gcc/ChangeLog

   1 2018-06-12  Martin Sebor  <msebor@redhat.com>
   2
   3         PR tree-optimization/85259
   4         * builtins.c (compute_objsize): Handle constant offsets.
   5         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
   6         true iff a warning has been issued.
   7         * gimple.h (gimple_nonartificial_location): New function.
   8         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
   9         gimple_nonartificial_location and handle -Wno-system-headers.
  10         (handle_builtin_stxncpy): Same.
  11
  12 2018-06-12  Martin Sebor  <msebor@redhat.com>
  13
  14         PR c/85931
  15         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
  16
  17 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
  18
  19         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
  20         BUILTIN_VEC_XST entries for pointer to double and long long.
  21
  22 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
  23
  24         PR target/85990
  25         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
  26         Update comments.
  27         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
  28         Likewise.
  29
  30 2018-06-12  Martin Liska  <mliska@suse.cz>
  31
  32         * doc/options.texi: Document IntegerRange.
  33
  34 2018-06-12  Martin Liska  <mliska@suse.cz>
  35
  36         * config/i386/i386.opt: Make MPX-related options as Deprecated.
  37         * opt-functions.awk: Handle Deprecated flag.
  38         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
  39         and report error.
  40         (read_cmdline_option): Report warning for a deprecated option.
  41         * opts.h (struct cl_option): Add new field cl_deprecated.
  42         (CL_ERR_DEPRECATED): New.
  43
  44 2018-06-12  Martin Liska  <mliska@suse.cz>
  45
  46         * doc/options.texi: Document Deprecated option flag.
  47
  48 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
  49
  50         * config/arc/arc-arch.h (arc_extras): New enum.
  51         (arc_cpu_t):Add field extra.
  52         (arc_cpu_types): Consider the extras.
  53         * config/arc/arc-cpus.def: Add extras info.
  54         * config/arc/arc-opts.h (processor_type): Consider extra field.
  55         * config/arc/arc.c (arc_override_options): Handle extra field.
  56
  57 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
  58
  59         * config/arc/arc-arch.h: Update ARC_OPTX macro.
  60         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
  61         field.
  62         * config/arc/arc.c (arc_init): Update pic warning.
  63         (irq_range): Update irq range parsing warnings.
  64         (arc_override_options): Update various warning messages.
  65         (arc_handle_aux_attribute): Likewise.
  66
  67 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
  68
  69         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
  70
  71 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
  72
  73         * doc/sourcebuild.texi: Document usage of line number 0 in verify
  74         compiler messages directives.
  75
  76 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
  77
  78         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
  79         * config/mips/mips-tables.opt: Regenerate.
  80         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
  81         mips64r6.
  82         * doc/invoke.texi: Document -march=i6500.
  83
  84 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
  85
  86         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
  87         (i6400_gpmul): Add cpu_unit.
  88         (i6400_gpdiv): Likewise.
  89         (i6400_msa_add_d): Update reservations.
  90         (i6400_msa_int_add) Likewise.
  91         (i6400_msa_short_logic3) Likewise.
  92         (i6400_msa_short_logic2) Likewise.
  93         (i6400_msa_short_logic) Likewise.
  94         (i6400_msa_move) Likewise.
  95         (i6400_msa_cmp) Likewise.
  96         (i6400_msa_short_float2) Likewise.
  97         (i6400_msa_div_d) Likewise.
  98         (i6400_msa_long_logic1) Likewise.
  99         (i6400_msa_long_logic2) Likewise.
 100         (i6400_msa_mult) Likewise.
 101         (i6400_msa_long_float2) Likewise.
 102         (i6400_msa_long_float4) Likewise.
 103         (i6400_msa_long_float5) Likewise.
 104         (i6400_msa_long_float8) Likewise.
 105         (i6400_fpu_fadd): Include frint type.
 106         (i6400_fpu_store): New define_insn_reservation.
 107         (i6400_fpu_load): Likewise.
 108         (i6400_fpu_move): Likewise.
 109         (i6400_fpu_fcmp): Likewise.
 110         (i6400_fpu_fmadd): Likewise.
 111         (i6400_int_mult): Include imul3nc type and update reservation.
 112         (i6400_int_div): Include idiv3 type and update reservation.
 113         (i6400_int_load): Update to check type not move_type.
 114         (i6400_int_store): Likewise.
 115         (i6400_int_prefetch): Set zero latency.
 116
 117 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
 118
 119         * gcc.c: Document new %@{...} sequence.
 120         (LINK_COMMAND_SPEC): Use it for the -L switches.
 121         (cpp_unique_options): Use it for the -I switches.
 122         (at_file_argbuf): New global variable.
 123         (in_at_file): Likewise.
 124         (alloc_args): Create at_file_argbuf.
 125         (clear_args): Truncate at_file_argbuf.
 126         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
 127         (open_at_file): New function.
 128         (close_at_file): Likewise.
 129         (create_at_file): Delete.
 130         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
 131         <'o'>: Likewise.
 132         <'@'>: New case.
 133         (validate_switches_from_spec): Deal with %@{...} sequence.
 134         (validate_switches): Likewise.
 135         (driver::finalize): Call clear_args.
 136
 137 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
 138
 139         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
 140
 141 2018-06-11  Martin Sebor  <msebor@redhat.com>
 142
 143         * doc/invoke.texi (-Wall): List -Wc++17-compat.
