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4822
4823         PR target/85177
4824         PR target/85255
4825         * config/i386/sse.md
4826         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4827         computation of the VEC_MERGE selector from mask.
4828         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4829         Fix decoding of the VEC_MERGE selector into mask.
4830
4831 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4832
4833         PR tree-optimization/85286
4834         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4835
4836 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4837
4838         * final.c (final_1): Set insn_last_address as well as
4839         insn_current_address.
4840
4841 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4842
4843         PR target/85173
4844         * explow.c (emit_stack_probe): Call validize_mem on memory location
4845         before passing it to gen_probe_stack.  Create address operand and
4846         legitimize it for the probe_stack_address case.
4847
4848 2018-04-09  Jan Hubicka  <jh@suse.cz>
4849
4850         PR lto/85078
4851         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4852         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4853         * tree.c (free_lang_data_in_type): Fix handling of binfos;
4854         walk basetypes.
4855         (free_lang_data): Rebuild type inheritance graph.
4856
4857 2018-04-09  Martin Sebor  <msebor@redhat.com>
4858
4859         * invoke.texi (-finline-small-functions): Mention other optimization
4860         options.
4861         (-findirect-inlining, -fpartial-inlining): Same.
4862         (-finline-functions-called-once): Same.
4863         (-freorder-blocks-and-partition): Same.
4864
4865 2018-04-09  Jan Hubicka  <jh@suse.cz>
4866
4867         PR rtl/84058
4868         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4869         jumps; choose last target that matches the criteria (i.e.
4870         no partition changes for non-crossing jumps).
4871         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4872         support for redirecting crossing jumps to non-crossing.
4873
4874 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
4875
4876         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4877         also for naked functions.
4878
4879 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
4880
4881         * config/arc/arc.md (add_shift): New pattern.
4882         (add_shift2): Likewise.
4883         (sub_shift): Likewise.
4884         (sub_shift_cmp0_noout): Likewise.
4885         (compare_si_ashiftsi): Likewise.
4886         (xbfu_cmp0_noout): New combine pattern.
4887         (xbfu_cmp0"): Likewise.
4888         (movsi_set_cc_insn): Place the predicable variant first.
4889         (commutative_binary_cmp0_noout): Remove clobber.
4890         (commutative_binary_cmp0): New pattern.
4891         (noncommutative_binary_cmp0): Likewise.
4892         (noncommutative_binary_cmp0_noout): Likewise.
4893         (noncommutative_binary_comparison_result_used): Removed.
4894         (rsub_cmp0): New pattern.
4895         (rsub_cmp0_noout): Likewise.
4896         (extzvsi): Changed, keep only meaningful variants.
4897         (SQH, SEZ): New iterators.
4898         (SQH_postfix): New mode attribute.
4899         (SEZ_prefix): New code attribute.
4900         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4901         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4902         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4903         of numerical value.
4904         (noncommutative_operator): Check the availability of barrel
4905         shifter option.
4906
4907 2018-04-09  Richard Biener  <rguenther@suse.de>
4908
4909         PR tree-optimization/85284
4910         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4911         Only use the niter constraining form of simple_iv when the exit
4912         is always executed.
4913
4914 2018-04-09  Tom de Vries  <tom@codesourcery.com>
4915
4916         PR target/84041
4917         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4918         (define_expand "*memory_barrier"): New define_expand.
4919         (define_insn "memory_barrier"): New insn.
4920
4921 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4922
4923         PR rtl-optimization/80463
4924         PR rtl-optimization/83972
4925         PR rtl-optimization/83480
4926
4927         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4928         correct producer for the insn.
4929         (tidy_control_flow): Fixup seqnos in case of debug insns.
4930
4931 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4932
4933         PR rtl-optimization/83913
4934
4935         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4936         different sched-times when merging exprs.
4937
4938 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4939
4940         PR rtl-optimization/83962
4941
4942         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4943         tidy_fallthru_edge and tidy_control_flow.
4944
4945 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4946
4947         PR rtl-optimization/83530
4948
4949         * sel-sched.c (force_next_insn): New global variable.
4950         (remove_insn_for_debug): When force_next_insn is true, also leave only
4951         next insn in the ready list.
4952         (sel_sched_region): When the region wasn't scheduled, make another pass
4953         over it with force_next_insn set to 1.
4954
4955 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
4956
4957         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4958         into tm_file.
4959         * config/nds32/constants.md (unspec_volatile_element): Add enum values
4960         for interrupt control.
4961         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4962         functions for interrupt control.
4963         * config/nds32/nds32-intrinsic.md: Likewise.
4964         * config/nds32/nds32_intrinsic.h: Likewise.
4965         * config/nds32/nds32.h (nds32_builtins): Likewise.
4966
4967 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
4968
4969         * config/nds32/nds32.c (nds32_init_machine_status,
4970         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4971         strict_aligned_p field.
4972         (nds32_expand_to_rtl_hook): New function.
4973         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4974         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4975
4976 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4977             Chung-Ju Wu  <jasonwucj@gmail.com>
4978
4979         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4980         * config/nds32/nds32-n7.md: New file.
4981         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4982         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4983         pipeline.
4984         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4985         * config/nds32/nds32.md (pipeline_model): Add n7.
4986         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4987         * config/nds32/pipelines.md: Include n7 settings.
4988
4989 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4990             Chung-Ju Wu  <jasonwucj@gmail.com>
4991
4992         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4993         * config/nds32/nds32-e8.md: New file.
4994         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4995         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4996         pipeline.
4997         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4998         * config/nds32/nds32.md (pipeline_model): Add e8.
4999         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5000         * config/nds32/pipelines.md: Include e8 settings.
5001
5002 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5003             Chung-Ju Wu  <jasonwucj@gmail.com>
5004
5005         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5006         * config/nds32/nds32-n8.md: New file.
5007         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5008         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5009         pipeline.
5010         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5011         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5012         * config/nds32/nds32.md (pipeline_model): Add n8.
5013         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5014         * config/nds32/pipelines.md: Include n8 settings.
5015
5016 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5017             Chung-Ju Wu  <jasonwucj@gmail.com>
5018
5019         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5020         * config/nds32/nds32-n9-2r1w.md: New file.
5021         * config/nds32/nds32-n9-3r2w.md: New file.
5022         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5023         nds32_register_ports): New or modify for cpu n9.
5024         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5025         pipeline.
5026         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5027         * config/nds32/nds32-utils.c: New file.
5028         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5029         TARGET_MUL_SLOW): Define.
5030         * config/nds32/nds32.md (pipeline_model): New attribute.
5031         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5032         New options that support cpu n9.
5033         * config/nds32/pipelines.md: Include n9 settings.
5034         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5035
5036 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
5037
5038         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5039         information if necessary.
5040         (output_cond_branch_compare_zero): Likewise.
5041         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5042         (nds32_target_alignment): Refine for alignment.
5043         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5044         (FUNCTION_BOUNDARY): Modify.
5045         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5046         align case.
5047         * config/nds32/nds32.opt (malways-align, malign-functions): New.
5048
5049 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
5050
5051         * config/nds32/constants.md (unspec_volatile_element): Add values for
5052         TLB operation and data prefetch.
5053         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5054         functions for TLB operation and data prefetch.
5055         * config/nds32/nds32-intrinsic.md: Likewise.
5056         * config/nds32/nds32_intrinsic.h: Likewise.
5057         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5058         (nds32_print_operand): Likewise.
5059         * config/nds32/nds32.h (nds32_builtins): Likewise.
5060
5061 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
5062         Andrew Pinski <pinsika@gcc.gnu.org>
5063
5064         PR middle-end/82976
5065         * match.pd: Use constant_boolean_node of correct type instead of
5066         boolean_true_node or boolean_false_node for simplifying
5067         pointer comparisons to zero.
5068
5069 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
5070
5071         PR tree-optimization/80021
5072         * tree.c (verify_type_variant): Make error call in verify_variant_match
5073         translatable and remove final full stop.
5074
5075 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5076
5077         * config/nds32/constants.md (unspec_volatile_element): Add
5078         UNSPEC_VOLATILE_EH_RETURN.
5079         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5080         nds32_output_stack_pop): Support dwarf exception handling process.
5081         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5082         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5083         exception handling process.
5084         (nds32_compute_stack_frame): Likewise.
5085         (nds32_return_addr_rtx): Likewise.
5086         (nds32_initial_elimination_offset): Likewise.
5087         (nds32_expand_prologue): Likewise.
5088         (nds32_expand_epilogue): Likewise.
5089         (nds32_dynamic_chain_address): New function.
5090         * config/nds32/nds32.h (machine_function): Add fields for dwarf
5091         exception handling.
5092         (DYNAMIC_CHAIN_ADDRESS): Define.
5093         (EH_RETURN_DATA_REGNO): Define.
5094         (EH_RETURN_STACKADJ_RTX): Define.
5095         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5096         patterns for dwarf exception handling.
5097
5098 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5099
5100         * config/nds32/nds32.h: Clean up obsolete macros.
5101
5102 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5103
5104         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5105         Add enum values for particular instructions.
5106         * config/nds32/nds32-intrinsic.c: Implementation of expanding
5107         particular intrinsic functions.
5108         * config/nds32/nds32-intrinsic.md: Likewise.
5109         * config/nds32/nds32_intrinsic.h: Likewise.
5110         * config/nds32/nds32.h (nds32_builtins): Likewise.
5111         * config/nds32/nds32.md (type): Add pbsad and pbsada.
5112         (btst, ave): New patterns for particular instructions.
5113
5114 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5115
5116         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5117         Add enum values for atomic load/store and memory sync.
5118         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5119         and memory sync.
5120         * config/nds32/nds32-intrinsic.md: Likewise.
5121         * config/nds32/nds32_intrinsic.h: Likewise.
5122         * config/nds32/nds32.h (nds32_builtins): Likewise.
5123
5124 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
5125
5126         PR tree-optimization/85257
5127         * fold-const.c (native_encode_vector): If not all elts could fit
5128         and off is -1, return 0 rather than offset.
5129         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5130         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5131         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
5132         adjust buffer in native_interpret_expr call.
5133
5134 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5135
5136         * config/nds32/constants.md (unspec_volatile_element): Add cache
5137         control enum values.
5138         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5139         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5140         * config/nds32/nds32.c (nds32_cctl_names): New.
5141         (nds32_print_operand): Handle cache control register names.
5142         * config/nds32/nds32.h (nds32_builtins): New enum values.
5143         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5144         macros.
5145         * config/nds32/nds32.md (type): Add mmu.
5146         * config/nds32/pipelines.md (simple_insn): Add mmu.
5147
5148 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5149
5150         * config/nds32/nds32.md (type): Remove call.
5151         * config/nds32/pipelines.md (simple_insn): Likewise.
5152
5153 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5154
5155         * config/nds32/constants.md (unspec_volatile_element): Add
5156         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5157         UNSPEC_VOLATILE_FMFCFG.
5158         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5159         description for fmfcfg and fmfcsr.
5160         (bdesc_1arg): Add fmtcsr.
5161         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5162         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5163         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5164         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5165         unspec_fmfcfg): New patterns.
5166         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5167         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5168         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5169         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5170         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5171         __nds32__fmfcfg): Define.
5172
5173 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5174
5175         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5176         intrinsic register names.
5177         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5178         intrinsic register enum values and macros.
5179
5180 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5181
5182         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5183         for load/store addressing form.
5184         (nds32_print_operand_address): Likewise.
5185
5186 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
5187
5188         PR target/85196
5189         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5190         based on LABEL_REF.  Remove useless assertion.
5191         (pic_address_needs_scratch): Fix formatting.
5192         (sparc_legitimize_pic_address): Minor tweaks.
5193         (sparc_delegitimize_address): Adjust assertion accordingly.
5194         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5195         into symbolic_operand.
5196         (movsi_high_pic_label_ref): Likewise.
5197         (movsi_lo_sum_pic_label_ref): Likewise.
5198         (movdi_pic_label_ref): Likewise.
5199         (movdi_high_pic_label_ref): Likewise.
5200         (movdi_lo_sum_pic_label_ref): Likewise.
5201
5202 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
5203
5204         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5205         custom LIB_SPEC setup.
5206
5207 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
5208             Kito Cheng  <kito.cheng@gmail.com>
5209
5210         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5211         * config/riscv/freebsd.h: New.
5212
5213 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5214
5215         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5216         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5217         file.
5218
5219 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5220             Kito Cheng  <kito.cheng@gmail.com>
5221
5222         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5223         nds32_output_call, nds32_symbol_binds_local_p): New functions.
5224         * config/nds32/nds32-protos.h (nds32_output_call,
5225         nds32_output_return): Declare.
5226         * config/nds32/nds32.md: Refine all the call and return patterns.
5227
5228 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
5229
5230         PR debug/85252
5231         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5232         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5233
5234         PR rtl-optimization/84872
5235         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5236         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5237         EDGE_CROSSING edge.
5238
5239 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
5240
5241         * expr.c (copy_blkmode_to_reg): Revert 254862.
5242         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5243
5244 2018-04-06  Richard Biener  <rguenther@suse.de>
5245
5246         PR middle-end/85244
5247         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5248         after seeing a component reference with an adjacent field.  Treat
5249         refs to arrays at struct end of external decls similar to
5250         refs to unconstrained commons.
5251
5252 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
5253
5254         PR sanitizer/85213
5255         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5256         look through SAVE_EXPRs with non-side-effects argument.  Adjust
5257         recursive calls.
5258         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5259         save_p here.
5260
5261 2018-04-06  Richard Biener  <rguenther@suse.de>
5262
5263         PR middle-end/85180
5264         * alias.c (find_base_term): New wrapper around find_base_term
5265         unwinding CSELIB_VAL_PTR changes.
5266         (find_base_term): Do not restore CSELIB_VAL_PTR during the
5267         recursion.
5268
5269 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5270
5271         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5272         instructions.
5273         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5274         constant definitions.
5275         ("nop"): lr 0,0 -> nopr r0
5276         ("nop_lr0", "nop_lr1"): New insn definitions.
5277
5278 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5279
5280         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5281         NDS32_V3PUSH_AVAILABLE_P macro.
5282
5283 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
5284             Chung-Ju Wu  <jasonwucj@gmail.com>
5285
5286         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5287         (nds32*-*-*): Add float and fpu_config into supported_defaults.
5288         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5289         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5290         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5291         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5292         * config/nds32/constraints.md: New constraints and checking for hard
5293         float configuration.
5294         * config/nds32/iterators.md: New mode iterator and attribute for hard
5295         float configuration.
5296         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5297         patterns.
5298         * config/nds32/nds32-fpu.md: New file.
5299         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5300         deal with hard float code generation.
5301         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5302         ARCH_V3S.
5303         (abi_type, float_reg_number): New enum type.
5304         * config/nds32/nds32-predicates.c: New predicates for hard float.
5305         * config/nds32/nds32-protos.h: Declare functions for hard float.
5306         * config/nds32/nds32.c: Implementation for hard float configuration.
5307         * config/nds32/nds32.h: Definitions for hard float configuration.
5308         * config/nds32/nds32.md: Include hard float machine description and
5309         modify patterns for hard float configuration.
5310         * config/nds32/nds32.opt: New options for hard float configuration.
5311         * config/nds32/predicates.md: New predicates for hard float
5312         configuration.
5313
5314 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
5315
5316         * common/config/nds32/nds32-common.c
5317         (nds32_option_optimization_table): Enable -mreleax-hint by default.
5318
5319 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
5320
5321         PR middle-end/85195
5322         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5323         CONSTRUCTOR_ELT (ctor, ...)->value.
5324
5325 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
5326
5327         PR target/85193
5328         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5329
5330 2018-04-05  Tom de Vries  <tom@codesourcery.com>
5331
5332         PR target/85204
5333         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5334         cond jump.
5335
5336 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
5337             Kito Cheng  <kito.cheng@gmail.com>
5338
5339         * config/nds32/constraints.md (U33): Fine-tune checking condition.
5340         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5341         * config/nds32/nds32.h (nds32_16bit_address_type): Add
5342         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5343
5344 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
5345             Kito Cheng  <kito.cheng@gmail.com>
5346
5347         * config/nds32/constraints.md (Ufe): New memory constraint.
5348         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5349         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5350         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5351         operands.
5352         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5353         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5354
5355 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5356
5357         * config/nds32/nds32.md: Use optimize_size in the condition for
5358         alu-shift instructions.
5359
5360 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5361
5362         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5363
5364 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5365
5366         * config/nds32/nds32.md (negsi2): Refine pattern.
5367
5368 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
5369             Chung-Ju Wu  <jasonwucj@gmail.com>
5370
5371         * config/nds32/iterators.md (shift_rotate): New code iterator.
5372         (shift): New code attribute.
5373         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5374         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5375         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5376         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5377         bit-wise operations.
5378         (andsi3, *andsi3): Ditto.
5379         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5380         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5381         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5382         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5383         nds32_ior_operand, nds32_xor_operand): New predicates.
5384
5385 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5386
5387         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5388         (addsi3, subsi3): ... this.
5389
5390 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5391
5392         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5393
5394 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5395
5396         * config/nds32/nds32.md: Adjust indention.
5397
5398 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
5399
5400         * config/nds32/nds32.md (feature): New attribute.
5401
5402 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5403
5404         * config/nds32/nds32.md (subtype): New attribute.
5405
5406 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
5407
5408         PR target/85203
5409         * config/arm/arm-builtins.c (arm_expand_builtin): Change
5410         expansion to perform a bitwise AND of the argument followed by a
5411         boolean negation of the result.
5412
5413 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
5414
5415         PR rtl-optimization/84878
5416         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5417         the basic block.  Assert the use reference is not artificial and that
5418         it has an associated insn.
5419
5420 2018-04-04  Michael Matz  <matz@suse.de>
5421
5422         * builtins.c (compute_objsize): Pass correct operand
5423         to array_at_struct_end_p.
5424
5425 2018-04-04  Richard Biener  <rguenther@suse.de>
5426
5427         PR lto/85176
5428         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5429         from contexts for DINFO_LEVEL_TERSE and below.
5430
5431 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5432
5433         * config/nds32/nds32-doubleword.md (move_<mode>): Require
5434         resiter_operand condition.
5435         * config/nds32/nds32.md (*move<mode>): Ditto.
5436
5437 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5438             Monk Chiang  <sh.chiang04@gmail.com>
5439
5440         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5441
5442 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5443
5444         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5445
5446 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5447             Kito Cheng  <kito.cheng@gmail.com>
5448
5449         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5450         nds32_cond_code_str, output_cond_branch,
5451         output_cond_branch_compare_zero, nds32_expand_cbranch,
5452         nds32_expand_cstore, nds32_expand_movcc,
5453         nds32_output_cbranchsi4_equality_zero,
5454         nds32_output_cbranchsi4_equality_reg,
5455         nds32_output_cbranchsi4_equality_reg_or_const_int,
5456         nds32_output_cbranchsi4_greater_less_zero: New functions.
5457         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5458         nds32_expand_cstore, nds32_expand_movcc,
5459         nds32_output_cbranchsi4_equality_zero,
5460         nds32_output_cbranchsi4_equality_reg,
5461         nds32_output_cbranchsi4_equality_reg_or_const_int,
5462         nds32_output_cbranchsi4_greater_less_zero): Declare.
5463         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5464         nds32_rimm11s_operand): New predicates.
5465         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5466         * config/nds32/nds32.md: Rewrite all the branch and conditional move
5467         patterns.
5468
5469 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5470
5471         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5472         * config/nds32/nds32.md: Ditto.
5473         * config/nds32/pipelines.md: Ditto.
5474
5475 2018-04-04  Richard Biener  <rguenther@suse.de>
5476
5477         PR tree-optimization/85168
5478         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5479         propagating abnormals.
5480
5481 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5482
5483         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5484
5485 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5486             Kito Cheng  <kito.cheng@gmail.com>
5487
5488         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5489         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5490         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5491         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5492         * config/nds32/nds32.md (sibcall_internal): New.
5493         (sibcall_register): Remove.
5494         (sibcall_immediate): Remove.
5495         (sibcall_value_internal): New.
5496         (sibcall_value_register): Remove.
5497         (sibcall_value_immediate): Remove.
5498         * config/nds32/predicates.md (nds32_general_register_operand): New.
5499         (nds32_call_address_operand): New.
5500
5501 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
5502
5503         PR rtl-optimization/85167
5504         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5505         bb_defs if *split_p, instead preinitialize it to NULL.
5506
5507         PR tree-optimization/85156
5508         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5509         evaluating the argument multiple times.
5510
5511 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
5512
5513         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5514         than vector.
5515         (_mm_cvtpd_ps): Likewise.
5516         (_mm_cvttpd_epi32): Likewise.
5517         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5518         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5519         vector, pixel, and bool following altivec.h include.
5520
5521 2018-04-03  Martin Sebor  <msebor@redhat.com>
5522
5523         * doc/extend.texi (Common Function Attributes): Clarify.
5524         (const attribute): Likewise.
5525         (pure attribute): Likewise.
5526
5527 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
5528
5529         PR target/85169
5530         * config/i386/i386.c (ix86_expand_vector_set): Use
5531         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
5532
5533 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
5534
5535         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5536         instructions when changing rounding bits to preserve precision bits
5537         in the x87 control word.
5538
5539 2018-04-03  Martin Liska  <mliska@suse.cz>
5540
5541         PR tree-optimization/82491
5542         * rtl.h (strip_offset_and_add): Replace += suboffset with
5543         poly_uint64 () + suboffset.
5544
5545 2018-03-29  Martin Liska  <mliska@suse.cz>
5546             Martin Jambor  <mjambor@suse.cz>
5547
5548         PR ipa/84947
5549         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5550         param_type is not an integral or pointer type.
5551
5552 2018-04-03  Richard Biener  <rguenther@suse.de>
5553
5554         * sese.h (recompute_all_dominators): Remove.
5555
5556 2018-04-02  Martin Sebor  <msebor@redhat.com>
5557
5558         * doc/invoke.texi (-Wrestrict): Fix typos.
5559
5560 2018-04-02  Jim Wilson  <jimw@sifive.com>
5561
5562         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5563         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5564         (<optab>di3, <optab>si3_extend): Likewise.
5565         (<optab>si3_mask, <optab>si3_mask_1): New.
5566         (<optab>di3_mask, <optab>di3_mask_1): New.
5567         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5568         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5569         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5570
5571 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
5572
5573         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5574         example.
5575
5576 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
5577
5578         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5579         (nds32_canonicalize_comparison): New function.
5580
5581 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5582             Kito Cheng  <kito.cheng@gmail.com>
5583             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
5584
5585         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5586         * config/nds32/constants.md (unspec_volatile_element): Add
5587         UNSPEC_VOLATILE_RELAX_GROUP.
5588         * config/nds32/nds32-relax-opt.c: New file.
5589         * config/nds32/nds32-predicates.c
5590         (nds32_symbol_load_store_p): New function.
5591         * config/nds32/nds32-protos.h
5592         (nds32_symbol_load_store_p): Declare function.
5593         (make_pass_nds32_relax_opt): Declare new rtl pass function.
5594         * config/nds32/nds32.c
5595         (nds32_register_pass): New function to register pass.
5596         (nds32_register_passes): New function to register passes.
5597         * config/nds32/nds32.md (relax_group): New pattern.
5598         * config/nds32/nds32.opt (mrelax-hint): New option.
5599         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5600
5601 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
5602
5603         * config/nds32/t-nds32: Modify files dependency.
5604
5605 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5606
5607         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5608         (PROFILE_HOOK): Define its implementation.
5609
5610 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5611
5612         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5613         type and 32-bit size.
5614
5615 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
5616
5617         PR middle-end/85090
5618         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5619         (V_128_256): New mode iterator.
5620         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5621         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5622         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5623         of V.
5624         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5625         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5626
5627 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
5628
5629         PR target/83315
5630         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5631         NaN inputs correctly.
5632
5633 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
5634
5635         PR target/80546
5636         * config/rs6000/vsx.md (??r): New mode attribute.
5637         (*vsx_mov<mode>_64bit): Use it.
5638         (*vsx_mov<mode>_32bit): Likewise.
5639
5640 2018-03-30  Martin Sebor  <msebor@redhat.com>
5641
5642         PR tree-optimization/84818
5643         * builtins.c (check_access): Use warning_n.
5644
5645 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
5646
5647         PR target/83822
5648         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5649         condition.
5650         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5651         condition.
5652
5653 2018-03-30  Julia Koval  <julia.koval@intel.com>
5654
5655         PR target/84413
5656         * x86-tune.def (movx, partial_reg_dependency): Enable for
5657         m_SKYLAKE_AVX512.
5658
5659 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
5660
5661         PR inline-asm/84985
5662         * lra-constraints.c (process_alt_operands): Move setting
5663         this_alternative_matches below.
5664
5665 2018-03-29  Martin Liska  <mliska@suse.cz>
5666
5667         PR lto/84995.
5668         * doc/invoke.texi: Document how LTO works with debug info.
5669         Describe auto-load support of binutils.  Mention 'x86-64'
5670         as valid option value of -march option.
5671
5672 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
5673
5674         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5675
5676         PR c/85094
5677         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5678         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5679         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5680         checking.
5681
5682 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
5683
5684         PR target/84912
5685         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5686         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5687         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5688         for RS6000_BTM_POWERPC64.
5689         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5690         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5691         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5692         definition.
5693         (DIVDE): Use it.
5694         (DIVDEU): Likewise.
5695
5696 2018-03-28 Carl Love  <cel@us.ibm.com>
5697
5698         Revert
5699         2017-09-27  Carl Love  <cel@us.ibm.com>
5700
5701         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5702         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5703         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5704         fctiw instruction.
5705
5706 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5707
5708         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5709         instead of __vector bool.
5710         (_mm_max_pu8): Likewise.
5711         (_mm_min_pi16): Likewise.
5712
5713 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
5714
5715         PR target/84912
5716         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5717         (DIVWEUO): Likewise.
5718         (DIVDEO): Likewise.
5719         (DIVDEUO): Likewise.
5720         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5721         DIVWEUO and DIVDEUO.
5722         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5723         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5724         (div_extend): Likewise.
5725         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5726         builtin function.
5727         (__builtin_divweuo): Likewise.
5728         (__builtin_divdeo): Likewise.
5729         (__builtin_divdeuo): Likewise.
5730
5731 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
5732
5733         PR target/85095
5734         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5735         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5736
5737         PR tree-optimization/82004
5738         * gimple-match-head.c (optimize_pow_to_exp): New function.
5739         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5740         Don't fold to exp if optimize_pow_to_exp is false.
5741
5742 2018-03-28  Martin Liska  <mliska@suse.cz>
5743
5744         PR other/84819
5745         * calls.c (initialize_argument_information): Fix trailing space.
5746         * common.opt: Fix typo and provide better explanation for
5747         -fsanitize-coverage option.
5748         * config/i386/i386.opt: Fix typo.
5749
5750 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
5751             Martin Liska  <mliska@suse.cz>
5752
5753         PR sanitizer/85081
5754         * gimplify.c (asan_poison_variable): Don't do the check for
5755         gimplify_omp_ctxp here.
5756         (gimplify_decl_expr): Do it here.
5757         (gimplify_target_expr): Likewise.
5758
5759 2018-03-28  Martin Liska  <mliska@suse.cz>
5760
5761         PR target/84988
5762         * config/i386/i386.c (ix86_function_arg_advance): Do not call
5763         chkp_type_bounds_count if MPX is not enabled.
5764
5765 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
5766
5767         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5768
5769 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5770
5771         PR target/84914
5772         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5773         function to create the function decl for complex long double
5774         multiply and divide for -mabi=ieeelongdouble.
5775         (init_float128_ieee): Call it.
5776
5777 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
5778
5779         PR target/85044
5780         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5781         -fcf-protection=branch -mibt.
5782         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5783
5784 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5785
5786         PR target/81863
5787         * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5788
5789 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
5790
5791         PR target/85056
5792         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5793         extern array declarations.
5794
5795 2018-03-27  Richard Biener  <rguenther@suse.de>
5796
5797         PR middle-end/84067
5798         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5799         explicit single_use checks.
5800
5801 2018-03-27  Richard Biener  <rguenther@suse.de>
5802
5803         PR tree-optimization/85082
5804         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5805         Valueize the VUSE.
5806
5807 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5808
5809         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5810         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5811         Turn on fasynchronous-unwind-tables and funwind-tables.
5812
5813 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
5814
5815         PR target/85073
5816         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5817         (*bmi_blsr_<mode>_ccz): Ditto.
5818
5819 2018-03-26  Tom de Vries  <tom@codesourcery.com>
5820
5821         PR tree-optimization/85063
5822         * omp-general.c (offloading_function_p): New function.  Factor out
5823         of ...
5824         * omp-offload.c (pass_omp_target_link::gate): ... here.
5825         * omp-general.h (offloading_function_p): Declare.
5826         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5827         with attribute omp declare target for offloading functions.
5828
5829 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
5830
5831         PR tree-optimization/84005
5832         * tree-data-ref.h (get_base_for_alignment): Declare.
5833         * tree-data-ref.c (get_base_for_alignment_1): New function.
5834         (get_base_for_alignment): Likewise.
5835         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5836         get_base_for_alignment to find a suitable base object, instead
5837         of always using drb->base_address.
5838
5839 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5840
5841         PR inline-asm/85022
5842         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5843         known size by default.
5844
5845 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
5846
5847         PR inline-asm/85030
5848         * lra-constraints.c (process_alt_operands): Don't match BLKmode
5849         and non BLKmode operands.
5850
5851 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5852
5853         PR target/85026
5854         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5855         Clean up attributes.
5856
5857 2018-03-23  Richard Biener  <rguenther@suse.de>
5858
5859         PR debug/85020
5860         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5861         we are going to emit early debug for LTO.
5862
5863 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5864
5865         PR inline-asm/85034
5866         * function.c (match_asm_constraints_1): Don't optimize if input
5867         doesn't satisfy general_operand predicate for output's mode.
5868
5869         PR inline-asm/85022
5870         * alias.c (write_dependence_p): Don't require for x_canonicalized
5871         non-VOIDmode if x has VOIDmode.
5872
5873         PR sanitizer/85029
5874         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5875         just don't try to optimize it rather than assert it never happens.
5876
5877 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5878
5879         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5880         macro expansions for definition of ST_INTERNAL_<mode> and
5881         LD_INTERNAL_<mode> builtins.
5882         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5883         Remove prototype.
5884         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5885         function.
5886         (altivec_expand_st_builtin): Likewise.
5887         (altivec_expand_builtin): Remove calls to deleted functions.
5888         (rs6000_address_for_altivec): Delete this function.
5889         * config/rs6000/vector.md: Remove expands for
5890         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5891
5892 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
5893
5894         PR target/84826
5895         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5896         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5897         re-computing once computed.
5898         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5899         (arm_init_machine_status): Initialize
5900         machine->static_chain_stack_bytes.
5901
5902 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5903
5904         PR target/84760
5905         * doc/extend.texi: Add four new prototypes for vec_ld.
5906         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5907         definitions for more logical presentation.
5908         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5909         entries for V1TI variants of __builtin_altivec_ld builtin.
5910         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5911         handling of V1TI variant of LVX icode pattern.
5912         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5913         (rs6000_gimple_fold_builtin): Likewise.
5914         (altivec_init_builtins): Add code to define
5915         __builtin_altivec_lvx_v1ti function.
5916
5917 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
5918
5919         PR inline-asm/84941
5920         * function.c (match_asm_constraints_1): Don't do the optimization
5921         if input isn't a REG, SUBREG, MEM or constant.
5922
5923 2018-03-22  Tom de Vries  <tom@codesourcery.com>
5924
5925         PR tree-optimization/84956
5926         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5927         bb_has_abnormal_pred.
5928
5929 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
5930
5931         PR sanitizer/85018
5932         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5933         DECL_INITIAL (decl) to decl at the end.
5934         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5935         adjust the comment.
5936
5937 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
5938
5939         * doc/extend.texi (__builtin_tgmath): Document when complex
5940         integer types are treated as _Complex _Float64.
5941
5942 2018-03-21  Tom de Vries  <tom@codesourcery.com>
5943
5944         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5945
5946 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5947
5948         PR tree-optimization/84960
5949         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5950         if it is ENTRY block, move them into single succ of ENTRY in that case.
5951
5952 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
5953
5954         PR tree-optimization/84811
5955         * poly-int.h (poly_span_traits): Remove the T3 parameter and
5956         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5957         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5958         (known_subrange_p): Update accordingly.  Cast each value involved
5959         in the size comparison, rather than casting the result of the
5960         subtraction.
5961
5962 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5963
5964         PR tree-optimization/84982
5965         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5966         by flipping the least significant bit rather than all bits from
5967         bitpos to bitpos + bitsize - 1.
5968
5969 2018-03-21  Nathan Sidwell  <nathan@acm.org>
5970
5971         * doc/extend.texi (Deprecated Features): Remove mention of
5972         long-deleted deprecations.
5973
5974 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5975
5976         PR jit/84288
5977         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5978         * configure: Regenerate.
5979
5980 2018-03-21  Tom de Vries  <tom@codesourcery.com>
5981
5982         PR tree-optimization/83126
5983         * tree-parloops.c (num_phis): New function.
5984         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5985
5986 2018-03-21  Nathan Sidwell  <nathan@acm.org>
5987
5988         * doc/extend.texi (Deprecated Features): Update deprecated flags,
5989         mention anon-struct/union members and trailing attributes.
5990
5991 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
5992
5993         PR tree-optimization/84969
5994         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5995         builtin memset partitions if they set different rhs values.
5996
5997 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5998
5999         PR rtl-optimization/84989
6000         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6001         VEC_DUPLICATE with scalar result mode.
6002
6003 2018-03-21  Martin Liska  <mliska@suse.cz>
6004
6005         PR ipa/84963
6006         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6007         not intended return statement.
6008
6009 2018-03-21  Martin Liska  <mliska@suse.cz>
6010
6011         PR target/84988
6012         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6013         (chkp_find_bound_slots_1): Limit number of iterations.
6014
6015 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
6016
6017         PR target/84838
6018         * Minor grammar fixes for x86 options.
6019
6020 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6021
6022         PR debug/84875
6023         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6024         holding REG_CFA_RESTORE notes, instead turn them into a USE.
6025
6026 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
6027
6028         PR target/83789
6029         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6030         (altivec_lvx_<mode>_1op): Likewise.
6031         (altivec_stvx_<mode>_2op): Likewise.
6032         (altivec_stvx_<mode>_1op): Likewise.
6033         (altivec_lvx_<VM2:mode>): New define_expand.
6034         (altivec_stvx_<VM2:mode>): Likewise.
6035         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6036         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6037         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6038         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6039         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6040         (rs6000_gen_lvx): Likewise.
6041         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6042         (altivec_expand_stv_builtin): Likewise.
6043         (altivec_expand_builtin): Likewise.
6044         * config/rs6000/vector.md: Likewise.
6045
6046 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6047
6048         PR target/82518
6049         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6050         BYTES_BIG_ENDIAN.
6051
6052 2018-03-20  Richard Biener  <rguenther@suse.de>
6053
6054         PR target/84986
6055         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6056         sign-conversions as zero, fall back to standard scalar_stmt
6057         cost for the rest.
6058
6059 2018-03-20  Martin Liska  <mliska@suse.cz>
6060
6061         PR ipa/84825
6062         * predict.c (rebuild_frequencies): Handle case when we have
6063         PROFILE_ABSENT, but flag_guess_branch_prob is false.
6064
6065 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6066
6067         PR target/84990
6068         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6069         flag_section_anchors.
6070         * varasm.c (use_blocks_for_decl_p): Remove hack for
6071         dw2_force_const_mem.
6072
6073         PR target/84845
6074         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6075         to ...
6076         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
6077         be created, use lowpart_subreg of operands[0] rather than operands[0]
6078         itself.
6079         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6080         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6081         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6082         and n constraint instead of aarch64_shift_imm_di and Usd.
6083         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6084         (*aarch64_<optab>_reg_minus<mode>3): ... this.
6085
6086 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
6087
6088         PR target/82989
6089         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6090         to favor GPR over NEON registers.
6091         (<shift>di3_neon): Likewise.
6092
6093 2018-03-20  Tom de Vries  <tom@codesourcery.com>
6094
6095         PR target/84952
6096         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6097         (nvptx_process_pars): Emit bar.sync asap and alap.
6098
6099 2018-03-20  Tom de Vries  <tom@codesourcery.com>
6100
6101         PR target/84954
6102         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6103         seen_label if seen_label is already set.
6104
6105 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6106
6107         PR target/84945
6108         * config/i386/i386.c (fold_builtin_cpu): For features above 31
6109         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6110         Use 1U instead of 1.  Formatting fixes.
6111
6112         PR c/84953
6113         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6114         instead of TREE_TYPE (s1) for the return value.
6115
6116 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
6117
6118         PR tree-optimization/84946
6119         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6120         bitsize + bitsize in poly_uint64 rather than poly_int64.
6121
6122         PR sanitizer/78651
6123         * dwarf2asm.c: Include fold-const.c.
6124         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6125         of decl rather than decl itself.
6126
6127         PR rtl-optimization/84643
6128         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6129
6130 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
6131
6132         PR sanitizer/78651
6133         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6134         calling assemble_variable.
6135
6136 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
6137
6138         PR target/81647
6139         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6140         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6141
6142 2018-03-19  Jim Wilson  <jimw@sifive.com>
6143
6144         PR bootstrap/84856
6145         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6146         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6147         (riscv_first_stack_step): Likewise.
6148         (riscv_option_override): Use STACK_BOUNDARY instead of
6149         MIN_STACK_BOUNDARY.
6150         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6151         MIN_STACK_BOUNDARY.
6152         (BIGGEST_ALIGNMENT): Set to 128.
6153         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6154         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6155         STACK_BOUNDARY.
6156
6157 2018-03-19  Richard Biener  <rguenther@suse.de>
6158
6159         PR tree-optimization/84933
6160         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6161         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6162
6163 2018-03-19  Richard Biener  <rguenther@suse.de>
6164
6165         PR tree-optimization/84859
6166         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6167         (cond_if_else_store_replacement): Perform sinking operation on
6168         single-store BBs regardless of MAX_STORES_TO_SINK setting.
6169         Generalize what a BB with a single eligible store is.
6170
6171 2018-03-19  Richard Biener  <rguenther@suse.de>
6172
6173         PR tree-optimization/84929
6174         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6175         chrec_is_positive against non-chrec arg.
6176
6177 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
6178
6179         PR target/84711
6180         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6181
6182 2018-03-18  Martin Liska  <mliska@suse.cz>
6183
6184         PR rtl-optimization/84635
6185         * regrename.c (build_def_use): Use matches_mode only when
6186         matches >= 0.
6187
6188 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
6189
6190         PR tree-optimization/84913
6191         * tree-vect-loop.c (vectorizable_reduction): Don't try to
6192         vectorize chains of COND_EXPRs.
6193
6194 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6195
6196         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6197
6198 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6199
6200         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6201
6202 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6203
6204         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6205
6206 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6207             Kito Cheng  <kito.cheng@gmail.com>
6208
6209         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6210         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6211         (nds32_adjust_reg_alloc_order): New function.
6212         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6213
6214 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
6215
6216         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6217         nds32_print_operand, nds32_print_operand_address): Use
6218         HOST_WIDE_INT_PRINT_DEC instead.
6219
6220 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6221
6222         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6223
6224 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
6225
6226         PR target/84902
6227         * config/i386/i386.c (initial_ix86_tune_features,
6228         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6229         unsigned long long.
6230         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6231         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6232         rather than 1u << ix86_tune.  Formatting fix.
6233         (ix86_option_override_internal): Change ix86_arch_mask from
6234         unsigned int to unsigned HOST_WIDE_INT, initialize to
6235         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6236         (ix86_function_specific_restore): Likewise.
6237
6238 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6239
6240         PR target/84899
6241         * postreload.c (reload_combine_recognize_pattern): Perform
6242         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6243         truncate_int_for_mode the result for the destination's mode.
6244
6245         PR c/84909
6246         * hsa-gen.c (mem_type_for_type): Fix comment typo.
6247         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6248         Likewise.
6249         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6250         Likewise.
6251
6252 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
6253
6254         PR target/84876
6255         * lra-assigns.c (lra_split_hard_reg_for): Don't use
6256         regno_allocno_class_array and sorted_pseudos.
6257         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6258         insns where regno is used.
6259
6260 2018-03-16  Martin Liska  <mliska@suse.cz>
6261
6262         PR ipa/84833
6263         * multiple_target.c (create_dispatcher_calls): Redirect
6264         reference in the symbol table.
6265
6266 2018-03-16  Martin Liska  <mliska@suse.cz>
6267
6268         PR ipa/84722
6269         * multiple_target.c (create_dispatcher_calls): Redirect also
6270         an alias.
6271
6272 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6273
6274         PR c++/79937
6275         PR c++/82410
6276         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6277         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6278         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6279
6280 2018-03-16  Julia Koval  <julia.koval@intel.com>
6281
6282         * doc/invoke.texi (Skylake Server): Add CLWB.
6283         Cannonlake): Remove CLWB.
6284
6285 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6286
6287         PR tree-optimization/84841
6288         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6289         1 << 3.
6290         (FLOAT_ONE_CONST_TYPE): Define.
6291         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6292         (sort_by_operand_rank): Put entries with higher constant_type last
6293         rather than first to match comments.
6294
6295 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
6296
6297         * config/nios2/nios2.md (movsi_internal): Fix thinko in
6298         split predicate.