 144         (Wno-class-memaccess): Add @opindex.
 145         (Wno-templates, Wno-multiple-inheritance): Same.
 146         (Wno-virtual-inheritance, Wno-namespaces): Same.
 147         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
 148         (Wno-format-overflow, Wno-format-truncation): Same.
 149         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
 150         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
 151         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
 152         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
 153         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
 154         (Wno-misspelled-isr): Same.
 155
 156 2018-06-11  Martin Sebor  <msebor@redhat.com>
 157
 158         * PR tree-optimization/86083
 159         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
 160
 161 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
 162
 163         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
 164
 165 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
 166
 167         PR target/85755
 168         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
 169         on the correct operand.
 170         (*movdi_internal64): Ditto.
 171
 172 2018-06-11  Martin Liska  <mliska@suse.cz>
 173
 174         PR tree-optimization/86089
 175         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
 176
 177 2018-06-11  Julia Koval  <julia.koval@intel.com>
 178
 179         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
 180         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
 181         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
 182
 183 2018-06-11  Olivier Hainque  <hainque@adacore.com>
 184
 185         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
 186         for Ada with strict dwarf2.
 187
 188 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
 189
 190         PR target/85755
 191         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
 192         addresses.
 193
 194 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
 195
 196         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
 197
 198 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
 199
 200         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
 201         TARGET_ELF.
 202
 203 2018-06-08  Martin Liska  <mliska@suse.cz>
 204
 205         * tree-cfg.h (debug_function): Fix argument type to match
 206         implementation.
 207
 208 2018-06-08  Martin Liska  <mliska@suse.cz>
 209
 210         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
 211         Remove usage of MPX-related (and removed) fields.
 212         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
 213
 214 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
 215
 216         * cfg.c (debug): Use TDF_NONE rather than 0.
 217         * cfghooks.c (debug): Likewise.
 218         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
 219         (struct dump_option_value_info): Convert to...
 220         (struct kv_pair): ...this template type.
 221         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
 222         rather than 0.
 223         (optinfo_verbosity_options): Likewise.
 224         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
 225         OPTGROUP_NONE.
 226         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
 227         than int for "optgroup_flags" param.
 228         (dump_generic_expr_loc): Use dump_flags_t rather than int for
 229         "dump_kind" param.
 230         (dump_dec): Likewise.
 231         (dump_finish): Use TDF_NONE rather than 0.
 232         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
 233         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
 234         than 0.  Update for change to option_ptr.
 235         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
 236         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
 237         0.  Update for changes to optinfo_verbosity_options and
 238         optgroup_options.
 239         (opt_info_switch_p): Convert optgroup_flags from int to
 240         optgroup_flags_t.
 241         (dump_basic_block): Use dump_flags_t rather than int
 242         for "dump_kind" param.
 243         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
 244         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
 245         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
 246         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
 247         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
 248         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
 249         TDF_NONE): Convert from macros to...
 250         (enum dump_flag): ...this new enum.
 251         (dump_flags_t): Update to use enum.
 252         (operator|, operator&, operator~, operator|=, operator&=):
 253         Implement for dump_flags_t.
 254         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
 255         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
 256         Convert from macros to...
 257         (enum optgroup_flag): ...this new enum.
 258         (optgroup_flags_t): New typedef.
 259         (operator|, operator|=): Implement for optgroup_flags_t.
 260         (struct dump_file_info): Convert field "alt_flags" to
 261         dump_flags_t.  Convert field "optgroup_flags" to
 262         optgroup_flags_t.
 263         (dump_basic_block): Use dump_flags_t rather than int for param.
 264         (dump_generic_expr_loc): Likewise.
 265         (dump_dec): Likewise.
 266         (dump_register): Convert param "optgroup_flags" to
 267         optgroup_flags_t.
 268         (opt_info_enable_passes): Likewise.
 269         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
 270         than 0.
 271         * gimple-pretty-print.c (debug): Likewise.
 272         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
 273         (merged_store_group::apply_stores): Likewise.
 274         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
 275         * gimple.c (verify_gimple_pp): Likewise.
 276         * graphite-poly.c (print_pbb_body): Likewise.
 277         * passes.c (pass_manager::register_one_dump_file): Convert
 278         local "optgroup_flags" to optgroup_flags_t.
 279         * print-tree.c (print_node): Use TDF_NONE rather than 0.
 280         (debug): Likewise.
 281         (debug_body): Likewise.
 282         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
 283         to optgroup_flags_t.
 284         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
 285         than 0.
 286         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
 287         (convert_mult_to_fma): Likewise.
 288         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
 289         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
 290         * tree-vect-data-refs.c (dump_lower_bound): Convert param
 291         "dump_kind" to dump_flags_t.
 292
 293 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
 294
 295         * config/rs6000/rs6000.c (min, max): Delete.
 296
 297 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
 298
 299         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
 300         -mabi=spe and -mabi=no-spe.
 301
 302 2018-06-08  Martin Liska  <mliska@suse.cz>
 303
 304         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
 305         where we expect an existing summary.
 306
 307 2018-06-08  Martin Liska  <mliska@suse.cz>
 308
 309         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
 310         * ipa-inline.h (estimate_edge_growth): Likewise.
 311
 312 2018-06-08  Martin Liska  <mliska@suse.cz>
 313
 314         * cgraph.c (function_version_hasher::hash): Use
 315         cgraph_node::get_uid ().
 316         (function_version_hasher::equal):
 317         * cgraph.h (cgraph_node::get_uid): New method.
 318         * ipa-inline.c (update_caller_keys): Use
 319         cgraph_node::get_uid ().
 320         (update_callee_keys): Likewise.
 321         * ipa-utils.c (searchc): Likewise.
 322         (ipa_reduced_postorder): Likewise.
 323         * lto-cgraph.c (input_node): Likewise.
 324         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
 325         * symbol-summary.h (symtab_insertion): Likewise.
 326         (symtab_removal): Likewise.
 327         (symtab_duplication): Likewise.
 328         * tree-pretty-print.c (dump_function_header): Likewise.
 329         * tree-sra.c (convert_callers_for_node): Likewise.
 330
 331 2018-06-08  Martin Liska  <mliska@suse.cz>
 332
 333         * cgraph.c (symbol_table::create_edge): Always assign a new
 334         unique number.
 335         (symbol_table::free_edge): Do not recycle numbers.
 336         * cgraph.h (cgraph_edge::get): New method.
 337         * symbol-summary.h (symtab_removal): Use it.
 338         (symtab_duplication): Likewise.
 339         (call_summary::hashable_uid): Remove.
 340
 341 2018-06-08  Martin Liska  <mliska@suse.cz>
 342
 343         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
 344         (initialize_growth_caches): Remove.
 345         (free_growth_caches): Likewise.
 346         (do_estimate_edge_time): Use edge_growth_cache.
 347         (do_estimate_edge_size): Likewise.
 348         (do_estimate_edge_hints): Likewise.
 349         * ipa-inline.c (reset_edge_caches): Likewise.
 350         (recursive_inlining): Likewise.
 351         (inline_small_functions): Likewise.
 352         * ipa-inline.h (initialize_growth_caches): Remove.
 353         (estimate_edge_size): Likewise.
 354         (estimate_edge_time): Likewise.
 355         (estimate_edge_hints): Likewise.
 356         (reset_edge_growth_cache): Likewise.
 357         * symbol-summary.h (call_summary::remove): New method.
 358
 359 2018-06-08  Martin Liska  <mliska@suse.cz>
 360
 361         * ipa-cp.c (class edge_clone_summary): New summary.
 362         (grow_edge_clone_vectors): Remove.
 363         (ipcp_edge_duplication_hook): Remove.
 364         (class edge_clone_summary_t): New call_summary class.
 365         (ipcp_edge_removal_hook): Remove.
 366         (edge_clone_summary_t::duplicate): New function.
 367         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
 368         (create_specialized_node): Likewise.
 369         (ipcp_driver): Initialize edge_clone_summaries and do not
 370         register hooks.
 371
 372 2018-06-08  Martin Liska  <mliska@suse.cz>
 373
 374         * symbol-summary.h (get): New function.
 375         (call_summary::m_initialize_when_cloning): New class member.
 376
 377 2018-06-08  Martin Liska  <mliska@suse.cz>
 378
 379         * cgraph.c (cgraph_node::remove): Do not recycle uid.
 380         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
 381         (symbol_table::allocate_cgraph_symbol): Do not set uid.
 382         * passes.c (uid_hash_t): Record removed_nodes by their uids.
 383         (remove_cgraph_node_from_order): Use the removed_nodes set.
 384         (do_per_function_toporder): Likwise.
 385         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
 386         instead of summary_uid.
 387         (symtab_removal): Likewise.
 388         (symtab_duplication): Likewise.
 389
 390 2018-06-08  Martin Liska  <mliska@suse.cz>
 391
 392         * ipa-cp.c (ipcp_store_bits_results): Use
 393         ipcp_transformation_sum.
 394         (ipcp_store_vr_results): Likewise.
 395         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
 396         to ...
 397         (ipcp_transformation_initialize): ... this.
 398         (ipa_set_node_agg_value_chain):
 399         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
 400         (write_ipcp_transformation_info): Likewise.
 401         (read_ipcp_transformation_info): Likewise.
 402         (ipcp_update_bits): Likewise.
 403         (ipcp_update_vr): Likewise.
 404         (ipcp_transform_function): Likewise.
 405         * ipa-prop.h: Rename ipcp_transformation_summary to
 406         ipcp_transformation.
 407         (class ipcp_transformation_t): New function summary.
 408         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
 409         (ipa_get_agg_replacements_for_node): Likewise.
 410
 411 2018-06-08  Martin Liska  <mliska@suse.cz>
 412
 413         * ipa-pure-const.c (struct funct_state_d): Do it class instead
 414         of struct.
 415         (class funct_state_summary_t): New function_summary class.
 416         (has_function_state): Remove.
 417         (get_function_state): Likewise.
 418         (set_function_state): Likewise.
 419         (add_new_function): Likewise.
 420         (funct_state_summary_t::insert): New function.
 421         (duplicate_node_data): Remove.
 422         (remove_node_data): Remove.
 423         (funct_state_summary_t::duplicate): New function.
 424         (register_hooks): Create new funct_state_summaries.
 425         (pure_const_generate_summary): Use it.
 426         (pure_const_write_summary): Likewise.
 427         (pure_const_read_summary): Likewise.