6299
6300 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
6301
6302         PR c++/79085
6303         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6304         check and use address of target always.
6305
6306 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
6307
6308         PR target/84574
6309         * config/i386/i386.c (indirect_thunk_needed): Update comments.
6310         (indirect_thunk_bnd_needed): Likewise.
6311         (indirect_thunks_used): Likewise.
6312         (indirect_thunks_bnd_used): Likewise.
6313         (indirect_return_needed): New.
6314         (indirect_return_bnd_needed): Likewise.
6315         (output_indirect_thunk_function): Add a bool argument for
6316         function return.
6317         (output_indirect_thunk_function): Don't generate alias for
6318         function return thunk.
6319         (ix86_code_end): Call output_indirect_thunk_function to generate
6320         function return thunks.
6321         (ix86_output_function_return): Set indirect_return_bnd_needed
6322         and indirect_return_needed instead of indirect_thunk_bnd_needed
6323         and indirect_thunk_needed.
6324
6325 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
6326
6327         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6328         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6329         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6330
6331 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
6332             Paul Hua <paul.hua.gm@gmail.com>
6333
6334         PR c/84852
6335         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6336
6337 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
6338
6339         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6340         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6341         resp. SFmode cases.
6342
6343 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
6344
6345         PR target/84711
6346         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6347         instead of GET_MODE_SIZE when comparing Units.
6348
6349 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
6350
6351         PR target/68256
6352         * varasm.c (hash_section): Return an unchangeble hash value
6353         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6354         Return !aarch64_can_use_per_function_literal_pools_p ().
6355
6356 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
6357
6358         PR target/84860
6359         * optabs.c (emit_conditional_move): Pass address of cmode's copy
6360         rather than address of cmode as last argument to prepare_cmp_insn.
6361
6362 2018-03-15  Julia Koval  <julia.koval@intel.com>
6363
6364         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6365         F_AVX512VNNI, F_AVX512BITALG): New.
6366
6367 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
6368
6369         PR target/83451
6370         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
6371         insn for floating-point loads and stores.
6372
6373 2018-03-14  Carl Love  <cel@us.ibm.com>
6374
6375         * config/rs6000/rs6000-c.c: Add macro definitions for
6376         ALTIVEC_BUILTIN_VEC_PERMXOR.
6377         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6378         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6379         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6380         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6381         UNSPEC_VPERMXOR.
6382         * config/doc/extend.texi: Add prototypes for vec_permxor.
6383
6384 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
6385
6386         PR c/84852
6387         * diagnostic-show-locus.c (class layout_point): Convert m_line
6388         from int to linenum_type.
6389         (line_span::comparator): Use linenum "compare" function when
6390         comparing line numbers.
6391         (test_line_span): New function.
6392         (layout_range::contains_point): Convert param "row" from int to
6393         linenum_type.
6394         (layout_range::intersects_line_p): Likewise.
6395         (layout::will_show_line_p): Likewise.
6396         (layout::print_source_line): Likewise.
6397         (layout::should_print_annotation_line_p): Likewise.
6398         (layout::print_annotation_line): Likewise.
6399         (layout::print_leading_fixits): Likewise.
6400         (layout::annotation_line_showed_range_p): Likewise.
6401         (struct line_corrections): Likewise for field m_row.
6402         (line_corrections::line_corrections): Likewise for param "row".
6403         (layout::print_trailing_fixits): Likewise.
6404         (layout::get_state_at_point): Likewise.
6405         (layout::get_x_bound_for_row): Likewise.
6406         (layout::print_line): Likewise.
6407         (diagnostic_show_locus): Likewise for locals "last_line" and
6408         "row".
6409         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6410         * input.c (selftest::test_linenum_comparisons): New function.
6411         (selftest::input_c_tests): Call it.
6412         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6413         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6414         * selftest.h (ASSERT_GT): New macro.
6415         (ASSERT_GT_AT): New macro.
6416         (ASSERT_LT): New macro.
6417         (ASSERT_LT_AT): New macro.
6418
6419 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
6420
6421         PR rtl-optimization/84780
6422         * combine.c (distribute_links): Don't make a link based on pc_rtx.
6423
6424 2018-03-14  Martin Liska  <mliska@suse.cz>
6425
6426         * tree.c (record_node_allocation_statistics): Use
6427         get_stats_node_kind.
6428         (get_stats_node_kind): New function extracted from
6429         record_node_allocation_statistics.
6430         (free_node): Use get_stats_node_kind.
6431
6432 2018-03-14  Richard Biener  <rguenther@suse.de>
6433
6434         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6435         that the value-set of ANTIC_IN doesn't grow.
6436
6437         Revert
6438         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6439         member.
6440         (BB_VISITED_WITH_VISITED_SUCCS): New define.
6441         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6442
6443 2018-03-14  Julia Koval  <julia.koval@intel.com>
6444
6445         * config.gcc (icelake-client, icelake-server): New.
6446         (icelake): Remove.
6447         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6448         (initial_ix86_arch_features): Ditto.
6449         (PTA_SKYLAKE): Add SGX.
6450         (PTA_ICELAKE): Remove.
6451         (PTA_ICELAKE_CLIENT): New.
6452         (PTA_ICELAKE_SERVER): New.
6453         (ix86_option_override_internal): Split up icelake on icelake client and
6454         icelake server.
6455         (get_builtin_code_for_version): Ditto.
6456         (fold_builtin_cpu): Ditto.
6457         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6458         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6459         * config/i386/i386.h (processor_type): Ditto.
6460         * doc/invoke.texi: Ditto.
6461
6462 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
6463
6464         PR sanitizer/83392
6465         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6466         INTEGER_CST offset, add it together with bitpos / 8 and
6467         sign extend based on POINTER_SIZE.
6468
6469         PR target/84844
6470         Revert
6471         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
6472
6473         PR target/78090
6474         * config/i386/constraints.md (Yc): New register constraint.
6475         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6476         Use Yc constraint for alternative 2 of operand 0.  Remove
6477         preferred_for_speed attribute.
6478
6479 2018-03-14  Richard Biener  <rguenther@suse.de>
6480
6481         PR tree-optimization/84830
6482         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6483         with the old one to avoid oscillations.
6484
6485 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
6486
6487         PR target/83712
6488         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6489         pseudos.
6490         (assign_by_spills): Return a flag of reload assignment failure.
6491         Do not process the reload assignment failures.  Do not spill other
6492         reload pseudos if they has the same reg class.  Update n if
6493         necessary.
6494         (lra_assign): Add a return arg.  Set up from the result of
6495         assign_by_spills call.
6496         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6497         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6498         usage_insns if it is not NULL.
6499         (spill_hard_reg_in_range): New function.
6500         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6501         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6502         function prototypes.
6503         (lra_assign): Change prototype.
6504         * lra.c (lra): Add code to deal with fails by splitting hard reg
6505         live ranges.
6506
6507 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
6508
6509         * config/riscv/riscv.opt (mrelax): New option.
6510         * config/riscv/riscv.c (riscv_file_start): Emit ".option
6511         "norelax" when riscv_mrelax is disabled.
6512         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6513
6514 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
6515
6516         PR target/84743
6517         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6518         reassociation for int modes.
6519
6520 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6521
6522         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6523         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6524         for big-endian.
6525         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6526         * config/aarch64/aarch64-sve.md
6527         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6528         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6529         (*extend<mode><Vwide>2): Rename to...
6530         (aarch64_sve_extend<mode><Vwide>2): ...this.
6531         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6532         renaming the old pattern to...
6533         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
6534         unsigned packs.
6535         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6536         define_expand, renaming the old pattern to...
6537         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6538         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6539         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6540         account when deciding which SVE instruction the optab should use.
6541         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6542
6543 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6544
6545         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6546         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6547         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6548         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6549         (tlsdesc_small_<mode>): Turn a define_expand and use
6550         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
6551         (tlsdesc_small_advsimd_<mode>): ...this.
6552         (tlsdesc_small_sve_<mode>): New pattern.
6553
6554 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6555
6556         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6557         (UNSPEC_UMUL_HIGHPART): New constants.
6558         (MUL_HIGHPART): New int iteraor.
6559         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6560         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6561         define_expand.
6562         (*<su>mul<mode>3_highpart): New define_insn.
6563
6564 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
6565
6566         PR lto/84805
6567         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6568         incomplete types.
6569
6570 2018-03-13  Martin Liska  <mliska@suse.cz>
6571
6572         PR ipa/84658.
6573         * (sem_item_optimizer::sem_item_optimizer): Initialize new
6574         vector.
6575         (sem_item_optimizer::~sem_item_optimizer): Release it.
6576         (sem_item_optimizer::merge_classes): Register variable aliases.
6577         (sem_item_optimizer::fixup_pt_set): New function.
6578         (sem_item_optimizer::fixup_points_to_sets): Likewise.
6579         * ipa-icf.h: Declare new variables and functions.
6580
6581 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
6582
6583         PR middle-end/84834
6584         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6585         integer_pow2p@2 and test integer_pow2p in condition.
6586         (A < 0 ? C : 0): Similarly for @1.
6587
6588         PR middle-end/84831
6589         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6590         characters starting at p contain '\0' character, don't look beyond
6591         that.
6592
6593         PR target/84827
6594         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6595         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6596
6597         PR target/84828
6598         * reg-stack.c (change_stack): Change update_end var from int to
6599         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6600         also call set_block_for_insn on the newly added insns and rescan.
6601
6602         PR target/84786
6603         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6604         on the last operand.
6605
6606         PR c++/84704
6607         * tree.c (stabilize_reference_1): Return save_expr (e) for
6608         STATEMENT_LIST even if it doesn't have side-effects.
6609
6610 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
6611
6612         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6613
6614 2018-03-12  Renlin Li  <renlin.li@arm.com>
6615
6616         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6617         aarch64_output_scalar_simd_mov_immediate.
6618
6619 2018-03-12  Martin Sebor  <msebor@redhat.com>
6620
6621         PR tree-optimization/83456
6622         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6623         for perfectly overlapping calls to memcpy.
6624         (gimple_fold_builtin_memory_chk): Same.
6625         (gimple_fold_builtin_strcpy): Handle no-warning.
6626         (gimple_fold_builtin_stxcpy_chk): Same.
6627         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6628
6629 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
6630
6631         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6632         parameter.  Use it for SFmode.
6633         (rs6000_function_arg_advance_1): Adjust.
6634         (rs6000_function_arg): Adjust.
6635         (rs6000_gimplify_va_arg): Pass false for that new parameter.
6636
6637 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
6638
6639         PR rtl-optimization/84169
6640         PR rtl-optimization/84780
6641         * combine.c (can_combine_p): Check for a 2-insn combination whether
6642         the destination register is used between the two insns, too.
6643
6644 2018-03-12  Richard Biener  <rguenther@suse.de>
6645
6646         PR tree-optimization/84803
6647         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6648         for refs DR analysis didn't process.
6649
6650 2018-03-12  Richard Biener  <rguenther@suse.de>
6651
6652         PR tree-optimization/84777
6653         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6654         force-vectorize loops ignore whether we are optimizing for size.
6655
6656 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6657
6658         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6659         (TARGET_MD_ASM_ADJUST): Define.
6660
6661 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
6662             Kito Cheng  <kito.cheng@gmail.com>
6663             Chung-Ju Wu  <jasonwucj@gmail.com>
6664
6665         * config/nds32/nds32.c (nds32_compute_stack_frame,
6666         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6667         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6668         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6669         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6670         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6671         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6672         * config/nds32/nds32.md (prologue, epilogue): Use macro
6673         NDS32_V3PUSH_AVAILABLE_P to do checking.
6674
6675 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
6676
6677         PR debug/58150
6678         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6679         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6680         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
6681         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
6682         addition of most attributes on !orig_type_die or the attribute not
6683         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6684
6685 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6686             Chung-Ju Wu  <jasonwucj@gmail.com>
6687
6688         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6689         __NDS32_VH__ macro.
6690         * config/nds32/nds32.opt (mvh): New option.
6691
6692 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6693             Chung-Ju Wu  <jasonwucj@gmail.com>
6694
6695         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6696         function.
6697         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6698         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6699         definition.
6700
6701 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6702             Chung-Ju Wu  <jasonwucj@gmail.com>
6703
6704         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6705         function.
6706         * config/nds32/nds32-multiple.md (strlensi): New pattern.
6707         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6708
6709 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
6710             Kito Cheng  <kito.cheng@gmail.com>
6711             Chung-Ju Wu  <jasonwucj@gmail.com>
6712
6713         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6714         UNSPEC_FFMISM and UNSPEC_FLMISM.
6715         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6716         for ffb, ffmism and flmism.
6717         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6718         (unspec_ffmism): Ditto.
6719         (unspec_flmism): Ditto.
6720         (nds32_expand_builtin_impl): Check if string extension is available.
6721         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6722         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6723
6724 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
6725
6726         Reverting patch:
6727         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6728
6729         PR target/83712
6730         * lra-assigns.c (assign_by_spills): Return a flag of reload
6731         assignment failure.  Do not process the reload assignment
6732         failures.  Do not spill other reload pseudos if they has the same
6733         reg class.
6734         (lra_assign): Add a return arg.  Set up from the result of
6735         assign_by_spills call.
6736         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6737         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6738         usage_insns if it is not NULL.
6739         (spill_hard_reg_in_range): New function.
6740         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6741         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6742         function prototypes.
6743         (lra_assign): Change prototype.
6744         * lra.c (lra): Add code to deal with fails by splitting hard reg
6745         live ranges.
6746
6747 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
6748
6749         PR target/84807
6750         * config/i386/i386.opt: Replace Enforcment with Enforcement.
6751
6752 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
6753
6754         PR debug/84620
6755         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6756         (dw_val_node): Add val_symbolic_view.
6757         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6758         (symview_upper_bound): New.
6759         (new_line_info_table): Initialize symviews_since_reset.
6760         (dwarf2out_source_line): Count symviews_since_reset and set
6761         symview_upper_bound.
6762         (dw_val_equal_p): Handle symview.
6763         (add_AT_symview): New.
6764         (print_dw_val): Handle symview.
6765         (attr_checksum, attr_checksum_ordered): Likewise.
6766         (same_dw_val_p, size_of_die): Likewise.
6767         (value_format, output_die): Likewise.
6768         (add_high_low_attributes): Use add_AT_symview for entry_view.
6769         (dwarf2out_finish): Reset symview_upper_bound, clear
6770         zero_view_p.
6771
6772 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
6773
6774         PR target/83969
6775         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6776         Add strict argument and use it.
6777         (rs6000_split_multireg_move): Update for new strict argument.
6778         (mem_operand_gpr): Disallow all non-offsettable addresses.
6779         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6780
6781 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
6782
6783         PR target/84772
6784         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6785         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6786         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6787
6788         PR c++/84767
6789         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6790         decl, use remap_type if we want to use the type.
6791
6792 2018-03-09  Martin Sebor  <msebor@redhat.com>
6793
6794         PR tree-optimization/84526
6795         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6796         Remove dead code.
6797         (builtin_access::generic_overlap): Be prepared to handle non-array
6798         base objects.
6799
6800 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
6801
6802         PR rtl-optimization/84682
6803         * lra-constraints.c (process_address_1): Check is_address flag
6804         for address constraints.
6805         (process_alt_operands): Likewise.
6806         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6807         preprocess_constraints.
6808         * recog.h (preprocess_constraints): Add oploc parameter.
6809         Adjust callers.
6810         * recog.c (preprocess_constraints): Test address_operand for
6811         CT_ADDRESS constraints.
6812
6813 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6814
6815         PR target/83712
6816         * lra-assigns.c (assign_by_spills): Return a flag of reload
6817         assignment failure.  Do not process the reload assignment
6818         failures.  Do not spill other reload pseudos if they has the same
6819         reg class.
6820         (lra_assign): Add a return arg.  Set up from the result of
6821         assign_by_spills call.
6822         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6823         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6824         usage_insns if it is not NULL.
6825         (spill_hard_reg_in_range): New function.
6826         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6827         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6828         function prototypes.
6829         (lra_assign): Change prototype.
6830         * lra.c (lra): Add code to deal with fails by splitting hard reg
6831         live ranges.
6832
6833 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6834
6835         PR target/83193
6836         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6837         Accept complain bool parameter.  Only emit errors if it is true.
6838         (arm_parse_cpu_option_name): Likewise.
6839         (arm_target_thumb_only): Adjust callers of the above.
6840         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6841         prototype to take a default true bool parameter.
6842         (arm_parse_arch_option_name): Likewise.
6843
6844 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
6845             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
6846
6847         PR jit/64089
6848         PR jit/84288
6849         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6850         * configure: Regenerate.
6851         * configure.ac ("linker --version-script option"): New.
6852         ("linker soname option"): New.
6853
6854 2018-03-09  Richard Biener  <rguenther@suse.de>
6855
6856         PR tree-optimization/84775
6857         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6858         immediate uses of predicate stmts and mark them modified.
6859
6860         Revert
6861         PR tree-optimization/84178
6862         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6863         to caller.
6864         (version_loop_for_if_conversion): Delay update_ssa call.
6865         (tree_if_conversion): Delay update_ssa until after predicate
6866         insertion.
6867
6868 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
6869
6870         PR target/84763
6871         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6872         when the function accesses prior frames.
6873
6874 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6875
6876         PR debug/84456
6877         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6878         gen_llsym, otherwise call maybe_gen_llsym.
6879
6880         PR inline-asm/84742
6881         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6882         has ',' character inside of it.
6883
6884 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6885
6886         PR target/84748
6887         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6888         as clobbering CC_REGNUM.
6889
6890 2018-03-08  Richard Biener  <rguenther@suse.de>
6891
6892         PR middle-end/84552
6893         * tree-scalar-evolution.c: Include tree-into-ssa.h.
6894         (follow_copies_to_constant): Do not follow SSA names registered
6895         for update.
6896
6897 2018-03-08  Richard Biener  <rguenther@suse.de>
6898
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 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
6907
6908         PR tree-optimization/84178
6909         * tree-if-conv.c (release_bb_predicate): Remove the
6910         the assertion that the stmts have NULL use_ops.
6911         Discard the statements, asserting that they haven't
6912         yet been added to a BB.
6913
6914 2018-03-08  Richard Biener  <rguenther@suse.de>
6915
6916         PR tree-optimization/84746
6917         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6918         (phi_translate): Pass in destination ANTIC_OUT set.
6919         (phi_translate_1): Likewise.  For a simplified result lookup
6920         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6921         (phi_translate_set): Adjust.
6922         (do_pre_regular_insertion): Likewise.
6923         (do_pre_partial_partial_insertion): Likewise.
6924
6925 2018-03-08  Martin Liska  <mliska@suse.cz>
6926
6927         PR gcov-profile/84735
6928         * doc/gcov.texi: Document usage of profile files.
6929         * gcov-io.h: Document changes in the format.
6930
6931 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
6932
6933         PR debug/84404
6934         PR debug/84408
6935         * dwarf2out.c (struct dw_line_info_table): Update comments for
6936         view == -1.
6937         (FORCE_RESET_NEXT_VIEW): New.
6938         (FORCE_RESETTING_VIEW_P): New.
6939         (RESETTING_VIEW_P): Check for -1 too.
6940         (ZERO_VIEW_P): Likewise.
6941         (new_line_info_table): Force-reset next view.
6942         (dwarf2out_begin_function): Likewise.
6943         (dwarf2out_source_line): Simplify zero_view_p initialization.
6944         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6945         view directly.  Omit view when omitting .loc at line 0.
6946
6947 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6948
6949         PR tree-optimization/84740
6950         * tree-switch-conversion.c (process_switch): Call build_constructors
6951         only if info.phi_count is non-zero.
6952
6953         PR tree-optimization/84739
6954         * tree-tailcall.c (find_tail_calls): Check call arguments against
6955         DECL_ARGUMENTS (current_function_decl) rather than
6956         DECL_ARGUMENTS (func) when checking for tail recursion.
6957
6958 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
6959
6960         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6961         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6962         Volker Reichelt's entry and add entries for people that perform
6963         GCC fuzzy testing and report numerous bugs.
6964
6965 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
6966
6967         PR target/82411
6968         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6969         readonly data in sdata, if that is disabled.
6970         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6971         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6972         -mreadonly-in-sdata option.
6973
6974 2018-03-07  Martin Sebor  <msebor@redhat.com>
6975
6976         PR tree-optimization/84468
6977         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6978         basic block when looking for nul assignment.
6979
6980 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
6981
6982         PR target/84277
6983         * except.h (output_function_exception_table): Adjust prototype.
6984         * except.c (output_function_exception_table): Remove FNNAME parameter
6985         and add SECTION parameter.  Ouput one part of the table at a time.
6986         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6987         the first part of the exception table and emit unwind directives.
6988         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6989         (i386_pe_seh_cold_init): Likewise.
6990         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6991         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6992         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6993         (ix86_output_call_insn): Emit a nop in one more case for SEH.
6994         * config/i386/winnt.c: Include except.h.
6995         (struct seh_frame_state): Add reg_offset, after_prologue and
6996         in_cold_section fields.
6997         (i386_pe_seh_end_prologue): Set seh->after_prologue.
6998         (i386_pe_seh_cold_init): New function.
6999         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7000         to seh->in_cold_section.
7001         (seh_emit_push): Record the offset of the push.
7002         (seh_emit_save): Record the offet of the save.
7003         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7004         Test seh->after_prologue to disregard the epilogue.
7005         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7006         (i386_pe_end_cold_function): New function.
7007
7008 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
7009
7010         PR fortran/84565
7011         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7012         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7013
7014         PR c++/84704
7015         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7016         on tmp_var.
7017         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7018         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7019
7020         PR middle-end/84723
7021         * multiple_target.c: Include tree-inline.h and intl.h.
7022         (expand_target_clones): Diagnose and fail if node->definition and
7023         !tree_versionable_function_p (node->decl).
7024
7025 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
7026
7027         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7028         sprint_ul.
7029         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7030         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7031         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7032
7033 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
7034
7035         PR target/84710
7036         * combine.c (try_combine): Use reg_or_subregno instead of handling
7037         just paradoxical SUBREGs and REGs.
7038
7039 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
7040
7041          * config/arc/arc.c (arc_finalize_pic): Remove function.
7042          (arc_must_save_register): We use single base PIC register, remove
7043          checks to save/restore the PIC register.
7044          (arc_expand_prologue): Likewise.
7045          * config/arc/arc-protos.h (arc_set_default_type_attributes):
7046          Remove.
7047          (arc_verify_short): Likewise.
7048          (arc_attr_type): Likewise.
7049          * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7050          (walk_stores): Likewise.
7051          (arc_address_cost): Make it static.
7052          (arc_verify_short): Likewise.
7053          (branch_dest): Likewise.
7054          (arc_attr_type): Likewise.
7055          * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7056          (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7057          (arc_final_prescan_insn): Remove inserting the nops due to
7058          hardware hazards.  It is done in reorg step.
7059          (insn_length_variant_t): Remove.
7060          (insn_length_parameters_t): Likewise.
7061          (arc_insn_length_parameters): Likewise.
7062          (arc_get_insn_variants): Likewise.
7063          * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7064
7065 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
7066
7067         PR inline-asm/84683
7068         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7069         assertion failure.
7070
7071         PR tree-optimization/84687
7072         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7073         on new_node->decl.
7074         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7075
7076 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7077
7078         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7079         Rename to ppc_speculation_barrier.
7080         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7081         __builtin_ppc_speculation_barrier.
7082
7083 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
7084
7085         PR target/84700
7086         * combine.c (combine_simplify_rtx): Don't try to simplify if
7087         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7088         are equal to x.
7089
7090 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
7091
7092         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7093         to 32 bytes when compiling for POWER9.
7094
7095 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
7096
7097         PR target/84564
7098         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7099         regparm >= 3 with no arg reg available also for calls with
7100         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
7101
7102         PR target/84524
7103         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7104         orig,vex.
7105         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
7106
7107 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
7108
7109         PR target/84264
7110         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7111
7112 2018-03-05  Richard Biener  <rguenther@suse.de>
7113
7114         PR tree-optimization/84486
7115         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7116         When inserting a __builtin_assume_aligned call set the LHS
7117         SSA name alignment info accordingly.
7118
7119 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
7120
7121         PR tree-optimization/84114
7122         * config/aarch64/aarch64.c (aarch64_reassociation_width)
7123         Avoid reassociation of FLOAT_MODE addition.
7124
7125 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
7126
7127         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7128         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7129         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7130         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7131         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7132         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7133         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7134         and -mwbnoinvd.
7135         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7136         __builtin_ia32_wbinvd): New builtins.
7137         (SPECIAL_ARGS2): New.
7138         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7139         (SPECIAL_ARGS2): New.
7140         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7141         (ix86_valid_target_attribute_inner_p): Ditto.
7142         (ix86_init_mmx_sse_builtins): Add special_args2.
7143         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7144         TARGET_WBNOINVD_P): New.
7145         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7146         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7147         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7148         * config/i386/immintrin.h (_wbinvd): New intrinsic.
7149         * config/i386/pconfigintrin.h: New file.
7150         * config/i386/wbnoinvdintrin.h: Ditto.
7151         * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7152         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7153
7154 2018-03-05  Richard Biener  <rguenther@suse.de>
7155
7156         PR tree-optimization/84670
7157         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7158         member.
7159         (BB_VISITED_WITH_VISITED_SUCCS): New define.
7160         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7161         (compute_antic_aux): Only assert the number of values in ANTIC_IN
7162         doesn't grow if all successors (recursively) were visited at least
7163         once.
7164
7165 2018-03-05  Richard Biener  <rguenther@suse.de>
7166
7167         PR tree-optimization/84650
7168         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7169         if executed in the loop pipeline.
7170
7171 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
7172
7173         * doc/configfiles.texi (Configuration Files): Move info about
7174         conditionalizing $target-protos.h to...
7175         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
7176         differs from $target-protos.h.
7177
7178 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
7179             Chung-Ju Wu  <jasonwucj@gmail.com>
7180
7181         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7182         * config/nds32/nds32-multiple.md (setmemsi): Define.
7183         * config/nds32/nds32-memory-manipulation.c
7184         (nds32_gen_dup_4_byte_to_word_value): New.
7185         (emit_setmem_word_loop): New.
7186         (emit_setmem_byte_loop): New.
7187         (nds32_expand_setmem_loop): New.
7188         (nds32_expand_setmem_loop_v3m): New.
7189         (nds32_expand_setmem_unroll): New.
7190         (nds32_expand_setmem): New.
7191
7192 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7193             Chung-Ju Wu  <jasonwucj@gmail.com>
7194
7195         * config/nds32/nds32-memory-manipulation.c
7196         (nds32_emit_load_store): New.
7197         (nds32_emit_post_inc_load_store): New.
7198         (nds32_emit_mem_move): New.
7199         (nds32_emit_mem_move_block): New.
7200         (nds32_expand_movmemsi_loop_unknown_size): New.
7201         (nds32_expand_movmemsi_loop_known_size): New.
7202         (nds32_expand_movmemsi_loop): New.
7203         (nds32_expand_movmemsi_unroll): New.
7204         (nds32_expand_movmemqi): Rename ...
7205         (nds32_expand_movmemsi): ... to this.
7206         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7207         (movmemsi): ... to this.
7208         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7209         (nds32_expand_movmemsi): ... to this.
7210
7211 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7212             Monk Chiang  <sh.chiang04@gmail.com>
7213             Chung-Ju Wu  <jasonwucj@gmail.com>
7214
7215         * config/nds32/nds32-protos.h
7216         (nds32_expand_load_multiple): New arguments.
7217         (nds32_expand_store_multiple): Ditto.
7218         (nds32_valid_multiple_load_store): Rename ...
7219         (nds32_valid_multiple_load_store_p): ... to this.
7220         * config/nds32/nds32-memory-manipulation.c
7221         (nds32_expand_load_multiple): Refine implementation.
7222         (nds32_expand_store_multiple): Ditto.
7223         * config/nds32/nds32-multiple.md
7224         (load_multiple): Update nds32_expand_load_multiple interface.
7225         (store_multiple): Update nds32_expand_store_multiple interface.
7226         * config/nds32/nds32-predicates.c
7227         (nds32_valid_multiple_load_store): Rename ...
7228         (nds32_valid_multiple_load_store_p): ... to this and refine
7229         implementation.
7230         * config/nds32/predicates.md
7231         (nds32_load_multiple_and_update_address_operation): New predicate.
7232         (nds32_store_multiple_and_update_address_operation): New predicate.
7233
7234 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7235             Chung-Ju Wu  <jasonwucj@gmail.com>
7236
7237         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7238         (combo): New attribute.
7239         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7240
7241 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
7242
7243         * config/nds32/nds32.opt: Change -mcmodel= default value.
7244
7245 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
7246             Monk Chiang  <sh.chiang04@gmail.com>
7247             Chung-Ju Wu  <jasonwucj@gmail.com>
7248
7249         * config/nds32/constants.md (unspec_element): New enum.
7250         * config/nds32/constraints.md (Umw): New constraint.
7251         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7252         * config/nds32/nds32-intrinsic.md: Likewise.
7253         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7254         (nds32_valid_smw_lwm_base_p): New.
7255         (nds32_output_smw_single_word): New.
7256         (nds32_output_lmw_single_word): New.
7257         (nds32_expand_unaligned_load): New.
7258         (nds32_expand_unaligned_store): New.
7259         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7260         (nds32_output_smw_single_word): Declare.
7261         (nds32_output_lmw_single_word): Declare.
7262         (nds32_expand_unaligned_load): Declare.
7263         (nds32_expand_unaligned_store): Declare.
7264         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7265         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7266         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7267         NDS32_BUILTIN_UASTORE_DW.
7268         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7269         predicate.
7270
7271 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
7272             Kito Cheng  <kito.cheng@gmail.com>
7273             Chung-Ju Wu  <jasonwucj@gmail.com>
7274
7275         * config/nds32/nds32-intrinsic.c
7276         (nds32_expand_builtin_null_ftype_reg): Delete.
7277         (nds32_expand_builtin_reg_ftype_imm): Ditto.
7278         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7279         (nds32_read_argument): New.
7280         (nds32_legitimize_target): Ditto.
7281         (nds32_legitimize_argument): Ditto.
7282         (nds32_check_constant_argument): Ditto.
7283         (nds32_expand_unop_builtin): Ditto.
7284         (nds32_expand_unopimm_builtin): Ditto.
7285         (nds32_expand_binop_builtin): Ditto.
7286         (nds32_builtin_decl_impl): Ditto.
7287         (builtin_description): Ditto.
7288         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7289         (nds32_init_builtins_impl): Ditto.
7290         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7291         (nds32_builtin_decl): New.
7292         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7293         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7294
7295 2018-03-02  Jeff Law  <law@redhat.com>
7296
7297         * reorg.c (stop_search_p): Handle DEBUG_INSN.
7298         (redundant_insn, fill_simple_delay_slots): Likewise.
7299         (fill_slots_from_thread): Likewise.
7300         * resource.c (mark_referenced_resources): Likewise.
7301         (mark_set_resources, find_dead_or_set_registers): Likewise.
7302
7303 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7304
7305         * substring-locations.h (format_warning_va): Formatting fix for
7306         ATTRIBUTE_GCC_DIAG.
7307         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7308         argument.
7309         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7310         * substring-locations.c: Include intl.h.
7311         (format_warning_va): Turned into small wrapper around
7312         format_warning_n_va, renamed to ...
7313         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7314         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7315         use ngettext.
7316         (format_warning_at_substring_n): New function.
7317         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7318         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
7319         format_warning_at_substring with just a shorter name instead of
7320         const function pointer.
7321         (fmtwarn_n): New function.
7322         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7323         appropriate, get rid of all the fmtstr temporaries, move conditionals
7324         with G_() wrapped string literals directly into fmtwarn arguments,
7325         cast dir.len to (int), formatting fixes.
7326
7327 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
7328
7329         * doc/invoke.texi: Remove "Cilk Plus" references.
7330
7331 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7332             Richard Biener  <rguenther@suse.de>
7333
7334         PR ipa/84628
7335         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7336         for error or warning attributes if CALL_FROM_THUNK_P is set.
7337         Formatting fixes.
7338
7339 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7340
7341         PR target/56540
7342         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7343         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7344
7345         PR target/56540
7346         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7347         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7348
7349         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7350         instead of -1U in last predictors element's probability member.
7351
7352 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
7353
7354         PR ipa/83983
7355         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7356         arguments if they are comparable.
7357
7358 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
7359
7360         PR tree-optimization/84634
7361         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7362         masks and masked_loop_p with a single loop_masks, making sure it's
7363         null for bb vectorization.
7364
7365 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
7366
7367         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7368         (vect_analyze_data_ref_access): Use loop->safe_len rather than
7369         loop->force_vectorize to check whether there is no alias.
7370
7371 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7372
7373         PR target/84614
7374         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7375         prototypes.
7376         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7377         comments.
7378         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7379         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7380         instead of a loop around prev_real_insn.
7381         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7382         prev_real_insn.
7383
7384         PR inline-asm/84625
7385         * config/i386/i386.c (ix86_print_operand): Use conditional
7386         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7387         zero vector.
7388
7389 2018-03-02  Richard Biener  <rguenther@suse.de>
7390
7391         PR tree-optimization/84427
7392         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7393         (bitmap_set_subtract_values): Rewrite to handle multiple
7394         exprs per value.
7395         (clean): Likewise.
7396         (prune_clobbered_mems): Likewise.
7397         (phi_translate): Take edge instead of pred/phiblock.
7398         (phi_translate_1): Likewise.
7399         (phi_translate_set): Likewise.  Insert all translated
7400         exprs for a value into the set, keeping possibly multiple
7401         expressions per value.
7402         (compute_antic_aux): Adjust for phi_translate changes.
7403         When intersecting union the expressions and prune those
7404         not in the final value set, keeping possibly multiple
7405         expressions per value.  Do not use value-insertion
7406         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7407         all expressions.  Add verification that the value-sets
7408         only shrink during iteration.
7409         (compute_partial_antic_aux): Adjust for the phi_translate changes.
7410         (do_pre_regular_insertion): Likewise.
7411         (do_pre_partial_partial_insertion): Likewise.
7412
7413 2018-03-02  Richard Biener  <rguenther@suse.de>
7414
7415         PR target/82005
7416         * config/darwin.c (saved_debug_info_level): New static global.
7417         (darwin_asm_lto_start): Disable debug info generation for LTO out.
7418         (darwin_asm_lto_end): Restore debug info generation settings.
7419
7420 2018-03-01  Martin Liska  <mliska@suse.cz>
7421
7422         PR sanitizer/82484
7423         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7424         volatile arguments.
7425
7426 2018-03-01  Richard Biener  <rguenther@suse.de>
7427
7428         PR debug/84645
7429         * dwarf2out.c (gen_variable_die): Properly handle late VLA
7430         type annotation with LTO when debug was disabled at compile-time.
7431
7432 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
7433
7434         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7435         XINT with INTVAL.
7436         (mips_final_postscan_insn): Likewise.
7437
7438 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
7439
7440         PR rtl-optimization/84528
7441         * alias.c (init_alias_target): Add commentary.
7442         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7443         a unique base value if the frame pointer is not eliminated
7444         to the stack pointer.
7445
7446 2018-03-01  Tom de Vries  <tom@codesourcery.com>
7447
7448         PR rtl-optimization/83327
7449         * lra-int.h (hard_regs_spilled_into): Declare.
7450         * lra.c (hard_regs_spilled_into): Define.
7451         (init_reg_info): Init hard_regs_spilled_into.
7452         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7453         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7454         (process_bb_lives): Handle hard_regs_spilled_into.
7455         (lra_create_live_ranges_1): Before doing liveness propagation, clear
7456         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7457
7458 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
7459
7460         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7461         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7462         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7463         * config/rs6000/aix72.h: New file.
7464
7465 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
7466
7467         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7468         instead of warning_at with conditional singular and plural messages
7469         where possible.
7470
7471         PR target/52991
7472         * stor-layout.c (update_alignment_for_field): For
7473         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7474         && !DECL_PACKED (field), do the alignment update, just use
7475         only desired_align instead of MAX (type_align, desired_align)
7476         as the alignment.
7477         (place_field): Don't do known_align < desired_align handling
7478         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7479         is non-NULL, instead do it after rli->prev_field handling and
7480         only if not within a bitfield word.  For DECL_PACKED (field)
7481         use type_align of BITS_PER_UNIT.
7482
7483 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
7484
7485         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7486         superfluous parentheses and trailing spaces.
7487
7488 2018-02-28  Richard Biener  <rguenther@suse.de>
7489
7490         PR tree-optimization/84584
7491         * graphite-scop-detection.c (scop_detection::add_scop): Discard
7492         SCoPs with fake exit edge.
7493
7494 2018-02-28  Martin Liska  <mliska@suse.cz>
7495
7496         PR testsuite/84597
7497         * timevar.c (timer::print): Fix format to properly print 100%
7498         values.
7499
7500 2018-02-28  Richard Biener  <rguenther@suse.de>
7501
7502         PR middle-end/84607
7503         * genmatch.c (capture_info::walk_match): Do not mark
7504         captured expressions without operands as expr_p given
7505         they act more like predicates and should be subject to
7506         "lost tail" side-effect preserving.
7507
7508 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
7509
7510         PR rtl-optimization/81611
7511         * auto-inc-dec.c (attempt_change): Move dead note from
7512         mem_insn if it's the next use of regno
7513         (find_address): Take address use of reg holding
7514         non-incremented value.  Add parm to limit search to the named
7515         reg only.
7516         (merge_in_block): Attempt to use a mem insn that is the next
7517         use of the original regno.
7518
7519 2018-02-27  Martin Sebor  <msebor@redhat.com>
7520
7521         PR c++/83871
7522         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7523         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7524
7525 2018-02-27  Martin Sebor  <msebor@redhat.com>
7526
7527         PR translation/84207
7528         * diagnostic-core.h (warning_n, error_n, inform_n): Change
7529         n argument to unsigned HOST_WIDE_INT.
7530         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7531         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
7532         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7533         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7534
7535 2018-02-27  Richard Biener  <rguenther@suse.de>
7536
7537         PR tree-optimization/84512
7538         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7539         Do not use the estimate returned from record_stmt_cost for
7540         the scalar iteration cost but sum properly using add_stmt_cost.
7541
7542 2018-02-27  Richard Biener  <rguenther@suse.de>
7543
7544         PR tree-optimization/84466
7545         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7546         Adjust last change to less strictly validate use operands.
7547
7548 2018-02-27  Martin Liska  <mliska@suse.cz>
7549
7550         PR gcov-profile/84548
7551         * gcov.c (process_file): Allow partial overlap and consider it
7552         also as group functions.
7553         (output_lines): Properly calculate range of lines for a group.
7554
7555 2018-02-27  Martin Liska  <mliska@suse.cz>
7556
7557         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7558         'ggc' suffixes.  Change first column width.
7559         (timer::print): Fix formatting of the column.
7560
7561 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
7562
7563         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7564         preserve inline entry blocks for the sake of debug inline
7565         entry point markers alone.
7566         (remove_unused_locals): Suggest in comments a better place to
7567         force the preservation of inline entry blocks that are
7568         otherwise unused, but do not preserve them.
7569
7570 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7571
7572         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7573
7574 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7575
7576         PR target/84039
7577         * config/i386/constraints.md (Bs): Replace
7578         ix86_indirect_branch_register with
7579         TARGET_INDIRECT_BRANCH_REGISTER.
7580         (Bw): Likewise.
7581         * config/i386/i386.md (indirect_jump): Likewise.
7582         (tablejump): Likewise.
7583         (*sibcall_memory): Likewise.
7584         (*sibcall_value_memory): Likewise.
7585         Peepholes of indirect call and jump via memory: Likewise.
7586         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7587         (*sibcall_value_GOT_32): Likewise.
7588         * config/i386/predicates.md (indirect_branch_operand): Likewise.
7589         (GOT_memory_operand): Likewise.
7590         (call_insn_operand): Likewise.
7591         (sibcall_insn_operand): Likewise.
7592         (GOT32_symbol_operand): Likewise.
7593         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7594
7595 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
7596
7597         PR rtl-optimization/83496
7598         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7599         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
7600         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7601         redundant insn, if any.
7602         (relax_delay_slots): Likewise.
7603         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7604
7605 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
7606
7607         PR tree-optimization/83965
7608         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7609         that grouped statements are part of a reduction chain.  Return
7610         true if the statement is not marked as a reduction itself but
7611         is part of a group.
7612         (vect_recog_dot_prod_pattern): Don't check whether the statement
7613         is part of a group here.
7614         (vect_recog_sad_pattern): Likewise.
7615         (vect_recog_widen_sum_pattern): Likewise.
7616
7617 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
7618
7619         PR debug/84545
7620         * final.c (rest_of_clean_state): Also look for calls inside sequences.
7621
7622 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7623
7624         PR target/84530
7625         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7626         the bool argument.
7627         (ix86_output_indirect_function_return): New prototype.
7628         (ix86_split_simple_return_pop_internal): Likewise.
7629         * config/i386/i386.c (indirect_return_via_cx): New.
7630         (indirect_return_via_cx_bnd): Likewise.
7631         (indirect_thunk_name): Handle return va CX_REG.
7632         (output_indirect_thunk_function): Create alias for
7633         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7634         (ix86_output_indirect_jmp): Remove the bool argument.
7635         (ix86_output_indirect_function_return): New function.
7636         (ix86_split_simple_return_pop_internal): Likewise.
7637         * config/i386/i386.md (*indirect_jump): Don't pass false
7638         to ix86_output_indirect_jmp.