 428         (propagate_pure_const): Likewise.
 429         (propagate_nothrow): Likewise.
 430         (dump_malloc_lattice): Likewise.
 431         (propagate_malloc): Likewise.
 432         (execute): Do not register hooks, just remove summary
 433         instead.
 434         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
 435         constructor.
 436
 437 2018-06-08  Martin Liska  <mliska@suse.cz>
 438
 439         * ipa-reference.c (remove_node_data): Remove.
 440         (duplicate_node_data): Likewise.
 441         (class ipa_ref_var_info_summary_t): New class.
 442         (class ipa_ref_opt_summary_t): Likewise.
 443         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
 444         (get_reference_optimization_summary): Use
 445         ipa_ref_opt_sum_summaries.
 446         (set_reference_vars_info): Remove.
 447         (set_reference_optimization_summary): Likewise.
 448         (ipa_init): Create summaries.
 449         (init_function_info): Use function summary.
 450         (ipa_ref_opt_summary_t::duplicate): New function.
 451         (ipa_ref_opt_summary_t::remove): New function.
 452         (get_read_write_all_from_node): Fix GNU coding style.
 453         (propagate): Use function summary.
 454         (write_node_summary_p): Fix GNU coding style.
 455         (stream_out_bitmap): Likewise.
 456         (ipa_reference_read_optimization_summary): Use function summary.
 457         (ipa_reference_c_finalize): Do not release hooks.
 458
 459 2018-06-08  Martin Liska  <mliska@suse.cz>
 460
 461         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
 462         (analyze_function_body): Extract multiple calls of get_create.
 463         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
 464         * ipa-inline.c (recursive_inlining): Use ::get method.
 465         * ipa-inline.h (estimate_edge_growth): Likewise.
 466
 467 2018-06-08  Martin Liska  <mliska@suse.cz>
 468
 469         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
 470         HSA_INVALID.
 471         (hsa_function_summary::hsa_function_summary): Use the new enum
 472         value.
 473         (hsa_gpu_implementation_p): Use hsa_summaries::get.
 474         * hsa-gen.c (hsa_get_host_function): Likewise.
 475         (get_brig_function_name): Likewise.
 476         * ipa-hsa.c (process_hsa_functions): Likewise.
 477         (ipa_hsa_write_summary): Likewise.
 478         * symbol-summary.h (symtab_duplication): Use ::get function/
 479         (get): New function.
 480
 481 2018-06-08  Martin Liska  <mliska@suse.cz>
 482
 483         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
 484         of get.
 485         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
 486         (hsa_register_kernel): Likewise.
 487         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
 488         * hsa-gen.c (hsa_get_host_function): Likewise.
 489         (get_brig_function_name): Likewise.
 490         (generate_hsa): Likewise.
 491         (pass_gen_hsail::execute): Likewise.
 492         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
 493         (devirtualization_time_bonus): Likewise.
 494         (ipcp_propagate_stage): Likewise.
 495         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
 496         (edge_set_predicate): Likewise.
 497         (evaluate_conditions_for_known_args): Likewise.
 498         (evaluate_properties_for_edge): Likewise.
 499         (ipa_fn_summary::reset): Likewise.
 500         (ipa_fn_summary_t::duplicate): Likewise.
 501         (dump_ipa_call_summary): Likewise.
 502         (ipa_dump_fn_summary): Likewise.
 503         (analyze_function_body): Likewise.
 504         (compute_fn_summary): Likewise.
 505         (estimate_edge_devirt_benefit): Likewise.
 506         (estimate_edge_size_and_time): Likewise.
 507         (estimate_calls_size_and_time): Likewise.
 508         (estimate_node_size_and_time): Likewise.
 509         (inline_update_callee_summaries): Likewise.
 510         (remap_edge_change_prob): Likewise.
 511         (remap_edge_summaries): Likewise.
 512         (ipa_merge_fn_summary_after_inlining): Likewise.
 513         (ipa_update_overall_fn_summary): Likewise.
 514         (read_ipa_call_summary): Likewise.
 515         (inline_read_section): Likewise.
 516         (write_ipa_call_summary): Likewise.
 517         (ipa_fn_summary_write): Likewise.
 518         (ipa_free_fn_summary): Likewise.
 519         * ipa-hsa.c (process_hsa_functions): Likewise.
 520         (ipa_hsa_write_summary): Likewise.
 521         (ipa_hsa_read_section): Likewise.
 522         * ipa-icf.c (sem_function::merge): Likewise.
 523         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
 524         (do_estimate_edge_time): Likewise.
 525         (estimate_size_after_inlining): Likewise.
 526         (estimate_growth): Likewise.
 527         (growth_likely_positive): Likewise.
 528         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
 529         (inline_call): Likewise.
 530         * ipa-inline.c (caller_growth_limits): Likewise.
 531         (can_inline_edge_p): Likewise.
 532         (can_inline_edge_by_limits_p): Likewise.
 533         (compute_uninlined_call_time): Likewise.
 534         (compute_inlined_call_time): Likewise.
 535         (want_inline_small_function_p): Likewise.
 536         (edge_badness): Likewise.
 537         (update_caller_keys): Likewise.
 538         (update_callee_keys): Likewise.
 539         (recursive_inlining): Likewise.
 540         (inline_small_functions): Likewise.