7639         (*tablejump_1): Likewise.
7640         (simple_return_pop_internal): Change it to define_insn_and_split.
7641         Call ix86_split_simple_return_pop_internal to split it for
7642         -mfunction-return=.
7643         (simple_return_indirect_internal): Call
7644         ix86_output_indirect_function_return instead of
7645         ix86_output_indirect_jmp.
7646
7647 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
7648
7649         PR bootstrap/84405
7650         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7651         memset and value initialization afterwards.
7652
7653 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
7654
7655         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7656
7657 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
7658
7659         PR target/84521
7660         * common/config/aarch64/aarch64-common.c
7661         (aarch_option_optimization_table[]): Switch
7662         off fomit-frame-pointer
7663
7664 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
7665             Chung-Ju Wu  <jasonwucj@gmail.com>
7666
7667         * config/nds32/nds32-multiple.md (load_multiple): Disallow
7668         volatile memory.
7669         (store_multiple): Ditto.
7670
7671 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
7672
7673         * config.gcc: Add --with-cpu support for nds32 target.
7674         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7675         * config/nds32/nds32.opt: Add -mcpu= option.
7676
7677 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
7678
7679         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7680         isel=yes): Warn for these deprecated options.
7681
7682 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
7683
7684         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7685         ISA_2_5_MASKS_EMBEDDED.
7686
7687 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
7688
7689         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7690         p->max as pointers rather than using iterative_hash_expr.
7691
7692 2018-02-23  Carl Love  <cel@us.ibm.com>
7693
7694         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7695         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7696         BU_P8V_OVERLOAD_2.
7697         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7698         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7699         P8V_BUILTIN_VEC_VUNSIGNED2.
7700
7701 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
7702
7703         PR target/81572
7704         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7705         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7706         LRA_UNKNOWN_ALT.
7707         * lra-constraints.c (curr_insn_transform): Set up
7708         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
7709         LRA_UNKNOWN_ALT.
7710         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7711         * lra-eliminations.c (spill_pseudos): Ditto.
7712         (process_insn_for_elimination): Ditto.
7713         * lra-lives.c (reg_early_clobber_p): Use the new macros.
7714         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7715         LRA_NON_CLOBBERED_ALT.
7716
7717 2018-02-22  Martin Sebor  <msebor@redhat.com>
7718
7719         PR tree-optimization/84480
7720         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7721         to maybe_diag_stxncpy_trunc.  Call it.
7722         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7723         from gimple_fold_builtin_strcpy.  Print inlining stack.
7724         (handle_builtin_stxncpy): Print inlining stack.
7725         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7726
7727 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
7728
7729         PR target/84176
7730         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7731         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7732         and -fcheck-pointer-bounds are used together.
7733         (indirect_thunk_prefix): New enum.
7734         (indirect_thunk_need_prefix): New function.
7735         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
7736         "_nt" instead of "_bnd" for NOTRACK prefix.
7737         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7738         (output_indirect_thunk_function): Likewise.
7739         (): Likewise.
7740         (ix86_code_end): Update output_indirect_thunk_function calls.
7741         (ix86_output_indirect_branch_via_reg): Replace
7742         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7743         (ix86_output_indirect_branch_via_push): Likewise.
7744         (ix86_output_function_return): Likewise.
7745         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7746         incompatible with -fcf-protection=branch and
7747         -fcheck-pointer-bounds.
7748
7749 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
7750
7751         PR target/83335
7752         * config/aarch64/aarch64.c (aarch64_print_address_internal):
7753         Change gcc_assert call to output_operand_lossage.
7754
7755 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
7756
7757         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7758
7759 2018-02-22  DJ Delorie  <dj@redhat.com>
7760             Sebastian Perta  <sebastian.perta@renesas.com>
7761             Oleg Endo  <olegendo@gcc.gnu.org>
7762
7763         * config/rx/rx.c (rx_rtx_costs): New function.
7764         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7765
7766 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7767
7768         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7769
7770 2018-02-22  Martin Liska  <mliska@suse.cz>
7771
7772         PR driver/83193
7773         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7774         Add "native" as a possible value.
7775
7776 2018-02-22  Martin Liska  <mliska@suse.cz>
7777
7778         PR driver/83193
7779         * config/i386/i386.c (ix86_option_override_internal):
7780         Add "native" as a possible value for -march and -mtune.
7781
7782 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
7783
7784         PR target/84502
7785         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7786         to all type variants.
7787
7788         PR tree-optimization/84503
7789         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7790         width as info->bitpos + info->bitsize - start.
7791         (merged_store_group::merge_overlapping): Simplify width computation.
7792         (check_no_overlap): New function.
7793         (imm_store_chain_info::try_coalesce_bswap): Compute expected
7794         start + width and last_order of the group, fail if check_no_overlap
7795         fails.
7796         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7797         to group if check_no_overlap fails.
7798
7799 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7800
7801         * config/rs6000/altivec.md: Delete contraint arguments to
7802         define_expand, define_split, and define_peephole2, and in
7803         define_insn_and_split if always unused.
7804         * config/rs6000/darwin.md: Ditto.
7805         * config/rs6000/dfp.md: Ditto.
7806         * config/rs6000/rs6000.md: Ditto.
7807         * config/rs6000/sync.md: Ditto.
7808         * config/rs6000/vector.md: Ditto.
7809         * config/rs6000/vsx.md: Ditto.
7810
7811 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7812
7813         * config/rs6000/altivec.md: Write output control strings as braced
7814         blocks instead of double-quoted strings.
7815         * config/rs6000/darwin.md: Ditto.
7816         * config/rs6000/rs6000.md: Ditto.
7817         * config/rs6000/vector.md: Ditto.
7818         * config/rs6000/vsx.md: Ditto.
7819
7820 2018-02-21  Jason Merrill  <jason@redhat.com>
7821
7822         PR c++/84314 - ICE with templates and fastcall attribute.
7823         * attribs.c (build_type_attribute_qual_variant): Remove assert.
7824
7825 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7826
7827         * ipa-cp.c (determine_versionability): Fix comment typos.
7828
7829 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7830
7831         PR c/84229
7832         * ipa-cp.c (determine_versionability): Do not version functions caling
7833         va_arg_pack.
7834
7835 2018-02-21  Martin Liska  <mliska@suse.cz>
7836
7837         PR driver/83193
7838         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7839         Add "native" as a possible value.
7840         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
7841         the macro when native cpu detection is available.
7842
7843 2018-02-21  Martin Liska  <mliska@suse.cz>
7844
7845         PR driver/83193
7846         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7847         Add "native" as a possible value.
7848         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7849         when native cpu detection is available.
7850
7851 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
7852             Martin Sebor  <msebor@redhat.com>
7853
7854         PR tree-optimization/84478
7855         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7856         false.
7857         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7858         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7859         support which is conservatively correct, for 2 only stay conservative
7860         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
7861         argument to the 2 argument get_range_strlen, adjust 6 arg
7862         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7863         false.
7864         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7865         (gimple_fold_builtin_strlen): Pass true as last argument to
7866         get_range_strlen.
7867
7868 2018-02-20  Martin Sebor  <msebor@redhat.com>
7869
7870         PR middle-end/84095
7871         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7872         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
7873         (builtin_memref::builtin_memref): Factor out parts into
7874         set_base_and_offset and call it.
7875
7876 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7877
7878         PR middle-end/84406
7879         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7880         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7881         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
7882         search at the associated MODE_INT.
7883
7884 2018-02-20  Jeff Law  <law@redhat.com>
7885
7886         PR middle-end/82123
7887         PR tree-optimization/81592
7888         PR middle-end/79257
7889         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7890         for range data rather than using global data.
7891         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7892         range data rather than using global data.
7893         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7894         pass it to children as needed.
7895         (struct directive::fmtresult): Similarly.
7896         (struct directive::set_width): Similarly.
7897         (struct directive::set_precision): Similarly.
7898         (format_integer, format_directive, parse_directive): Similarly.
7899         (format_none): Accept unnamed vr_values parameter.
7900         (format_percent, format_floating, format_character): Similarly.
7901         (format_string, format_plain): Similarly.
7902         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7903         the EVRP range analyzer for range data rather than using global data.
7904         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7905         gimple-ssa-evrp-analyze.h
7906         (class sprintf_dom_walker): Add after_dom_children member function.
7907         Add evrp_range_analyzer member.
7908         (sprintf_dom_walker::before_dom_children): Call into the EVRP
7909         range analyzer as needed.
7910         (sprintf_dom_walker::after_dom_children): New member function.
7911         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7912         if not optimizing.
7913         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7914         (evrp_range_analyzer::pop_to_marker): Likewise.
7915
7916 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7917
7918         PR tree-optimization/84419
7919         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7920         with the required type if its current type is compatible but
7921         different.
7922
7923 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
7924
7925         PR middle-end/82004
7926         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7927         after vectorization.
7928
7929 2018-02-20  Martin Liska  <mliska@suse.cz>
7930
7931         PR driver/83193
7932         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7933         possible values if we don't have a hint.
7934
7935 2018-02-20  Martin Liska  <mliska@suse.cz>
7936
7937         PR c/84310
7938         PR target/79747
7939         * final.c (shorten_branches): Build align_tab array with one
7940         more element.
7941         * opts.c (finish_options): Add alignment option limit check.
7942         (MAX_CODE_ALIGN): Likewise.
7943         (MAX_CODE_ALIGN_VALUE): Likewise.
7944         * doc/invoke.texi: Document maximum allowed option value for
7945         all -falign-* options.
7946
7947 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
7948
7949         PR target/84146
7950         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7951         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7952         * var-tracking.c (emit_note_insn_var_location): Remove all references
7953         to NOTE_INSN_CALL_ARG_LOCATION.
7954         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7955         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7956         Use copy_rtx_if_shared.
7957         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7958         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7959         (dwarf2out_var_location): Remove handling of
7960         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7961         on call_insn.
7962         * final.c (final_scan_insn): Remove all references to
7963         NOTE_INSN_CALL_ARG_LOCATION.
7964         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
7965         before dumping final insns.
7966         * except.c (emit_note_eh_region_end): Remove all references to
7967         NOTE_INSN_CALL_ARG_LOCATION.
7968         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7969         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7970         * config/arc/arc.c (hwloop_optimize): Likewise.
7971         * config/arm/arm.c (create_fix_barrier): Likewise.
7972         * config/s390/s390.c (s390_chunkify_start): Likewise.
7973         * config/sh/sh.c (find_barrier): Likewise.
7974         * config/i386/i386.c (rest_of_insert_endbranch,
7975         ix86_seh_fixup_eh_fallthru): Likewise.
7976         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7977         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7978         * config/frv/frv.c (frv_function_prologue): Likewise.
7979         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
7980         reg note.
7981         (note_outside_basic_block_p): Remove all references to
7982         NOTE_INSN_CALL_ARG_LOCATION.
7983         * gengtype.c (adjust_field_rtx_def): Likewise.
7984         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7985         Likewise.
7986         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7987         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7988
7989         PR c++/84444
7990         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7991         is ADDR_EXPR.
7992
7993         PR tree-optimization/84452
7994         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7995         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7996         is NULL.
7997
7998 2018-02-19  Martin Liska  <mliska@suse.cz>
7999
8000         PR sanitizer/82183
8001         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8002
8003 2018-02-19  Martin Liska  <mliska@suse.cz>
8004             Richard Sandiford  <richard.sandiford@linaro.org>
8005
8006         PR tree-optimization/82491
8007         * gimple-fold.c (get_base_constructor): Make earlier bail out
8008         to prevent ubsan.
8009
8010 2018-02-19  Carl Love  <cel@us.ibm.com>
8011
8012         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8013         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8014         BU_P8V_OVERLOAD_1.
8015         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8016         P8V_BUILTIN_VEC_NEG.
8017
8018 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
8019
8020         * config/rl78/rl78.md (movdf): New define expand.
8021
8022 2018-02-19  Martin Liska  <mliska@suse.cz>
8023
8024         PR other/80589
8025         * doc/invoke.texi: Fix typo.
8026         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8027
8028 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
8029
8030         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8031         handle rs6000_single_float and rs6000_double_float specially for
8032         e500 family CPUs.
8033
8034 2018-02-16  Jeff Law  <law@redhat.com>
8035
8036         * config/rx/rx.c (add_pop_cfi_notes): New function.;
8037         (pop_regs): Use it.
8038
8039 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
8040
8041         PR ipa/84425
8042         * ipa-inline.c (inline_small_functions): Fix a typo.
8043
8044 2018-02-16  Nathan Sidwell  <nathan@acm.org>
8045
8046         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8047
8048 2018-02-16  Carl Love  <cel@us.ibm.com>
8049
8050         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8051         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8052         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8053         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8054         expansion to P8V_BUILTIN_VEC_FLOAT2.
8055
8056 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
8057
8058         PR rtl-optimization/70023
8059         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8060         src_regno into account.
8061
8062 2018-02-16  Carl Love  <cel@us.ibm.com>
8063
8064         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8065         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8066         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8067         * config/rs6000/rs6000.c: Remove case statements for
8068         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8069         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8070         and P9V_BUILTIN_VEC_VINSERT4B.
8071         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8072         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8073         * config/rs6000/vsx.md:
8074         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8075         vec_insert4b.
8076
8077 2018-02-16  Carl Love  <cel@us.ibm.com>
8078
8079         * config/rs6000/altivec.h: Add builtin names vec_extract4b
8080         vec_insert4b.
8081         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8082         definitions.
8083         * config/rs6000/rs6000-c.c: Add the definitions for
8084         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8085         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8086         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8087         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
8088         definition for insert4b and define insn *insert3b_internal.
8089         * doc/extend.texi: Add documentation for vec_extract4b.
8090
8091 2018-02-16  Nathan Sidwell  <nathan@acm.org>
8092
8093         * doc/extend.texi (Backwards Compatibility): Mention friend
8094         injection.  Note for-scope is deprecated.
8095         * doc/invoke.texi (-ffriend-injection): Deprecate.
8096
8097 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
8098
8099         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8100         that moved to I2, also allow destinations that are a paradoxical
8101         subreg (instead of a normal reg).
8102
8103 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
8104
8105         PR target/83831
8106         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8107         to QImode.
8108
8109 2018-02-16  Richard Biener  <rguenther@suse.de>
8110
8111         PR tree-optimization/84037
8112         PR tree-optimization/84016
8113         PR target/82862
8114         * config/i386/i386.c (ix86_builtin_vectorization_cost):
8115         Adjust vec_construct for the fact we need additional higher latency
8116         128bit inserts for AVX256 and AVX512 vector builds.
8117         (ix86_add_stmt_cost): Scale vector construction cost for
8118         elementwise loads.
8119
8120 2018-02-16  Richard Biener  <rguenther@suse.de>
8121
8122         PR tree-optimization/84417
8123         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8124         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8125         (non_rewritable_lvalue_p): Likewise, use poly-ints.
8126
8127 2018-02-16  Martin Liska  <mliska@suse.cz>
8128
8129         PR sanitizer/84307
8130         * internal-fn.def (ASAN_CHECK): Set proper flags.
8131         (ASAN_MARK): Likewise.
8132
8133 2018-02-16  Julia Koval  <julia.koval@intel.com>
8134
8135         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8136         from PTA_CANNONLAKE.
8137
8138 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
8139
8140         PR target/84272
8141         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8142         Use ++iter rather than iter++ for std::list iterators.
8143         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
8144         defer deleting them until all nodes in the forest are processed.  Do
8145         free even leaf nodes.  Change to_process into auto_vec.
8146
8147         PR bootstrap/84405
8148         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8149         * vec.h (vec_default_construct): Use memset instead of placement new
8150         if BROKEN_VALUE_INITIALIZATION is defined.
8151         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8152         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8153         is defined.
8154
8155         PR rtl-optimization/83723
8156         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8157         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
8158         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
8159         recursive calls.
8160         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8161         callers.
8162         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8163
8164 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
8165
8166         PR rtl-optimization/81443
8167         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8168         from inner REGs to paradoxical SUBREGs.
8169
8170 2018-02-16  Richard Biener  <rguenther@suse.de>
8171
8172         PR tree-optimization/84399
8173         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8174         For operands we can analyze at their definition make sure we can
8175         analyze them at each use as well.
8176
8177 2018-02-16  Richard Biener  <rguenther@suse.de>
8178
8179         PR tree-optimization/84190
8180         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8181         volatile accesses if the decl isn't volatile.
8182
8183 2018-02-15  Jason Merrill  <jason@redhat.com>
8184
8185         PR c++/84314 - ICE with templates and fastcall attribute.
8186         * attribs.c (build_type_attribute_qual_variant): Don't clobber
8187         TYPE_CANONICAL on an existing type.
8188
8189 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
8190
8191         PR tree-optimization/84383
8192         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8193         dstoff nor call operand_equal_p if dstbase is NULL.
8194
8195         PR tree-optimization/84334
8196         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8197         also a CONSTANT_CLASS_P, punt.
8198
8199 2018-02-14  Jim Wilson  <jimw@sifive.com>
8200
8201         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8202         first SMALL_OPERAND check.  New local min_second_step.  Move assert
8203         to where locals are set.  Add TARGET_RVC support.
8204         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8205
8206 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
8207
8208         * doc/invoke.texi: Correct -Wformat-overflow code sample.
8209
8210 2018-02-14  Martin Sebor  <msebor@redhat.com>
8211
8212         PR tree-optimization/83698
8213         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8214         arrays constrain the offset range to their bounds.
8215         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8216         (builtin_access::overlap): Avoid setting the size of overlap if it's
8217         already been set.
8218         (maybe_diag_overlap): Also consider arrays when deciding what values
8219         of offsets to include in diagnostics.
8220
8221 2018-02-14  Martin Sebor  <msebor@redhat.com>
8222
8223         PR c/84108
8224         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8225         that correspond to the kind of a declaration.
8226
8227 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
8228
8229         PR target/83984
8230         * config/pa/pa.md: Load address of PIC label using the linkage table
8231         if the label is nonlocal.
8232
8233 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8234
8235         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8236         warning message if user requests -maltivec=be.
8237         * doc/invoke.texi: Document deprecation of -maltivec=be.
8238
8239 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
8240
8241         PR target/84220
8242         * config/rs6000/rs6000-c.c: Update definitions for
8243         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8244         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8245
8246 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
8247
8248         PR target/84239
8249         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8250         add _get_ssp intrinsics. Remove argument from
8251         __builtin_ia32_rdssp[d|q].
8252         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8253         * config/i386/i386-builtin.def: Remove argument from
8254         __builtin_ia32_rdssp[d|q].
8255         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8256         ix86_expand_special_args_builtin for _rdssp[d|q].
8257         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8258         Clear register before usage.
8259         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8260         Add documentation for new _get_ssp and _inc_ssp intrinsics.
8261
8262 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
8263
8264         PR tree-optimization/84357
8265         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8266         operand 1 of an ARRAY_REF too.
8267
8268 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
8269
8270         PR target/83831
8271         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8272         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8273         declarations.
8274         (set_of_reg): New struct.
8275         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8276         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8277         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8278         functions.
8279         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8280         Split into bitclr, bitset, bitinvert patterns if appropriate.
8281         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8282         use rx_fuse_in_memory_bitop.
8283         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8284         to named insn, correct maximum insn length.
8285
8286 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
8287
8288         PR target/79242
8289         * machmode.def: Define a complex mode for PARTIAL_INT.
8290         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8291         MODE_PARTIAL_INT.
8292         * doc/rtl.texi: Document CSPImode.
8293         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8294         handling.
8295         (msp430_hard_regno_nregs_with_padding): Likewise.
8296
8297 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
8298
8299         PR target/84279
8300         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8301
8302 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
8303
8304         PR rtl-optimization/84169
8305         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
8306         we generated a parallel as new i3 and we split that to new i2 and i3
8307         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
8308         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
8309         those to i2, not i1.  Partially rewrite this scan code.
8310
8311 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
8312
8313         PR c/82210
8314         * stor-layout.c (place_field): For variable length fields, adjust
8315         offset_align afterwards not just based on the field's alignment,
8316         but also on the size.
8317
8318         PR middle-end/84309
8319         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8320         of exps and logs in the use_exp2 case.
8321
8322 2018-02-13  Jeff Law  <law@redhat.com>
8323
8324         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8325         entry for "vector".
8326
8327         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8328         ARGS as unused.
8329
8330 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
8331
8332         PR debug/84342
8333         PR debug/84319
8334         * common.opt (gas-loc-support, gas-locview-support): New.
8335         (ginline-points, ginternal-reset-location-views): New.
8336         * doc/invoke.texi: Document them.  Use @itemx where intended.
8337         (gvariable-location-views): Adjust.
8338         * target.def (reset_location_view): New.
8339         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8340         (TARGET_RESET_LOCATION_VIEW): New.
8341         * doc/tm.texi: Rebuilt.
8342         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8343         (dwarf2out_default_as_locview_support): New.
8344         (output_asm_line_debug_info): Use option variables.
8345         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8346         (output_loc_list): Likewise.
8347         (add_high_low_attributes): Check option variables.
8348         Don't output entry view attribute in strict mode.
8349         (gen_inlined_subroutine_die): Check option variables.
8350         (dwarf2out_inline_entry): Likewise.
8351         (init_sections_and_labels): Likewise.
8352         (dwarf2out_early_finish): Likewise.
8353         (maybe_reset_location_view): New, from...
8354         (dwarf2out_var_location): ... here.  Call it.
8355         * debug.h (dwarf2out_default_as_loc_support): Declare.
8356         (dwarf2out_default_as_locview_support): Declare.
8357         * hooks.c (hook_int_rtx_insn_0): New.
8358         * hooks.h (hook_int_rtx_insn_0): Declare.
8359         * toplev.c (process_options): Take -gas-loc-support and
8360         -gas-locview-support from dwarf2out.  Enable
8361         -gvariable-location-views by default only with locview
8362         assembler support.  Enable -ginternal-reset-location-views by
8363         default only if the target defines the corresponding hook.
8364         Enable -ginline-points by default if location views are
8365         enabled; force it disabled if statement frontiers are
8366         disabled.
8367         * tree-inline.c (expand_call_inline): Check option variables.
8368         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8369
8370 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
8371
8372         PR tree-optimization/84321
8373         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8374         handling.  Also check whether the anti-range contains any values
8375         that satisfy the mask; switch to a VR_RANGE if not.
8376
8377 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
8378
8379         PR sanitizer/84340
8380         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8381
8382 2018-02-13  Martin Jambor  <mjambor@suse.cz>
8383
8384         PR c++/83990
8385         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8386         of call statements, also set location of a load to a temporary.
8387
8388 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
8389
8390         * config/rl78/rl78.c (add_vector_labels): New function.
8391         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8392         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8393         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8394         which checks that no arguments are passed.
8395         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8396         * doc/extend.texi: Documentation for the new attribute.
8397
8398 2018-02-13  Andreas Schwab  <schwab@suse.de>
8399
8400         * config/riscv/linux.h (CPP_SPEC): Define.
8401
8402 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
8403
8404         PR target/84335
8405         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8406         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8407         OPTION_MASK_ISA_AES as first argument to def_builtin_const
8408         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8409         instead of OPTION_MASK_ISA_PCLMUL as first argument to
8410         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8411         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8412         temporarily for AES and PCLMUL builtins.
8413
8414         PR tree-optimization/84339
8415         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8416         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8417         Formatting fixes.
8418
8419         PR middle-end/84309
8420         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8421         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8422         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8423         inline function.
8424         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8425         inline function.
8426         * omp-simd-clone.h: New file.
8427         * omp-simd-clone.c: Include omp-simd-clone.h.
8428         (expand_simd_clones): No longer static.
8429         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8430         cgraph.h and omp-simd-clone.h.
8431         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8432         (vect_recog_widen_shift_pattern): Formatting fix.
8433         (vect_pattern_recog_1): Don't check optab for calls.
8434
8435         PR target/84336
8436         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8437         operands[2] into a REG before using gen_lowpart on it.
8438
8439 2018-02-12  Jeff Law  <law@redhat.com>
8440
8441         PR target/83760
8442         * config/sh/sh.c (find_barrier): Consider a sibling call
8443         a barrier as well.
8444
8445         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8446         successfully back substituting a reg.
8447
8448 2018-02-12  Richard Biener  <rguenther@suse.de>
8449
8450         PR tree-optimization/84037
8451         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8452         parameter, move visited init to caller.
8453         (vect_slp_analyze_operations): Separate cost from validity
8454         check, initialize visited once for all instances.
8455         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8456         for all instances.
8457         * tree-vect-stmts.c (vect_model_simple_cost): Make early
8458         out an assert.
8459         (vect_model_promotion_demotion_cost): Likewise.
8460         (vectorizable_bswap): Guard cost modeling with !slp_node
8461         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8462         SLP stmts.
8463         (vectorizable_call): Likewise.
8464         (vectorizable_conversion): Likewise.
8465         (vectorizable_assignment): Likewise.
8466         (vectorizable_shift): Likewise.
8467         (vectorizable_operation): Likewise.
8468         (vectorizable_store): Likewise.
8469         (vectorizable_load): Likewise.
8470         (vectorizable_condition): Likewise.
8471         (vectorizable_comparison): Likewise.
8472
8473 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
8474
8475         PR sanitizer/84307
8476         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8477         (ASAN_MARK): Fix fnspec to account for return value, change pointer
8478         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8479
8480 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
8481
8482         PR middle-end/83665
8483         * params.def (inline-min-speedup): Increase from 8 to 15.
8484         (max-inline-insns-auto): Decrease from 40 to 30.
8485         * ipa-split.c (consider_split): Add some buffer for function to
8486         be considered inlining candidate.
8487         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8488         default values.
8489
8490 2018-02-12  Richard Biener  <rguenther@suse.de>
8491
8492         PR tree-optimization/84037
8493         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8494         matched stmts if we cannot swap the non-matched ones.
8495
8496 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
8497
8498         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8499         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8500         _mm_maskz_scalef_round_ss): New intrinsics.
8501         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8502         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8503         __builtin_ia32_scalefss_round): Remove.
8504         (__builtin_ia32_scalefsd_mask_round,
8505         __builtin_ia32_scalefss_mask_round): New intrinsics.
8506         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8507         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8508         ((match_operand:VF_128 2 "<round_nimm_predicate>"
8509         "<round_constraint>")): Changed to ...
8510         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8511         "<round_scalar_constraint>")): ... this.
8512         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8513         %0, %1, %2<round_op3>}"): Changed to ...
8514         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8515         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8516         %2<round_scalar_mask_op3>}"): ... this.
8517         * config/i386/subst.md (round_scalar_nimm_predicate): New.
8518
8519 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
8520
8521         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8522         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8523         (_mm_maskz_sqrt_round_ss): New intrinsics.
8524         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8525         (__builtin_ia32_sqrtsd_mask_round)
8526         (__builtin_ia32_sqrtss_mask_round): New builtins.
8527         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8528         (__builtin_ia32_sqrtss_round): Remove.
8529         (__builtin_ia32_sqrtsd_mask_round)
8530         (__builtin_ia32_sqrtss_mask_round): New builtins.
8531         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8532         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8533         ((match_operand:VF_128 1 "vector_operand"
8534         "xBm,<round_constraint>")): Changed to ...
8535         ((match_operand:VF_128 1 "vector_operand"
8536         "xBm,<round_scalar_constraint>")): ... this.
8537         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8538         %0, %2, %<iptr>1<round_op3>}): Changed to ...
8539         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8540         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8541         %<iptr>1<round_scalar_mask_op3>}): ... this.
8542         ((set_attr "prefix" "<round_prefix>")): Changed to ...
8543         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8544
8545 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
8546
8547         PR target/84266
8548         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8549         Cast vec_cmpeq result to correct type.
8550         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8551         Cast vec_cmpgt result to correct type.
8552
8553 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
8554
8555         * final.c (final_scan_insn_1): Renamed from...
8556         (final_scan_insn): ... this.  New wrapper, to recover
8557         seen from the outermost call in recursive ones.
8558         * config/sparc/sparc.c (output_return): Drop seen from call.
8559         (output_sibcall): Likewise.
8560         * config/visium/visium.c (output_branch): Likewise.
8561
8562 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
8563
8564         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8565         function label.
8566
8567 2018-02-10  Alan Modra  <amodra@gmail.com>
8568
8569         PR target/84300
8570         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8571         Specify LR as an input.
8572
8573 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
8574
8575         PR sanitizer/83987
8576         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8577         remove_member_access_dummy_vars): New functions.
8578         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8579         lower_omp_1, execute_lower_omp): Use them.
8580
8581         PR rtl-optimization/84308
8582         * shrink-wrap.c (spread_components): Release todo vector.
8583
8584 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
8585
8586         PR rtl-optimization/57193
8587         * ira-color.c (struct allocno_color_data): Add member
8588         conflict_allocno_hard_prefs.
8589         (update_conflict_allocno_hard_prefs): New.
8590         (bucket_allocno_compare_func): Add a preference based on
8591         conflict_allocno_hard_prefs.
8592         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8593         (color_allocnos): Remove a dead code.  Initiate
8594         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
8595
8596 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
8597
8598         PR target/84226
8599         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8600         constraint from =wa to wa.  Avoid a subreg on the output operand,
8601         instead use a pseudo and subreg it in a move.
8602         (p9_xxbrd_<mode>): Changed to ...
8603         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8604         (p9_xxbrd_v2df): New expander.
8605         (p9_xxbrw_<mode>): Changed to ...
8606         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8607         (p9_xxbrw_v4sf): New expander.
8608
8609 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
8610
8611         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8612
8613 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
8614
8615         PR target/83926
8616         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8617         multiply in 32-bit mode.
8618         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8619         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8620         mode.
8621
8622 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
8623
8624         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8625         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8626         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8627         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8628
8629 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
8630
8631         PR lto/84213
8632         * dwarf2out.c (is_trivial_indirect_ref): New function.
8633         (dwarf2out_late_global_decl): Do not generate a location
8634         attribute for variables that have a non-trivial DECL_VALUE_EXPR
8635         and that are not defined in the current unit.
8636
8637 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
8638
8639         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8640         instead of a libcall for UNORDERED.
8641
8642 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
8643
8644         PR target/82641
8645         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8646         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8647
8648 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8649
8650         PR target/PR84295
8651         * config/s390/s390.c (s390_set_current_function): Invoke
8652         s390_indirect_branch_settings also if fndecl didn't change.
8653
8654 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
8655
8656         * config/rs6000/rs6000.md (blockage): Set length to zero.
8657
8658 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
8659
8660         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8661
8662 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
8663
8664         PR sanitizer/84285
8665         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8666         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8667         -static-lib*san.
8668
8669         PR debug/84252
8670         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8671         PARALLEL incoming that failed vt_get_decl_and_offset check.
8672
8673         PR middle-end/84237
8674         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8675         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8676         TREE_READONLY bit.
8677         (get_variable_section): For decls in named .bss* sections pass true as
8678         second argument to bss_initializer_p.
8679
8680 2018-02-09  Marek Polacek  <polacek@redhat.com>
8681             Jakub Jelinek  <jakub@redhat.com>
8682
8683         PR c++/83659
8684         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8685         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
8686         Sync some changes from cxx_fold_indirect_ref.
8687
8688 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
8689
8690         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8691         markers.
8692         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8693         (BLOCK_INLINE_ENTRY_LABEL): New.
8694         (dwarf2out_var_location): Disregard inline entry markers.
8695         (inline_entry_data): New struct.
8696         (inline_entry_data_hasher): New hashtable type.
8697         (inline_entry_data_hasher::hash): New.
8698         (inline_entry_data_hasher::equal): New.
8699         (inline_entry_data_table): New variable.
8700         (add_high_low_attributes): Add DW_AT_entry_pc and
8701         DW_AT_GNU_entry_view attributes if a pending entry is found
8702         in inline_entry_data_table.  Add old entry_pc attribute only
8703         if debug nonbinding markers are disabled.
8704         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8705         markers are enabled.
8706         (block_within_block_p, dwarf2out_inline_entry): New.
8707         (dwarf2out_finish): Check that no entries remained in
8708         inline_entry_data_table.
8709         * final.c (reemit_insn_block_notes): Handle inline entry notes.
8710         (final_scan_insn, notice_source_line): Likewise.
8711         (rest_of_clean_state): Skip inline entry markers.
8712         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8713         markers.
8714         * gimple.c (gimple_build_debug_inline_entry): New.
8715         * gimple.h (enum gimple_debug_subcode): Add
8716         GIMPLE_DEBUG_INLINE_ENTRY.
8717         (gimple_build_debug_inline_entry): Declare.
8718         (gimple_debug_inline_entry_p): New.
8719         (gimple_debug_nonbind_marker_p): Adjust.
8720         * insn-notes.def (INLINE_ENTRY): New.
8721         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8722         inline entry marker notes.
8723         (print_insn): Likewise.
8724         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8725         (INSN_DEBUG_MARKER_KIND): Likewise.
8726         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8727         * tree-inline.c (expand_call_inline): Build and insert
8728         debug_inline_entry stmt.
8729         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8730         inline entry blocks early, if nonbind markers are enabled.
8731         (dump_scope_block): Dump fragment info.
8732         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8733         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8734         (gimple_build_debug_inline_entry): New.
8735         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8736         Enable/disable inline entry points too.
8737         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8738         (DEBUG_INSN): Describe inline entry markers.
8739
8740         * common.opt (gvariable-location-views): New.
8741         (gvariable-location-views=incompat5): New.
8742         * config.in: Rebuilt.
8743         * configure: Rebuilt.
8744         * configure.ac: Test assembler for view support.
8745         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8746         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8747         * dwarf2out.c (var_loc_view): New typedef.
8748         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8749         (dwarf2out_locviews_in_attribute): New.
8750         (dwarf2out_locviews_in_loclist): New.
8751         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8752         (enum dw_line_info_opcode): Add LI_adv_address.
8753         (struct dw_line_info_table): Add view.
8754         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8755         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8756         (zero_view_p): New variable.
8757         (ZERO_VIEW_P): New macro.
8758         (output_asm_line_debug_info): New.
8759         (struct var_loc_node): Add view.
8760         (add_AT_view_list, AT_loc_list): New.
8761         (add_var_loc_to_decl): Add view param.  Test it against last.
8762         (new_loc_list): Add view params.  Record them.
8763         (AT_loc_list_ptr): Handle loc and view lists.
8764         (view_list_to_loc_list_val_node): New.
8765         (print_dw_val): Handle dw_val_class_view_list.
8766         (size_of_die): Likewise.
8767         (value_format): Likewise.
8768         (loc_list_has_views): New.
8769         (gen_llsym): Set vl_symbol too.
8770         (maybe_gen_llsym, skip_loc_list_entry): New.
8771         (dwarf2out_maybe_output_loclist_view_pair): New.
8772         (output_loc_list): Output view list or entries too.
8773         (output_view_list_offset): New.
8774         (output_die): Handle dw_val_class_view_list.
8775         (output_dwarf_version): New.
8776         (output_compilation_unit_header): Use it.
8777         (output_skeleton_debug_sections): Likewise.
8778         (output_rnglists, output_line_info): Likewise.
8779         (output_pubnames, output_aranges): Update version comments.
8780         (output_one_line_info_table): Output view numbers in asm comments.
8781         (dw_loc_list): Determine current endview, pass it to new_loc_list.
8782         Call maybe_gen_llsym.
8783         (loc_list_from_tree_1): Adjust.
8784         (add_AT_location_description): Create view list attribute if
8785         needed, check it's absent otherwise.
8786         (convert_cfa_to_fb_loc_list): Adjust.
8787         (maybe_emit_file): Call output_asm_line_debug_info for test.
8788         (dwarf2out_var_location): Reset views as needed.  Precompute
8789         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
8790         attribute.  Set view.
8791         (new_line_info_table): Reset next view.
8792         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8793         (dwarf2out_source_line): Likewise.  Output view resets and labels to
8794         the assembler, or select appropriate line info opcodes.
8795         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8796         (optimize_string_length): Catch it.  Adjust.
8797         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
8798         dw_val_class_view_list, and remove it if no longer needed.
8799         (hash_loc_list): Hash view numbers.
8800         (loc_list_hasher::equal): Compare them.
8801         (optimize_location_lists): Check whether a view list symbol is
8802         needed, and whether the locview attribute is present, and
8803         whether they match.  Remove the locview attribute if no longer
8804         needed.
8805         (index_location_lists): Call skip_loc_list_entry for test.
8806         (dwarf2out_finish): Call output_asm_line_debug_info for test.
8807         Use output_dwarf_version.
8808         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8809         (struct dw_val_node): Add val_view_list.
8810         * final.c (SEEN_NEXT_VIEW): New.
8811         (set_next_view_needed): New.
8812         (clear_next_view_needed): New.
8813         (maybe_output_next_view): New.
8814         (final_start_function): Rename to...
8815         (final_start_function_1): ... this.  Take pointer to FIRST,
8816         add SEEN parameter.  Emit param bindings in the initial view.
8817         (final_start_function): Reintroduce SEEN-less interface.
8818         (final): Rename to...
8819         (final_1): ... this.  Take SEEN parameter.  Output final pending
8820         next view at the end.
8821         (final): Reintroduce seen-less interface.
8822         (final_scan_insn): Output pending next view before switching
8823         sections or ending a block.  Mark the next view as needed when
8824         outputting variable locations.  Notify debug backend of section
8825         changes, and of location view changes.
8826         (rest_of_handle_final): Adjust.
8827         * toplev.c (process_options): Autodetect value for debug variable
8828         location views option.  Warn on incompat5 without -gdwarf-5.
8829         * doc/invoke.texi (gvariable-location-views): New.
8830         (gvariable-location-views=incompat5): New.
8831         (gno-variable-location-views): New.
8832
8833 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
8834
8835         PR tree-optimization/84136
8836         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8837         that the result of find_edge is non-NULL.
8838
8839 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8840
8841         PR target/83008
8842         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8843         storing integer register in SImode.  Fix cost of 256 and 512
8844         byte aligned SSE register store.
8845
8846 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8847
8848         * config/i386/i386.c (ix86_multiplication_cost): Fix
8849         multiplication cost for TARGET_AVX512DQ.
8850
8851 2018-02-08  Marek Polacek  <polacek@redhat.com>
8852
8853         PR tree-optimization/84238
8854         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8855         get_range_strlen.
8856
8857 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8858
8859         PR tree-optimization/84265
8860         * tree-vect-stmts.c (vectorizable_store): Don't treat
8861         VMAT_CONTIGUOUS accesses as grouped.
8862         (vectorizable_load): Likewise.
8863
8864 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8865
8866         PR tree-optimization/81635
8867         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8868         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8869         (test_round_for_mask): New functions.
8870         (wide_int_cc_tests): Call test_round_for_mask.
8871         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8872         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8873         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8874         range returned by get_range_info.
8875
8876 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
8877
8878         PR ipa/81360
8879         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8880         * symtab.c: Include builtins.h
8881         (symtab_node::output_to_lto_symbol_table_p): Move here
8882         from lto-streamer-out.c:output_symbol_p.
8883         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8884         (output_symbol_p): Move all logic to symtab.c
8885         (produce_symtab): Update.
8886
8887 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8888
8889         * config/s390/s390-opts.h (enum indirect_branch): Define.
8890         * config/s390/s390-protos.h (s390_return_addr_from_memory)
8891         (s390_indirect_branch_via_thunk)
8892         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8893         (enum s390_indirect_branch_type): Define.
8894         * config/s390/s390.c (struct s390_frame_layout, struct
8895         machine_function): Remove.
8896         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8897         (indirect_branch_table_label_no, indirect_branch_table_name):
8898         Define variables.
8899         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8900         (enum s390_indirect_branch_option): Define.
8901         (s390_return_addr_from_memory): New function.
8902         (s390_handle_string_attribute): New function.
8903         (s390_attribute_table): Add new attribute handler.
8904         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8905         (s390_indirect_branch_via_thunk): New function.
8906         (s390_indirect_branch_via_inline_thunk): New function.
8907         (s390_function_ok_for_sibcall): When jumping via thunk disallow
8908         sibling call optimization for non z10 compiles.
8909         (s390_emit_call): Force indirect branch target to be a single
8910         register.  Add r1 clobber for non-z10 compiles.
8911         (s390_emit_epilogue): Emit return jump via return_use expander.
8912         (s390_reorg): Handle JUMP_INSNs as execute targets.
8913         (s390_option_override_internal): Perform validity checks for the
8914         new command line options.
8915         (s390_indirect_branch_attrvalue): New function.
8916         (s390_indirect_branch_settings): New function.
8917         (s390_set_current_function): Invoke s390_indirect_branch_settings.
8918         (s390_output_indirect_thunk_function):  New function.
8919         (s390_code_end): Implement target hook.
8920         (s390_case_values_threshold): Implement target hook.
8921         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8922         macros.
8923         * config/s390/s390.h (struct s390_frame_layout)
8924         (struct machine_function): Move here from s390.c.
8925         (TARGET_INDIRECT_BRANCH_NOBP_RET)
8926         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8927         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8928         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8929         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8930         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8931         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8932         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8933         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8934         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8935         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8936         (mnemonic attribute): Add values which aren't recognized
8937         automatically.
8938         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8939         pattern for branch conversion.  Fix mnemonic attribute.
8940         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8941         indirect branch via thunk if requested.
8942         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8943         ("*indirect_jump"): Disable for branch conversion using out of
8944         line thunks.
8945         ("indirect_jump_via_thunk<mode>_z10")
8946         ("indirect_jump_via_thunk<mode>")
8947         ("indirect_jump_via_inlinethunk<mode>_z10")
8948         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8949         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8950         ("casesi_jump_via_inlinethunk<mode>_z10")
8951         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8952         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8953         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8954         ("*indirect2_jump"): Disable for branch conversion.
8955         ("casesi_jump"): Turn into expander and expand patterns for branch
8956         conversion.
8957         ("return_use"): New expander.
8958         ("*return"): Emit return via thunk and rename it to ...
8959         ("*return<mode>"): ... this one.
8960         * config/s390/s390.opt: Add new options and and enum for the
8961         option values.
8962
8963 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8964
8965         * lra-constraints.c (match_reload): Unconditionally use
8966         gen_lowpart_SUBREG, rather than selecting between that
8967         and equivalent gen_rtx_SUBREG code.
8968
8969 2018-02-08  Richard Biener  <rguenther@suse.de>
8970
8971         PR tree-optimization/84233
8972         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8973         changed flag instead of boguously re-using phi_inserted.
8974
8975 2018-02-08  Martin Jambor  <mjambor@suse.cz>
8976
8977         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8978         static local variables.
8979
8980 2018-02-08  Richard Biener  <rguenther@suse.de>
8981
8982         PR tree-optimization/84278
8983         * tree-vect-stmts.c (vectorizable_store): When looking for
8984         smaller vector types to perform grouped strided loads/stores
8985         make sure the mode is supported by the target.
8986         (vectorizable_load): Likewise.
8987
8988 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
8989
8990         * config/aarch64/aarch64.c (aarch64_components_for_bb):
8991         Increase LDP/STP opportunities by adding adjacent callee-saves.
8992
8993 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
8994
8995         PR rtl-optimization/84068
8996         PR rtl-optimization/83459
8997         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8998
8999 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
9000
9001         PR tree-optimization/84224
9002         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9003         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9004         non-zero arguments.
9005
9006 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
9007
9008         PR target/84113
9009         * config/rs6000/altivec.md (*restore_world): Remove LR use.
9010         * config/rs6000/predicates.md (restore_world_operation): Adjust op
9011         count, remove one USE.
9012
9013 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
9014
9015         * doc/install.texi (Configuration): Document the
9016         --with-long-double-format={ibm,ieee} PowerPC configuration
9017         options.
9018
9019         PR target/84154
9020         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9021         Convert from define_expand to be define_insn_and_split.  Rework
9022         float/double/_Float128 conversions to QI/HI/SImode to work with
9023         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
9024         conversions to QI/HImode types did a store and then a load to
9025         truncate the value.  For conversions to VSX registers, don't split
9026         the insn, instead emit the code directly.  Use the code iterator
9027         any_fix to combine signed and unsigned conversions.
9028         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9029         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9030         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9031         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9032         (fix_<mode>di2_hw): Likewise.
9033         (fixuns_<mode>di2_hw): Likewise.
9034         (fix_<mode>si2_hw): Likewise.
9035         (fixuns_<mode>si2_hw): Likewise.
9036         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9037         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9038         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9039         fix<uns>_trunc<SFDF:mode>si2_p8.
9040         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
9041         used.
9042         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9043         (fix<uns>_<mode>_mem): Likewise.
9044         (fctiw<u>z_<mode>_mem): Likewise.
9045         (fix<uns>_<mode>_mem): Likewise.
9046         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9047         the register allocator from doing a direct move to the GPRs to do
9048         a store, and instead use the ISA 3.0 store byte/half-word from
9049         vector register instruction.  For IEEE 128-bit floating point,
9050         also optimize stores of 32-bit ints.
9051         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9052
9053 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
9054
9055         * genextract.c (push_pathstr_operand): New function to support
9056         [a-zA-Z].
9057         (walk_rtx): Call push_pathstr_operand.
9058         (print_path): Support [a-zA-Z].
9059
9060 2018-02-07  Richard Biener  <rguenther@suse.de>
9061
9062         PR tree-optimization/84037
9063         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9064         (cse_and_gimplify_to_preheader): Declare.
9065         (vect_get_place_in_interleaving_chain): Likewise.
9066         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9067         ivexpr_map.
9068         (_loop_vec_info::~_loop_vec_info): Delete it.
9069         (cse_and_gimplify_to_preheader): New function.
9070         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9071         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9072         (vectorizable_load): Likewise.  For grouped stores always base
9073         the IV on the first element.
9074         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9075         condition before gimplifying.
9076
9077 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
9078
9079         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9080         *DIV_EXPR and *MOD_EXPR.
9081
9082 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
9083
9084         PR target/84248
9085         * config/i386/i386.c (ix86_option_override_internal): Mask out
9086         the CF_SET bit when checking -fcf-protection.
9087
9088 2018-02-07  Tom de Vries  <tom@codesourcery.com>
9089
9090         PR libgomp/84217
9091         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9092         enough.
9093
9094 2018-02-07  Richard Biener  <rguenther@suse.de>
9095
9096         PR tree-optimization/84204
9097         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9098         this place.
9099
9100         PR tree-optimization/84205
9101         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9102         special-case isl_ast_op_zdiv_r.
9103
9104         PR tree-optimization/84223
9105         * graphite-scop-detection.c (gather_bbs::before_dom_children):
9106         Only add conditions from within the region.
9107         (gather_bbs::after_dom_children): Adjust.
9108
9109 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
9110
9111         PR target/84209
9112         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9113         * config/avr/avr.md: Only post-reload split REG-REG moves if
9114         either register is GENERAL_REG_P.
9115
9116 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
9117
9118         PR tree-optimization/84235
9119         * tree-ssa-scopedtables.c
9120         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9121         if the subtraction is performed in floating point type where NaNs are
9122         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9123         build 1.  Formatting fix.
9124
9125 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
9126
9127         PR target/84146
9128         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9129         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9130         and skip it regardless of bb boundaries.  Use CALL_P macro,
9131         don't test INSN_P (insn) together with CALL_P or JUMP_P check
9132         unnecessarily, formatting fix.
9133
9134 2018-02-06  Michael Collison  <michael.collison@arm.com>
9135
9136         * config/arm/thumb2.md:
9137         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9138         (*thumb_mov_notscc): Ditto.
9139
9140 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
9141
9142         PR target/84154
9143         * config/rs6000/rs6000.md (su code attribute): Use "u" for
9144         unsigned_fix, not "s".
9145
9146 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9147
9148         * configure.ac (gcc_fn_eh_frame_ro): New function.
9149         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9150         correct .eh_frame permissions.
9151         * configure: Regenerate.
9152
9153 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
9154
9155         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9156         irrelevant options.
9157
9158 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9159
9160         * config/rs6000/rs6000.c (rs6000_option_override_internal):
9161         Display warning message for -mno-speculate-indirect-jumps.
9162
9163 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
9164
9165         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9166         Undocumented.
9167         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9168
9169 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
9170
9171         PR tree-optimization/84225
9172         * tree-eh.c (find_trapping_overflow): Only call
9173         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9174
9175 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
9176
9177         PR target/84145
9178         * config/i386/i386.c: Reimplement the check of possible options
9179         -mibt/-mshstk conbination. Change error messages.
9180         * doc/invoke.texi: Fix a typo: remove extra '='.
9181
9182 2018-02-06  Marek Polacek  <polacek@redhat.com>
9183
9184         PR tree-optimization/84228
9185         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9186
9187 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
9188
9189         PR target/82641
9190         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9191         emitted arch directives.
9192         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9193         __ARM_FEATURE_COPROC before changing architectures.
9194
9195 2018-02-06  Richard Biener  <rguenther@suse.de>
9196
9197         * config/i386/i386.c (print_reg): Fix typo.
9198         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9199
9200 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
9201
9202         * configure: Regenerate.
9203
9204 2018-02-05  Martin Sebor  <msebor@redhat.com>
9205
9206         PR tree-optimization/83369
9207         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9208         inlining context.
9209
9210 2018-02-05  Martin Liska  <mliska@suse.cz>
9211
9212         * doc/invoke.texi: Cherry-pick upstream r323995.
9213
9214 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
9215
9216         * ira.c (ira_init_register_move_cost): Adjust comment.
9217
9218 2018-02-05  Martin Liska  <mliska@suse.cz>
9219
9220         PR gcov-profile/84137
9221         * doc/gcov.texi: Fix typo in documentation.
9222
9223 2018-02-05  Martin Liska  <mliska@suse.cz>
9224
9225         PR gcov-profile/83879
9226         * doc/gcov.texi: Document necessity of --dynamic-list-data when
9227         using dlopen functionality.
9228
9229 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
9230
9231         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9232         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9233         _mm_maskz_range_ss, _mm_mask_range_round_ss,
9234         _mm_maskz_range_round_ss): New intrinsics.
9235         (__builtin_ia32_rangesd128_round)
9236         (__builtin_ia32_rangess128_round): Remove.
9237         (__builtin_ia32_rangesd128_mask_round,
9238         __builtin_ia32_rangess128_mask_round): New builtins.
9239         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9240         __builtin_ia32_rangess128_round): Remove.
9241         (__builtin_ia32_rangesd128_mask_round,
9242         __builtin_ia32_rangess128_mask_round): New builtins.
9243         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9244         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9245         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9246         "<round_saeonly_constraint>")): Changed to ...
9247         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9248         "<round_saeonly_scalar_constraint>")): ... this.
9249         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9250         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9251         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9252         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9253         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9254
9255 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
9256
9257         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9258         options.
9259         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9260         Remove all values except native, 8540 and 8548.
9261
9262 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
9263
9264         * config/i386/i386.c (ix86_output_function_return): Pass
9265         INVALID_REGNUM, instead of -1, as invalid register number to
9266         indirect_thunk_name and output_indirect_thunk.
9267
9268 2018-02-02  Julia Koval  <julia.koval@intel.com>
9269
9270         * config.gcc: Add -march=icelake.
9271         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9272         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9273         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9274         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9275         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9276         (processor_target_table): Add icelake.
9277         (ix86_option_override_internal): Handle new PTAs.
9278         (get_builtin_code_for_version): Handle icelake.
9279         (M_INTEL_COREI7_ICELAKE): New.
9280         (fold_builtin_cpu): Handle icelake.
9281         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9282         * doc/invoke.texi: Add -march=icelake.
9283
9284 2018-02-02  Julia Koval  <julia.koval@intel.com>
9285
9286         * config/i386/i386.c (ix86_option_override_internal): Change flags type
9287         to wide_int_bitmask.
9288         * wide-int-bitmask.h: New.
9289
9290 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
9291
9292         PR target/84066
9293         * config/i386/i386.md: Replace Pmode with word_mode in
9294         builtin_setjmp_setup and builtin_longjmp to support x32.
9295
9296 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
9297
9298         PR target/56010
9299         PR target/83743
9300         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9301         #include "opts.h".
9302         (rs6000_supported_cpu_names): New static variable.
9303         (linux_cpu_translation_table): Likewise.
9304         (elf_platform) <cpu>: Define new static variable and use it.
9305         Translate kernel AT_PLATFORM name to canonical name if needed.
9306         Error if platform name is unknown.
9307
9308 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
9309
9310         PR target/84089
9311         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9312
9313 2018-02-01  Jeff Law  <law@redhat.com>
9314
9315         PR target/84128
9316         * config/i386/i386.c (release_scratch_register_on_entry): Add new
9317         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
9318         the scratch if RELEASE_VIA_POP is false.
9319         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9320         If we have to save a temporary register, decrement SIZE appropriately.
9321         Pass new arguments to release_scratch_register_on_entry.
9322         (ix86_adjust_stack_and_probe): Likewise.
9323         (ix86_emit_probe_stack_range): Pass new arguments to
9324         release_scratch_register_on_entry.
9325
9326 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
9327
9328         PR rtl-optimization/84157
9329         * combine.c (change_zero_ext): Use REG_P predicate in
9330         front of HARD_REGISTER_P predicate.
9331
9332 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
9333
9334         * config/avr/avr.c (avr_option_override): Move disabling of
9335         -fdelete-null-pointer-checks to...
9336         * common/config/avr/avr-common.c (avr_option_optimization_table):
9337         ...here.
9338
9339 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9340
9341         PR tree-optimization/81635
9342         * tree-data-ref.c (split_constant_offset_1): For types that
9343         wrap on overflow, try to use range info to prove that wrapping
9344         cannot occur.
9345
9346 2018-02-01  Renlin Li  <renlin.li@arm.com>
9347
9348         PR target/83370
9349         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9350         TAILCALL_ADDR_REGS.
9351         (aarch64_register_move_cost): Likewise.
9352         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9353         TAILCALL_ADDR_REGS.
9354         (REG_CLASS_NAMES): Likewise.
9355         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9356         TAILCALL_ADDR_REGS. Remove IP registers.
9357         * config/aarch64/aarch64.md (Ucs): Update register constraint.
9358
9359 2018-02-01  Richard Biener  <rguenther@suse.de>
9360
9361         * domwalk.h (dom_walker::dom_walker): Add additional constructor
9362         for specifying RPO order and allow NULL for that.
9363         * domwalk.c (dom_walker::dom_walker): Likewise.
9364         (dom_walker::walk): Handle NULL RPO order.
9365         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9366         in RPO order.
9367         (rewrite_update_dom_walker): Likewise.
9368         (mark_def_dom_walker): Likewise.
9369
9370 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9371
9372         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9373         (aarch64_maybe_expand_sve_subreg_move): Declare.
9374         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9375         * config/aarch64/predicates.md (aarch64_any_register_operand): New
9376         predicate.
9377         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9378         that are semantically a reverse operation.
9379         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9380         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9381         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9382         functions.
9383         (aarch64_can_change_mode_class): For big-endian, forbid changes
9384         between two SVE modes if they have different element sizes.
9385
9386 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9387
9388         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9389         the TImode handling for big-endian targets.
9390
9391 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9392
9393         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9394         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
9395         not just bytes.
9396         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9397         Remove BSWAP handing for big-endian targets and use the form of
9398         LD1RQ appropariate for the mode.
9399
9400 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9401
9402         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9403         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9404         duplicated element.
9405
9406 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9407
9408         PR tearget/83845
9409         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9410         check for operands that need to go through aarch64_sve_reload_be.
9411
9412 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
9413
9414         PR tree-optimization/81661
9415         PR tree-optimization/84117
9416         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9417         * tree-eh.c: Include gimplify.h.
9418         (find_trapping_overflow, replace_trapping_overflow,
9419         rewrite_to_non_trapping_overflow): New functions.
9420         * tree-vect-loop.c: Include tree-eh.h.
9421         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9422         * tree-data-ref.c: Include tree-eh.h.
9423         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9424
9425 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
9426
9427         PR rtl-optimization/84123
9428         * combine.c (change_zero_ext): Check if hard register satisfies
9429         can_change_dest_mode before calling gen_lowpart_SUBREG.
9430
9431 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
9432
9433         PR target/82444
9434         * ira.c (ira_init_register_move_cost): Remove assert.
9435
9436 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9437
9438         PR rtl-optimization/84071
9439         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9440         * doc/tm.texi: Regenerate.
9441
9442 2018-01-31  Richard Biener  <rguenther@suse.de>
9443
9444         PR tree-optimization/84132
9445         * tree-data-ref.c (analyze_miv_subscript): Properly
9446         check whether evolution_function_is_affine_multivariate_p
9447         before calling gcd_of_steps_may_divide_p.
9448
9449 2018-01-31  Julia Koval  <julia.koval@intel.com>
9450
9451         PR target/83618
9452         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9453         * config/i386/i386.md (rdpid_rex64) New.
9454         (rdpid): Make 32bit only.
9455
9456 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
9457
9458         PR lto/84105
9459         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9460         an IDENTIFIER_NODE for FUNCTION_TYPE's.
9461
9462 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9463
9464         Revert
9465         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
9466
9467         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9468
9469 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9470
9471         PR rtl-optimization/84071
9472         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9473         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9474
9475 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
9476
9477         * config/arc/arc.c (arc_handle_aux_attribute): New function.
9478         (arc_attribute_table): Add 'aux' attribute.
9479         (arc_in_small_data_p): Consider aux like variables.
9480         (arc_is_aux_reg_p): New function.
9481         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9482         (arc_get_aux_arg): New function.
9483         (prepare_move_operands): Handle aux-register access.
9484         (arc_handle_aux_attribute): New function.
9485         * doc/extend.texi (ARC Variable attributes): Add subsection.
9486
9487 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
9488
9489         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9490         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9491         (arc_attribute_table): Add 'uncached' attribute.
9492         (arc_print_operand): Print '.di' flag for uncached memory
9493         accesses.
9494         (arc_in_small_data_p): Do not consider for small data the uncached
9495         types.
9496         (arc_is_uncached_mem_p): New function.
9497         * config/arc/predicates.md (compact_store_memory_operand): Check
9498         for uncached memory accesses.
9499         (nonvol_nonimm_operand): Likewise.
9500         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9501
9502 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
9503
9504         PR c/84100
9505         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9506         falign-loops=): Add Optimization flag.
9507
9508 2018-01-30  Jeff Law  <law@redhat.com>
9509
9510         PR target/84064
9511         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9512         INT_REGISTERS_SAVED.  Check it prior to calling
9513         get_scratch_register_on_entry.
9514         (ix86_adjust_stack_and_probe): Similarly.
9515         (ix86_emit_probe_stack_range): Similarly.
9516         (ix86_expand_prologue): Corresponding changes.
9517
9518 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9519
9520         PR target/40411
9521         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9522         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9523
9524 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
9525
9526         PR target/84112
9527         * lra-constraints.c (curr_insn_transform): Process AND in the
9528         address.
9529
9530 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
9531
9532         PR rtl-optimization/83986
9533         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9534         dependence against last_pending_memory_flush in addition to
9535         pending_jump_insns.
9536
9537 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
9538
9539         PR tree-optimization/81611
9540         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9541         copies.
9542
9543 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
9544
9545         PR target/83758
9546         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9547         a reg rtx.
9548
9549 2018-01-30  Richard Biener  <rguenther@suse.de>
9550             Jakub Jelinek  <jakub@redhat.com>
9551
9552         PR tree-optimization/84111
9553         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9554         inner loops added during recursion, as they don't have up-to-date
9555         SSA form.
9556
9557 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
9558
9559         PR ipa/81360
9560         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9561         (can_inline_edge_by_limits_p): ... here.
9562         (can_early_inline_edge_p, check_callers,
9563         update_caller_keys, update_callee_keys, recursive_inlining,
9564         add_new_edges_to_heap, speculation_useful_p,
9565         inline_small_functions,
9566         inline_small_functions, flatten_function,
9567         inline_to_all_callers_1): Update.
9568
9569 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
9570
9571         * profile-count.c (profile_count::combine_with_ipa_count): Handle
9572         zeros correctly.
9573
9574 2018-01-30  Richard Biener  <rguenther@suse.de>
9575
9576         PR tree-optimization/83008
9577         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9578         invariant and constant vector uses in stmts when they need
9579         more than one stmt.
9580
9581 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9582
9583         PR bootstrap/84017
9584         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9585         * configure: Regenerate.
9586
9587 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9588
9589         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9590         pattern.
9591         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9592         Use gen_rtx_REG rather than gen_lowpart.
9593
9594 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9595
9596         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9597         rather than 0 when creating partial subregs.
9598
9599 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9600
9601         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9602         of usage.
9603
9604 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
9605
9606         PR target/81550
9607         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9608         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9609         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9610         flags.  This restores the settings used before the 2017-07-24.
9611         Turning off pre increment/decrement/modify allows IVOPTS to
9612         optimize DF/SF loops where the index is an int.
9613
9614 2018-01-29  Richard Biener  <rguenther@suse.de>
9615             Kelvin Nilsen  <kelvin@gcc.gnu.org>
9616
9617         PR bootstrap/80867
9618         * tree-vect-stmts.c (vectorizable_call): Don't call
9619         targetm.vectorize_builtin_md_vectorized_function if callee is
9620         NULL.
9621
9622 2018-01-22  Carl Love  <cel@us.ibm.com>
9623
9624         * doc/extend.tex: Fix typo in second arg in
9625         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9626
9627 2018-01-29  Richard Biener  <rguenther@suse.de>
9628
9629         PR tree-optimization/84086
9630         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9631         (flush_ssaname_freelist): When SSA names were released reset
9632         the SCEV hash table.
9633
9634 2018-01-29  Richard Biener  <rguenther@suse.de>
9635
9636         PR tree-optimization/84057
9637         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9638         removed paths when removing edges.
9639
9640 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
9641
9642         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9643         -mfunction-return=@var{choice}.
9644
9645 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9646
9647         PR diagnostic/84034
9648         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9649         Handle CR like TAB.
9650         (layout::print_source_line): Likewise.
9651         (test_get_line_width_without_trailing_whitespace): Add test cases.
9652
9653 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
9654
9655         PR middle-end/84040
9656         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9657         debug insns.
9658
9659 2018-01-26  Jim Wilson  <jimw@sifive.com>
9660
9661         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9662
9663         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9664         specified.
9665
9666 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9667
9668         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9669         and CMP + SUB-immediate -> SUBS.
9670
9671 2018-01-26  Martin Sebor  <msebor@redhat.com>
9672
9673         PR tree-optimization/83896
9674         * tree-ssa-strlen.c (get_string_len): Rename...
9675         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
9676         Avoid assuming length is constant.
9677         (handle_char_store): Use HOST_WIDE_INT for string length.
9678
9679 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
9680
9681         PR target/81763
9682         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9683         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9684
9685 2018-01-26  Richard Biener  <rguenther@suse.de>
9686
9687         PR rtl-optimization/84003
9688         * dse.c (record_store): Only record redundant stores when
9689         the earlier store aliases at least all accesses the later one does.
9690
9691 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
9692
9693         PR rtl-optimization/83985
9694         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9695         REG_CFA_RESTORE insns.
9696         (delete_unmarked_insns): Don't ignore separate shrink wrapping
9697         REG_CFA_RESTORE insns here.
9698
9699         PR c/83989
9700         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9701         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9702
9703 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9704
9705         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9706         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9707         (arc_init): Likewise.
9708         (arc_override_options): Likewise.
9709         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9710         value.
9711         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9712         support.
9713         * config/arc/arc.h (TARGET_DBNZ): Define.
9714         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9715         properly set the tune attribute.
9716         (dbnz): Use TARGET_DBNZ guard.
9717         * config/arc/arc.opt (mtune): Add core3 option.
9718
9719 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9720
9721         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9722         recognize new pic like addresses.
9723         (arc_delegitimize_address): Clean up.
9724
9725 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9726
9727         * config/arc/arc-arches.def: Option mrf16 valid for all
9728         architectures.
9729         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9730         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9731         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9732         * config/arc/arc-tables.opt: Regenerate.
9733         * config/arc/arc.c (arc_conditional_register_usage): Handle
9734         reduced register file case.
9735         (arc_file_start): Set must have build attributes.
9736         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9737         mrf16 option value.
9738         * config/arc/arc.opt (mrf16): Add new option.
9739         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9740         * config/arc/genmultilib.awk: Handle new mrf16 option.
9741         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9742         * config/arc/t-multilib: Regenerate.
9743         * doc/invoke.texi (ARC Options): Document mrf16 option.
9744
9745 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9746
9747         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9748         * config/arc/arc.c (arc_handle_secure_attribute): New function.
9749         (arc_attribute_table): Add 'secure_call' attribute.
9750         (arc_print_operand): Print secure call operand.
9751         (arc_function_ok_for_sibcall): Don't optimize tail calls when
9752         secure.
9753         (arc_is_secure_call_p): New function.  * config/arc/arc.md
9754         (call_i): Add support for sjli instruction.
9755         (call_value_i): Likewise.
9756         * config/arc/constraints.md (Csc): New constraint.
9757
9758 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9759             John Eric Martin  <John.Martin@emmicro-us.com>
9760
9761         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9762         * config/arc/arc.c (_arc_jli_section): New struct.
9763         (arc_jli_section): New type.
9764         (rc_jli_sections): New static variable.
9765         (arc_handle_jli_attribute): New function.
9766         (arc_attribute_table): Add jli_always and jli_fixed attribute.
9767         (arc_file_end): New function.
9768         (TARGET_ASM_FILE_END): Define.
9769         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9770         (arc_add_jli_section): New function.
9771         (jli_call_scan): Likewise.
9772         (arc_reorg): Call jli_call_scan.
9773         (arc_output_addsi): Remove 'S' from printing asm operand.
9774         (arc_is_jli_call_p): New function.
9775         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9776         operand.
9777         (movhi_insn): Likewise.
9778         (movsi_insn): Likewise.
9779         (movsi_set_cc_insn): Likewise.
9780         (loadqi_update): Likewise.
9781         (load_zeroextendqisi_update): Likewise.
9782         (load_signextendqisi_update): Likewise.
9783         (loadhi_update): Likewise.
9784         (load_zeroextendhisi_update): Likewise.
9785         (load_signextendhisi_update): Likewise.
9786         (loadsi_update): Likewise.
9787         (loadsf_update): Likewise.
9788         (movsicc_insn): Likewise.
9789         (bset_insn): Likewise.
9790         (bxor_insn): Likewise.
9791         (bclr_insn): Likewise.
9792         (bmsk_insn): Likewise.
9793         (bicsi3_insn): Likewise.
9794         (cmpsi_cc_c_insn): Likewise.
9795         (movsi_ne): Likewise.
9796         (movsi_cond_exec): Likewise.
9797         (clrsbsi2): Likewise.
9798         (norm_f): Likewise.
9799         (normw): Likewise.
9800         (swap): Likewise.
9801         (divaw): Likewise.
9802         (flag): Likewise.
9803         (sr): Likewise.
9804         (kflag): Likewise.
9805         (ffs): Likewise.
9806         (ffs_f): Likewise.
9807         (fls): Likewise.
9808         (call_i): Remove 'S' asm letter, add jli instruction.
9809         (call_value_i): Likewise.
9810         * config/arc/arc.op (mjli-always): New option.
9811         * config/arc/constraints.md (Cji): New constraint.
9812         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9813         operand.
9814         (subsf3_fpx): Likewise.
9815         (mulsf3_fpx): Likewise.
9816         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9817         asm operand.
9818         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9819         function attrbutes.
9820         * doc/invoke.texi (ARC): Document mjli-always option.
9821
9822 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
9823
9824         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9825         avoid addition with 0 and use incw and decw where possible.
9826
9827 2018-01-26  Richard Biener  <rguenther@suse.de>
9828
9829         PR tree-optimization/81082
9830         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9831         association if it requires casting to unsigned.
9832         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9833         from fold_plusminus_mult_expr to catch important cases late when
9834         range info is available.
9835
9836 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9837
9838         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9839         * configure.ac (hidden_linkonce): New test.
9840         * configure: Regenerate.
9841         * config.in: Regenerate.
9842
9843 2018-01-26  Julia Koval  <julia.koval@intel.com>
9844
9845         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9846         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9847         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9848         _mm_mask_bitshuffle_epi64_mask): Fix type.
9849         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9850         USI_FTYPE_V4DI_V4DI_USI): Remove.
9851         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9852         __builtin_ia32_vpshufbitqmb256_mask,
9853         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9854         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9855         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9856
9857 2018-01-26  Alan Modra  <amodra@gmail.com>
9858
9859         PR target/84033
9860         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9861         UNSPEC_VBPERMQ.  Sort other unspecs.
9862
9863 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
9864
9865         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9866
9867 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
9868
9869         PR middle-end/83055
9870         * predict.c (drop_profile): Do not push/pop cfun; update also
9871         node->count.
9872         (handle_missing_profiles): Fix logic looking for zero profiles.
9873
9874 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
9875
9876         PR middle-end/83977
9877         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9878         on functions with #pragma omp declare simd or functions with simd
9879         attribute.
9880         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9881         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9882         Remove trailing \n from warning_at calls.
9883
9884 2018-01-25  Tom de Vries  <tom@codesourcery.com>
9885
9886         PR target/84028
9887         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9888         for neutered workers.
9889
9890 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
9891
9892         PR target/68467
9893         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9894         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9895
9896 2018-01-24  Jeff Law  <law@redhat.com>
9897
9898         PR target/83994
9899         * i386.c (get_probe_interval): Move to earlier point.
9900         (ix86_compute_frame_layout): If -fstack-clash-protection and
9901         the frame is larger than the probe interval, then use pushes
9902         to save registers rather than reg->mem moves.
9903         (ix86_expand_prologue): Remove conditional for int_registers_saved
9904         assertion.
9905
9906 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
9907
9908         PR target/84014
9909         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9910         min/max for never referenced object.
9911
9912 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
9913
9914         PR middle-end/83977
9915         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9916         here.
9917         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9918         attributes from DECL_ATTRIBUTES (decl) without affecting
9919         DECL_ATTRIBUTES (current_function_decl).
9920         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9921         functions with non-NULL DECL_ABSTRACT_ORIGIN.
9922
9923 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
9924
9925         PR tree-optimization/83979
9926         * fold-const.c (fold_comparison): Use constant_boolean_node
9927         instead of boolean_{true,false}_node.
9928
9929 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
9930
9931         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9932         with zero counts.
9933
9934 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9935
9936         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9937         Simplify the clause that sets the length attribute.
9938         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9939         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9940         clause that sets the length attribute.
9941         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9942
9943 2018-01-24  Tom de Vries  <tom@codesourcery.com>
9944
9945         PR target/83589
9946         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9947         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9948         Add strict parameter.
9949         (prevent_branch_around_nothing): Insert dummy insn between branch to
9950         label and label with no ptx insn inbetween.
9951         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9952
9953 2018-01-24  Tom de Vries  <tom@codesourcery.com>
9954
9955         PR target/81352
9956         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9957         for neutered threads in warp.
9958         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9959
9960 2018-01-24  Richard Biener  <rguenther@suse.de>
9961
9962         PR tree-optimization/83176
9963         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9964         operands.
9965
9966 2018-01-24  Richard Biener  <rguenther@suse.de>
9967
9968         PR tree-optimization/82819
9969         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9970         code generating pluses that are no-ops in the target precision.
9971
9972 2018-01-24  Richard Biener  <rguenther@suse.de>
9973
9974         PR middle-end/84000
9975         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9976
9977 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
9978
9979         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9980         to merge probabilities.
9981         * predict.c (probably_never_executed): Also mark as cold functions
9982         with global 0 profile and guessed local profile.
9983         * profile-count.c (profile_probability::combine_with_count): New
9984         member function.
9985         * profile-count.h (profile_probability::operator*,
9986         profile_probability::operator*=, profile_probability::operator/,
9987         profile_probability::operator/=): Reduce precision to adjusted
9988         and set value to guessed on contradictory divisions.
9989         (profile_probability::combine_with_freq): Remove.
9990         (profile_probability::combine_wiht_count): Declare.
9991         (profile_count::force_nonzero):: Set to adjusted.
9992         (profile_count::probability_in):: Set quality to adjusted.
9993         * tree-ssa-tail-merge.c (replace_block_by): Use
9994         combine_with_count.
9995
9996 2018-01-23  Andrew Waterman  <andrew@sifive.com>
9997             Jim Wilson  <jimw@sifive.com>
9998
9999         * config/riscv/riscv.c (riscv_stack_boundary): New.
10000         (riscv_option_override): Set riscv_stack_boundary.  Handle
10001         riscv_preferred_stack_boundary_arg.
10002         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10003         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10004         (STACK_BOUNDARY): Set to riscv_stack_boundary.
10005         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10006         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10007         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10008
10009 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
10010
10011         PR target/83905
10012         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10013         of struct ix86_frame.
10014         (ix86_expand_epilogue): Likewise.  Add a local variable for
10015         the reg_save_offset field in struct ix86_frame.
10016
10017 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
10018
10019         PR tree-optimization/82604
10020         * tree-loop-distribution.c (enum partition_kind): New enum item
10021         PKIND_PARTIAL_MEMSET.
10022         (partition_builtin_p): Support above new enum item.
10023         (generate_code_for_partition): Ditto.
10024         (compute_access_range): Differentiate cases that equality can be
10025         proven at all loops, the innermost loops or no loops.
10026         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10027         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
10028         (finalize_partitions, distribute_loop): Don't fuse partition of
10029         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10030         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10031         parloop is enabled.
10032
10033 2018-01-23  Martin Liska  <mliska@suse.cz>
10034
10035         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10036         order to ignore the predictor.
10037         (PRED_POLYMORPHIC_CALL): Likewise.
10038         (PRED_RECURSIVE_CALL): Likewise.
10039
10040 2018-01-23  Martin Liska  <mliska@suse.cz>
10041
10042         * tree-profile.c (tree_profiling): Print function header to
10043         aware reader which function we are working on.
10044         * value-prof.c (gimple_find_values_to_profile): Do not print
10045         not interesting value histograms.
10046
10047 2018-01-23  Martin Liska  <mliska@suse.cz>
10048
10049         * profile-count.h (enum profile_quality): Add
10050         profile_uninitialized as the first value. Do not number values
10051         as they are zero based.
10052         (profile_count::verify): Update sanity check.
10053         (profile_probability::verify): Likewise.
10054
10055 2018-01-23  Nathan Sidwell  <nathan@acm.org>
10056
10057         * doc/invoke.texi (ffor-scope): Deprecate.
10058
10059 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10060
10061         PR tree-optimization/83510
10062         * domwalk.c (set_all_edges_as_executable): New function.
10063         (dom_walker::dom_walker): Convert bool param
10064         "skip_unreachable_blocks" to enum reachability.  Move setup of
10065         edge flags to set_all_edges_as_executable and only do it when
10066         reachability is REACHABLE_BLOCKS.
10067         * domwalk.h (enum dom_walker::reachability): New enum.
10068         (dom_walker::dom_walker): Convert bool param
10069         "skip_unreachable_blocks" to enum reachability.
10070         (set_all_edges_as_executable): New decl.
10071         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
10072         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10073         "reachability".
10074         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10075         but converting true to REACHABLE_BLOCKS.
10076         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10077         * tree-vrp.c
10078         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10079         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10080         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10081         REACHABLE_BLOCKS.
10082         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10083         if check_all_array_refs will be called.
10084
10085 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10086
10087         * tree.c (selftest::test_location_wrappers): Add more test
10088         coverage.
10089
10090 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10091
10092         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10093         (selftest::test_bit_in_range): Likewise.
10094
10095 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
10096
10097         PR testsuite/83888
10098         * doc/sourcebuild.texi (vect_float): Say that the selector
10099         only describes the situation when -funsafe-math-optimizations is on.
10100         (vect_float_strict): Document.
10101
10102 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
10103
10104         PR tree-optimization/83965
10105         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10106         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10107         instead of checking only for a reduction.
10108         (vect_recog_widen_sum_pattern): Likewise.
10109
10110 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
10111
10112         * predict.c (probably_never_executed): Only use precise profile info.
10113         (compute_function_frequency): Skip after inlining hack since we now
10114         have quality checking.
10115
10116 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
10117
10118         * profile-count.h (profile_probability::very_unlikely,
10119         profile_probability::unlikely, profile_probability::even): Set
10120         precision to guessed.
10121
10122 2018-01-23  Richard Biener  <rguenther@suse.de>
10123
10124         PR tree-optimization/83963
10125         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10126         Properly terminate dominator walk when crossing the exit edge not
10127         when visiting its source block.
10128
10129 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
10130
10131         PR c++/83918
10132         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10133         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10134
10135 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
10136
10137         PR tree-optimization/83957
10138         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
10139         semicolon after for body surrounded by braces.
10140
10141         PR tree-optimization/83081
10142         * profile-count.h (profile_probability::split): New method.
10143         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10144         Use profile_probability::split.
10145         (do_compare_rtx_and_jump): Fix adjustment of probabilities
10146         when splitting a single conditional jump into 2.
10147
10148 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
10149
10150         PR tree-optimization/69452
10151         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10152         decl.
10153
10154 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10155
10156         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10157         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10158         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10159
10160 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10161
10162         * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10163         * config/rl78/rl78.md (movdi): New define_expand.
10164         * config/rl78/rl78.c (rl78_split_movdi): New function.
10165
10166 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
10167
10168         PR target/83862
10169         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10170         no longer used.
10171         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10172         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10173         128-bit to produce an UNSPEC move to get the double word with the
10174         signbit and then a shift directly to do signbit.
10175         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10176         implementation with a new version that just does either a direct
10177         move or a regular move.  Move memory interface to separate insns.
10178         Move insns so they are next to the expander.
10179         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10180         with signbit move.  Split big and little endian case.
10181         (signbit<mode>2_dm_mem_le): Likewise.
10182         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10183         (signbit<mode>2_dm2): Likewise.
10184
10185 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10186
10187         * config/rl78/rl78.md (anddi3): New define_expand.
10188
10189 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10190
10191         * config/rl78/rl78.md (umindi3): New define_expand.
10192
10193 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10194
10195         * config/rl78/rl78.md (smindi3): New define_expand.
10196
10197 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10198
10199         * config/rl78/rl78.md (smaxdi3): New define_expand.
10200
10201 2018-01-22  Carl Love  <cel@us.ibm.com>
10202
10203         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10204         LVX_V1TI): Add macro expansion.
10205         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10206         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10207         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10208         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10209         Change check to determine if the instruction is a byte reversing
10210         entry.  Fix typo in comment.
10211         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10212         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10213         Add def_builtin calls for new builtins.
10214         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10215         Add define_insn expansion.
10216
10217 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10218
10219         * config/rl78/rl78.md (umaxdi3): New define_expand.
10220
10221 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10222
10223         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10224         for non-QImode registers.
10225
10226 2018-01-22  Richard Biener  <rguenther@suse.de>
10227
10228         PR tree-optimization/83963
10229         * graphite-scop-detection.c (scop_detection::get_sese): Delay
10230         including the loop exit block.
10231         (scop_detection::merge_sese): Likewise.
10232         (scop_detection::add_scop): Do it here instead.
10233
10234 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10235
10236         * doc/sourcebuild.texi (arm_softfloat): Document.
10237
10238 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
10239
10240         PR gcc/77734
10241         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10242         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10243         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10244
10245 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10246             David Edelsohn  <dje.gcc@gmail.com>
10247
10248         PR target/83946
10249         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10250         Change "crset eq" to "crset 2".
10251         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10252         (*call_indirect_aix<mode>_nospec): Likewise.
10253         (*call_value_indirect_aix<mode>_nospec): Likewise.
10254         (*call_indirect_elfv2<mode>_nospec): Likewise.
10255         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10256         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10257         change assembly output from . to $.
10258         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10259         (indirect_jump<mode>_nospec): Change assembly output from . to $.
10260         (*tablejump<mode>_internal1_nospec): Likewise.
10261
10262 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
10263
10264         PR target/80870
10265         * config/sh/sh_optimize_sett_clrt.cc:
10266         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10267
10268 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
10269
10270         PR tree-optimization/83940
10271         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10272         offset_dt to vect_constant_def rather than vect_unknown_def_type.
10273         (vect_check_load_store_mask): Add a mask_dt_out parameter and
10274         use it to pass back the definition type.
10275         (vect_check_store_rhs): Likewise rhs_dt_out.
10276         (vect_build_gather_load_calls): Add a mask_dt argument and use
10277         it instead of a call to vect_is_simple_use.
10278         (vectorizable_store): Update calls to vect_check_load_store_mask
10279         and vect_check_store_rhs.  Use the dt returned by the latter instead
10280         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
10281         instead of calls to vect_is_simple_use.  Pass the scalar rather
10282         than the vector operand to vect_is_simple_use when handling
10283         second and subsequent copies of an rhs value.
10284         (vectorizable_load): Update calls to vect_check_load_store_mask
10285         and vect_build_gather_load_calls.  Use the cached mask_dt and
10286         gs_info.offset_dt instead of calls to vect_is_simple_use.
10287
10288 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
10289
10290         PR middle-end/83945
10291         * tree-emutls.c: Include gimplify.h.
10292         (lower_emutls_2): New function.
10293         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10294         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10295         it before further processing.
10296
10297         PR target/83930
10298         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10299         UINTVAL (trueop1) instead of INTVAL (op1).
10300
10301 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
10302
10303         PR debug/81570
10304         PR debug/83728
10305         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10306         INCOMING_FRAME_SP_OFFSET if not defined.
10307         (scan_trace): Add ENTRY argument.  If true and
10308         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10309         emit a note to adjust the CFA offset.
10310         (create_cfi_notes): Adjust scan_trace callers.
10311         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10312         INCOMING_FRAME_SP_OFFSET in the CIE.
10313         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10314         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10315         Likewise.
10316         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10317         * doc/tm.texi: Regenerated.
10318
10319 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
10320
10321         PR rtl-optimization/83147
10322         * lra-constraints.c (remove_inheritance_pseudos): Use
10323         lra_substitute_pseudo_within_insn.
10324
10325 2018-01-19  Tom de Vries  <tom@codesourcery.com>
10326             Cesar Philippidis  <cesar@codesourcery.com>
10327
10328         PR target/83920
10329         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10330
10331 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
10332
10333         PR target/83790
10334         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10335         spaces for function labels.
10336
10337 2018-01-19  Martin Liska  <mliska@suse.cz>
10338
10339         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10340         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10341         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10342         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10343         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10344         (PRED_CONST_RETURN): Change from 69 to 65.
10345         (PRED_NULL_RETURN): Change from 91 to 71.
10346         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10347         (PRED_LOOP_GUARD): Change from 66 to 73.
10348
10349 2018-01-19  Martin Liska  <mliska@suse.cz>
10350
10351         * predict.c (predict_insn_def): Add new assert.
10352         (struct branch_predictor): Change type to signed integer.
10353         (test_prediction_value_range): Amend test to cover
10354         PROB_UNINITIALIZED.
10355         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10356         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10357         (PRED_LOOP_ITERATIONS_MAX): Likewise.