 541         (inline_to_all_callers_1): Likewise.
 542         (dump_overall_stats): Likewise.
 543         (early_inline_small_functions): Likewise.
 544         (early_inliner): Likewise.
 545         * ipa-inline.h (estimate_edge_growth): Likewise.
 546         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
 547         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
 548         * ipa-prop.h (IPA_NODE_REF): Likewise.
 549         (IPA_EDGE_REF): Likewise.
 550         * ipa-pure-const.c (malloc_candidate_p): Likewise.
 551         (propagate_malloc): Likewise.
 552         * ipa-split.c (execute_split_functions): Likewise.
 553         * symbol-summary.h: Rename get to get_create.
 554         (get): Likewise.
 555         (get_create): Likewise.
 556         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
 557
 558 2018-06-08  Martin Liska  <mliska@suse.cz>
 559
 560         * symbol-summary.h (release): Move definition out of class
 561         declaration.
 562         (symtab_removal): Likewise.
 563         (symtab_duplication): Likewise.
 564
 565 2018-06-08  Martin Liska  <mliska@suse.cz>
 566
 567         * symbol-summary.h (function_summary): Move constructor
 568         implementation out of class declaration.
 569         (release): Likewise.
 570         (symtab_insertion): Likewise.
 571         (symtab_removal): Likewise.
 572         (symtab_duplication): Likewise.
 573         (get): Likewise.
 574
 575 2018-06-08  Martin Liska  <mliska@suse.cz>
 576
 577         * Makefile.in: Remove support for MPX (macros, related functions,
 578         fields in cgraph_node, ...).
 579         * builtin-types.def (BT_BND): Likewise.
 580         (BT_FN_BND_CONST_PTR): Likewise.
 581         (BT_FN_CONST_PTR_BND): Likewise.
 582         (BT_FN_VOID_PTR_BND): Likewise.
 583         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
 584         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
 585         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
 586         (expand_builtin_mempcpy_with_bounds): Likewise.
 587         (expand_builtin_memset_with_bounds): Likewise.
 588         (expand_builtin_memset_args): Likewise.
 589         (std_expand_builtin_va_start): Likewise.
 590         (expand_builtin): Likewise.
 591         (expand_builtin_with_bounds): Likewise.
 592         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
 593         (DEF_LIB_BUILTIN_CHKP): Likewise.
 594         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
 595         (DEF_CHKP_BUILTIN): Likewise.
 596         (BUILT_IN_MEMCPY): Likewise.
 597         (BUILT_IN_MEMMOVE): Likewise.
 598         (BUILT_IN_MEMPCPY): Likewise.
 599         (BUILT_IN_MEMSET): Likewise.
 600         (BUILT_IN_STPCPY): Likewise.
 601         (BUILT_IN_STRCAT): Likewise.
 602         (BUILT_IN_STRCHR): Likewise.
 603         (BUILT_IN_STRCPY): Likewise.
 604         (BUILT_IN_STRLEN): Likewise.
 605         (BUILT_IN_MEMCPY_CHK): Likewise.
 606         (BUILT_IN_MEMMOVE_CHK): Likewise.
 607         (BUILT_IN_MEMPCPY_CHK): Likewise.
 608         (BUILT_IN_MEMSET_CHK): Likewise.
 609         (BUILT_IN_STPCPY_CHK): Likewise.
 610         (BUILT_IN_STRCAT_CHK): Likewise.
 611         (BUILT_IN_STRCPY_CHK): Likewise.
 612         * calls.c (store_bounds): Likewise.
 613         (emit_call_1): Likewise.
 614         (special_function_p): Likewise.
 615         (maybe_warn_nonstring_arg): Likewise.
 616         (initialize_argument_information): Likewise.
 617         (finalize_must_preallocate): Likewise.
 618         (compute_argument_addresses): Likewise.
 619         (expand_call): Likewise.
 620         * cfgexpand.c (expand_call_stmt): Likewise.
 621         (expand_return): Likewise.
 622         (expand_gimple_stmt_1): Likewise.
 623         (pass_expand::execute): Likewise.
 624         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
 625         (cgraph_node::remove): Likewise.
 626         (cgraph_node::dump): Likewise.
 627         (cgraph_node::verify_node): Likewise.
 628         * cgraph.h (chkp_function_instrumented_p): Likewise.
 629         (symtab_node::get_alias_target): Likewise.
 630         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
 631         (cgraph_local_p): Likewise.
 632         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
 633         (cgraph_edge::rebuild_references): Likewise.
 634         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
 635         (walk_polymorphic_call_targets): Likewise.
 636         (cgraph_node::expand_thunk): Likewise.
 637         (symbol_table::output_weakrefs): Likewise.
 638         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
 639         (ix86_handle_option): Likewise.
 640         * config/i386/constraints.md: Likewise.
 641         * config/i386/i386-builtin-types.def (BND): Likewise.
 642         (VOID): Likewise.
 643         (PVOID): Likewise.
 644         (ULONG): Likewise.
 645         * config/i386/i386-builtin.def (BDESC_END): Likewise.
 646         (BDESC_FIRST): Likewise.
 647         (BDESC): Likewise.
 648         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
 649         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
 650         * config/i386/i386.c (enum reg_class): Likewise.
 651         (ix86_target_string): Likewise.
 652         (ix86_option_override_internal): Likewise.