10358         (PRED_LOOP_IV_COMPARE): Likewise.
10359         * predict.h (PROB_UNINITIALIZED): Define new constant.
10360
10361 2018-01-19  Martin Liska  <mliska@suse.cz>
10362
10363         * predict.c (dump_prediction): Add new format for
10364         analyze_brprob.py script which is enabled with -details
10365         suboption.
10366         * profile-count.h (precise_p): New function.
10367
10368 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
10369
10370         PR tree-optimization/83922
10371         * tree-vect-loop.c (vect_verify_full_masking): Return false if
10372         there are no statements that need masking.
10373         (vect_active_double_reduction_p): New function.
10374         (vect_analyze_loop_operations): Use it when handling phis that
10375         are not in the loop header.
10376
10377 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
10378
10379         PR tree-optimization/83914
10380         * tree-vect-loop.c (vectorizable_induction): Don't convert
10381         init_expr or apply the peeling adjustment for inductions
10382         that are nested within the vectorized loop.
10383
10384 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10385
10386         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10387         instead of NEG.
10388
10389 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
10390
10391         PR sanitizer/81715
10392         PR testsuite/83882
10393         * function.h (gimplify_parameters): Add gimple_seq * argument.
10394         * function.c: Include gimple.h and options.h.
10395         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10396         for the added local temporaries if needed.
10397         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10398         if there are any parameter cleanups, wrap whole body into a
10399         try/finally with the cleanups.
10400
10401 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
10402
10403         PR target/82964
10404         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10405         Use GET_MODE_CLASS for scalar floating point.
10406
10407 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
10408
10409         PR ipa/82256
10410         patch by PaX Team
10411         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10412         Fix call of call_cgraph_insertion_hooks.
10413
10414 2018-01-18  Martin Sebor  <msebor@redhat.com>
10415
10416         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10417
10418 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
10419
10420         PR ipa/83619
10421         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10422         frequencies.
10423
10424 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
10425
10426         PR other/70268
10427         * common.opt: (-ffile-prefix-map): New option.
10428         * opts.c (common_handle_option): Defer it.
10429         * opts-global.c (handle_common_deferred_options): Handle it.
10430         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10431         * file-prefix-map.h: New file.
10432         (remap_debug_filename, add_debug_prefix_map): ...here.
10433         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10434         * final.c (debug_prefix_map, add_debug_prefix_map
10435         remap_debug_filename): Move to...
10436         * file-prefix-map.c: New file.
10437         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10438         generalize, get rid of alloca(), use strrchr() instead of strchr().
10439         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10440         Implement in terms of add_prefix_map().
10441         (remap_macro_filename, remap_debug_filename): Implement in term of
10442         remap_filename().
10443         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10444         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10445         * dbxout.c: Include file-prefix-map.h.
10446         * varasm.c: Likewise.
10447         * vmsdbgout.c: Likewise.
10448         * xcoffout.c: Likewise.
10449         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10450         * doc/cppopts.texi (-fmacro-prefix-map): Document.
10451         * doc/invoke.texi (-ffile-prefix-map): Document.
10452         (-fdebug-prefix-map): Update description.
10453
10454 2018-01-18  Martin Liska  <mliska@suse.cz>
10455
10456         * config/i386/i386.c (indirect_thunk_name): Document that also
10457         lfence is emitted.
10458         (output_indirect_thunk): Document why both instructions
10459         (pause and lfence) are generated.
10460
10461 2018-01-18  Richard Biener  <rguenther@suse.de>
10462
10463         PR tree-optimization/83887
10464         * graphite-scop-detection.c
10465         (scop_detection::get_nearest_dom_with_single_entry): Remove.
10466         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10467         (scop_detection::merge_sese): Re-implement with a flood-fill
10468         algorithm that properly finds a SESE region if it exists.
10469
10470 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
10471
10472         PR c/61240
10473         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10474         pointer_diff optimizations use view_convert instead of convert.
10475
10476 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10477
10478         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10479         Generate different code for -mno-speculate-indirect-jumps.
10480         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10481         (*call_indirect_aix<mode>): Disable for
10482         -mno-speculate-indirect-jumps.
10483         (*call_indirect_aix<mode>_nospec): New define_insn.
10484         (*call_value_indirect_aix<mode>): Disable for
10485         -mno-speculate-indirect-jumps.
10486         (*call_value_indirect_aix<mode>_nospec): New define_insn.
10487         (*sibcall_nonlocal_sysv<mode>): Generate different code for
10488         -mno-speculate-indirect-jumps.
10489         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10490
10491 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
10492
10493         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10494         long double type, set the flags for noting the default long double
10495         type, even if we don't pass or return a long double type.
10496
10497 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
10498
10499         PR ipa/83051
10500         * ipa-inline.c (flatten_function): Do not overwrite final inlining
10501         failure.
10502
10503 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
10504
10505         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10506         support for merge[hl].
10507         (fold_mergehl_helper): New helper function.
10508         (tree-vector-builder.h): New #include for tree_vector_builder usage.
10509         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10510         (altivec_vmrglw_direct): Add xxmrglw insn.
10511
10512 2018-01-17  Andrew Waterman  <andrew@sifive.com>
10513
10514         * config/riscv/riscv.c (riscv_conditional_register_usage): If
10515         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10516
10517 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
10518
10519         PR lto/83121
10520         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10521         call the lto_location_cache before reading the
10522         DECL_SOURCE_LOCATION of the types.
10523
10524 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
10525             Richard Sandiford  <richard.sandiford@linaro.org>
10526
10527         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10528         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10529         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10530         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10531         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10532         Add declaration.
10533         * config/aarch64/constraints.md (aarch64_movti_operand):
10534         Limit immediates.
10535         * config/aarch64/predicates.md (Uti): Add new constraint.
10536
10537 2018-01-17 Carl Love  <cel@us.ibm.com>
10538
10539         * config/rs6000/vsx.md (define_expand xl_len_r,
10540         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10541         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10542         lxvll.
10543         (define_expand, define_insn): Move the shift left from  the
10544         define_insn to the define_expand for lxvl and stxvl instructions.
10545         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10546         and XL_LEN_R definitions to PURE.
10547
10548 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
10549
10550         * config/i386/i386.c (indirect_thunk_name): Declare regno
10551         as unsigned int.  Compare regno with INVALID_REGNUM.
10552         (output_indirect_thunk): Ditto.
10553         (output_indirect_thunk_function): Ditto.
10554         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
10555         in the call to output_indirect_thunk_function.
10556
10557 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
10558
10559         PR middle-end/83884
10560         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10561         rather than the size of inner_type to determine the stack slot size
10562         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10563
10564 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
10565
10566         PR target/83546
10567         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10568         to PTA_SILVERMONT.
10569
10570 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
10571
10572         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10573         endian Linux systems to optionally enable multilibs for selecting
10574         the long double type if the user configured an explicit type.
10575         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10576         have no long double multilibs if not defined.
10577         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10578         warn if the user used -mabi={ieee,ibm}longdouble and we built
10579         multilibs for long double.
10580         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10581         appropriate multilib option.
10582         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10583         multilib options.
10584         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10585         for building long double multilibs.
10586         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10587
10588 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
10589
10590         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10591         copies.
10592
10593         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10594         64 bits.
10595         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10596         128 bits.
10597
10598         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10599         variables.
10600
10601         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10602         return value.
10603
10604 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
10605
10606         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10607         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10608
10609 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10610
10611         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10612         different rtl trees depending on TARGET_64BIT.
10613         (rs6000_gen_lvx): Likewise.
10614
10615 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
10616
10617         * config/visium/visium.md (nop): Tweak comment.
10618         (hazard_nop): Likewise.
10619
10620 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10621
10622         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10623         -mspeculate-indirect-jumps.
10624         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10625         for -mno-speculate-indirect-jumps.
10626         (*call_indirect_elfv2<mode>_nospec): New define_insn.
10627         (*call_value_indirect_elfv2<mode>): Disable for
10628         -mno-speculate-indirect-jumps.
10629         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10630         (indirect_jump): Emit different RTL for
10631         -mno-speculate-indirect-jumps.
10632         (*indirect_jump<mode>): Disable for
10633         -mno-speculate-indirect-jumps.
10634         (*indirect_jump<mode>_nospec): New define_insn.
10635         (tablejump): Emit different RTL for
10636         -mno-speculate-indirect-jumps.
10637         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10638         (tablejumpsi_nospec): New define_expand.
10639         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10640         (tablejumpdi_nospec): New define_expand.
10641         (*tablejump<mode>_internal1): Disable for
10642         -mno-speculate-indirect-jumps.
10643         (*tablejump<mode>_internal1_nospec): New define_insn.
10644         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10645         option.
10646
10647 2018-01-16  Artyom Skrobov tyomitch@gmail.com
10648
10649         * caller-save.c (insert_save): Drop unnecessary parameter.  All
10650         callers updated.
10651
10652 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10653             Richard Biener  <rguenth@suse.de>
10654
10655         PR libgomp/83590
10656         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10657         return early, inline manually is_gimple_sizepos.  Make sure if we
10658         call gimplify_expr we don't end up with a gimple constant.
10659         * tree.c (variably_modified_type_p): Don't return true for
10660         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
10661         * gimplify.h (is_gimple_sizepos): Remove.
10662
10663 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10664
10665         PR tree-optimization/83857
10666         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10667         vectorizable_live_operation for pure SLP statements.
10668         (vectorizable_live_operation): Handle PHIs.
10669
10670 2018-01-16  Richard Biener  <rguenther@suse.de>
10671
10672         PR tree-optimization/83867
10673         * tree-vect-stmts.c (vect_transform_stmt): Precompute
10674         nested_in_vect_loop_p since the scalar stmt may get invalidated.
10675
10676 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10677
10678         PR c/83844
10679         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10680         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10681         If off is not INTEGER_CST, issue a may not be aligned warning
10682         rather than isn't aligned.  Use isn%'t rather than isn't.
10683         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10684         into MULT_EXPR.
10685         <case MULT_EXPR>: Improve the case when bottom and one of the
10686         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10687         operand, in that case check if the other operand is multiple of
10688         bottom divided by the INTEGER_CST operand.
10689
10690 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10691
10692         PR target/83858
10693         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10694         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10695         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10696         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10697         * config/pa/pa.c (pa_function_arg_advance): Likewise.
10698         (pa_function_arg, pa_arg_partial_bytes): Likewise.
10699         (pa_function_arg_size): New function.
10700
10701 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10702
10703         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10704         in a separate statement.
10705
10706 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10707
10708         PR tree-optimization/83847
10709         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10710         group gathers and scatters.
10711
10712 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10713
10714         PR rtl-optimization/86620
10715         * params.def (max-sched-ready-insns): Bump minimum value to 1.
10716
10717         PR rtl-optimization/83213
10718         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10719         to last if both are JUMP_INSNs.
10720
10721         PR tree-optimization/83843
10722         * gimple-ssa-store-merging.c
10723         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10724         store_immediate_info for bswap/nop orig_stores.
10725
10726 2018-01-15  Andrew Waterman  <andrew@sifive.com>
10727
10728         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10729         !TARGET_MUL.
10730         <UDIV>: Increase cost if !TARGET_DIV.
10731
10732 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
10733
10734         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10735         (define_attr "cr_logical_3op"): New.
10736         (cceq_ior_compare): Adjust.
10737         (cceq_ior_compare_complement): Adjust.
10738         (*cceq_rev_compare): Adjust.
10739         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10740         (is_cracked_insn): Adjust.
10741         (insn_must_be_first_in_group): Adjust.
10742         * config/rs6000/40x.md: Adjust.
10743         * config/rs6000/440.md: Adjust.
10744         * config/rs6000/476.md: Adjust.
10745         * config/rs6000/601.md: Adjust.
10746         * config/rs6000/603.md: Adjust.
10747         * config/rs6000/6xx.md: Adjust.
10748         * config/rs6000/7450.md: Adjust.
10749         * config/rs6000/7xx.md: Adjust.
10750         * config/rs6000/8540.md: Adjust.
10751         * config/rs6000/cell.md: Adjust.
10752         * config/rs6000/e300c2c3.md: Adjust.
10753         * config/rs6000/e500mc.md: Adjust.
10754         * config/rs6000/e500mc64.md: Adjust.
10755         * config/rs6000/e5500.md: Adjust.
10756         * config/rs6000/e6500.md: Adjust.
10757         * config/rs6000/mpc.md: Adjust.
10758         * config/rs6000/power4.md: Adjust.
10759         * config/rs6000/power5.md: Adjust.
10760         * config/rs6000/power6.md: Adjust.
10761         * config/rs6000/power7.md: Adjust.
10762         * config/rs6000/power8.md: Adjust.
10763         * config/rs6000/power9.md: Adjust.
10764         * config/rs6000/rs64.md: Adjust.
10765         * config/rs6000/titan.md: Adjust.
10766
10767 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10768
10769         * config/i386/predicates.md (indirect_branch_operand): Rewrite
10770         ix86_indirect_branch_register logic.
10771
10772 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10773
10774         * config/i386/constraints.md (Bs): Update
10775         ix86_indirect_branch_register check.  Don't check
10776         ix86_indirect_branch_register with GOT_memory_operand.
10777         (Bw): Likewise.
10778         * config/i386/predicates.md (GOT_memory_operand): Don't check
10779         ix86_indirect_branch_register here.
10780         (GOT32_symbol_operand): Likewise.
10781
10782 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10783
10784         * config/i386/predicates.md (constant_call_address_operand):
10785         Rewrite ix86_indirect_branch_register logic.
10786         (sibcall_insn_operand): Likewise.
10787
10788 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10789
10790         * config/i386/constraints.md (Bs): Replace
10791         ix86_indirect_branch_thunk_register with
10792         ix86_indirect_branch_register.
10793         (Bw): Likewise.
10794         * config/i386/i386.md (indirect_jump): Likewise.
10795         (tablejump): Likewise.
10796         (*sibcall_memory): Likewise.
10797         (*sibcall_value_memory): Likewise.
10798         Peepholes of indirect call and jump via memory: Likewise.
10799         * config/i386/i386.opt: Likewise.
10800         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10801         (GOT_memory_operand): Likewise.
10802         (call_insn_operand): Likewise.
10803         (sibcall_insn_operand): Likewise.
10804         (GOT32_symbol_operand): Likewise.
10805
10806 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10807
10808         PR middle-end/83837
10809         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10810         type rather than type addr's type points to.
10811         (expand_omp_atomic_mutex): Likewise.
10812         (expand_omp_atomic): Likewise.
10813
10814 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10815
10816         PR target/83839
10817         * config/i386/i386.c (output_indirect_thunk_function): Use
10818         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10819         for  __x86_return_thunk.
10820
10821 2018-01-15  Richard Biener  <rguenther@suse.de>
10822
10823         PR middle-end/83850
10824         * expmed.c (extract_bit_field_1): Fix typo.
10825
10826 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10827
10828         PR target/83687
10829         * config/arm/iterators.md (VF): New mode iterator.
10830         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10831         Remove integer-related logic from pattern.
10832         (neon_vabd<mode>_3): Likewise.
10833
10834 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10835
10836         PR middle-end/82694
10837         * common.opt (fstrict-overflow): No longer an alias.
10838         (fwrapv-pointer): New option.
10839         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10840         also for pointer types based on flag_wrapv_pointer.
10841         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10842         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10843         opts->x_flag_wrapv got set.
10844         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10845         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10846         POINTER_TYPE_OVERFLOW_UNDEFINED.
10847         * match.pd: Likewise in address comparison pattern.
10848         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10849
10850 2018-01-15  Richard Biener  <rguenther@suse.de>
10851
10852         PR lto/83804
10853         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10854         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
10855         Reset type names to their identifier if their TYPE_DECL doesn't
10856         have linkage (and thus is used for ODR and devirt).
10857         (save_debug_info_for_decl): Remove.
10858         (save_debug_info_for_type): Likewise.
10859         (add_tree_to_fld_list): Adjust.
10860         * tree-pretty-print.c (dump_generic_node): Make dumping of
10861         type names more robust.
10862
10863 2018-01-15  Richard Biener  <rguenther@suse.de>
10864
10865         * BASE-VER: Bump to 8.0.1.
10866
10867 2018-01-14  Martin Sebor  <msebor@redhat.com>
10868
10869         PR other/83508
10870         * builtins.c (check_access): Avoid warning when the no-warning bit
10871         is set.
10872
10873 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
10874
10875         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10876         * ira-color (allocno_hard_regs_compare): Likewise.
10877
10878 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
10879
10880         PR target/83013
10881         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10882         Use .pushsection/.popsection.
10883
10884 2018-01-14  Martin Sebor  <msebor@redhat.com>
10885
10886         PR c++/81327
10887         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10888
10889 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
10890
10891         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10892         entry from extra_headers.
10893         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10894         extra_headers, make the list bitwise identical to the i?86-*-* one.
10895
10896 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10897
10898         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10899         -mcmodel=large with -mindirect-branch=thunk,
10900         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10901         -mfunction-return=thunk-extern.
10902         * doc/invoke.texi: Document -mcmodel=large is incompatible with
10903         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10904         -mfunction-return=thunk and -mfunction-return=thunk-extern.
10905
10906 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10907
10908         * config/i386/i386.c (print_reg): Print the name of the full
10909         integer register without '%'.
10910         (ix86_print_operand): Handle 'V'.
10911          * doc/extend.texi: Document 'V' modifier.
10912
10913 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10914
10915         * config/i386/constraints.md (Bs): Disallow memory operand for
10916         -mindirect-branch-register.
10917         (Bw): Likewise.
10918         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10919         (GOT_memory_operand): Likewise.
10920         (call_insn_operand): Likewise.
10921         (sibcall_insn_operand): Likewise.
10922         (GOT32_symbol_operand): Likewise.
10923         * config/i386/i386.md (indirect_jump): Call convert_memory_address
10924         for -mindirect-branch-register.
10925         (tablejump): Likewise.
10926         (*sibcall_memory): Likewise.
10927         (*sibcall_value_memory): Likewise.
10928         Disallow peepholes of indirect call and jump via memory for
10929         -mindirect-branch-register.
10930         (*call_pop): Replace m with Bw.
10931         (*call_value_pop): Likewise.
10932         (*sibcall_pop_memory): Replace m with Bs.
10933         * config/i386/i386.opt (mindirect-branch-register): New option.
10934         * doc/invoke.texi: Document -mindirect-branch-register option.
10935
10936 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10937
10938         * config/i386/i386-protos.h (ix86_output_function_return): New.
10939         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10940         set function_return_type.
10941         (indirect_thunk_name): Add ret_p to indicate thunk for function
10942         return.
10943         (output_indirect_thunk_function): Pass false to
10944         indirect_thunk_name.
10945         (ix86_output_indirect_branch_via_reg): Likewise.
10946         (ix86_output_indirect_branch_via_push): Likewise.
10947         (output_indirect_thunk_function): Create alias for function
10948         return thunk if regno < 0.
10949         (ix86_output_function_return): New function.
10950         (ix86_handle_fndecl_attribute): Handle function_return.
10951         (ix86_attribute_table): Add function_return.
10952         * config/i386/i386.h (machine_function): Add
10953         function_return_type.
10954         * config/i386/i386.md (simple_return_internal): Use
10955         ix86_output_function_return.
10956         (simple_return_internal_long): Likewise.
10957         * config/i386/i386.opt (mfunction-return=): New option.
10958         (indirect_branch): Mention -mfunction-return=.
10959         * doc/extend.texi: Document function_return function attribute.
10960         * doc/invoke.texi: Document -mfunction-return= option.
10961
10962 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10963
10964         * config/i386/i386-opts.h (indirect_branch): New.
10965         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10966         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10967         with local indirect jump when converting indirect call and jump.
10968         (ix86_set_indirect_branch_type): New.
10969         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10970         (indirectlabelno): New.
10971         (indirect_thunk_needed): Likewise.
10972         (indirect_thunk_bnd_needed): Likewise.
10973         (indirect_thunks_used): Likewise.
10974         (indirect_thunks_bnd_used): Likewise.
10975         (INDIRECT_LABEL): Likewise.
10976         (indirect_thunk_name): Likewise.
10977         (output_indirect_thunk): Likewise.
10978         (output_indirect_thunk_function): Likewise.
10979         (ix86_output_indirect_branch_via_reg): Likewise.
10980         (ix86_output_indirect_branch_via_push): Likewise.
10981         (ix86_output_indirect_branch): Likewise.
10982         (ix86_output_indirect_jmp): Likewise.
10983         (ix86_code_end): Call output_indirect_thunk_function if needed.
10984         (ix86_output_call_insn): Call ix86_output_indirect_branch if
10985         needed.
10986         (ix86_handle_fndecl_attribute): Handle indirect_branch.
10987         (ix86_attribute_table): Add indirect_branch.
10988         * config/i386/i386.h (machine_function): Add indirect_branch_type
10989         and has_local_indirect_jump.
10990         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10991         to true.
10992         (tablejump): Likewise.
10993         (*indirect_jump): Use ix86_output_indirect_jmp.
10994         (*tablejump_1): Likewise.
10995         (simple_return_indirect_internal): Likewise.
10996         * config/i386/i386.opt (mindirect-branch=): New option.
10997         (indirect_branch): New.
10998         (keep): Likewise.
10999         (thunk): Likewise.
11000         (thunk-inline): Likewise.
11001         (thunk-extern): Likewise.
11002         * doc/extend.texi: Document indirect_branch function attribute.
11003         * doc/invoke.texi: Document -mindirect-branch= option.
11004
11005 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
11006
11007         PR ipa/83051
11008         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11009
11010 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
11011
11012         * ipa-inline.c (want_inline_small_function_p): Return false if
11013         inlining has already failed with CIF_FINAL_ERROR.
11014         (update_caller_keys): Call want_inline_small_function_p before
11015         can_inline_edge_p.
11016         (update_callee_keys): Likewise.
11017
11018 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11019
11020         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11021         New function.
11022         (rs6000_quadword_masked_address_p): Likewise.
11023         (quad_aligned_load_p): Likewise.
11024         (quad_aligned_store_p): Likewise.
11025         (const_load_sequence_p): Add comment to describe the outer-most loop.
11026         (mimic_memory_attributes_and_flags): New function.
11027         (rs6000_gen_stvx): Likewise.
11028         (replace_swapped_aligned_store): Likewise.
11029         (rs6000_gen_lvx): Likewise.
11030         (replace_swapped_aligned_load): Likewise.
11031         (replace_swapped_load_constant): Capitalize argument name in
11032         comment describing this function.
11033         (rs6000_analyze_swaps): Add a third pass to search for vector loads
11034         and stores that access quad-word aligned addresses and replace
11035         with stvx or lvx instructions when appropriate.
11036         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11037         New function prototype.
11038         (rs6000_quadword_masked_address_p): Likewise.
11039         (rs6000_gen_lvx): Likewise.
11040         (rs6000_gen_stvx): Likewise.
11041         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11042         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11043         when memory address is aligned.
11044         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11045         this split to select lvx instruction when memory address is aligned.
11046         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11047         instruction when memory address is aligned.
11048         (*vsx_le_perm_load_v16qi): Likewise.
11049         (four unnamed splitters): Modify to select the stvx instruction
11050         when memory is aligned.
11051
11052 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
11053
11054         * predict.c (determine_unlikely_bbs): Handle correctly BBs
11055         which appears in the queue multiple times.
11056
11057 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11058             Alan Hayward  <alan.hayward@arm.com>
11059             David Sherwood  <david.sherwood@arm.com>
11060
11061         * tree-vectorizer.h (vec_lower_bound): New structure.
11062         (_loop_vec_info): Add check_nonzero and lower_bounds.
11063         (LOOP_VINFO_CHECK_NONZERO): New macro.
11064         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11065         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11066         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11067         fields.  Make seg_len the distance travelled, not including the
11068         access size.
11069         (dr_direction_indicator): Declare.
11070         (dr_zero_step_indicator): Likewise.
11071         (dr_known_forward_stride_p): Likewise.
11072         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11073         tree-ssanames.h.
11074         (runtime_alias_check_p): Allow runtime alias checks with
11075         variable strides.
11076         (operator ==): Compare access_size and align.
11077         (prune_runtime_alias_test_list): Rework for new distinction between
11078         the access_size and seg_len.
11079         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
11080         segment lengths.
11081         (get_segment_min_max): New function.
11082         (create_intersect_range_checks): Use it.
11083         (dr_step_indicator): New function.
11084         (dr_direction_indicator): Likewise.
11085         (dr_zero_step_indicator): Likewise.
11086         (dr_known_forward_stride_p): Likewise.
11087         * tree-loop-distribution.c (data_ref_segment_size): Return
11088         DR_STEP * (niters - 1).
11089         (compute_alias_check_pairs): Update call to the dr_with_seg_len
11090         constructor.
11091         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11092         (vect_preserves_scalar_order_p): New function, split out from...
11093         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
11094         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11095         (vect_vfa_access_size): New function.
11096         (vect_vfa_align): Likewise.
11097         (vect_compile_time_alias): Take access_size_a and access_b arguments.
11098         (dump_lower_bound): New function.
11099         (vect_check_lower_bound): Likewise.
11100         (vect_small_gap_p): Likewise.
11101         (vectorizable_with_step_bound_p): Likewise.
11102         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11103         depencies if the vectorization factor is 1.  Convert the checks
11104         for nonzero steps into checks on the bounds of DR_STEP.  Try using
11105         a bunds check for variable steps if the minimum required step is
11106         relatively small. Update calls to the dr_with_seg_len
11107         constructor and to vect_compile_time_alias.
11108         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11109         function.
11110         (vect_loop_versioning): Call it.
11111         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11112         when retrying.
11113         (vect_estimate_min_profitable_iters): Account for any bounds checks.
11114
11115 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11116             Alan Hayward  <alan.hayward@arm.com>
11117             David Sherwood  <david.sherwood@arm.com>
11118
11119         * doc/sourcebuild.texi (vect_scatter_store): Document.
11120         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11121         optabs.
11122         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11123         Document.
11124         * genopinit.c (main): Add supports_vec_scatter_store and
11125         supports_vec_scatter_store_cached to target_optabs.
11126         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11127         IFN_MASK_SCATTER_STORE.
11128         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11129         functions.
11130         * internal-fn.h (internal_store_fn_p): Declare.
11131         (internal_fn_stored_value_index): Likewise.
11132         * internal-fn.c (scatter_store_direct): New macro.
11133         (expand_scatter_store_optab_fn): New function.
11134         (direct_scatter_store_optab_supported_p): New macro.
11135         (internal_store_fn_p): New function.
11136         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11137         IFN_MASK_SCATTER_STORE.
11138         (internal_fn_mask_index): Likewise.
11139         (internal_fn_stored_value_index): New function.
11140         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11141         for scatter stores.
11142         * optabs-query.h (supports_vec_scatter_store_p): Declare.
11143         * optabs-query.c (supports_vec_scatter_store_p): New function.
11144         * tree-vectorizer.h (vect_get_store_rhs): Declare.
11145         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11146         true for scatter stores.
11147         (vect_gather_scatter_fn_p): Handle scatter stores too.
11148         (vect_check_gather_scatter): Consider using scatter stores if
11149         supports_vec_scatter_store_p.
11150         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11151         scatter stores too.
11152         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11153         internal_fn_stored_value_index.
11154         (check_load_store_masking): Handle scatter stores too.
11155         (vect_get_store_rhs): Make public.
11156         (vectorizable_call): Use internal_store_fn_p.
11157         (vectorizable_store): Handle scatter store internal functions.
11158         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11159         when deciding whether the end of the group has been reached.
11160         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11161         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11162         (mask_scatter_store<mode>): New insns.
11163
11164 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11165             Alan Hayward  <alan.hayward@arm.com>
11166             David Sherwood  <david.sherwood@arm.com>
11167
11168         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11169         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11170         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11171         function.
11172         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11173         Use vect_truncate_gather_scatter_offset if we can't treat the
11174         operation as a normal gather load or scatter store.
11175         (get_group_load_store_type): Take the gather_scatter_info
11176         as argument.  Try using a gather load or scatter store for
11177         single-element groups.
11178         (get_load_store_type): Update calls to get_group_load_store_type
11179         and vect_use_strided_gather_scatters_p.
11180
11181 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11182             Alan Hayward  <alan.hayward@arm.com>
11183             David Sherwood  <david.sherwood@arm.com>
11184
11185         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11186         optional tree argument.
11187         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11188         null target hooks.
11189         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11190         but continue to use the current value as a fallback.
11191         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11192         to compare the updates.
11193         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11194         (get_load_store_type): Use it when handling a strided access.
11195         (vect_get_strided_load_store_ops): New function.
11196         (vect_get_data_ptr_increment): Likewise.
11197         (vectorizable_load): Handle strided gather loads.  Always pass
11198         a step to vect_create_data_ref_ptr and bump_vector_ptr.
11199
11200 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11201             Alan Hayward  <alan.hayward@arm.com>
11202             David Sherwood  <david.sherwood@arm.com>
11203
11204         * doc/md.texi (gather_load@var{m}): Document.
11205         (mask_gather_load@var{m}): Likewise.
11206         * genopinit.c (main): Add supports_vec_gather_load and
11207         supports_vec_gather_load_cached to target_optabs.
11208         * optabs-tree.c (init_tree_optimization_optabs): Use
11209         ggc_cleared_alloc to allocate target_optabs.
11210         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11211         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11212         functions.
11213         * internal-fn.h (internal_load_fn_p): Declare.
11214         (internal_gather_scatter_fn_p): Likewise.
11215         (internal_fn_mask_index): Likewise.
11216         (internal_gather_scatter_fn_supported_p): Likewise.
11217         * internal-fn.c (gather_load_direct): New macro.
11218         (expand_gather_load_optab_fn): New function.
11219         (direct_gather_load_optab_supported_p): New macro.
11220         (direct_internal_fn_optab): New function.
11221         (internal_load_fn_p): Likewise.
11222         (internal_gather_scatter_fn_p): Likewise.
11223         (internal_fn_mask_index): Likewise.
11224         (internal_gather_scatter_fn_supported_p): Likewise.
11225         * optabs-query.c (supports_at_least_one_mode_p): New function.
11226         (supports_vec_gather_load_p): Likewise.
11227         * optabs-query.h (supports_vec_gather_load_p): Declare.
11228         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11229         and memory_type field.
11230         (NUM_PATTERNS): Bump to 15.
11231         * tree-vect-data-refs.c: Include internal-fn.h.
11232         (vect_gather_scatter_fn_p): New function.
11233         (vect_describe_gather_scatter_call): Likewise.
11234         (vect_check_gather_scatter): Try using internal functions for
11235         gather loads.  Recognize existing calls to a gather load function.
11236         (vect_analyze_data_refs): Consider using gather loads if
11237         supports_vec_gather_load_p.
11238         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11239         (vect_get_gather_scatter_offset_type): Likewise.
11240         (vect_convert_mask_for_vectype): Likewise.
11241         (vect_add_conversion_to_patterm): Likewise.
11242         (vect_try_gather_scatter_pattern): Likewise.
11243         (vect_recog_gather_scatter_pattern): New pattern recognizer.
11244         (vect_vect_recog_func_ptrs): Add it.
11245         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11246         internal_fn_mask_index and internal_gather_scatter_fn_p.
11247         (check_load_store_masking): Take the gather_scatter_info as an
11248         argument and handle gather loads.
11249         (vect_get_gather_scatter_ops): New function.
11250         (vectorizable_call): Check internal_load_fn_p.
11251         (vectorizable_load): Likewise.  Handle gather load internal
11252         functions.
11253         (vectorizable_store): Update call to check_load_store_masking.
11254         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11255         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11256         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11257         (aarch64_gather_scale_operand_d): New predicates.
11258         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11259         (mask_gather_load<mode>): New insns.
11260
11261 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11262             Alan Hayward  <alan.hayward@arm.com>
11263             David Sherwood  <david.sherwood@arm.com>
11264
11265         * optabs.def (fold_left_plus_optab): New optab.
11266         * doc/md.texi (fold_left_plus_@var{m}): Document.
11267         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11268         * internal-fn.c (fold_left_direct): Define.
11269         (expand_fold_left_optab_fn): Likewise.
11270         (direct_fold_left_optab_supported_p): Likewise.
11271         * fold-const-call.c (fold_const_fold_left): New function.
11272         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11273         * tree-parloops.c (valid_reduction_p): New function.
11274         (gather_scalar_reductions): Use it.
11275         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11276         (vect_finish_replace_stmt): Declare.
11277         * tree-vect-loop.c (fold_left_reduction_fn): New function.
11278         (needs_fold_left_reduction_p): New function, split out from...
11279         (vect_is_simple_reduction): ...here.  Accept reductions that
11280         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11281         (vect_force_simple_reduction): Also store the reduction type in
11282         the assignment's STMT_VINFO_REDUC_TYPE.
11283         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11284         (merge_with_identity): New function.
11285         (vect_expand_fold_left): Likewise.
11286         (vectorize_fold_left_reduction): Likewise.
11287         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
11288         scalar phi in place for it.  Check for target support and reject
11289         cases that would reassociate the operation.  Defer the transform
11290         phase to vectorize_fold_left_reduction.
11291         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11292         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11293         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11294
11295 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11296
11297         * tree-if-conv.c (predicate_mem_writes): Remove redundant
11298         call to ifc_temp_var.
11299
11300 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11301             Alan Hayward  <alan.hayward@arm.com>
11302             David Sherwood  <david.sherwood@arm.com>
11303
11304         * target.def (legitimize_address_displacement): Take the original
11305         offset as a poly_int.
11306         * targhooks.h (default_legitimize_address_displacement): Update
11307         accordingly.
11308         * targhooks.c (default_legitimize_address_displacement): Likewise.
11309         * doc/tm.texi: Regenerate.
11310         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11311         as an argument, moving assert of ad->disp == ad->disp_term to...
11312         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
11313         Try calling targetm.legitimize_address_displacement before expanding
11314         the address rather than afterwards, and adjust for the new interface.
11315         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11316         Match the new hook interface.  Handle SVE addresses.
11317         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11318         new hook interface.
11319
11320 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11321
11322         * Makefile.in (OBJS): Add early-remat.o.
11323         * target.def (select_early_remat_modes): New hook.
11324         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11325         * doc/tm.texi: Regenerate.
11326         * targhooks.h (default_select_early_remat_modes): Declare.
11327         * targhooks.c (default_select_early_remat_modes): New function.
11328         * timevar.def (TV_EARLY_REMAT): New timevar.
11329         * passes.def (pass_early_remat): New pass.
11330         * tree-pass.h (make_pass_early_remat): Declare.
11331         * early-remat.c: New file.
11332         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11333         function.
11334         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11335
11336 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11337             Alan Hayward  <alan.hayward@arm.com>
11338             David Sherwood  <david.sherwood@arm.com>
11339
11340         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11341         vfm1 with a bound_epilog parameter.
11342         (vect_do_peeling): Update calls accordingly, and move the prologue
11343         call earlier in the function.  Treat the base bound_epilog as 0 for
11344         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
11345         this base when peeling for gaps.
11346         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11347         with fully-masked loops.
11348         (vect_estimate_min_profitable_iters): Handle the single peeled
11349         iteration in that case.
11350
11351 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11352             Alan Hayward  <alan.hayward@arm.com>
11353             David Sherwood  <david.sherwood@arm.com>
11354
11355         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11356         single-element interleaving even if the size is not a power of 2.
11357         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11358         accesses for single-element interleaving if the group size is
11359         not a power of 2.
11360
11361 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11362             Alan Hayward  <alan.hayward@arm.com>
11363             David Sherwood  <david.sherwood@arm.com>
11364
11365         * doc/md.texi (fold_extract_last_@var{m}): Document.
11366         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11367         * optabs.def (fold_extract_last_optab): New optab.
11368         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11369         * internal-fn.c (fold_extract_direct): New macro.
11370         (expand_fold_extract_optab_fn): Likewise.
11371         (direct_fold_extract_optab_supported_p): Likewise.
11372         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11373         * tree-vect-loop.c (vect_model_reduction_cost): Handle
11374         EXTRACT_LAST_REDUCTION.
11375         (get_initial_def_for_reduction): Do not create an initial vector
11376         for EXTRACT_LAST_REDUCTION reductions.
11377         (vectorizable_reduction): Leave the scalar phi in place for
11378         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
11379         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
11380         epilogue code for EXTRACT_LAST_REDUCTION and defer the
11381         transform phase to vectorizable_condition.
11382         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11383         split out from...
11384         (vect_finish_stmt_generation): ...here.
11385         (vect_finish_replace_stmt): New function.
11386         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11387         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11388         pattern.
11389         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11390
11391 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11392             Alan Hayward  <alan.hayward@arm.com>
11393             David Sherwood  <david.sherwood@arm.com>
11394
11395         * doc/md.texi (extract_last_@var{m}): Document.
11396         * optabs.def (extract_last_optab): New optab.
11397         * internal-fn.def (EXTRACT_LAST): New internal function.
11398         * internal-fn.c (cond_unary_direct): New macro.
11399         (expand_cond_unary_optab_fn): Likewise.
11400         (direct_cond_unary_optab_supported_p): Likewise.
11401         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11402         loops using EXTRACT_LAST.
11403         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11404         (extract_last_<mode>): ...this optab.
11405         (vec_extract<mode><Vel>): Update accordingly.
11406
11407 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11408             Alan Hayward  <alan.hayward@arm.com>
11409             David Sherwood  <david.sherwood@arm.com>
11410
11411         * target.def (empty_mask_is_expensive): New hook.
11412         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11413         * doc/tm.texi: Regenerate.
11414         * targhooks.h (default_empty_mask_is_expensive): Declare.
11415         * targhooks.c (default_empty_mask_is_expensive): New function.
11416         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11417         if the target says that empty masks are expensive.
11418         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11419         New function.
11420         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11421
11422 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11423             Alan Hayward  <alan.hayward@arm.com>
11424             David Sherwood  <david.sherwood@arm.com>
11425
11426         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11427         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11428         (vect_use_loop_mask_for_alignment_p): New function.
11429         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11430         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11431         niters_skip argument.  Make sure that the first niters_skip elements
11432         of the first iteration are inactive.
11433         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11434         Update call to vect_set_loop_masks_directly.
11435         (get_misalign_in_elems): New function, split out from...
11436         (vect_gen_prolog_loop_niters): ...here.
11437         (vect_update_init_of_dr): Take a code argument that specifies whether
11438         the adjustment should be added or subtracted.
11439         (vect_update_init_of_drs): Likewise.
11440         (vect_prepare_for_masked_peels): New function.
11441         (vect_do_peeling): Skip prologue peeling if we're using a mask
11442         instead.  Update call to vect_update_inits_of_drs.
11443         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11444         mask_skip_niters.
11445         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11446         alignment.  Do not include the number of peeled iterations in
11447         the minimum threshold in that case.
11448         (vectorizable_induction): Adjust the start value down by
11449         LOOP_VINFO_MASK_SKIP_NITERS iterations.
11450         (vect_transform_loop): Call vect_prepare_for_masked_peels.
11451         Take the number of skipped iterations into account when calculating
11452         the loop bounds.
11453         * tree-vect-stmts.c (vect_gen_while_not): New function.
11454
11455 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11456             Alan Hayward  <alan.hayward@arm.com>
11457             David Sherwood  <david.sherwood@arm.com>
11458
11459         * doc/sourcebuild.texi (vect_fully_masked): Document.
11460         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11461         default value to 0.
11462         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11463         split out from...
11464         (vect_analyze_loop_2): ...here. Don't check the vectorization
11465         factor against the number of loop iterations if the loop is
11466         fully-masked.
11467
11468 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11469             Alan Hayward  <alan.hayward@arm.com>
11470             David Sherwood  <david.sherwood@arm.com>
11471
11472         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11473         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11474         (dump_groups): Update accordingly.
11475         (iv_use::mem_type): New member variable.
11476         (address_p): New function.
11477         (record_use): Add a mem_type argument and initialize the new
11478         mem_type field.
11479         (record_group_use): Add a mem_type argument.  Use address_p.
11480         Remove obsolete null checks of base_object.  Update call to record_use.
11481         (find_interesting_uses_op): Update call to record_group_use.
11482         (find_interesting_uses_cond): Likewise.
11483         (find_interesting_uses_address): Likewise.
11484         (get_mem_type_for_internal_fn): New function.
11485         (find_address_like_use): Likewise.
11486         (find_interesting_uses_stmt): Try find_address_like_use before
11487         calling find_interesting_uses_op.
11488         (addr_offset_valid_p): Use the iv mem_type field as the type
11489         of the addressed memory.
11490         (add_autoinc_candidates): Likewise.
11491         (get_address_cost): Likewise.
11492         (split_small_address_groups_p): Use address_p.
11493         (split_address_groups): Likewise.
11494         (add_iv_candidate_for_use): Likewise.
11495         (autoinc_possible_for_pair): Likewise.
11496         (rewrite_groups): Likewise.
11497         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11498         (determine_group_iv_cost): Update after split of USE_ADDRESS.
11499         (get_alias_ptr_type_for_ptr_address): New function.
11500         (rewrite_use_address): Rewrite address uses in calls that were
11501         identified by find_address_like_use.
11502
11503 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11504             Alan Hayward  <alan.hayward@arm.com>
11505             David Sherwood  <david.sherwood@arm.com>
11506
11507         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11508         TARGET_MEM_REFs.
11509         * gimple-expr.h (is_gimple_addressable: Likewise.
11510         * gimple-expr.c (is_gimple_address): Likewise.
11511         * internal-fn.c (expand_call_mem_ref): New function.
11512         (expand_mask_load_optab_fn): Use it.
11513         (expand_mask_store_optab_fn): Likewise.
11514
11515 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11516             Alan Hayward  <alan.hayward@arm.com>
11517             David Sherwood  <david.sherwood@arm.com>
11518
11519         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11520         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11521         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11522         (cond_umax@var{mode}): Document.
11523         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11524         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11525         (cond_umin_optab, cond_umax_optab): New optabs.
11526         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11527         (COND_IOR, COND_XOR): New internal functions.
11528         * internal-fn.h (get_conditional_internal_fn): Declare.
11529         * internal-fn.c (cond_binary_direct): New macro.
11530         (expand_cond_binary_optab_fn): Likewise.
11531         (direct_cond_binary_optab_supported_p): Likewise.
11532         (get_conditional_internal_fn): New function.
11533         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11534         Cope with reduction statements that are vectorized as calls rather
11535         than assignments.
11536         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11537         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11538         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11539         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11540         (UNSPEC_COND_EOR): New unspecs.
11541         (optab): Add mappings for them.
11542         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11543         (sve_int_op, sve_fp_op): New int attributes.
11544
11545 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11546             Alan Hayward  <alan.hayward@arm.com>
11547             David Sherwood  <david.sherwood@arm.com>
11548
11549         * optabs.def (while_ult_optab): New optab.
11550         * doc/md.texi (while_ult@var{m}@var{n}): Document.
11551         * internal-fn.def (WHILE_ULT): New internal function.
11552         * internal-fn.h (direct_internal_fn_supported_p): New override
11553         that takes two types as argument.
11554         * internal-fn.c (while_direct): New macro.
11555         (expand_while_optab_fn): New function.
11556         (convert_optab_supported_p): Likewise.
11557         (direct_while_optab_supported_p): New macro.
11558         * wide-int.h (wi::udiv_ceil): New function.
11559         * tree-vectorizer.h (rgroup_masks): New structure.
11560         (vec_loop_masks): New typedef.
11561         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11562         and fully_masked_p.
11563         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11564         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11565         (vect_max_vf): New function.
11566         (slpeel_make_loop_iterate_ntimes): Delete.
11567         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11568         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11569         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11570         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11571         internal-fn.h, stor-layout.h and optabs-query.h.
11572         (vect_set_loop_mask): New function.
11573         (add_preheader_seq): Likewise.
11574         (add_header_seq): Likewise.
11575         (interleave_supported_p): Likewise.
11576         (vect_maybe_permute_loop_masks): Likewise.
11577         (vect_set_loop_masks_directly): Likewise.
11578         (vect_set_loop_condition_masked): Likewise.
11579         (vect_set_loop_condition_unmasked): New function, split out from
11580         slpeel_make_loop_iterate_ntimes.
11581         (slpeel_make_loop_iterate_ntimes): Rename to..
11582         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
11583         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11584         (vect_do_peeling): Update call accordingly.
11585         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11586         loops.
11587         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11588         mask_compare_type, can_fully_mask_p and fully_masked_p.
11589         (release_vec_loop_masks): New function.
11590         (_loop_vec_info): Use it to free the loop masks.
11591         (can_produce_all_loop_masks_p): New function.
11592         (vect_get_max_nscalars_per_iter): Likewise.
11593         (vect_verify_full_masking): Likewise.
11594         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11595         retries, and free the mask rgroups before retrying.  Check loop-wide
11596         reasons for disallowing fully-masked loops.  Make the final decision
11597         about whether use a fully-masked loop or not.
11598         (vect_estimate_min_profitable_iters): Do not assume that peeling
11599         for the number of iterations will be needed for fully-masked loops.
11600         (vectorizable_reduction): Disable fully-masked loops.
11601         (vectorizable_live_operation): Likewise.
11602         (vect_halve_mask_nunits): New function.
11603         (vect_double_mask_nunits): Likewise.
11604         (vect_record_loop_mask): Likewise.
11605         (vect_get_loop_mask): Likewise.
11606         (vect_transform_loop): Handle the case in which the final loop
11607         iteration might handle a partial vector.  Call vect_set_loop_condition
11608         instead of slpeel_make_loop_iterate_ntimes.
11609         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11610         (check_load_store_masking): New function.
11611         (prepare_load_store_mask): Likewise.
11612         (vectorizable_store): Handle fully-masked loops.
11613         (vectorizable_load): Likewise.
11614         (supportable_widening_operation): Use vect_halve_mask_nunits for
11615         booleans.
11616         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11617         (vect_gen_while): New function.
11618         * config/aarch64/aarch64.md (umax<mode>3): New expander.
11619         (aarch64_uqdec<mode>): New insn.
11620
11621 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11622             Alan Hayward  <alan.hayward@arm.com>
11623             David Sherwood  <david.sherwood@arm.com>
11624
11625         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11626         (reduc_xor_scal_optab): New optabs.
11627         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11628         (reduc_xor_scal_@var{m}): Document.
11629         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11630         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11631         internal functions.
11632         * fold-const-call.c (fold_const_call): Handle them.
11633         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11634         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11635         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11636         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11637         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11638         (UNSPEC_XORV): New unspecs.
11639         (optab): Add entries for them.
11640         (BITWISEV): New int iterator.
11641         (bit_reduc_op): New int attributes.
11642
11643 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11644             Alan Hayward  <alan.hayward@arm.com>
11645             David Sherwood  <david.sherwood@arm.com>
11646
11647         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11648         * internal-fn.def (VEC_SHL_INSERT): New internal function.
11649         * optabs.def (vec_shl_insert_optab): New optab.
11650         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11651         (duplicate_and_interleave): Likewise.
11652         * tree-vect-loop.c: Include internal-fn.h.
11653         (neutral_op_for_slp_reduction): New function, split out from
11654         get_initial_defs_for_reduction.
11655         (get_initial_def_for_reduction): Handle option 2 for variable-length
11656         vectors by loading the neutral value into a vector and then shifting
11657         the initial value into element 0.
11658         (get_initial_defs_for_reduction): Replace the code argument with
11659         the neutral value calculated by neutral_op_for_slp_reduction.
11660         Use gimple_build_vector for constant-length vectors.
11661         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11662         but the first group_size elements have a neutral value.
11663         Use duplicate_and_interleave otherwise.
11664         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11665         Update call to get_initial_defs_for_reduction.  Handle SLP
11666         reductions for variable-length vectors by creating one vector
11667         result for each scalar result, with the elements associated
11668         with other scalar results stubbed out with the neutral value.
11669         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11670         Require IFN_VEC_SHL_INSERT for double reductions on
11671         variable-length vectors, or SLP reductions that have
11672         a neutral value.  Require can_duplicate_and_interleave_p
11673         support for variable-length unchained SLP reductions if there
11674         is no neutral value, such as for MIN/MAX reductions.  Also require
11675         the number of vector elements to be a multiple of the number of
11676         SLP statements when doing variable-length unchained SLP reductions.
11677         Update call to vect_create_epilog_for_reduction.
11678         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11679         and remove initial values.
11680         (duplicate_and_interleave): Make public.
11681         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11682         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11683
11684 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11685             Alan Hayward  <alan.hayward@arm.com>
11686             David Sherwood  <david.sherwood@arm.com>
11687
11688         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11689         (can_duplicate_and_interleave_p): New function.
11690         (vect_get_and_check_slp_defs): Take the vector of statements
11691         rather than just the current one.  Remove excess parentheses.
11692         Restriction rejectinon of vect_constant_def and vect_external_def
11693         for variable-length vectors to boolean types, or types for which
11694         can_duplicate_and_interleave_p is false.
11695         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11696         (duplicate_and_interleave): New function.
11697         (vect_get_constant_vectors): Use gimple_build_vector for
11698         constant-length vectors and suitable variable-length constant
11699         vectors.  Use duplicate_and_interleave for other variable-length
11700         vectors.  Don't defer the update when inserting new statements.
11701
11702 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11703             Alan Hayward  <alan.hayward@arm.com>
11704             David Sherwood  <david.sherwood@arm.com>
11705
11706         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11707         min_profitable_iters doesn't go negative.
11708
11709 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11710             Alan Hayward  <alan.hayward@arm.com>
11711             David Sherwood  <david.sherwood@arm.com>
11712
11713         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11714         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11715         * optabs.def (vec_mask_load_lanes_optab): New optab.
11716         (vec_mask_store_lanes_optab): Likewise.
11717         * internal-fn.def (MASK_LOAD_LANES): New internal function.
11718         (MASK_STORE_LANES): Likewise.
11719         * internal-fn.c (mask_load_lanes_direct): New macro.
11720         (mask_store_lanes_direct): Likewise.
11721         (expand_mask_load_optab_fn): Handle masked operations.
11722         (expand_mask_load_lanes_optab_fn): New macro.
11723         (expand_mask_store_optab_fn): Handle masked operations.
11724         (expand_mask_store_lanes_optab_fn): New macro.
11725         (direct_mask_load_lanes_optab_supported_p): Likewise.
11726         (direct_mask_store_lanes_optab_supported_p): Likewise.
11727         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11728         parameter.
11729         (vect_load_lanes_supported): Likewise.
11730         * tree-vect-data-refs.c (strip_conversion): New function.
11731         (can_group_stmts_p): Likewise.
11732         (vect_analyze_data_ref_accesses): Use it instead of checking
11733         for a pair of assignments.
11734         (vect_store_lanes_supported): Take a masked_p parameter.
11735         (vect_load_lanes_supported): Likewise.
11736         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11737         vect_store_lanes_supported and vect_load_lanes_supported.
11738         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11739         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11740         parameter.  Don't allow gaps for masked accesses.
11741         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
11742         and vect_load_lanes_supported.
11743         (get_load_store_type): Take a masked_p parameter and update
11744         call to get_group_load_store_type.
11745         (vectorizable_store): Update call to get_load_store_type.
11746         Handle IFN_MASK_STORE_LANES.
11747         (vectorizable_load): Update call to get_load_store_type.
11748         Handle IFN_MASK_LOAD_LANES.
11749
11750 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11751             Alan Hayward  <alan.hayward@arm.com>
11752             David Sherwood  <david.sherwood@arm.com>
11753
11754         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11755         modes for SVE.
11756         * config/aarch64/aarch64-protos.h
11757         (aarch64_sve_struct_memory_operand_p): Declare.
11758         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11759         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11760         (VPRED, vpred): Handle SVE structure modes.
11761         * config/aarch64/constraints.md (Utx): New constraint.
11762         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11763         (aarch64_sve_struct_nonimmediate_operand): New predicates.
11764         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11765         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11766         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11767         structure modes.  Split into pieces after RA.
11768         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11769         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11770         New patterns.
11771         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11772         SVE structure modes.
11773         (aarch64_classify_address): Likewise.
11774         (sizetochar): Move earlier in file.
11775         (aarch64_print_operand): Handle SVE register lists.
11776         (aarch64_array_mode): New function.
11777         (aarch64_sve_struct_memory_operand_p): Likewise.
11778         (TARGET_ARRAY_MODE): Redefine.
11779
11780 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11781             Alan Hayward  <alan.hayward@arm.com>
11782             David Sherwood  <david.sherwood@arm.com>
11783
11784         * target.def (array_mode): New target hook.
11785         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11786         * doc/tm.texi: Regenerate.
11787         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11788         * hooks.c (hook_optmode_mode_uhwi_none): New function.
11789         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11790         targetm.array_mode.
11791         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
11792         type sizes.
11793
11794 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11795             Alan Hayward  <alan.hayward@arm.com>
11796             David Sherwood  <david.sherwood@arm.com>
11797
11798         * fold-const.c (fold_binary_loc): Check the argument types
11799         rather than the result type when testing for a vector operation.
11800
11801 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11802
11803         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11804         * doc/tm.texi: Regenerate.
11805
11806 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11807             Alan Hayward  <alan.hayward@arm.com>
11808             David Sherwood  <david.sherwood@arm.com>
11809
11810         * doc/invoke.texi (-msve-vector-bits=): Document new option.
11811         (sve): Document new AArch64 extension.
11812         * doc/md.texi (w): Extend the description of the AArch64
11813         constraint to include SVE vectors.
11814         (Upl, Upa): Document new AArch64 predicate constraints.
11815         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11816         enum.
11817         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11818         (msve-vector-bits=): New option.
11819         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11820         SVE when these are disabled.
11821         (sve): New extension.
11822         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11823         modes.  Adjust their number of units based on aarch64_sve_vg.
11824         (MAX_BITSIZE_MODE_ANY_MODE): Define.
11825         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11826         aarch64_addr_query_type.
11827         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11828         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11829         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11830         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11831         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11832         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11833         (aarch64_simd_imm_zero_p): Delete.
11834         (aarch64_check_zero_based_sve_index_immediate): Declare.
11835         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11836         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11837         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11838         (aarch64_sve_float_mul_immediate_p): Likewise.
11839         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11840         rather than an rtx.
11841         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11842         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11843         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11844         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11845         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11846         (aarch64_regmode_natural_size): Likewise.
11847         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11848         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11849         left one place.
11850         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11851         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11852         for VG and the SVE predicate registers.
11853         (V_ALIASES): Add a "z"-prefixed alias.
11854         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11855         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11856         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11857         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11858         (REG_CLASS_NAMES): Add entries for them.
11859         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
11860         and the predicate registers.
11861         (aarch64_sve_vg): Declare.
11862         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11863         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11864         (REGMODE_NATURAL_SIZE): Define.
11865         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11866         SVE macros.
11867         * config/aarch64/aarch64.c: Include cfgrtl.h.
11868         (simd_immediate_info): Add a constructor for series vectors,
11869         and an associated step field.
11870         (aarch64_sve_vg): New variable.
11871         (aarch64_dbx_register_number): Handle VG and the predicate registers.
11872         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11873         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11874         (VEC_ANY_DATA, VEC_STRUCT): New constants.
11875         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11876         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11877         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11878         (aarch64_get_mask_mode): New functions.
11879         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11880         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11881         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
11882         predicate modes and predicate registers.  Explicitly restrict
11883         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
11884         to store a vector mode if it is recognized by
11885         aarch64_classify_vector_mode.
11886         (aarch64_regmode_natural_size): New function.
11887         (aarch64_hard_regno_caller_save_mode): Return the original mode
11888         for predicates.
11889         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11890         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11891         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11892         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11893         functions.
11894         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
11895         does not overlap dest if the function is frame-related.  Handle
11896         SVE constants.
11897         (aarch64_split_add_offset): New function.
11898         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11899         them aarch64_add_offset.
11900         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11901         and update call to aarch64_sub_sp.
11902         (aarch64_add_cfa_expression): New function.
11903         (aarch64_expand_prologue): Pass extra temporary registers to the
11904         functions above.  Handle the case in which we need to emit new
11905         DW_CFA_expressions for registers that were originally saved
11906         relative to the stack pointer, but now have to be expressed
11907         relative to the frame pointer.
11908         (aarch64_output_mi_thunk): Pass extra temporary registers to the
11909         functions above.
11910         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
11911         IP0 and IP1 values for SVE frames.
11912         (aarch64_expand_vec_series): New function.
11913         (aarch64_expand_sve_widened_duplicate): Likewise.
11914         (aarch64_expand_sve_const_vector): Likewise.
11915         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11916         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
11917         into the register, rather than emitting a SET directly.
11918         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11919         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11920         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11921         (offset_9bit_signed_scaled_p): New functions.
11922         (aarch64_replicate_bitmask_imm): New function.
11923         (aarch64_bitmask_imm): Use it.
11924         (aarch64_cannot_force_const_mem): Reject expressions involving
11925         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
11926         (aarch64_classify_index): Handle SVE indices, by requiring
11927         a plain register index with a scale that matches the element size.
11928         (aarch64_classify_address): Handle SVE addresses.  Assert that
11929         the mode of the address is VOIDmode or an integer mode.
11930         Update call to aarch64_classify_symbol.
11931         (aarch64_classify_symbolic_expression): Update call to
11932         aarch64_classify_symbol.
11933         (aarch64_const_vec_all_in_range_p): New function.
11934         (aarch64_print_vector_float_operand): Likewise.
11935         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
11936         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
11937         and the FP immediates 1.0 and 0.5.
11938         (aarch64_print_address_internal): Handle SVE addresses.
11939         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11940         (aarch64_regno_regclass): Handle predicate registers.
11941         (aarch64_secondary_reload): Handle big-endian reloads of SVE
11942         data modes.
11943         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11944         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11945         (aarch64_convert_sve_vector_bits): New function.
11946         (aarch64_override_options): Use it to handle -msve-vector-bits=.
11947         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11948         rather than an rtx.
11949         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11950         Handle SVE vector and predicate modes.  Accept VL-based constants
11951         that need only one temporary register, and VL offsets that require
11952         no temporary registers.
11953         (aarch64_conditional_register_usage): Mark the predicate registers
11954         as fixed if SVE isn't available.
11955         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11956         Return true for SVE vector and predicate modes.
11957         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11958         rather than an unsigned int.  Handle SVE modes.
11959         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
11960         SVE modes.
11961         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11962         if SVE is enabled.
11963         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11964         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11965         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11966         (aarch64_sve_float_mul_immediate_p): New functions.
11967         (aarch64_sve_valid_immediate): New function.
11968         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11969         Explicitly reject structure modes.  Check for INDEX constants.
11970         Handle PTRUE and PFALSE constants.
11971         (aarch64_check_zero_based_sve_index_immediate): New function.
11972         (aarch64_simd_imm_zero_p): Delete.
11973         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11974         vector modes.  Accept constants in the range of CNT[BHWD].
11975         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11976         ask for an Advanced SIMD mode.
11977         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11978         (aarch64_simd_vector_alignment): Handle SVE predicates.
11979         (aarch64_vectorize_preferred_vector_alignment): New function.
11980         (aarch64_simd_vector_alignment_reachable): Use it instead of
11981         the vector size.
11982         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11983         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11984         functions.
11985         (MAX_VECT_LEN): Delete.
11986         (expand_vec_perm_d): Add a vec_flags field.
11987         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11988         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11989         (aarch64_evpc_ext): Don't apply a big-endian lane correction
11990         for SVE modes.
11991         (aarch64_evpc_rev): Rename to...
11992         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
11993         (aarch64_evpc_rev_global): New function.
11994         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11995         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11996         MAX_VECT_LEN.
11997         (aarch64_evpc_sve_tbl): New function.
11998         (aarch64_expand_vec_perm_const_1): Update after rename of
11999         aarch64_evpc_rev.  Handle SVE permutes too, trying
12000         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12001         than aarch64_evpc_tbl.
12002         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12003         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12004         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12005         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12006         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12007         (aarch64_expand_sve_vcond): New functions.
12008         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12009         of aarch64_vector_mode_p.
12010         (aarch64_dwarf_poly_indeterminate_value): New function.
12011         (aarch64_compute_pressure_classes): Likewise.
12012         (aarch64_can_change_mode_class): Likewise.
12013         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12014         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12015         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12016         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12017         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12018         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12019         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12020         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12021         constraints.
12022         (Dn, Dl, Dr): Accept const as well as const_vector.
12023         (Dz): Likewise.  Compare against CONST0_RTX.
12024         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12025         of "vector" where appropriate.
12026         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12027         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12028         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12029         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12030         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12031         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12032         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12033         (v_int_equiv): Extend to SVE modes.
12034         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12035         mode attributes.
12036         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12037         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12038         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12039         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12040         (SVE_COND_FP_CMP): New int iterators.
12041         (perm_hilo): Handle the new unpack unspecs.
12042         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12043         attributes.
12044         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12045         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12046         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12047         (aarch64_equality_operator, aarch64_constant_vector_operand)
12048         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12049         (aarch64_sve_nonimmediate_operand): Likewise.
12050         (aarch64_sve_general_operand): Likewise.
12051         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12052         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12053         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12054         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12055         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12056         (aarch64_sve_float_arith_immediate): Likewise.
12057         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12058         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12059         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12060         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12061         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12062         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12063         (aarch64_sve_float_arith_operand): Likewise.
12064         (aarch64_sve_float_arith_with_sub_operand): Likewise.
12065         (aarch64_sve_float_mul_operand): Likewise.
12066         (aarch64_sve_vec_perm_operand): Likewise.
12067         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12068         (aarch64_mov_operand): Accept const_poly_int and const_vector.
12069         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12070         as well as const_vector.
12071         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12072         in file.  Use CONST0_RTX and CONSTM1_RTX.
12073         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
12074         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12075         Use aarch64_simd_imm_zero.
12076         * config/aarch64/aarch64-sve.md: New file.
12077         * config/aarch64/aarch64.md: Include it.
12078         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12079         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12080         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12081         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12082         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12083         (sve): New attribute.
12084         (enabled): Disable instructions with the sve attribute unless
12085         TARGET_SVE.
12086         (movqi, movhi): Pass CONST_POLY_INT operaneds through
12087         aarch64_expand_mov_immediate.
12088         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12089         CNT[BHSD] immediates.
12090         (movti): Split CONST_POLY_INT moves into two halves.
12091         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12092         Split additions that need a temporary here if the destination
12093         is the stack pointer.
12094         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12095         (*add<mode>3_poly_1): New instruction.
12096         (set_clobber_cc): New expander.
12097
12098 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12099
12100         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12101         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
12102         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12103         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12104         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
12105         Change innermode from fixed_mode_size to machine_mode.
12106         (simplify_subreg): Update call accordingly.  Handle a constant-sized
12107         subreg of a variable-length CONST_VECTOR.
12108
12109 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12110             Alan Hayward  <alan.hayward@arm.com>
12111             David Sherwood  <david.sherwood@arm.com>
12112
12113         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12114         (add_offset_to_base): New function, split out from...
12115         (create_mem_ref): ...here.  When handling a scale other than 1,
12116         check first whether the address is valid without the offset.
12117         Add it into the base if so, leaving the index and scale as-is.
12118
12119 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
12120
12121         PR c++/83778
12122         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12123         fold_for_warn before checking if arg2 is INTEGER_CST.
12124
12125 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
12126
12127         * config/rs6000/predicates.md (load_multiple_operation): Delete.
12128         (store_multiple_operation): Delete.
12129         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12130         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12131         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12132         guarded by TARGET_STRING.
12133         (rs6000_output_load_multiple): Delete.
12134         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12135         OPTION_MASK_STRING / TARGET_STRING handling.
12136         (print_operand) <'N', 'O'>: Add comment that these are unused now.
12137         (const rs6000_opt_masks) <"string">: Change mask to 0.
12138         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12139         (MASK_STRING): Delete.
12140         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12141         parts.  Simplify.
12142         (load_multiple): Delete.
12143         (*ldmsi8): Delete.
12144         (*ldmsi7): Delete.
12145         (*ldmsi6): Delete.
12146         (*ldmsi5): Delete.
12147         (*ldmsi4): Delete.
12148         (*ldmsi3): Delete.
12149         (store_multiple): Delete.
12150         (*stmsi8): Delete.
12151         (*stmsi7): Delete.
12152         (*stmsi6): Delete.
12153         (*stmsi5): Delete.
12154         (*stmsi4): Delete.
12155         (*stmsi3): Delete.
12156         (movmemsi_8reg): Delete.
12157         (corresponding unnamed define_insn): Delete.
12158         (movmemsi_6reg): Delete.
12159         (corresponding unnamed define_insn): Delete.
12160         (movmemsi_4reg): Delete.
12161         (corresponding unnamed define_insn): Delete.
12162         (movmemsi_2reg): Delete.
12163         (corresponding unnamed define_insn): Delete.
12164         (movmemsi_1reg): Delete.
12165         (corresponding unnamed define_insn): Delete.
12166         * config/rs6000/rs6000.opt (mno-string): New.
12167         (mstring): Replace by deprecation warning stub.
12168         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12169
12170 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
12171
12172         * regrename.c (regrename_do_replace): If replacing the same
12173         reg multiple times, try to reuse last created gen_raw_REG.
12174
12175         PR debug/81155
12176         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12177         main to workaround a bug in GDB.
12178
12179 2018-01-12  Tom de Vries  <tom@codesourcery.com>
12180
12181         PR target/83737
12182         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12183
12184 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
12185
12186         PR rtl-optimization/80481
12187         * ira-color.c (get_cap_member): New function.
12188         (allocnos_conflict_by_live_ranges_p): Use it.
12189         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12190         (setup_slot_coalesced_allocno_live_ranges): Ditto.
12191
12192 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
12193
12194         PR target/83628
12195         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12196         (*saddl_se_1): Ditto.
12197         (*ssubsi_1): Ditto.
12198         (*ssubl_se_1): Ditto.
12199
12200 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12201
12202         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12203         rather than wi::to_widest for DR_INITs.
12204         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12205         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12206         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12207         INTEGER_CSTs.
12208         (vect_analyze_group_access_1): Note that here.
12209
12210 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12211
12212         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12213         polynomial type sizes.
12214
12215 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12216
12217         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12218         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12219         (gimple_add_tmp_var): Likewise.
12220
12221 2018-01-12  Martin Liska  <mliska@suse.cz>
12222
12223         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12224         (gimple_alloc_sizes): Likewise.
12225         (dump_gimple_statistics): Use PRIu64 in printf format.
12226         * gimple.h: Change uint64_t to int.
12227
12228 2018-01-12  Martin Liska  <mliska@suse.cz>
12229
12230         * tree-core.h: Use uint64_t instead of int.
12231         * tree.c (tree_node_counts): Likewise.
12232         (tree_node_sizes): Likewise.
12233         (dump_tree_statistics): Use PRIu64 in printf format.
12234
12235 2018-01-12  Martin Liska  <mliska@suse.cz>
12236
12237         * Makefile.in: As qsort_chk is implemented in vec.c, add
12238         vec.o to linkage of gencfn-macros.
12239         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12240         passing the info to record_node_allocation_statistics.
12241         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12242         and pass the info.
12243         * ggc-common.c (struct ggc_usage): Add operator== and use
12244         it in operator< and compare function.
12245         * mem-stats.h (struct mem_usage): Likewise.
12246         * vec.c (struct vec_usage): Remove operator< and compare
12247         function. Can be simply inherited.
12248
12249 2018-01-12  Martin Jambor  <mjambor@suse.cz>
12250
12251         PR target/81616
12252         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12253         * tree-ssa-math-opts.c: Include domwalk.h.
12254         (convert_mult_to_fma_1): New function.
12255         (fma_transformation_info): New type.
12256         (fma_deferring_state): Likewise.
12257         (cancel_fma_deferring): New function.
12258         (result_of_phi): Likewise.
12259         (last_fma_candidate_feeds_initial_phi): Likewise.
12260         (convert_mult_to_fma): Added deferring logic, split actual
12261         transformation to convert_mult_to_fma_1.
12262         (math_opts_dom_walker): New type.
12263         (math_opts_dom_walker::after_dom_children): New method, body moved
12264         here from pass_optimize_widening_mul::execute, added deferring logic
12265         bits.
12266         (pass_optimize_widening_mul::execute): Moved most of code to
12267         math_opts_dom_walker::after_dom_children.
12268         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12269         * config/i386/i386.c (ix86_option_override_internal): Added
12270         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12271
12272 2018-01-12  Richard Biener  <rguenther@suse.de>
12273
12274         PR debug/83157
12275         * dwarf2out.c (gen_variable_die): Do not reset old_die for
12276         inline instance vars.
12277
12278 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
12279
12280         PR target/81819
12281         * config/rx/rx.c (rx_is_restricted_memory_address):
12282         Handle SUBREG case.
12283
12284 2018-01-12  Richard Biener  <rguenther@suse.de>
12285
12286         PR tree-optimization/80846
12287         * target.def (split_reduction): New target hook.
12288         * targhooks.c (default_split_reduction): New function.
12289         * targhooks.h (default_split_reduction): Declare.
12290         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12291         target requests first reduce vectors by combining low and high
12292         parts.
12293         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12294         (get_vectype_for_scalar_type_and_size): Export.
12295         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12296         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12297         * doc/tm.texi: Regenerate.
12298         * config/i386/i386.c (ix86_split_reduction): Implement
12299         TARGET_VECTORIZE_SPLIT_REDUCTION.
12300
12301 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
12302
12303         PR target/83368
12304         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12305         in PIC mode except for TARGET_VXWORKS_RTP.
12306         * config/sparc/sparc.c: Include cfgrtl.h.
12307         (TARGET_INIT_PIC_REG): Define.
12308         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12309         (sparc_pic_register_p): New predicate.
12310         (sparc_legitimate_address_p): Use it.
12311         (sparc_legitimize_pic_address): Likewise.
12312         (sparc_delegitimize_address): Likewise.
12313         (sparc_mode_dependent_address_p): Likewise.
12314         (gen_load_pcrel_sym): Remove 4th parameter.
12315         (load_got_register): Adjust call to above.  Remove obsolete stuff.
12316         (sparc_expand_prologue): Do not call load_got_register here.
12317         (sparc_flat_expand_prologue): Likewise.
12318         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12319         (sparc_use_pseudo_pic_reg): New function.
12320         (sparc_init_pic_reg): Likewise.
12321         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12322         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12323
12324 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
12325
12326         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12327         Add item for branch_cost.
12328
12329 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
12330
12331         PR rtl-optimization/83565
12332         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12333         not extend the result to a larger mode for rotate operations.
12334         (num_sign_bit_copies1): Likewise.
12335
12336 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12337
12338         PR target/40411
12339         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12340         -symbolic.
12341         Use values-Xc.o for -pedantic.
12342         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12343
12344 2018-01-12  Martin Liska  <mliska@suse.cz>
12345
12346         PR ipa/83054
12347         * ipa-devirt.c (final_warning_record::grow_type_warnings):
12348         New function.
12349         (possible_polymorphic_call_targets): Use it.
12350         (ipa_devirt): Likewise.
12351
12352 2018-01-12  Martin Liska  <mliska@suse.cz>
12353
12354         * profile-count.h (enum profile_quality): Use 0 as invalid
12355         enum value of profile_quality.
12356
12357 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
12358
12359         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12360         -mext-string options.
12361
12362 2018-01-12  Richard Biener  <rguenther@suse.de>
12363
12364         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12365         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12366         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12367         Likewise.
12368         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12369
12370 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
12371
12372         * configure.ac (--with-long-double-format): Add support for the
12373         configuration option to change the default long double format on
12374         PowerPC systems.
12375         * config.gcc (powerpc*-linux*-*): Likewise.
12376         * configure: Regenerate.
12377         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12378         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12379         used without modification.
12380
12381 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
12382
12383         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12384         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12385         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12386         MISC_BUILTIN_SPEC_BARRIER.
12387         (rs6000_init_builtins): Likewise.
12388         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12389         enum value.
12390         (speculation_barrier): New define_insn.
12391         * doc/extend.texi: Document __builtin_speculation_barrier.
12392
12393 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
12394
12395         PR target/83203
12396         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12397         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12398         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12399         iterators.
12400         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
12401         integral modes instead of "ss" and "sd".
12402         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12403         vectors with 32-bit and 64-bit elements.
12404         (vecdupssescalarmodesuffix): New mode attribute.
12405         (vec_dup<mode>): Use it.
12406
12407 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
12408
12409         PR target/83330
12410         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12411         frame if argument is passed on stack.
12412
12413 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
12414
12415         PR target/82682
12416         * ree.c (combine_reaching_defs): Optimize also
12417         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12418         reg2=any_extend(exp); reg1=reg2;, formatting fix.
12419
12420 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
12421
12422         PR middle-end/83189
12423         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12424
12425 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
12426
12427         PR middle-end/83718
12428         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12429         after they are computed.
12430
12431 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
12432
12433         PR tree-optimization/83695
12434         * gimple-loop-linterchange.cc
12435         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12436         reset cached scev information after interchange.
12437         (pass_linterchange::execute): Remove call to scev_reset_htab.
12438
12439 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12440
12441         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12442         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12443         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12444         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12445         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12446         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12447         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12448         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12449         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12450         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12451         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12452         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12453         (V_lane_reg): Likewise.
12454         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12455         New define_expand.
12456         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12457         (vfmal_lane_low<mode>_intrinsic,
12458         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12459         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12460         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12461         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12462         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12463         vfmsl_lane_high<mode>_intrinsic): New define_insns.
12464
12465 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12466
12467         * config/arm/arm-cpus.in (fp16fml): New feature.
12468         (ALL_SIMD): Add fp16fml.
12469         (armv8.2-a): Add fp16fml as an option.
12470         (armv8.3-a): Likewise.
12471         (armv8.4-a): Add fp16fml as part of fp16.
12472         * config/arm/arm.h (TARGET_FP16FML): Define.
12473         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12474         when appropriate.
12475         * config/arm/arm-modes.def (V2HF): Define.
12476         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12477         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12478         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12479         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12480         vfmsl_low, vfmsl_high): New set of builtins.
12481         * config/arm/iterators.md (PLUSMINUS): New code iterator.
12482         (vfml_op): New code attribute.
12483         (VFMLHALVES): New int iterator.
12484         (VFML, VFMLSEL): New mode attributes.
12485         (V_reg): Define mapping for V2HF.
12486         (V_hi, V_lo): New mode attributes.
12487         (VF_constraint): Likewise.
12488         (vfml_half, vfml_half_selector): New int attributes.
12489         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12490         define_expand.
12491         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12492         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12493         New define_insn.
12494         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12495         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12496         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12497         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
12498         documentation.
12499         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12500         Document new effective target and option set.
12501
12502 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12503
12504         * config/arm/arm-cpus.in (armv8_4): New feature.
12505         (ARMv8_4a): New fgroup.
12506         (armv8.4-a): New arch.
12507         * config/arm/arm-tables.opt: Regenerate.
12508         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12509         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12510         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12511         Add matching rules for -march=armv8.4-a and extensions.
12512         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12513
12514 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
12515
12516         PR target/81821
12517         * config/rx/rx.md (BW): New mode attribute.
12518         (sync_lock_test_and_setsi): Add mode suffix to insn output.
12519
12520 2018-01-11  Richard Biener  <rguenther@suse.de>
12521
12522         PR tree-optimization/83435
12523         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12524         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12525         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12526
12527 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12528             Alan Hayward  <alan.hayward@arm.com>
12529             David Sherwood  <david.sherwood@arm.com>
12530
12531         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12532         field.
12533         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
12534         (aarch64_print_address_internal): Use it to check for a zero offset.
12535
12536 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12537             Alan Hayward  <alan.hayward@arm.com>
12538             David Sherwood  <david.sherwood@arm.com>
12539
12540         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12541         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12542         Return a poly_int64 rather than a HOST_WIDE_INT.
12543         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12544         rather than a HOST_WIDE_INT.
12545         * config/aarch64/aarch64.h (aarch64_frame): Protect with
12546         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
12547         hard_fp_offset, frame_size, initial_adjust, callee_offset and
12548         final_offset from HOST_WIDE_INT to poly_int64.
12549         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12550         to_constant when getting the number of units in an Advanced SIMD
12551         mode.
12552         (aarch64_builtin_vectorized_function): Check for a constant number
12553         of units.
12554         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12555         GET_MODE_SIZE.
12556         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12557         attribute instead of GET_MODE_NUNITS.
12558         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12559         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12560         GET_MODE_SIZE for fixed-size registers.
12561         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12562         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12563         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12564         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12565         (aarch64_print_operand, aarch64_print_address_internal)
12566         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12567         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12568         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12569         Handle polynomial GET_MODE_SIZE.
12570         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
12571         wider than SImode without modification.
12572         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12573         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12574         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12575         passing and returning SVE modes.
12576         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12577         rather than GEN_INT.
12578         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12579         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12580         (aarch64_allocate_and_probe_stack_space): Likewise.
12581         (aarch64_layout_frame): Cope with polynomial offsets.
12582         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12583         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
12584         polynomial offsets.
12585         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12586         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12587         poly_int64 rather than a HOST_WIDE_INT.
12588         (aarch64_get_separate_components, aarch64_process_components)
12589         (aarch64_expand_prologue, aarch64_expand_epilogue)
12590         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12591         (aarch64_anchor_offset): New function, split out from...
12592         (aarch64_legitimize_address): ...here.
12593         (aarch64_builtin_vectorization_cost): Handle polynomial
12594         TYPE_VECTOR_SUBPARTS.
12595         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12596         GET_MODE_NUNITS.
12597         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12598         number of elements from the PARALLEL rather than the mode.
12599         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12600         rather than GET_MODE_BITSIZE.
12601         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12602         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12603         (aarch64_expand_vec_perm_const_1): Handle polynomial
12604         d->perm.length () and d->perm elements.
12605         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
12606         Apply to_constant to d->perm elements.
12607         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12608         polynomial CONST_VECTOR_NUNITS.
12609         (aarch64_move_pointer): Take amount as a poly_int64 rather
12610         than an int.
12611         (aarch64_progress_pointer): Avoid temporary variable.
12612         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12613         the mode attribute instead of GET_MODE.
12614
12615 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12616             Alan Hayward  <alan.hayward@arm.com>
12617             David Sherwood  <david.sherwood@arm.com>
12618
12619         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12620         x exists before using it.
12621         (aarch64_add_constant_internal): Rename to...
12622         (aarch64_add_offset_1): ...this.  Replace regnum with separate
12623         src and dest rtxes.  Handle the case in which they're different,
12624         including when the offset is zero.  Replace scratchreg with an rtx.
12625         Use 2 additions if there is no spare register into which we can
12626         move a 16-bit constant.
12627         (aarch64_add_constant): Delete.
12628         (aarch64_add_offset): Replace reg with separate src and dest
12629         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
12630         Use aarch64_add_offset_1.
12631         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12632         an rtx rather than an int.  Take the delta as a poly_int64
12633         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
12634         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12635         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12636         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12637         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12638         and aarch64_add_sp.
12639         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12640         aarch64_add_constant.
12641
12642 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12643
12644         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12645         Use scalar_float_mode.
12646
12647 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12648
12649         * config/aarch64/aarch64-simd.md
12650         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12651         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12652         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12653         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12654         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12655         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12656         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12657         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12658         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12659         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12660
12661 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12662
12663         PR target/83514
12664         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12665         targ_options->x_arm_arch_string is non NULL.
12666
12667 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
12668
12669         * config/aarch64/aarch64.h
12670         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
12671
12672 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
12673
12674         PR target/82096
12675         * expmed.c (emit_store_flag_force): Swap if const op0
12676         and change VOIDmode to mode of op0.
12677
12678 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12679
12680         PR rtl-optimization/83761
12681         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12682         than bytes to mode_for_size.
12683
12684 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12685
12686         PR middle-end/83189
12687         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12688         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12689         profile.
12690
12691 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12692
12693         PR middle-end/83575
12694         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12695         when in layout mode.
12696         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12697         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12698         partition fixup.
12699
12700 2018-01-10  Michael Collison  <michael.collison@arm.com>
12701
12702         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12703         * config/aarch64/aarch64-option-extension.def: Add
12704         AARCH64_OPT_EXTENSION of 'fp16fml'.
12705         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12706         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12707         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12708         * config/aarch64/constraints.md (Ui7): New constraint.
12709         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12710         (VFMLA_SEL_W): Ditto.
12711         (f16quad): Ditto.
12712         (f16mac1): Ditto.
12713         (VFMLA16_LOW): New int iterator.
12714         (VFMLA16_HIGH): Ditto.
12715         (UNSPEC_FMLAL): New unspec.
12716         (UNSPEC_FMLSL): Ditto.
12717         (UNSPEC_FMLAL2): Ditto.
12718         (UNSPEC_FMLSL2): Ditto.
12719         (f16mac): New code attribute.
12720         * config/aarch64/aarch64-simd-builtins.def
12721         (aarch64_fmlal_lowv2sf): Ditto.
12722         (aarch64_fmlsl_lowv2sf): Ditto.
12723         (aarch64_fmlalq_lowv4sf): Ditto.
12724         (aarch64_fmlslq_lowv4sf): Ditto.
12725         (aarch64_fmlal_highv2sf): Ditto.
12726         (aarch64_fmlsl_highv2sf): Ditto.
12727         (aarch64_fmlalq_highv4sf): Ditto.
12728         (aarch64_fmlslq_highv4sf): Ditto.
12729         (aarch64_fmlal_lane_lowv2sf): Ditto.
12730         (aarch64_fmlsl_lane_lowv2sf): Ditto.
12731         (aarch64_fmlal_laneq_lowv2sf): Ditto.
12732         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12733         (aarch64_fmlalq_lane_lowv4sf): Ditto.
12734         (aarch64_fmlsl_lane_lowv4sf): Ditto.
12735         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12736         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12737         (aarch64_fmlal_lane_highv2sf): Ditto.
12738         (aarch64_fmlsl_lane_highv2sf): Ditto.
12739         (aarch64_fmlal_laneq_highv2sf): Ditto.
12740         (aarch64_fmlsl_laneq_highv2sf): Ditto.
12741         (aarch64_fmlalq_lane_highv4sf): Ditto.
12742         (aarch64_fmlsl_lane_highv4sf): Ditto.
12743         (aarch64_fmlalq_laneq_highv4sf): Ditto.
12744         (aarch64_fmlsl_laneq_highv4sf): Ditto.
12745         * config/aarch64/aarch64-simd.md:
12746         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12747         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12748         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12749         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12750         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12751         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12752         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12753         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12754         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12755         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12756         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12757         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12758         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12759         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12760         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12761         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12762         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12763         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12764         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12765         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12766         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12767         (vfmlsl_low_u32): Ditto.
12768         (vfmlalq_low_u32): Ditto.
12769         (vfmlslq_low_u32): Ditto.