 653         (ix86_conditional_register_usage): Likewise.
 654         (ix86_valid_target_attribute_inner_p): Likewise.
 655         (ix86_set_indirect_branch_type): Likewise.
 656         (ix86_set_current_function): Likewise.
 657         (ix86_function_arg_regno_p): Likewise.
 658         (init_cumulative_args): Likewise.
 659         (ix86_function_arg_advance): Likewise.
 660         (ix86_function_arg): Likewise.
 661         (ix86_pass_by_reference): Likewise.
 662         (ix86_function_value_regno_p): Likewise.
 663         (ix86_function_value_1): Likewise.
 664         (ix86_function_value_bounds): Likewise.
 665         (ix86_return_in_memory): Likewise.
 666         (ix86_setup_incoming_vararg_bounds): Likewise.
 667         (ix86_va_start): Likewise.
 668         (indirect_thunk_need_prefix): Likewise.
 669         (print_reg): Likewise.
 670         (ix86_print_operand): Likewise.
 671         (ix86_expand_call): Likewise.
 672         (ix86_output_function_return): Likewise.
 673         (reg_encoded_number): Likewise.
 674         (BDESC_VERIFYS): Likewise.
 675         (ix86_init_mpx_builtins): Likewise.
 676         (ix86_init_builtins): Likewise.
 677         (ix86_emit_cmove): Likewise.
 678         (ix86_emit_move_max): Likewise.
 679         (ix86_expand_builtin): Likewise.
 680         (ix86_builtin_mpx_function): Likewise.
 681         (ix86_get_arg_address_for_bt): Likewise.
 682         (ix86_load_bounds): Likewise.
 683         (ix86_store_bounds): Likewise.
 684         (ix86_load_returned_bounds): Likewise.
 685         (ix86_store_returned_bounds): Likewise.
 686         (ix86_class_likely_spilled_p): Likewise.
 687         (ix86_hard_regno_mode_ok): Likewise.
 688         (x86_order_regs_for_local_alloc): Likewise.
 689         (ix86_mitigate_rop): Likewise.
 690         (ix86_bnd_prefixed_insn_p): Likewise.
 691         (ix86_mpx_bound_mode): Likewise.
 692         (ix86_make_bounds_constant): Likewise.
 693         (ix86_initialize_bounds): Likewise.
 694         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
 695         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
 696         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
 697         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
 698         (TARGET_CHKP_BOUND_MODE): Likewise.
 699         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
 700         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
 701         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
 702         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
 703         * config/i386/i386.h (TARGET_MPX): Likewise.
 704         (TARGET_MPX_P): Likewise.
 705         (VALID_BND_REG_MODE): Likewise.
 706         (FIRST_BND_REG): Likewise.
 707         (LAST_BND_REG): Likewise.
 708         (enum reg_class): Likewise.
 709         (BND_REG_P): Likewise.
 710         (BND_REGNO_P): Likewise.
 711         (BNDmode): Likewise.
 712         (ADJUST_INSN_LENGTH): Likewise.
 713         * config/i386/i386.md: Likewise.
 714         * config/i386/i386.opt: Likewise.
 715         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
 716         (defined): Likewise.
 717         (LINK_MPX): Likewise.
 718         (MPX_SPEC): Likewise.
 719         (LIBMPX_SPEC): Likewise.
 720         (LIBMPXWRAPPERS_SPEC): Likewise.
 721         (CHKP_SPEC): Likewise.
 722         * config/i386/predicates.md: Likewise.
 723         * dbxout.c (dbxout_type): Likewise.
 724         * doc/extend.texi: Likewise.
 725         * doc/invoke.texi: Likewise.
 726         * doc/md.texi: Likewise.
 727         * doc/tm.texi: Likewise.
 728         * doc/tm.texi.in: Likewise.
 729         * dwarf2out.c (is_base_type): Likewise.
 730         (gen_formal_types_die): Likewise.
 731         (gen_subprogram_die): Likewise.
 732         (gen_type_die_with_usage): Likewise.
 733         (gen_decl_die): Likewise.
 734         (dwarf2out_late_global_decl): Likewise.
 735         * expr.c (expand_assignment): Likewise.
 736         (emit_storent_insn): Likewise.
 737         (store_expr_with_bounds): Likewise.
 738         (store_expr): Likewise.
 739         (expand_expr_real_1): Likewise.
 740         * expr.h (store_expr_with_bounds): Likewise.
 741         * function.c (use_register_for_decl): Likewise.
 742         (struct bounds_parm_data): Likewise.
 743         (assign_parms_augmented_arg_list): Likewise.
 744         (assign_parm_find_entry_rtl): Likewise.
 745         (assign_parm_is_stack_parm): Likewise.
 746         (assign_parm_load_bounds): Likewise.
 747         (assign_bounds): Likewise.
 748         (assign_parms): Likewise.
 749         (expand_function_start): Likewise.
 750         * gcc.c (CHKP_SPEC): Likewise.
 751         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
 752         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
 753         (wrestrict_dom_walker::check_call): Likewise.
 754         * gimple.c (gimple_build_call_from_tree): Likewise.
 755         * gimple.h (enum gf_mask): Likewise.
 756         (gimple_call_with_bounds_p): Likewise.
 757         (gimple_call_set_with_bounds): Likewise.