12770         (vfmlal_high_u32): Ditto.
12771         (vfmlsl_high_u32): Ditto.
12772         (vfmlalq_high_u32): Ditto.
12773         (vfmlslq_high_u32): Ditto.
12774         (vfmlal_lane_low_u32): Ditto.
12775         (vfmlsl_lane_low_u32): Ditto.
12776         (vfmlal_laneq_low_u32): Ditto.
12777         (vfmlsl_laneq_low_u32): Ditto.
12778         (vfmlalq_lane_low_u32): Ditto.
12779         (vfmlslq_lane_low_u32): Ditto.
12780         (vfmlalq_laneq_low_u32): Ditto.
12781         (vfmlslq_laneq_low_u32): Ditto.
12782         (vfmlal_lane_high_u32): Ditto.
12783         (vfmlsl_lane_high_u32): Ditto.
12784         (vfmlal_laneq_high_u32): Ditto.
12785         (vfmlsl_laneq_high_u32): Ditto.
12786         (vfmlalq_lane_high_u32): Ditto.
12787         (vfmlslq_lane_high_u32): Ditto.
12788         (vfmlalq_laneq_high_u32): Ditto.
12789         (vfmlslq_laneq_high_u32): Ditto.
12790         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12791         (AARCH64_FL_FOR_ARCH8_4): New.
12792         (AARCH64_ISA_F16FML): New ISA flag.
12793         (TARGET_F16FML): New feature flag for fp16fml.
12794         (doc/invoke.texi): Document new fp16fml option.
12795
12796 2018-01-10  Michael Collison  <michael.collison@arm.com>
12797
12798         * config/aarch64/aarch64-builtins.c:
12799         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12800         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12801         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12802         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12803         (AARCH64_ISA_SHA3): New ISA flag.
12804         (TARGET_SHA3): New feature flag for sha3.
12805         * config/aarch64/iterators.md (sha512_op): New int attribute.
12806         (CRYPTO_SHA512): New int iterator.
12807         (UNSPEC_SHA512H): New unspec.
12808         (UNSPEC_SHA512H2): Ditto.
12809         (UNSPEC_SHA512SU0): Ditto.
12810         (UNSPEC_SHA512SU1): Ditto.
12811         * config/aarch64/aarch64-simd-builtins.def
12812         (aarch64_crypto_sha512hqv2di): New builtin.
12813         (aarch64_crypto_sha512h2qv2di): Ditto.
12814         (aarch64_crypto_sha512su0qv2di): Ditto.
12815         (aarch64_crypto_sha512su1qv2di): Ditto.
12816         (aarch64_eor3qv8hi): Ditto.
12817         (aarch64_rax1qv2di): Ditto.
12818         (aarch64_xarqv2di): Ditto.
12819         (aarch64_bcaxqv8hi): Ditto.
12820         * config/aarch64/aarch64-simd.md:
12821         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12822         (aarch64_crypto_sha512su0qv2di): Ditto.
12823         (aarch64_crypto_sha512su1qv2di): Ditto.
12824         (aarch64_eor3qv8hi): Ditto.
12825         (aarch64_rax1qv2di): Ditto.
12826         (aarch64_xarqv2di): Ditto.
12827         (aarch64_bcaxqv8hi): Ditto.
12828         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12829         (vsha512h2q_u64): Ditto.
12830         (vsha512su0q_u64): Ditto.
12831         (vsha512su1q_u64): Ditto.
12832         (veor3q_u16): Ditto.
12833         (vrax1q_u64): Ditto.
12834         (vxarq_u64): Ditto.
12835         (vbcaxq_u16): Ditto.
12836         * config/arm/types.md (crypto_sha512): New type attribute.
12837         (crypto_sha3): Ditto.
12838         (doc/invoke.texi): Document new sha3 option.
12839
12840 2018-01-10  Michael Collison  <michael.collison@arm.com>
12841
12842         * config/aarch64/aarch64-builtins.c:
12843         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12844         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12845         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12846         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12847         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12848         (AARCH64_ISA_SM4): New ISA flag.
12849         (TARGET_SM4): New feature flag for sm4.
12850         * config/aarch64/aarch64-simd-builtins.def
12851         (aarch64_sm3ss1qv4si): Ditto.
12852         (aarch64_sm3tt1aq4si): Ditto.
12853         (aarch64_sm3tt1bq4si): Ditto.
12854         (aarch64_sm3tt2aq4si): Ditto.
12855         (aarch64_sm3tt2bq4si): Ditto.
12856         (aarch64_sm3partw1qv4si): Ditto.
12857         (aarch64_sm3partw2qv4si): Ditto.
12858         (aarch64_sm4eqv4si): Ditto.
12859         (aarch64_sm4ekeyqv4si): Ditto.
12860         * config/aarch64/aarch64-simd.md:
12861         (aarch64_sm3ss1qv4si): Ditto.
12862         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12863         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12864         (aarch64_sm4eqv4si): Ditto.
12865         (aarch64_sm4ekeyqv4si): Ditto.
12866         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12867         (sm3part_op): Ditto.
12868         (CRYPTO_SM3TT): Ditto.
12869         (CRYPTO_SM3PART): Ditto.
12870         (UNSPEC_SM3SS1): New unspec.
12871         (UNSPEC_SM3TT1A): Ditto.
12872         (UNSPEC_SM3TT1B): Ditto.
12873         (UNSPEC_SM3TT2A): Ditto.
12874         (UNSPEC_SM3TT2B): Ditto.
12875         (UNSPEC_SM3PARTW1): Ditto.
12876         (UNSPEC_SM3PARTW2): Ditto.
12877         (UNSPEC_SM4E): Ditto.
12878         (UNSPEC_SM4EKEY): Ditto.
12879         * config/aarch64/constraints.md (Ui2): New constraint.
12880         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12881         * config/arm/types.md (crypto_sm3): New type attribute.
12882         (crypto_sm4): Ditto.
12883         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12884         (vsm3tt1aq_u32): Ditto.
12885         (vsm3tt1bq_u32): Ditto.
12886         (vsm3tt2aq_u32): Ditto.
12887         (vsm3tt2bq_u32): Ditto.
12888         (vsm3partw1q_u32): Ditto.
12889         (vsm3partw2q_u32): Ditto.
12890         (vsm4eq_u32): Ditto.
12891         (vsm4ekeyq_u32): Ditto.
12892         (doc/invoke.texi): Document new sm4 option.
12893
12894 2018-01-10  Michael Collison  <michael.collison@arm.com>
12895
12896         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12897         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12898         (AARCH64_FL_FOR_ARCH8_4): New.
12899         (AARCH64_FL_V8_4): New flag.
12900         (doc/invoke.texi): Document new armv8.4-a option.
12901
12902 2018-01-10  Michael Collison  <michael.collison@arm.com>
12903
12904         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12905         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12906         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12907         * config/aarch64/aarch64-option-extension.def: Add
12908         AARCH64_OPT_EXTENSION of 'sha2'.
12909         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12910         (crypto): Disable sha2 and aes if crypto disabled.
12911         (crypto): Enable aes and sha2 if enabled.
12912         (simd): Disable sha2 and aes if simd disabled.
12913         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12914         New flags.
12915         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12916         (TARGET_SHA2): New feature flag for sha2.
12917         (TARGET_AES): New feature flag for aes.
12918         * config/aarch64/aarch64-simd.md:
12919         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12920         conditional on TARGET_AES.
12921         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12922         (aarch64_crypto_sha1hsi): Make pattern conditional
12923         on TARGET_SHA2.
12924         (aarch64_crypto_sha1hv4si): Ditto.
12925         (aarch64_be_crypto_sha1hv4si): Ditto.
12926         (aarch64_crypto_sha1su1v4si): Ditto.
12927         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12928         (aarch64_crypto_sha1su0v4si): Ditto.
12929         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12930         (aarch64_crypto_sha256su0v4si): Ditto.
12931         (aarch64_crypto_sha256su1v4si): Ditto.
12932         (doc/invoke.texi): Document new aes and sha2 options.
12933
12934 2018-01-10  Martin Sebor  <msebor@redhat.com>
12935
12936         PR tree-optimization/83781
12937         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12938         as string arrays.
12939
12940 2018-01-11  Martin Sebor  <msebor@gmail.com>
12941             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12942
12943         PR tree-optimization/83501
12944         PR tree-optimization/81703
12945
12946         * tree-ssa-strlen.c (get_string_cst): Rename...
12947         (get_string_len): ...to this.  Handle global constants.
12948         (handle_char_store): Adjust.
12949
12950 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
12951             Jim Wilson  <jimw@sifive.com>
12952
12953         * config/riscv/riscv-protos.h (riscv_output_return): New.
12954         * config/riscv/riscv.c (struct machine_function): New naked_p field.
12955         (riscv_attribute_table, riscv_output_return),
12956         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12957         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12958         (riscv_compute_frame_info): Only compute frame->mask if not a naked
12959         function.
12960         (riscv_expand_prologue): Add early return for naked function.
12961         (riscv_expand_epilogue): Likewise.
12962         (riscv_function_ok_for_sibcall): Return false for naked function.
12963         (riscv_set_current_function): New.
12964         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12965         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12966         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12967         * doc/extend.texi (RISC-V Function Attributes): New.
12968
12969 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
12970
12971         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12972         check for 128-bit long double before checking TCmode.
12973         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12974         128-bit long doubles before checking TFmode or TCmode.
12975         (FLOAT128_IBM_P): Likewise.
12976
12977 2018-01-10  Martin Sebor  <msebor@redhat.com>
12978
12979         PR tree-optimization/83671
12980         * builtins.c (c_strlen): Unconditionally return zero for the empty
12981         string.
12982         Use -Warray-bounds for warnings.
12983         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12984         for non-constant array indices with COMPONENT_REF, arrays of
12985         arrays, and pointers to arrays.
12986         (gimple_fold_builtin_strlen): Determine and set length range for
12987         non-constant character arrays.
12988
12989 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
12990
12991         PR middle-end/81897
12992         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12993         empty blocks.
12994
12995 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
12996
12997         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12998
12999 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
13000
13001         PR target/83399
13002         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13003         VECTOR_MEM_ALTIVEC_OR_VSX_P.
13004         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13005         indexed_or_indirect_operand predicate.
13006         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13007         (*vsx_le_perm_load_v8hi): Likewise.
13008         (*vsx_le_perm_load_v16qi): Likewise.
13009         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13010         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13011         (*vsx_le_perm_store_v8hi): Likewise.
13012         (*vsx_le_perm_store_v16qi): Likewise.
13013         (eight unnamed splitters): Likewise.
13014
13015 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
13016
13017         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13018         * config/rs6000/emmintrin.h: Likewise.
13019         * config/rs6000/mmintrin.h: Likewise.
13020         * config/rs6000/xmmintrin.h: Likewise.
13021
13022 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
13023
13024         PR c++/43486
13025         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13026         "public_flag".
13027         * tree.c (tree_nop_conversion): Return true for location wrapper
13028         nodes.
13029         (maybe_wrap_with_location): New function.
13030         (selftest::check_strip_nops): New function.
13031         (selftest::test_location_wrappers): New function.
13032         (selftest::tree_c_tests): Call it.
13033         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13034         (maybe_wrap_with_location): New decl.
13035         (EXPR_LOCATION_WRAPPER_P): New macro.
13036         (location_wrapper_p): New inline function.
13037         (tree_strip_any_location_wrapper): New inline function.
13038
13039 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
13040
13041         PR target/83735
13042         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13043         stack_realign_offset for the largest alignment of stack slot
13044         actually used.
13045         (ix86_find_max_used_stack_alignment): New function.
13046         (ix86_finalize_stack_frame_flags): Use it.  Set
13047         max_used_stack_alignment if we don't realign stack.
13048         * config/i386/i386.h (machine_function): Add
13049         max_used_stack_alignment.
13050
13051 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
13052
13053         * config/arm/arm.opt (-mbranch-cost): New option.
13054         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13055         account.
13056
13057 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
13058
13059         PR target/83629
13060         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13061         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13062
13063 2018-01-10  Richard Biener  <rguenther@suse.de>
13064
13065         PR debug/83765
13066         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13067         early out so it also covers the case where we have a non-NULL
13068         origin.
13069
13070 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
13071
13072         PR tree-optimization/83753
13073         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13074         for non-strided grouped accesses if the number of elements is 1.
13075
13076 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
13077
13078         PR target/81616
13079         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13080         * i386.h (TARGET_USE_GATHER): Define.
13081         * x86-tune.def (X86_TUNE_USE_GATHER): New.
13082
13083 2018-01-10  Martin Liska  <mliska@suse.cz>
13084
13085         PR bootstrap/82831
13086         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13087         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13088         partitioning.
13089         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13090         CLEANUP_NO_PARTITIONING is not set.
13091
13092 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
13093
13094         * doc/rtl.texi: Remove documentation of (const ...) wrappers
13095         for vectors, as a partial revert of r254296.
13096         * rtl.h (const_vec_p): Delete.
13097         (const_vec_duplicate_p): Don't test for vector CONSTs.
13098         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13099         * expmed.c (make_tree): Likewise.
13100
13101         Revert:
13102         * common.md (E, F): Use CONSTANT_P instead of checking for
13103         CONST_VECTOR.
13104         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13105         checking for CONST_VECTOR.
13106
13107 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13108
13109         PR middle-end/83575
13110         * predict.c (force_edge_cold): Handle in more sane way edges
13111         with no prediction.
13112
13113 2018-01-09  Carl Love  <cel@us.ibm.com>
13114
13115         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13116         V4SI, V4SF types.
13117         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13118         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13119         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13120         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
13121         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13122         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
13123         * config/rs6000/rs6000-protos.h: Add extern defition for
13124         rs6000_generate_float2_double_code.
13125         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13126         function.
13127         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13128         (float2_v2df): Add define_expand.
13129
13130 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
13131
13132         PR target/83628
13133         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13134         op_mode in the force_to_mode call.
13135
13136 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13137
13138         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13139         instead of checking each element individually.
13140         (aarch64_evpc_uzp): Likewise.
13141         (aarch64_evpc_zip): Likewise.
13142         (aarch64_evpc_ext): Likewise.
13143         (aarch64_evpc_rev): Likewise.
13144         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13145         instead of checking each element individually.  Return true without
13146         generating rtl if
13147         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13148         whether all selected elements come from the same input, instead of
13149         checking each element individually.  Remove calls to gen_rtx_REG,
13150         start_sequence and end_sequence and instead assert that no rtl is
13151         generated.
13152
13153 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13154
13155         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13156         order of HIGH and CONST checks.
13157
13158 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13159
13160         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13161         if the destination isn't an SSA_NAME.
13162
13163 2018-01-09  Richard Biener  <rguenther@suse.de>
13164
13165         PR tree-optimization/83668
13166         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13167         move prologue...
13168         (canonicalize_loop_form): ... here, renamed from ...
13169         (canonicalize_loop_closed_ssa_form): ... this and amended to
13170         swap successor edges for loop exit blocks to make us use
13171         the RPO order we need for initial schedule generation.
13172
13173 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
13174
13175         PR tree-optimization/64811
13176         * match.pd: When optimizing comparisons with Inf, avoid
13177         introducing or losing exceptions from comparisons with NaN.
13178
13179 2018-01-09  Martin Liska  <mliska@suse.cz>
13180
13181         PR sanitizer/82517
13182         * asan.c (shadow_mem_size): Add gcc_assert.
13183
13184 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
13185
13186         Don't save registers in main().
13187
13188         PR target/83738
13189         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13190         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13191         * config/avr/avr.c (avr_set_current_function): Don't error if
13192         naked, OS_task or OS_main are specified at the same time.
13193         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13194         OS_main.
13195         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13196         attribute.
13197         * common/config/avr/avr-common.c (avr_option_optimization_table):
13198         Switch on -mmain-is-OS_task for optimizing compilations.
13199
13200 2018-01-09  Richard Biener  <rguenther@suse.de>
13201
13202         PR tree-optimization/83572
13203         * graphite.c: Include cfganal.h.
13204         (graphite_transform_loops): Connect infinite loops to exit
13205         and remove fake edges at the end.
13206
13207 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13208
13209         * ipa-inline.c (edge_badness): Revert accidental checkin.
13210
13211 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13212
13213         PR ipa/80763
13214         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13215         symbols; not inline clones.
13216
13217 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
13218
13219         PR target/83507
13220         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13221         hard registers.  Formatting fixes.
13222
13223         PR preprocessor/83722
13224         * gcc.c (try_generate_repro): Pass
13225         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13226         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13227         do_report_bug.
13228
13229 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
13230             Kito Cheng  <kito.cheng@gmail.com>
13231
13232         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13233         (riscv_leaf_function_p): Delete.
13234         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13235
13236 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13237
13238         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13239         function.
13240         (do_ifelse): New function.
13241         (do_isel): New function.
13242         (do_sub3): New function.
13243         (do_add3): New function.
13244         (do_load_mask_compare): New function.
13245         (do_overlap_load_compare): New function.
13246         (expand_compare_loop): New function.
13247         (expand_block_compare): Call expand_compare_loop() when appropriate.
13248         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13249         option description.
13250         (-mblock-compare-inline-loop-limit): New option.
13251
13252 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13253
13254         PR target/83677
13255         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13256         Reverse order of second and third operands in first alternative.
13257         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13258         of first and second elements in UNSPEC_VPERMR vector.
13259         (altivec_expand_vec_perm_le): Likewise.
13260
13261 2018-01-08  Jeff Law  <law@redhat.com>
13262
13263         PR rtl-optimizatin/81308
13264         * tree-switch-conversion.c (cfg_altered): New file scoped static.
13265         (process_switch): If group_case_labels makes a change, then set
13266         cfg_altered.
13267         (pass_convert_switch::execute): If a switch is converted, then
13268         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
13269
13270         PR rtl-optimization/81308
13271         * recog.c (split_all_insns): Conditionally cleanup the CFG after
13272         splitting insns.
13273
13274 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
13275
13276         PR target/83663 - Revert r255946
13277         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13278         generation for cases where splatting a value is not useful.
13279         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13280         across a vec_duplicate and a paradoxical subreg forming a vector
13281         mode to a vec_concat.
13282
13283 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13284
13285         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13286         -march=armv8.3-a variants.
13287         * config/arm/t-multilib: Likewise.
13288         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
13289
13290 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13291
13292         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13293         to generate rtl.
13294         (cceq_ior_compare_complement): Give it a name so I can use it, and
13295         change boolean_or_operator predicate to boolean_operator so it can
13296         be used to generate a crand.
13297         (eqne): New code iterator.
13298         (bd/bd_neg): New code_attrs.
13299         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13300         a single define_insn.
13301         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13302         decrement (bdnzt/bdnzf/bdzt/bdzf).
13303         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13304         with the new names of the branch decrement patterns, and added the
13305         names of the branch decrement conditional patterns.
13306
13307 2018-01-08  Richard Biener  <rguenther@suse.de>
13308
13309         PR tree-optimization/83563
13310         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13311         cache.
13312
13313 2018-01-08  Richard Biener  <rguenther@suse.de>
13314
13315         PR middle-end/83713
13316         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13317
13318 2018-01-08  Richard Biener  <rguenther@suse.de>
13319
13320         PR tree-optimization/83685
13321         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13322         references to abnormals.
13323
13324 2018-01-08  Richard Biener  <rguenther@suse.de>
13325
13326         PR lto/83719
13327         * dwarf2out.c (output_indirect_strings): Handle empty
13328         skeleton_debug_str_hash.
13329         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13330
13331 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
13332
13333         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13334         (emit_store_direct): Likewise.
13335         (arc_trampoline_adjust_address): Likewise.
13336         (arc_asm_trampoline_template): New function.
13337         (arc_initialize_trampoline): Use asm_trampoline_template.
13338         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13339         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13340         * config/arc/arc.md (flush_icache): Delete pattern.
13341
13342 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
13343
13344         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13345         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13346         munaligned-access.
13347
13348 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13349
13350         PR target/83681
13351         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13352         by not USED_FOR_TARGET.
13353         (make_pass_resolve_sw_modes): Likewise.
13354
13355 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13356
13357         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13358         USED_FOR_TARGET.
13359
13360 2018-01-08  Richard Biener  <rguenther@suse.de>
13361
13362         PR middle-end/83580
13363         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13364
13365 2018-01-08  Richard Biener  <rguenther@suse.de>
13366
13367         PR middle-end/83517
13368         * match.pd ((t * 2) / 2) -> t): Add missing :c.
13369
13370 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
13371
13372         PR middle-end/81897
13373         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13374         basic blocks with a small number of successors.
13375         (convert_control_dep_chain_into_preds): Improve handling of
13376         forwarder blocks.
13377         (dump_predicates): Split apart into...
13378         (dump_pred_chain): ...here...
13379         (dump_pred_info): ...and here.
13380         (can_one_predicate_be_invalidated_p): Add debugging printfs.
13381         (can_chain_union_be_invalidated_p): Improve check for invalidation
13382         of paths.
13383         (uninit_uses_cannot_happen): Avoid unnecessary if
13384         convert_control_dep_chain_into_preds yielded nothing.
13385
13386 2018-01-06  Martin Sebor  <msebor@redhat.com>
13387
13388         PR tree-optimization/83640
13389         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13390         subtracting negative offset from size.
13391         (builtin_access::overlap): Adjust offset bounds of the access to fall
13392         within the size of the object if possible.
13393
13394 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
13395
13396         PR rtl-optimization/83699
13397         * expmed.c (extract_bit_field_1): Restrict the vector usage of
13398         extract_bit_field_as_subreg to cases in which the extracted
13399         value is also a vector.
13400
13401         * lra-constraints.c (process_alt_operands): Test for the equivalence
13402         substitutions when detecting a possible reload cycle.
13403
13404 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
13405
13406         PR debug/83480
13407         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13408         by default if flag_selective_schedling{,2}.  Formatting fixes.
13409
13410         PR rtl-optimization/83682
13411         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13412         if it has non-VECTOR_MODE element mode.
13413         (vec_duplicate_p): Likewise.
13414
13415         PR middle-end/83694
13416         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13417         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13418
13419 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
13420
13421         PR target/83604
13422         * config/i386/i386-builtin.def
13423         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13424         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13425         Require also OPTION_MASK_ISA_AVX512F in addition to
13426         OPTION_MASK_ISA_GFNI.
13427         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13428         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13429         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13430         to OPTION_MASK_ISA_GFNI.
13431         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13432         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13433         OPTION_MASK_ISA_AVX512BW.
13434         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13435         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13436         addition to OPTION_MASK_ISA_GFNI.
13437         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13438         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13439         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13440         to OPTION_MASK_ISA_GFNI.
13441         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13442         a requirement for all ISAs rather than any of them with a few
13443         exceptions.
13444         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13445         processing.
13446         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13447         bitmasks to be enabled with 3 exceptions, instead of requiring any
13448         enabled ISA with lots of exceptions.
13449         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13450         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13451         Change avx512bw in isa attribute to avx512f.
13452         * config/i386/sgxintrin.h: Add license boilerplate.
13453         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
13454         to __AVX512F__ and __AVX512VL to __AVX512VL__.
13455         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13456         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13457         defined.
13458         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13459         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13460         temporarily sse2 rather than sse if not enabled already.
13461
13462         PR target/83604
13463         * config/i386/sse.md (VI248_VLBW): Rename to ...
13464         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
13465         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13466         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13467         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13468         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13469         mode iterator instead of VI248_VLBW.
13470
13471 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
13472
13473         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13474         (record_modified): Skip clobbers; add debug output.
13475         (param_change_prob): Use sreal frequencies.
13476
13477 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
13478
13479         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13480         punt for user-aligned variables.
13481
13482 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
13483
13484         * tree-chrec.c (chrec_contains_symbols): Return true for
13485         POLY_INT_CST.
13486
13487 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
13488
13489         PR target/82439
13490         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13491         of (x|y) == x for BICS pattern.
13492
13493 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
13494
13495         PR tree-optimization/83605
13496         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13497         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13498         can throw.
13499
13500 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13501
13502         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13503         * config/epiphany/rtems.h: New file.
13504
13505 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13506             Uros Bizjak  <ubizjak@gmail.com>
13507
13508         PR target/83554
13509         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13510         QIreg_operand instead of register_operand predicate.
13511         * config/i386/i386.c (ix86_rop_should_change_byte_p,
13512         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13513         comments instead of -fmitigate[-_]rop.
13514
13515 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13516
13517         PR bootstrap/81926
13518         * cgraphunit.c (symbol_table::compile): Switch to text_section
13519         before calling assembly_start debug hook.
13520         * run-rtl-passes.c (run_rtl_passes): Likewise.
13521         Include output.h.
13522
13523 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13524
13525         * tree-vrp.c (extract_range_from_binary_expr_1): Check
13526         range_int_cst_p rather than !symbolic_range_p before calling
13527         extract_range_from_multiplicative_op_1.
13528
13529 2018-01-04  Jeff Law  <law@redhat.com>
13530
13531         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13532         redundant test in assertion.
13533
13534 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13535
13536         * doc/rtl.texi: Document machine_mode wrapper classes.
13537
13538 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13539
13540         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13541         using tree_to_uhwi.
13542
13543 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13544
13545         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13546         the VEC_PERM_EXPR fold to fail.
13547
13548 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13549
13550         PR debug/83585
13551         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13552         to switched_sections.
13553
13554 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13555
13556         PR target/83680
13557         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13558         test for d.testing.
13559
13560 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
13561
13562         PR target/83387
13563         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13564         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13565
13566 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13567
13568         PR debug/83666
13569         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13570         is BLKmode and bitpos not zero or mode change is needed.
13571
13572 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13573
13574         PR target/83675
13575         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13576         TARGET_VIS2.
13577
13578 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
13579
13580         PR target/83628
13581         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13582         instead of MULT rtx.  Update all corresponding splitters.
13583         (*saddl_se): Ditto.
13584         (*ssub<modesuffix>): Ditto.
13585         (*ssubl_se): Ditto.
13586         (*cmp_sadd_di): Update split patterns.
13587         (*cmp_sadd_si): Ditto.
13588         (*cmp_sadd_sidi): Ditto.
13589         (*cmp_ssub_di): Ditto.
13590         (*cmp_ssub_si): Ditto.
13591         (*cmp_ssub_sidi): Ditto.
13592         * config/alpha/predicates.md (const23_operand): New predicate.
13593         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13594         Look for ASHIFT, not MULT inner operand.
13595         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13596
13597 2018-01-04  Martin Liska  <mliska@suse.cz>
13598
13599         PR gcov-profile/83669
13600         * gcov.c (output_intermediate_file): Add version to intermediate
13601         gcov file.
13602         * doc/gcov.texi: Document new field 'version' in intermediate
13603         file format. Fix location of '-k' option of gcov command.
13604
13605 2018-01-04  Martin Liska  <mliska@suse.cz>
13606
13607         PR ipa/82352
13608         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13609
13610 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13611
13612         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13613
13614 2018-01-03  Martin Sebor  <msebor@redhat.com>
13615
13616         PR tree-optimization/83655
13617         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13618         checking calls with invalid arguments.
13619
13620 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13621
13622         * tree-vect-stmts.c (vect_get_store_rhs): New function.
13623         (vectorizable_mask_load_store): Delete.
13624         (vectorizable_call): Return false for masked loads and stores.
13625         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
13626         instead of gimple_assign_rhs1.
13627         (vectorizable_load): Handle IFN_MASK_LOAD.
13628         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13629
13630 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13631
13632         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13633         split out from..,
13634         (vectorizable_mask_load_store): ...here.
13635         (vectorizable_load): ...and here.
13636
13637 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13638
13639         * tree-vect-stmts.c (vect_build_all_ones_mask)
13640         (vect_build_zero_merge_argument): New functions, split out from...
13641         (vectorizable_load): ...here.
13642
13643 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13644
13645         * tree-vect-stmts.c (vect_check_store_rhs): New function,
13646         split out from...
13647         (vectorizable_mask_load_store): ...here.
13648         (vectorizable_store): ...and here.
13649
13650 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13651
13652         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13653         split out from...
13654         (vectorizable_mask_load_store): ...here.
13655
13656 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13657
13658         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13659         (vect_model_store_cost): Take a vec_load_store_type instead of a
13660         vect_def_type.
13661         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13662         (vect_model_store_cost): Take a vec_load_store_type instead of a
13663         vect_def_type.
13664         (vectorizable_mask_load_store): Update accordingly.
13665         (vectorizable_store): Likewise.
13666         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13667
13668 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13669
13670         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13671         IFN_MASK_LOAD calls here rather than...
13672         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13673
13674 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13675             Alan Hayward  <alan.hayward@arm.com>
13676             David Sherwood  <david.sherwood@arm.com>
13677
13678         * expmed.c (extract_bit_field_1): For vector extracts,
13679         fall back to extract_bit_field_as_subreg if vec_extract
13680         isn't available.
13681
13682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13683             Alan Hayward  <alan.hayward@arm.com>
13684             David Sherwood  <david.sherwood@arm.com>
13685
13686         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13687         they are variable or constant sized.
13688         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13689         slots for constant-sized data.
13690
13691 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13692             Alan Hayward  <alan.hayward@arm.com>
13693             David Sherwood  <david.sherwood@arm.com>
13694
13695         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13696         handling COND_EXPRs with boolean comparisons, try to find a better
13697         basis for the mask type than the boolean itself.
13698
13699 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13700
13701         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13702         is calculated and how it can be overridden.
13703         * genmodes.c (max_bitsize_mode_any_mode): New variable.
13704         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13705         if defined.
13706         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13707         if nonzero.
13708
13709 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13710             Alan Hayward  <alan.hayward@arm.com>
13711             David Sherwood  <david.sherwood@arm.com>
13712
13713         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13714         Remove the mode argument.
13715         (aarch64_simd_valid_immediate): Remove the mode and inverse
13716         arguments.
13717         * config/aarch64/iterators.md (bitsize): New iterator.
13718         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13719         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13720         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13721         aarch64_simd_valid_immediate.
13722         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13723         (aarch64_reg_or_bic_imm): Likewise.
13724         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13725         with an insn_type enum and msl with a modifier_type enum.
13726         Replace element_width with a scalar_mode.  Change the shift
13727         to unsigned int.  Add constructors for scalar_float_mode and
13728         scalar_int_mode elements.
13729         (aarch64_vect_float_const_representable_p): Delete.
13730         (aarch64_can_const_movi_rtx_p)
13731         (aarch64_simd_scalar_immediate_valid_for_move)
13732         (aarch64_simd_make_constant): Update call to
13733         aarch64_simd_valid_immediate.
13734         (aarch64_advsimd_valid_immediate_hs): New function.
13735         (aarch64_advsimd_valid_immediate): Likewise.
13736         (aarch64_simd_valid_immediate): Remove mode and inverse
13737         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
13738         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13739         and aarch64_float_const_representable_p on the result.
13740         (aarch64_output_simd_mov_immediate): Remove mode argument.
13741         Update call to aarch64_simd_valid_immediate and use of
13742         simd_immediate_info.
13743         (aarch64_output_scalar_simd_mov_immediate): Update call
13744         accordingly.
13745
13746 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13747             Alan Hayward  <alan.hayward@arm.com>
13748             David Sherwood  <david.sherwood@arm.com>
13749
13750         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13751         (mode_nunits): Likewise CONST_MODE_NUNITS.
13752         * machmode.def (ADJUST_NUNITS): Document.
13753         * genmodes.c (mode_data::need_nunits_adj): New field.
13754         (blank_mode): Update accordingly.
13755         (adj_nunits): New variable.
13756         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13757         parameter.
13758         (emit_mode_size_inline): Set need_bytesize_adj for all modes
13759         listed in adj_nunits.
13760         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13761         listed in adj_nunits.  Don't emit case statements for such modes.
13762         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13763         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
13764         nothing if adj_nunits is nonnull.
13765         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13766         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13767         (emit_mode_fbit): Update use of print_maybe_const_decl.
13768         (emit_move_size): Likewise.  Treat the array as non-const
13769         if adj_nunits.
13770         (emit_mode_adjustments): Handle adj_nunits.
13771
13772 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13773
13774         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13775         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13776         (VECTOR_MODES): Use it.
13777         (make_vector_modes): Take the prefix as an argument.
13778
13779 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13780             Alan Hayward  <alan.hayward@arm.com>
13781             David Sherwood  <david.sherwood@arm.com>
13782
13783         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13784         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13785         for MODE_VECTOR_BOOL.
13786         * machmode.def (VECTOR_BOOL_MODE): Document.
13787         * genmodes.c (VECTOR_BOOL_MODE): New macro.
13788         (make_vector_bool_mode): New function.
13789         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13790         MODE_VECTOR_BOOL.
13791         * lto-streamer-in.c (lto_input_mode_table): Likewise.
13792         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13793         Likewise.
13794         * stor-layout.c (int_mode_for_mode): Likewise.
13795         * tree.c (build_vector_type_for_mode): Likewise.
13796         * varasm.c (output_constant_pool_2): Likewise.
13797         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13798         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
13799         for MODE_VECTOR_BOOL.
13800         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13801         of mode class checks.
13802         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13803         instead of a list of mode class checks.
13804         (expand_vector_scalar_condition): Likewise.
13805         (type_for_widest_vector_mode): Handle BImode as an inner mode.
13806
13807 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13808             Alan Hayward  <alan.hayward@arm.com>
13809             David Sherwood  <david.sherwood@arm.com>
13810
13811         * machmode.h (mode_size): Change from unsigned short to
13812         poly_uint16_pod.
13813         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13814         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13815         or if measurement_type is not polynomial.
13816         (fixed_size_mode::includes_p): Check for constant-sized modes.
13817         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13818         return a poly_uint16 rather than an unsigned short.
13819         (emit_mode_size): Change the type of mode_size from unsigned short
13820         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
13821         (emit_mode_adjustments): Cope with polynomial vector sizes.
13822         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13823         for GET_MODE_SIZE.
13824         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13825         for GET_MODE_SIZE.
13826         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13827         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13828         * caller-save.c (setup_save_areas): Likewise.
13829         (replace_reg_with_saved_mem): Likewise.
13830         * calls.c (emit_library_call_value_1): Likewise.
13831         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13832         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13833         (gen_lowpart_for_combine): Likewise.
13834         * convert.c (convert_to_integer_1): Likewise.
13835         * cse.c (equiv_constant, cse_insn): Likewise.
13836         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13837         (cselib_subst_to_values): Likewise.
13838         * dce.c (word_dce_process_block): Likewise.
13839         * df-problems.c (df_word_lr_mark_ref): Likewise.
13840         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13841         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13842         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13843         (rtl_for_decl_location): Likewise.
13844         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13845         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13846         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13847         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13848         (expand_expr_real_1): Likewise.
13849         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13850         (pad_below): Likewise.
13851         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13852         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13853         * ira.c (get_subreg_tracking_sizes): Likewise.
13854         * ira-build.c (ira_create_allocno_objects): Likewise.
13855         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13856         (ira_sort_regnos_for_alter_reg): Likewise.
13857         * ira-costs.c (record_operand_costs): Likewise.
13858         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13859         (resolve_simple_move): Likewise.
13860         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13861         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13862         (lra_constraints): Likewise.
13863         (CONST_POOL_OK_P): Reject variable-sized modes.
13864         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13865         (add_pseudo_to_slot, lra_spill): Likewise.
13866         * omp-low.c (omp_clause_aligned_alignment): Likewise.
13867         * optabs-query.c (get_best_extraction_insn): Likewise.
13868         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13869         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13870         (expand_mult_highpart, valid_multiword_target_p): Likewise.
13871         * recog.c (offsettable_address_addr_space_p): Likewise.
13872         * regcprop.c (maybe_mode_change): Likewise.
13873         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13874         * regrename.c (build_def_use): Likewise.
13875         * regstat.c (dump_reg_info): Likewise.
13876         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13877         (find_reloads, find_reloads_subreg_address): Likewise.
13878         * reload1.c (eliminate_regs_1): Likewise.
13879         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13880         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13881         (simplify_binary_operation_1, simplify_subreg): Likewise.
13882         * targhooks.c (default_function_arg_padding): Likewise.
13883         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13884         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13885         (verify_gimple_assign_ternary): Likewise.
13886         * tree-inline.c (estimate_move_cost): Likewise.
13887         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13888         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13889         (get_address_cost_ainc): Likewise.
13890         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13891         (vect_supportable_dr_alignment): Likewise.
13892         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13893         (vectorizable_reduction): Likewise.
13894         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13895         (vectorizable_operation, vectorizable_load): Likewise.
13896         * tree.c (build_same_sized_truth_vector_type): Likewise.
13897         * valtrack.c (cleanup_auto_inc_dec): Likewise.
13898         * var-tracking.c (emit_note_insn_var_location): Likewise.
13899         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13900         (ADDR_VEC_ALIGN): Likewise.
13901
13902 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13903             Alan Hayward  <alan.hayward@arm.com>
13904             David Sherwood  <david.sherwood@arm.com>
13905
13906         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13907         unsigned short.
13908         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13909         or if measurement_type is polynomial.
13910         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13911         * combine.c (make_extraction): Likewise.
13912         * dse.c (find_shift_sequence): Likewise.
13913         * dwarf2out.c (mem_loc_descriptor): Likewise.
13914         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13915         (extract_bit_field, extract_low_bits): Likewise.
13916         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13917         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13918         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13919         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13920         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13921         * reload.c (find_reloads): Likewise.
13922         * reload1.c (alter_reg): Likewise.
13923         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13924         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13925         * tree-if-conv.c (predicate_mem_writes): Likewise.
13926         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13927         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13928         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13929         * valtrack.c (dead_debug_insert_temp): Likewise.
13930         * varasm.c (mergeable_constant_section): Likewise.
13931         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13932
13933 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13934             Alan Hayward  <alan.hayward@arm.com>
13935             David Sherwood  <david.sherwood@arm.com>
13936
13937         * expr.c (expand_assignment): Cope with polynomial mode sizes
13938         when assigning to a CONCAT.
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_precision): Change from unsigned short to
13945         poly_uint16_pod.
13946         (mode_to_precision): Return a poly_uint16 rather than an unsigned
13947         short.
13948         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13949         or if measurement_type is not polynomial.
13950         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
13951         in which the mode is already known to be a scalar_int_mode.
13952         * genmodes.c (emit_mode_precision): Change the type of mode_precision
13953         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
13954         initializer.
13955         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13956         for GET_MODE_PRECISION.
13957         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13958         for GET_MODE_PRECISION.
13959         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13960         as polynomial.
13961         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13962         (expand_field_assignment, make_extraction): Likewise.
13963         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13964         (get_last_value): Likewise.
13965         * convert.c (convert_to_integer_1): Likewise.
13966         * cse.c (cse_insn): Likewise.
13967         * expr.c (expand_expr_real_1): Likewise.
13968         * lra-constraints.c (simplify_operand_subreg): Likewise.
13969         * optabs-query.c (can_atomic_load_p): Likewise.
13970         * optabs.c (expand_atomic_load): Likewise.
13971         (expand_atomic_store): Likewise.
13972         * ree.c (combine_reaching_defs): Likewise.
13973         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13974         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13975         * tree.h (type_has_mode_precision_p): Likewise.
13976         * ubsan.c (instrument_si_overflow): Likewise.
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         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13983         polynomial numbers of units.
13984         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13985         (valid_vector_subparts_p): New function.
13986         (build_vector_type): Remove temporary shim and take the number
13987         of units as a poly_uint64 rather than an int.
13988         (build_opaque_vector_type): Take the number of units as a
13989         poly_uint64 rather than an int.
13990         * tree.c (build_vector_from_ctor): Handle polynomial
13991         TYPE_VECTOR_SUBPARTS.
13992         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13993         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13994         (build_vector_from_val): If the number of units is variable,
13995         use build_vec_duplicate_cst for constant operands and
13996         VEC_DUPLICATE_EXPR otherwise.
13997         (make_vector_type): Remove temporary is_constant ().
13998         (build_vector_type, build_opaque_vector_type): Take the number of
13999         units as a poly_uint64 rather than an int.
14000         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14001         VECTOR_CST_NELTS.
14002         * cfgexpand.c (expand_debug_expr): Likewise.
14003         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14004         (store_constructor, expand_expr_real_1): Likewise.
14005         (const_scalar_mask_from_tree): Likewise.
14006         * fold-const-call.c (fold_const_reduction): Likewise.
14007         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14008         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14009         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14010         (fold_relational_const): Likewise.
14011         (native_interpret_vector): Likewise.  Change the size from an
14012         int to an unsigned int.
14013         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14014         TYPE_VECTOR_SUBPARTS.
14015         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14016         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14017         duplicating a non-constant operand into a variable-length vector.
14018         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14019         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14020         * ipa-icf.c (sem_variable::equals): Likewise.
14021         * match.pd: Likewise.
14022         * omp-simd-clone.c (simd_clone_subparts): Likewise.
14023         * print-tree.c (print_node): Likewise.
14024         * stor-layout.c (layout_type): Likewise.
14025         * targhooks.c (default_builtin_vectorization_cost): Likewise.
14026         * tree-cfg.c (verify_gimple_comparison): Likewise.
14027         (verify_gimple_assign_binary): Likewise.
14028         (verify_gimple_assign_ternary): Likewise.
14029         (verify_gimple_assign_single): Likewise.
14030         * tree-pretty-print.c (dump_generic_node): Likewise.
14031         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14032         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14033         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14034         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14035         (vect_shift_permute_load_chain): Likewise.
14036         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14037         (expand_vector_condition, optimize_vector_constructor): Likewise.
14038         (lower_vec_perm, get_compute_type): Likewise.