 758         * gimplify.c (gimplify_init_constructor): Likewise.
 759         * ipa-cp.c (initialize_node_lattices): Likewise.
 760         (propagate_constants_across_call): Likewise.
 761         (find_more_scalar_values_for_callers_subset): Likewise.
 762         * ipa-hsa.c (process_hsa_functions): Likewise.
 763         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
 764         * ipa-icf.c (sem_function::merge): Likewise.
 765         * ipa-inline.c (early_inliner): Likewise.
 766         * ipa-pure-const.c (warn_function_noreturn): Likewise.
 767         (warn_function_cold): Likewise.
 768         (propagate_pure_const): Likewise.
 769         * ipa-ref.h (enum GTY): Likewise.
 770         * ipa-split.c (find_retbnd): Likewise.
 771         (consider_split): Likewise.
 772         (split_function): Likewise.
 773         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
 774         * ipa.c (walk_polymorphic_call_targets): Likewise.
 775         (symbol_table::remove_unreachable_nodes): Likewise.
 776         (process_references): Likewise.
 777         (cgraph_build_static_cdtor_1): Likewise.
 778         * lto-cgraph.c (lto_output_node): Likewise.
 779         (output_refs): Likewise.
 780         (compute_ltrans_boundary): Likewise.
 781         (input_overwrite_node): Likewise.
 782         (input_node): Likewise.
 783         (input_cgraph_1): Likewise.
 784         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
 785         * passes.c (pass_manager::execute_early_local_passes): Likewise.
 786         (class pass_chkp_instrumentation_passes): Likewise.
 787         (make_pass_chkp_instrumentation_passes): Likewise.
 788         * passes.def: Likewise.
 789         * rtl.h (struct GTY): Likewise.
 790         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
 791         * stor-layout.c (layout_type): Likewise.
 792         * symtab.c: Likewise.
 793         * target.def: Likewise.
 794         * targhooks.c (default_chkp_bound_type): Likewise.
 795         (default_chkp_bound_mode): Likewise.
 796         (default_builtin_chkp_function): Likewise.
 797         (default_chkp_function_value_bounds): Likewise.
 798         (default_chkp_make_bounds_constant): Likewise.
 799         (default_chkp_initialize_bounds): Likewise.
 800         * targhooks.h (default_chkp_bound_type): Likewise.
 801         (default_chkp_bound_mode): Likewise.
 802         (default_builtin_chkp_function): Likewise.
 803         (default_chkp_function_value_bounds): Likewise.
 804         (default_chkp_make_bounds_constant): Likewise.
 805         (default_chkp_initialize_bounds): Likewise.
 806         * toplev.c (compile_file): Likewise.
 807         (process_options): Likewise.
 808         * tree-core.h (DEF_BUILTIN): Likewise.
 809         (DEF_BUILTIN_CHKP): Likewise.
 810         * tree-inline.c (declare_return_variable): Likewise.
 811         (remap_gimple_stmt): Likewise.
 812         (copy_bb): Likewise.
 813         (initialize_inlined_parameters): Likewise.
 814         (expand_call_inline): Likewise.
 815         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
 816         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
 817         (make_pass_ipa_chkp_produce_thunks): Likewise.
 818         (make_pass_chkp): Likewise.
 819         (make_pass_chkp_opt): Likewise.
 820         (make_pass_chkp_instrumentation_passes): Likewise.
 821         * tree-pretty-print.c (dump_generic_node): Likewise.
 822         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
 823         * tree-ssa-dce.c (propagate_necessity): Likewise.
 824         (eliminate_unnecessary_stmts): Likewise.
 825         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
 826         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
 827         * tree-ssa-sccvn.h: Likewise.
 828         * tree-ssa-strlen.c (get_string_length): Likewise.
 829         (valid_builtin_call): Likewise.
 830         (adjust_last_stmt): Likewise.
 831         (handle_builtin_strchr): Likewise.
 832         (handle_builtin_strcpy): Likewise.
 833         (handle_builtin_stxncpy): Likewise.
 834         (handle_builtin_memcpy): Likewise.
 835         (handle_builtin_strcat): Likewise.
 836         (strlen_check_and_optimize_stmt): Likewise.
 837         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
 838         * tree-streamer-in.c: Likewise.
 839         * tree-streamer.c (record_common_node): Likewise.
 840         * tree.c (tree_code_size): Likewise.
 841         (wide_int_to_tree_1): Likewise.
 842         (type_contains_placeholder_1): Likewise.
 843         (build_common_tree_nodes): Likewise.
 844         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
 845         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
 846         (POINTER_BOUNDS_P): Likewise.
 847         (BOUNDED_TYPE_P): Likewise.
 848         (BOUNDED_P): Likewise.
 849         (CALL_WITH_BOUNDS_P): Likewise.
 850         (pointer_bounds_type_node): Likewise.
 851         * value-prof.c (gimple_ic): Likewise.
 852         * var-tracking.c (vt_add_function_parameters): Likewise.
 853         * varasm.c (make_decl_rtl): Likewise.
 854         (assemble_start_function): Likewise.
 855         (output_constant): Likewise.
 856         (maybe_assemble_visibility): Likewise.
 857         * varpool.c (ctor_for_folding): Likewise.
 858         * chkp-builtins.def: Remove.
 859         * ipa-chkp.c: Remove.