14039         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14040         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14041         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14042         (vect_recog_mask_conversion_pattern): Likewise.
14043         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14044         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14045         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14046         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14047         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14048         (vectorizable_shift, vectorizable_operation, vectorizable_store)
14049         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14050         (supportable_widening_operation): Likewise.
14051         (supportable_narrowing_operation): Likewise.
14052         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14053         Likewise.
14054         * varasm.c (output_constant): Likewise.
14055
14056 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14057             Alan Hayward  <alan.hayward@arm.com>
14058             David Sherwood  <david.sherwood@arm.com>
14059
14060         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14061         so that both the length == 3 and length != 3 cases set up their
14062         own permute vectors.  Add comments explaining why we know the
14063         number of elements is constant.
14064         (vect_permute_load_chain): Likewise.
14065
14066 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14067             Alan Hayward  <alan.hayward@arm.com>
14068             David Sherwood  <david.sherwood@arm.com>
14069
14070         * machmode.h (mode_nunits): Change from unsigned char to
14071         poly_uint16_pod.
14072         (ONLY_FIXED_SIZE_MODES): New macro.
14073         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14074         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14075         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14076         New typedefs.
14077         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14078         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14079         or if measurement_type is not polynomial.
14080         * genmodes.c (ZERO_COEFFS): New macro.
14081         (emit_mode_nunits_inline): Make mode_nunits_inline return a
14082         poly_uint16.
14083         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14084         Use ZERO_COEFFS when emitting initializers.
14085         * data-streamer.h (bp_pack_poly_value): New function.
14086         (bp_unpack_poly_value): Likewise.
14087         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14088         for GET_MODE_NUNITS.
14089         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14090         for GET_MODE_NUNITS.
14091         * tree.c (make_vector_type): Remove temporary shim and make
14092         the real function take the number of units as a poly_uint64
14093         rather than an int.
14094         (build_vector_type_for_mode): Handle polynomial nunits.
14095         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14096         * emit-rtl.c (const_vec_series_p_1): Likewise.
14097         (gen_rtx_CONST_VECTOR): Likewise.
14098         * fold-const.c (test_vec_duplicate_folding): Likewise.
14099         * genrecog.c (validate_pattern): Likewise.
14100         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14101         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14102         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14103         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14104         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14105         * rtlanal.c (subreg_get_info): Likewise.
14106         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14107         (vect_grouped_load_supported): Likewise.
14108         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14109         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14110         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14111         (simplify_const_unary_operation, simplify_binary_operation_1)
14112         (simplify_const_binary_operation, simplify_ternary_operation)
14113         (test_vector_ops_duplicate, test_vector_ops): Likewise.
14114         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14115         instead of CONST_VECTOR_NUNITS.
14116         * varasm.c (output_constant_pool_2): Likewise.
14117         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14118         explicit-encoded elements in the XVEC for variable-length vectors.
14119
14120 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14121
14122         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14123
14124 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14125             Alan Hayward  <alan.hayward@arm.com>
14126             David Sherwood  <david.sherwood@arm.com>
14127
14128         * coretypes.h (fixed_size_mode): Declare.
14129         (fixed_size_mode_pod): New typedef.
14130         * builtins.h (target_builtins::x_apply_args_mode)
14131         (target_builtins::x_apply_result_mode): Change type to
14132         fixed_size_mode_pod.
14133         * builtins.c (apply_args_size, apply_result_size, result_vector)
14134         (expand_builtin_apply_args_1, expand_builtin_apply)
14135         (expand_builtin_return): Update accordingly.
14136
14137 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14138
14139         * cse.c (hash_rtx_cb): Hash only the encoded elements.
14140         * cselib.c (cselib_hash_rtx): Likewise.
14141         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14142         CONST_VECTOR encoding.
14143
14144 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14145             Jeff Law  <law@redhat.com>
14146
14147         PR target/83641
14148         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14149         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14150         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14151         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14152
14153         PR target/83641
14154         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14155         explicitly probe *sp in a noreturn function if there were any callee
14156         register saves or frame pointer is needed.
14157
14158 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14159
14160         PR debug/83621
14161         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14162         BLKmode for ternary, binary or unary expressions.
14163
14164         PR debug/83645
14165         * var-tracking.c (delete_vta_debug_insn): New inline function.
14166         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14167         insns from get_insns () to NULL instead of each bb separately.
14168         Use delete_vta_debug_insn.  No longer static.
14169         (vt_debug_insns_local, variable_tracking_main_1): Adjust
14170         delete_vta_debug_insns callers.
14171         * rtl.h (delete_vta_debug_insns): Declare.
14172         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14173         instead of variable_tracking_main.
14174
14175 2018-01-03  Martin Sebor  <msebor@redhat.com>
14176
14177         PR tree-optimization/83603
14178         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14179         arguments past the endof the argument list in functions declared
14180         without a prototype.
14181         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14182         Avoid checking when arguments are null.
14183
14184 2018-01-03  Martin Sebor  <msebor@redhat.com>
14185
14186         PR c/83559
14187         * doc/extend.texi (attribute const): Fix a typo.
14188         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14189         issuing -Wsuggest-attribute for void functions.
14190
14191 2018-01-03  Martin Sebor  <msebor@redhat.com>
14192
14193         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14194         offset_int::from instead of wide_int::to_shwi.
14195         (maybe_diag_overlap): Remove assertion.
14196         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14197         * gimple-ssa-sprintf.c (format_directive): Same.
14198         (parse_directive): Same.
14199         (sprintf_dom_walker::compute_format_length): Same.
14200         (try_substitute_return_value): Same.
14201
14202 2018-01-03  Jeff Law  <law@redhat.com>
14203
14204         PR middle-end/83654
14205         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14206         non-constant residual for zero at runtime and avoid probing in
14207         that case.  Reorganize code for trailing problem to mirror handling
14208         of the residual.
14209
14210 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14211
14212         PR tree-optimization/83501
14213         * tree-ssa-strlen.c (get_string_cst): New.
14214         (handle_char_store): Call get_string_cst.
14215
14216 2018-01-03  Martin Liska  <mliska@suse.cz>
14217
14218         PR tree-optimization/83593
14219         * tree-ssa-strlen.c: Include tree-cfg.h.
14220         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14221         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14222         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14223         to false.
14224         (strlen_dom_walker::before_dom_children): Call
14225         gimple_purge_dead_eh_edges. Dump tranformation with details
14226         dump flags.
14227         (strlen_dom_walker::before_dom_children): Update call by adding
14228         new argument cleanup_eh.
14229         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14230
14231 2018-01-03  Martin Liska  <mliska@suse.cz>
14232
14233         PR ipa/83549
14234         * cif-code.def (VARIADIC_THUNK): New enum value.
14235         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14236         thunks.
14237
14238 2018-01-03  Jan Beulich  <jbeulich@suse.com>
14239
14240         * sse.md (mov<mode>_internal): Tighten condition for when to use
14241         vmovdqu<ssescalarsize> for TI and OI modes.
14242
14243 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14244
14245         Update copyright years.
14246
14247 2018-01-03  Martin Liska  <mliska@suse.cz>
14248
14249         PR ipa/83594
14250         * ipa-visibility.c (function_and_variable_visibility): Skip
14251         functions with noipa attribure.
14252
14253 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14254
14255         * gcc.c (process_command): Update copyright notice dates.
14256         * gcov-dump.c (print_version): Ditto.
14257         * gcov.c (print_version): Ditto.
14258         * gcov-tool.c (print_version): Ditto.
14259         * gengtype.c (create_file): Ditto.
14260         * doc/cpp.texi: Bump @copying's copyright year.
14261         * doc/cppinternals.texi: Ditto.
14262         * doc/gcc.texi: Ditto.
14263         * doc/gccint.texi: Ditto.
14264         * doc/gcov.texi: Ditto.
14265         * doc/install.texi: Ditto.
14266         * doc/invoke.texi: Ditto.
14267
14268 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14269
14270         * vector-builder.h (vector_builder::m_full_nelts): Change from
14271         unsigned int to poly_uint64.
14272         (vector_builder::full_nelts): Update prototype accordingly.
14273         (vector_builder::new_vector): Likewise.
14274         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14275         (vector_builder::operator ==): Likewise.
14276         (vector_builder::finalize): Likewise.
14277         * int-vector-builder.h (int_vector_builder::int_vector_builder):
14278         Take the number of elements as a poly_uint64 rather than an
14279         unsigned int.
14280         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14281         from unsigned int to poly_uint64.
14282         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14283         (vec_perm_indices::new_vector): Likewise.
14284         (vec_perm_indices::length): Likewise.
14285         (vec_perm_indices::nelts_per_input): Likewise.
14286         (vec_perm_indices::input_nelts): Likewise.
14287         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14288         number of elements per input as a poly_uint64 rather than an
14289         unsigned int.  Use the original encoding for variable-length
14290         vectors, rather than clamping each individual element.
14291         For the second and subsequent elements in each pattern,
14292         clamp the step and base before clamping their sum.
14293         (vec_perm_indices::series_p): Handle polynomial element counts.
14294         (vec_perm_indices::all_in_range_p): Likewise.
14295         (vec_perm_indices_to_tree): Likewise.
14296         (vec_perm_indices_to_rtx): Likewise.
14297         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14298         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14299         (tree_vector_builder::new_binary_operation): Handle polynomial
14300         element counts.  Return false if we need to know the number
14301         of elements at compile time.
14302         * fold-const.c (fold_vec_perm): Punt if the number of elements
14303         isn't known at compile time.
14304
14305 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14306
14307         * vec-perm-indices.h (vec_perm_builder): Change element type
14308         from HOST_WIDE_INT to poly_int64.
14309         (vec_perm_indices::element_type): Update accordingly.
14310         (vec_perm_indices::clamp): Handle polynomial element_types.
14311         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14312         (vec_perm_indices::all_in_range_p): Likewise.
14313         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14314         than shwi trees.
14315         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14316         polynomial vec_perm_indices element types.
14317         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14318         * fold-const.c (fold_vec_perm): Likewise.
14319         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14320         * tree-vect-generic.c (lower_vec_perm): Likewise.
14321         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14322         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14323         element type to HOST_WIDE_INT.
14324
14325 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14326             Alan Hayward  <alan.hayward@arm.com>
14327             David Sherwood  <david.sherwood@arm.com>
14328
14329         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14330         rather than an int.  Use plus_constant.
14331         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14332         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14333
14334 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14335             Alan Hayward  <alan.hayward@arm.com>
14336             David Sherwood  <david.sherwood@arm.com>
14337
14338         * calls.c (emit_call_1, expand_call): Change struct_value_size from
14339         a HOST_WIDE_INT to a poly_int64.
14340
14341 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14342             Alan Hayward  <alan.hayward@arm.com>
14343             David Sherwood  <david.sherwood@arm.com>
14344
14345         * calls.c (load_register_parameters): Cope with polynomial
14346         mode sizes.  Require a constant size for BLKmode parameters
14347         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
14348         forces a parameter to be padded at the lsb end in order to
14349         fill a complete number of words, require the parameter size
14350         to be ordered wrt UNITS_PER_WORD.
14351
14352 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14353             Alan Hayward  <alan.hayward@arm.com>
14354             David Sherwood  <david.sherwood@arm.com>
14355
14356         * reload1.c (spill_stack_slot_width): Change element type
14357         from unsigned int to poly_uint64_pod.
14358         (alter_reg): Treat mode sizes as polynomial.
14359
14360 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14361             Alan Hayward  <alan.hayward@arm.com>
14362             David Sherwood  <david.sherwood@arm.com>
14363
14364         * reload.c (complex_word_subreg_p): New function.
14365         (reload_inner_reg_of_subreg, push_reload): Use it.
14366
14367 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14368             Alan Hayward  <alan.hayward@arm.com>
14369             David Sherwood  <david.sherwood@arm.com>
14370
14371         * lra-constraints.c (process_alt_operands): Reject matched
14372         operands whose sizes aren't ordered.
14373         (match_reload): Refer to this check here.
14374
14375 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14376             Alan Hayward  <alan.hayward@arm.com>
14377             David Sherwood  <david.sherwood@arm.com>
14378
14379         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14380         that the mode size is in the set {1, 2, 4, 8, 16}.
14381
14382 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14383             Alan Hayward  <alan.hayward@arm.com>
14384             David Sherwood  <david.sherwood@arm.com>
14385
14386         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14387         Use plus_constant instead of gen_rtx_PLUS.
14388
14389 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14390             Alan Hayward  <alan.hayward@arm.com>
14391             David Sherwood  <david.sherwood@arm.com>
14392
14393         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14394         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14395         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14396         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14397         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14398         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14399         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14400         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14401         * config/i386/i386.c (ix86_push_rounding): ...this new function.
14402         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14403         a poly_int64.
14404         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14405         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14406         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14407         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14408         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14409         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14410         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14411         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14412         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14413         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14414         function.
14415         * expr.c (emit_move_resolve_push): Treat the input and result
14416         of PUSH_ROUNDING as a poly_int64.
14417         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14418         (emit_push_insn): Likewise.
14419         * lra-eliminations.c (mark_not_eliminable): Likewise.
14420         * recog.c (push_operand): Likewise.
14421         * reload1.c (elimination_effects): Likewise.
14422         * rtlanal.c (nonzero_bits1): Likewise.
14423         * calls.c (store_one_arg): Likewise.  Require the padding to be
14424         known at compile time.
14425
14426 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14427             Alan Hayward  <alan.hayward@arm.com>
14428             David Sherwood  <david.sherwood@arm.com>
14429
14430         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14431         Use plus_constant instead of gen_rtx_PLUS.
14432
14433 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14434             Alan Hayward  <alan.hayward@arm.com>
14435             David Sherwood  <david.sherwood@arm.com>
14436
14437         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14438         rather than an int.
14439
14440 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14441             Alan Hayward  <alan.hayward@arm.com>
14442             David Sherwood  <david.sherwood@arm.com>
14443
14444         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14445         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14446         via stack temporaries.  Treat the mode size as polynomial too.
14447
14448 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14449             Alan Hayward  <alan.hayward@arm.com>
14450             David Sherwood  <david.sherwood@arm.com>
14451
14452         * expr.c (expand_expr_real_2): When handling conversions involving
14453         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14454         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
14455         as a poly_uint64 too.
14456
14457 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14458             Alan Hayward  <alan.hayward@arm.com>
14459             David Sherwood  <david.sherwood@arm.com>
14460
14461         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14462
14463 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14464             Alan Hayward  <alan.hayward@arm.com>
14465             David Sherwood  <david.sherwood@arm.com>
14466
14467         * combine.c (can_change_dest_mode): Handle polynomial
14468         REGMODE_NATURAL_SIZE.
14469         * expmed.c (store_bit_field_1): Likewise.
14470         * expr.c (store_constructor): Likewise.
14471         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14472         and polynomial REGMODE_NATURAL_SIZE.
14473         (gen_lowpart_common): Likewise.
14474         * reginfo.c (record_subregs_of_mode): Likewise.
14475         * rtlanal.c (read_modify_subreg_p): Likewise.
14476
14477 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14478             Alan Hayward  <alan.hayward@arm.com>
14479             David Sherwood  <david.sherwood@arm.com>
14480
14481         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14482         numbers of elements.
14483
14484 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14485             Alan Hayward  <alan.hayward@arm.com>
14486             David Sherwood  <david.sherwood@arm.com>
14487
14488         * match.pd: Cope with polynomial numbers of vector elements.
14489
14490 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14491             Alan Hayward  <alan.hayward@arm.com>
14492             David Sherwood  <david.sherwood@arm.com>
14493
14494         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14495         in a POINTER_PLUS_EXPR.
14496
14497 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14498             Alan Hayward  <alan.hayward@arm.com>
14499             David Sherwood  <david.sherwood@arm.com>
14500
14501         * omp-simd-clone.c (simd_clone_subparts): New function.
14502         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14503         (ipa_simd_modify_function_body): Likewise.
14504
14505 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14506             Alan Hayward  <alan.hayward@arm.com>
14507             David Sherwood  <david.sherwood@arm.com>
14508
14509         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14510         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14511         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14512         (expand_vector_condition, vector_element): Likewise.
14513         (subparts_gt): New function.
14514         (get_compute_type): Use subparts_gt.
14515         (count_type_subparts): Delete.
14516         (expand_vector_operations_1): Use subparts_gt instead of
14517         count_type_subparts.
14518
14519 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14520             Alan Hayward  <alan.hayward@arm.com>
14521             David Sherwood  <david.sherwood@arm.com>
14522
14523         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14524         (vect_compile_time_alias): ...this new function.  Do the calculation
14525         on poly_ints rather than trees.
14526         (vect_prune_runtime_alias_test_list): Update call accordingly.
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         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14533         numbers of units.
14534         (vect_schedule_slp_instance): Likewise.
14535
14536 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14537             Alan Hayward  <alan.hayward@arm.com>
14538             David Sherwood  <david.sherwood@arm.com>
14539
14540         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14541         constant and extern definitions for variable-length vectors.
14542         (vect_get_constant_vectors): Note that the number of units
14543         is known to be constant.
14544
14545 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14546             Alan Hayward  <alan.hayward@arm.com>
14547             David Sherwood  <david.sherwood@arm.com>
14548
14549         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14550         of units as polynomial.  Choose between WIDE and NARROW based
14551         on multiple_p.
14552
14553 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14554             Alan Hayward  <alan.hayward@arm.com>
14555             David Sherwood  <david.sherwood@arm.com>
14556
14557         * tree-vect-stmts.c (simd_clone_subparts): New function.
14558         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14559
14560 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14561             Alan Hayward  <alan.hayward@arm.com>
14562             David Sherwood  <david.sherwood@arm.com>
14563
14564         * tree-vect-stmts.c (vectorizable_call): Treat the number of
14565         vectors as polynomial.  Use build_index_vector for
14566         IFN_GOMP_SIMD_LANE.
14567
14568 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14569             Alan Hayward  <alan.hayward@arm.com>
14570             David Sherwood  <david.sherwood@arm.com>
14571
14572         * tree-vect-stmts.c (get_load_store_type): Treat the number of
14573         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14574         for variable-length vectors.
14575         (vectorizable_mask_load_store): Treat the number of units as
14576         polynomial, asserting that it is constant if the condition has
14577         already been enforced.
14578         (vectorizable_store, vectorizable_load): Likewise.
14579
14580 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14581             Alan Hayward  <alan.hayward@arm.com>
14582             David Sherwood  <david.sherwood@arm.com>
14583
14584         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14585         of units as polynomial.  Punt if we can't tell at compile time
14586         which vector contains the final result.
14587
14588 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14589             Alan Hayward  <alan.hayward@arm.com>
14590             David Sherwood  <david.sherwood@arm.com>
14591
14592         * tree-vect-loop.c (vectorizable_induction): Treat the number
14593         of units as polynomial.  Punt on SLP inductions.  Use an integer
14594         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
14595         cast of such a series for variable-length floating-point
14596         reductions.
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.h (build_index_vector): Declare.
14603         * tree.c (build_index_vector): New function.
14604         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14605         of units as polynomial, forcibly converting it to a constant if
14606         vectorizable_reduction has already enforced the condition.
14607         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
14608         to create a {1,2,3,...} vector.
14609         (vectorizable_reduction): Treat the number of units as polynomial.
14610         Choose vectype_in based on the largest scalar element size rather
14611         than the smallest number of units.  Enforce the restrictions
14612         relied on above.
14613
14614 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14615             Alan Hayward  <alan.hayward@arm.com>
14616             David Sherwood  <david.sherwood@arm.com>
14617
14618         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14619         number of units as polynomial.
14620
14621 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14622             Alan Hayward  <alan.hayward@arm.com>
14623             David Sherwood  <david.sherwood@arm.com>
14624
14625         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14626         * target.def (autovectorize_vector_sizes): Return the vector sizes
14627         by pointer, using vector_sizes rather than a bitmask.
14628         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14629         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14630         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14631         Likewise.
14632         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14633         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14634         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14635         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14636         * omp-general.c (omp_max_vf): Likewise.
14637         * omp-low.c (omp_clause_aligned_alignment): Likewise.
14638         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14639         * tree-vect-loop.c (vect_analyze_loop): Likewise.
14640         * tree-vect-slp.c (vect_slp_bb): Likewise.
14641         * doc/tm.texi: Regenerate.
14642         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14643         to a poly_uint64.
14644         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14645         the vector size as a poly_uint64 rather than an unsigned int.
14646         (current_vector_size): Change from an unsigned int to a poly_uint64.
14647         (get_vectype_for_scalar_type): Update accordingly.
14648         * tree.h (build_truth_vector_type): Take the size and number of
14649         units as a poly_uint64 rather than an unsigned int.
14650         (build_vector_type): Add a temporary overload that takes
14651         the number of units as a poly_uint64 rather than an unsigned int.
14652         * tree.c (make_vector_type): Likewise.
14653         (build_truth_vector_type): Take the number of units as a poly_uint64
14654         rather than an unsigned int.
14655
14656 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14657             Alan Hayward  <alan.hayward@arm.com>
14658             David Sherwood  <david.sherwood@arm.com>
14659
14660         * target.def (get_mask_mode): Take the number of units and length
14661         as poly_uint64s rather than unsigned ints.
14662         * targhooks.h (default_get_mask_mode): Update accordingly.
14663         * targhooks.c (default_get_mask_mode): Likewise.
14664         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14665         * doc/tm.texi: Regenerate.
14666
14667 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14668             Alan Hayward  <alan.hayward@arm.com>
14669             David Sherwood  <david.sherwood@arm.com>
14670
14671         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14672         * omp-general.c (omp_max_vf): Likewise.
14673         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14674         (expand_omp_simd): Handle polynomial safelen.
14675         * omp-low.c (omplow_simd_context): Add a default constructor.
14676         (omplow_simd_context::max_vf): Change from int to poly_uint64.
14677         (lower_rec_simd_input_clauses): Update accordingly.
14678         (lower_rec_input_clauses): Likewise.
14679
14680 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14681             Alan Hayward  <alan.hayward@arm.com>
14682             David Sherwood  <david.sherwood@arm.com>
14683
14684         * tree-vectorizer.h (vect_nunits_for_cost): New function.
14685         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14686         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14687         (vect_analyze_slp_cost): Likewise.
14688         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14689         (vect_model_load_cost): Likewise.
14690
14691 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14692             Alan Hayward  <alan.hayward@arm.com>
14693             David Sherwood  <david.sherwood@arm.com>
14694
14695         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14696         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14697         from an unsigned int * to a poly_uint64_pod *.
14698         (calculate_unrolling_factor): New function.
14699         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
14700
14701 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14702             Alan Hayward  <alan.hayward@arm.com>
14703             David Sherwood  <david.sherwood@arm.com>
14704
14705         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14706         from an unsigned int to a poly_uint64.
14707         (_loop_vec_info::slp_unrolling_factor): Likewise.
14708         (_loop_vec_info::vectorization_factor): Change from an int
14709         to a poly_uint64.
14710         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14711         (vect_get_num_vectors): New function.
14712         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14713         (vect_get_num_copies): Use vect_get_num_vectors.
14714         (vect_analyze_data_ref_dependences): Change max_vf from an int *
14715         to an unsigned int *.
14716         (vect_analyze_data_refs): Change min_vf from an int * to a
14717         poly_uint64 *.
14718         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14719         than an unsigned HOST_WIDE_INT.
14720         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14721         (vect_analyze_data_ref_dependence): Change max_vf from an int *
14722         to an unsigned int *.
14723         (vect_analyze_data_ref_dependences): Likewise.
14724         (vect_compute_data_ref_alignment): Handle polynomial vf.
14725         (vect_enhance_data_refs_alignment): Likewise.
14726         (vect_prune_runtime_alias_test_list): Likewise.
14727         (vect_shift_permute_load_chain): Likewise.
14728         (vect_supportable_dr_alignment): Likewise.
14729         (dependence_distance_ge_vf): Take the vectorization factor as a
14730         poly_uint64 rather than an unsigned HOST_WIDE_INT.
14731         (vect_analyze_data_refs): Change min_vf from an int * to a
14732         poly_uint64 *.
14733         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14734         vfm1 as a poly_uint64 rather than an int.  Make the same change
14735         for the returned bound_scalar.
14736         (vect_gen_vector_loop_niters): Handle polynomial vf.
14737         (vect_do_peeling): Likewise.  Update call to
14738         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14739         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14740         be constant.
14741         * tree-vect-loop.c (vect_determine_vectorization_factor)
14742         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14743         (vect_get_known_peeling_cost): Likewise.
14744         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14745         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14746         (vect_transform_loop): Likewise.  Use the lowest possible VF when
14747         updating the upper bounds of the loop.
14748         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
14749         rather than an int.
14750         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14751         polynomial unroll factors.
14752         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14753         (vect_make_slp_decision): Likewise.
14754         (vect_supported_load_permutation_p): Likewise, and polynomial
14755         vf too.
14756         (vect_analyze_slp_cost): Handle polynomial vf.
14757         (vect_slp_analyze_node_operations): Likewise.
14758         (vect_slp_analyze_bb_1): Likewise.
14759         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14760         than an unsigned HOST_WIDE_INT.
14761         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14762         (vectorizable_load): Handle polynomial vf.
14763         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14764         a poly_uint64.
14765         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14766
14767 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14768             Alan Hayward  <alan.hayward@arm.com>
14769             David Sherwood  <david.sherwood@arm.com>
14770
14771         * match.pd: Handle bit operations involving three constants
14772         and try to fold one pair.
14773
14774 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14775
14776         * tree-vect-loop-manip.c: Include gimple-fold.h.
14777         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14778         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
14779         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14780         Add a path that uses a step of VF instead of 1, but disable it
14781         for now.
14782         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14783         and niters_no_overflow parameters.  Update calls to
14784         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14785         Create a new SSA name if the latter choses to use a ste other
14786         than zero, and return it via niters_vector_mult_vf_var.
14787         * tree-vect-loop.c (vect_transform_loop): Update calls to
14788         vect_do_peeling, vect_gen_vector_loop_niters and
14789         slpeel_make_loop_iterate_ntimes.
14790         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14791         (vect_gen_vector_loop_niters): Update declarations after above changes.
14792
14793 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
14794
14795         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14796         128-bit round to integer instructions.
14797         (ceil<mode>2): Likewise.
14798         (btrunc<mode>2): Likewise.
14799         (round<mode>2): Likewise.
14800
14801 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14802
14803         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14804         unaligned VSX load/store on P8/P9.
14805         (expand_block_clear): Allow the use of unaligned VSX
14806         load/store on P8/P9.
14807
14808 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14809
14810         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14811         New function.
14812         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14813         swap associated with both a load and a store.
14814
14815 2018-01-02  Andrew Waterman  <andrew@sifive.com>
14816
14817         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14818         * config/riscv/riscv.md (clear_cache): Use it.
14819
14820 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
14821
14822         * web.c: Remove out-of-date comment.
14823
14824 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14825
14826         * expr.c (fixup_args_size_notes): Check that any existing
14827         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14828         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14829         (emit_single_push_insn): ...here.
14830
14831 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14832
14833         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14834         (const_vector_encoded_nelts): New function.
14835         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14836         (const_vector_int_elt, const_vector_elt): Declare.
14837         * emit-rtl.c (const_vector_int_elt_1): New function.
14838         (const_vector_elt): Likewise.
14839         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14840         of CONST_VECTOR_ELT.
14841
14842 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14843
14844         * expr.c: Include rtx-vector-builder.h.
14845         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14846         directly on the tree encoding.
14847         (const_vector_from_tree): Likewise.
14848         * optabs.c: Include rtx-vector-builder.h.
14849         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14850         sequence of "u" values.
14851         * vec-perm-indices.c: Include rtx-vector-builder.h.
14852         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14853         directly on the vec_perm_indices encoding.
14854
14855 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14856
14857         * doc/rtl.texi (const_vector): Describe new encoding scheme.
14858         * Makefile.in (OBJS): Add rtx-vector-builder.o.
14859         * rtx-vector-builder.h: New file.
14860         * rtx-vector-builder.c: Likewise.
14861         * rtl.h (rtx_def::u2): Add a const_vector field.
14862         (CONST_VECTOR_NPATTERNS): New macro.
14863         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14864         (CONST_VECTOR_DUPLICATE_P): Likewise.
14865         (CONST_VECTOR_STEPPED_P): Likewise.
14866         (CONST_VECTOR_ENCODED_ELT): Likewise.
14867         (const_vec_duplicate_p): Check for a duplicated vector encoding.
14868         (unwrap_const_vec_duplicate): Likewise.
14869         (const_vec_series_p): Check for a non-duplicated vector encoding.
14870         Say that the function only returns true for integer vectors.
14871         * emit-rtl.c: Include rtx-vector-builder.h.
14872         (gen_const_vec_duplicate_1): Delete.
14873         (gen_const_vector): Call gen_const_vec_duplicate instead of
14874         gen_const_vec_duplicate_1.
14875         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14876         (gen_const_vec_duplicate): Use rtx_vector_builder.
14877         (gen_const_vec_series): Likewise.
14878         (gen_rtx_CONST_VECTOR): Likewise.
14879         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14880         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14881         Build a new vector rather than modifying a CONST_VECTOR in-place.
14882         (handle_special_swappables): Update call accordingly.
14883         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14884         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14885         Build a new vector rather than modifying a CONST_VECTOR in-place.
14886         (handle_special_swappables): Update call accordingly.
14887
14888 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14889
14890         * simplify-rtx.c (simplify_const_binary_operation): Use
14891         CONST_VECTOR_ELT instead of XVECEXP.
14892
14893 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14894
14895         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14896         the selector elements to be different from the data elements
14897         if the selector is a VECTOR_CST.
14898         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14899         ssizetype for the selector.
14900
14901 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14902
14903         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14904         before testing each element individually.
14905         * tree-vect-generic.c (lower_vec_perm): Likewise.
14906
14907 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14908
14909         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14910         * selftest-run-tests.c (selftest::run_tests): Call it.
14911         * vector-builder.h (vector_builder::operator ==): New function.
14912         (vector_builder::operator !=): Likewise.
14913         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14914         (vec_perm_indices::all_from_input_p): New function.
14915         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14916         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14917         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14918         instead of reading the VECTOR_CST directly.  Detect whether both
14919         vector inputs are the same before constructing the vec_perm_indices,
14920         and update the number of inputs argument accordingly.  Use the
14921         utility functions added above.  Only construct sel2 if we need to.
14922
14923 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14924
14925         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14926         the broadcast of the low byte.
14927         (expand_mult_highpart): Use an explicit encoding for the permutes.
14928         * optabs-query.c (can_mult_highpart_p): Likewise.
14929         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14930         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14931         (vectorizable_bswap): Likewise.
14932         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14933         explicit encoding for the power-of-2 permutes.
14934         (vect_permute_store_chain): Likewise.
14935         (vect_grouped_load_supported): Likewise.
14936         (vect_permute_load_chain): Likewise.
14937
14938 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14939
14940         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14941         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14942         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14943         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14944         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14945         (vect_gen_perm_mask_any): Likewise.
14946
14947 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14948
14949         * int-vector-builder.h: New file.
14950         * vec-perm-indices.h: Include int-vector-builder.h.
14951         (vec_perm_indices): Redefine as an int_vector_builder.
14952         (auto_vec_perm_indices): Delete.
14953         (vec_perm_builder): Redefine as a stand-alone class.
14954         (vec_perm_indices::vec_perm_indices): New function.
14955         (vec_perm_indices::clamp): Likewise.
14956         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14957         (vec_perm_indices::new_vector): New function.
14958         (vec_perm_indices::new_expanded_vector): Update for new
14959         vec_perm_indices class.
14960         (vec_perm_indices::rotate_inputs): New function.
14961         (vec_perm_indices::all_in_range_p): Operate directly on the
14962         encoded form, without computing elided elements.
14963         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14964         encoding.  Update for new vec_perm_indices class.
14965         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14966         the given vec_perm_builder.
14967         (expand_vec_perm_var): Update vec_perm_builder constructor.
14968         (expand_mult_highpart): Use vec_perm_builder instead of
14969         auto_vec_perm_indices.
14970         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14971         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
14972         or double series encoding as appropriate.
14973         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14974         vec_perm_indices instead of auto_vec_perm_indices.
14975         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14976         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14977         (vect_permute_store_chain): Likewise.
14978         (vect_grouped_load_supported): Likewise.
14979         (vect_permute_load_chain): Likewise.
14980         (vect_shift_permute_load_chain): Likewise.
14981         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14982         (vect_transform_slp_perm_load): Likewise.
14983         (vect_schedule_slp_instance): Likewise.
14984         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14985         (vectorizable_mask_load_store): Likewise.
14986         (vectorizable_bswap): Likewise.
14987         (vectorizable_store): Likewise.
14988         (vectorizable_load): Likewise.
14989         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14990         vec_perm_indices instead of auto_vec_perm_indices.  Use
14991         tree_to_vec_perm_builder to read the vector from a tree.
14992         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14993         vec_perm_builder instead of a vec_perm_indices.
14994         (have_whole_vector_shift): Use vec_perm_builder and
14995         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
14996         truncation to calc_vec_perm_mask_for_shift.
14997         (vect_create_epilog_for_reduction): Likewise.
14998         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14999         from auto_vec_perm_indices to vec_perm_indices.
15000         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15001         instead of changing individual elements.
15002         (aarch64_vectorize_vec_perm_const): Use new_vector to install
15003         the vector in d.perm.
15004         * config/arm/arm.c (expand_vec_perm_d::perm): Change
15005         from auto_vec_perm_indices to vec_perm_indices.
15006         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15007         instead of changing individual elements.
15008         (arm_vectorize_vec_perm_const): Use new_vector to install
15009         the vector in d.perm.
15010         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15011         Update vec_perm_builder constructor.
15012         (rs6000_expand_interleave): Likewise.
15013         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15014         (rs6000_expand_interleave): Likewise.
15015
15016 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15017
15018         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15019         to qimode could truncate the indices.
15020         * optabs.c (expand_vec_perm_var): Likewise.
15021
15022 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15023
15024         * Makefile.in (OBJS): Add vec-perm-indices.o.
15025         * vec-perm-indices.h: New file.
15026         * vec-perm-indices.c: Likewise.
15027         * target.h (vec_perm_indices): Replace with a forward class
15028         declaration.
15029         (auto_vec_perm_indices): Move to vec-perm-indices.h.
15030         * optabs.h: Include vec-perm-indices.h.
15031         (expand_vec_perm): Delete.
15032         (selector_fits_mode_p, expand_vec_perm_var): Declare.
15033         (expand_vec_perm_const): Declare.
15034         * target.def (vec_perm_const_ok): Replace with...
15035         (vec_perm_const): ...this new hook.
15036         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15037         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15038         * doc/tm.texi: Regenerate.
15039         * optabs.def (vec_perm_const): Delete.
15040         * doc/md.texi (vec_perm_const): Likewise.
15041         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15042         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15043         expand_vec_perm for constant permutation vectors.  Assert that
15044         the mode of variable permutation vectors is the integer equivalent
15045         of the mode that is being permuted.
15046         * optabs-query.h (selector_fits_mode_p): Declare.
15047         * optabs-query.c: Include vec-perm-indices.h.
15048         (selector_fits_mode_p): New function.
15049         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15050         is defined, instead of checking whether the vec_perm_const_optab
15051         exists.  Use targetm.vectorize.vec_perm_const instead of
15052         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
15053         fit in the vector mode before using a variable permute.
15054         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15055         vec_perm_indices instead of an rtx.
15056         (expand_vec_perm): Replace with...
15057         (expand_vec_perm_const): ...this new function.  Take the selector
15058         as a vec_perm_indices rather than an rtx.  Also take the mode of
15059         the selector.  Update call to shift_amt_for_vec_perm_mask.
15060         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15061         Use vec_perm_indices::new_expanded_vector to expand the original
15062         selector into bytes.  Check whether the indices fit in the vector
15063         mode before using a variable permute.
15064         (expand_vec_perm_var): Make global.
15065         (expand_mult_highpart): Use expand_vec_perm_const.
15066         * fold-const.c: Includes vec-perm-indices.h.
15067         * tree-ssa-forwprop.c: Likewise.
15068         * tree-vect-data-refs.c: Likewise.
15069         * tree-vect-generic.c: Likewise.
15070         * tree-vect-loop.c: Likewise.
15071         * tree-vect-slp.c: Likewise.
15072         * tree-vect-stmts.c: Likewise.
15073         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15074         Delete.
15075         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15076         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15077         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15078         (aarch64_vectorize_vec_perm_const): ...this new function.
15079         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15080         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15081         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15082         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15083         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15084         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15085         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15086         into...
15087         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
15088         check for NEON modes.
15089         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15090         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15091         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15092         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15093         into...
15094         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
15095         the old VEC_PERM_CONST conditions.
15096         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15097         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15098         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15099         (ia64_vectorize_vec_perm_const_ok): Merge into...
15100         (ia64_vectorize_vec_perm_const): ...this new function.
15101         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15102         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15103         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15104         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15105         * config/mips/mips.c (mips_expand_vec_perm_const)
15106         (mips_vectorize_vec_perm_const_ok): Merge into...
15107         (mips_vectorize_vec_perm_const): ...this new function.
15108         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15109         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15110         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15111         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15112         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15113         (rs6000_expand_vec_perm_const): Delete.
15114         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15115         Delete.
15116         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15117         (altivec_expand_vec_perm_const_le): Take each operand individually.
15118         Operate on constant selectors rather than rtxes.
15119         (altivec_expand_vec_perm_const): Likewise.  Update call to
15120         altivec_expand_vec_perm_const_le.
15121         (rs6000_expand_vec_perm_const): Delete.
15122         (rs6000_vectorize_vec_perm_const_ok): Delete.
15123         (rs6000_vectorize_vec_perm_const): New function.
15124         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15125         an element count and rtx array.
15126         (rs6000_expand_extract_even): Update call accordingly.
15127         (rs6000_expand_interleave): Likewise.
15128         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15129         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15130         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15131         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15132         (rs6000_expand_vec_perm_const): Delete.
15133         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15134         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15135         (altivec_expand_vec_perm_const_le): Take each operand individually.
15136         Operate on constant selectors rather than rtxes.
15137         (altivec_expand_vec_perm_const): Likewise.  Update call to
15138         altivec_expand_vec_perm_const_le.
15139         (rs6000_expand_vec_perm_const): Delete.
15140         (rs6000_vectorize_vec_perm_const_ok): Delete.
15141         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
15142         reference to the SPE evmerge intructions.
15143         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15144         an element count and rtx array.
15145         (rs6000_expand_extract_even): Update call accordingly.
15146         (rs6000_expand_interleave): Likewise.
15147         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15148         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15149         new function.
15150         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15151
15152 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15153
15154         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15155         vector mode and that that mode matches the mode of the data
15156         being permuted.
15157         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15158         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
15159         directly using expand_vec_perm_1 when forcing selectors into
15160         registers.
15161         (expand_vec_perm_var): New function, split out from expand_vec_perm.
15162
15163 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15164
15165         * optabs-query.h (can_vec_perm_p): Delete.
15166         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15167         * optabs-query.c (can_vec_perm_p): Split into...
15168         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15169         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15170         particular selector is valid.
15171         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15172         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15173         (vect_grouped_load_supported): Likewise.
15174         (vect_shift_permute_load_chain): Likewise.
15175         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15176         (vect_transform_slp_perm_load): Likewise.
15177         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15178         (vectorizable_bswap): Likewise.
15179         (vect_gen_perm_mask_checked): Likewise.
15180         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
15181         implementations of variable permutation vectors into account
15182         when deciding which selector to use.
15183         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15184         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15185         with a false third argument.
15186         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15187         to test whether the constant selector is valid and can_vec_perm_var_p
15188         to test whether a variable selector is valid.
15189
15190 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15191
15192         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15193         * optabs-query.c (can_vec_perm_p): Likewise.
15194         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15195         instead of vec_perm_indices.
15196         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15197         (vect_gen_perm_mask_checked): Likewise,
15198         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15199         (vect_gen_perm_mask_checked): Likewise,
15200
15201 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15202
15203         * optabs-query.h (qimode_for_vec_perm): Declare.
15204         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15205         (qimode_for_vec_perm): ...this new function.
15206         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15207
15208 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
15209
15210         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15211         does not have a conditional at the top.
15212
15213 2018-01-02  Richard Biener  <rguenther@suse.de>
15214
15215         * ipa-inline.c (big_speedup_p): Fix expression.
15216
15217 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
15218
15219         PR target/81616
15220         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15221         for generic 4->6.
15222
15223 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
15224
15225         PR target/81616
15226         Generic tuning.
15227         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15228         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15229         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15230         cond_taken_branch_cost 3->4.
15231
15232 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
15233
15234         PR tree-optimization/83581
15235         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15236         TODO_cleanup_cfg if any changes have been made.
15237
15238         PR middle-end/83608
15239         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15240         convert_modes if target mode has the right side, but different mode
15241         class.
15242
15243         PR middle-end/83609
15244         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15245         last argument when extracting from CONCAT.  If either from_real or
15246         from_imag is NULL, use expansion through memory.  If result is not
15247         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15248         the parts directly to inner mode, if even that fails, use expansion
15249         through memory.
15250
15251         PR middle-end/83623
15252         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15253         check for bswap in mode rather than HImode and use that in expand_unop
15254         too.
15255 \f
15256 Copyright (C) 2018 Free Software Foundation, Inc.
15257
15258 Copying and distribution of this file, with or without modification,
15259 are permitted in any medium without royalty provided the copyright
15260 notice and this notice are preserved.