 860         * ipa-chkp.h: Remove.
 861         * rtl-chkp.c: Remove.
 862         * rtl-chkp.h: Remove.
 863         * tree-chkp-opt.c: Remove.
 864         * tree-chkp.c: Remove.
 865         * tree-chkp.h: Remove.
 866
 867 2018-06-07  Carl Love  <cel@us.ibm.com>
 868
 869         * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
 870         vextract_fp_from_shortl): Add BE support.
 871
 872 2018-06-07  Paul Koning  <ni1d@arrl.net>
 873
 874         * compare-elim.c (try_merge_compare): Don't merge compare if
 875         address contains a side effect.
 876         (try_eliminate_compare): Likewise.
 877
 878 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
 879
 880         * config.gcc: Support "tremont".
 881         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
 882         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
 883         PROCESSOR_TREMONT.
 884         * config/i386/i386.c (m_TREMONT): Define.
 885         (processor_target_table): Add "tremont".
 886         (PTA_TREMONT): Define.
 887         (ix86_lea_outperforms): Add TARGET_TREMONT.
 888         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
 889         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
 890         and M_INTEL_GOLDMONT_PLUS.
 891         (fold_builtin_cpu): Add "tremont".
 892         (ix86_add_stmt_cost): Add TARGET_TREMONT.
 893         (ix86_option_override_internal): Add "tremont".
 894         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
 895         (processor_type): Add PROCESSOR_TREMONT.
 896         * config/i386/x86-tune.def: Add m_TREMONT.
 897         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
 898
 899 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
 900
 901         * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
 902         symbol defined for msp430i* devices to be lower case.
 903
 904         gcc/testsuite/gcc.target/msp430/
 905         * msp430i-device-symbol.c: New test.
 906         * msp430f-device-symbol.c: New test.
 907         * msp430.h: New test header file.
 908
 909 2018-06-07  Richard Biener  <rguenther@suse.de>
 910
 911         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
 912         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
 913         Properly wrap signed arithmetic if overflow wraps.
 914
 915 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
 916
 917         PR tree-optimization/69615
 918         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
 919         of a cast from a same precision integral SSA_NAME in a bb dominated
 920         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
 921         cast to utype if rhs2 has already a compatible type.
 922
 923 2018-06-07  Richard Biener  <rguenther@suse.de>
 924
 925         PR tree-optimization/85935
 926         * graphite-scop-detection.c (find_params_in_bb): Analyze
 927         condition operands with respect to the correct loop.  Assert
 928         the analysis doesn't fail.
 929
 930 2018-06-04  Carl Love  <cel@us.ibm.com>
 931
 932         * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
 933         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
 934         as it is slightly cheaper.
 935         (first_match_or_eos_index_<mode>):
 936         Calculate index using natural element order.
 937         (first_match_index_<mode>):
 938         Calculate index using natural element order.
 939         (first_match_or_eos_index_<mode>):
 940         Calculate index using natural order.
 941         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
 942         for BE and LE modes.
 943         * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
 944         P9V_BUILTIN_VCLZLSBB_V16QI.
 945         * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
 946         specific.
 947
 948 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 949
 950         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
 951         indentation and line wrap for many prototypes.  Add missing
 952         @smallexample directives around block of prototypes for vec_xl and
 953         vec_xst.
 954
 955 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
 956
 957         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
 958         track if we pass or return IEEE 128-bit floating point.
 959         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
 960         C++ mangling that is compatible with GCC 8.1.
 961         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
 962         (init_cumulative_args): Note if we pass or return IEEE 128-bit
 963         floating point types.
 964         (rs6000_function_arg_advance_1): Likewise.
 965         (rs6000_mangle_type): Optionally generate mangled names that match
 966         what GCC 8.1 generated for IEEE 128-bit floating point types.
 967         (rs6000_globalize_decl_name): If we have an external function that
 968         passes or returns IEEE 128-bit types, generate a weak reference
 969         from the mangled name used in GCC 8.1 to the current mangled
 970         name.
 971         (rs6000_init_builtins): Make __ibm128 use the long double type if
 972         long double is IBM extended double.  Make __float128 use the long
 973         double type if long double is IEEE 128-bit.
 974
 975         PR target/85657
 976         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
 977         macro for __ibm128 built-in functions.
 978         (PACK_IF): Add __ibm128 pack/unpack functions.
 979         (UNPACK_IF): Likewise.
 980         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
 981         enable long double built-in functions if long double is IEEE
 982         128-bit floating point.
 983         (rs6000_invalid_builtin): Update long double built-in function
 984         error message.
 985         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
 986         functions, adjust the built-in function to use the long double
 987         built-in function if __ibm128 and long double are the same type.
 988         * doc/extend.texi (PowerPC builtins): Update documention for
 989         __builtin_{,un}pack_longdouble.  Add documentation for
 990         __builtin_{,un}pack_ibm128.
 991
 992 2018-06-06  Jim Wilson  <jimw@sifive.com>
 993
 994         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
 995         (struct machine_function): New field interrupt_mode.
 996         (riscv_handle_type_attribute): New function.  Add forward declaration.
 997         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
 998         (riscv_expand_epilogue): Check interrupt_mode field.
 999         (riscv_set_current_function): Check interrupt attribute args and
1000         set interrupt_mode field.
1001         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1002         (riscv_sret, riscv_uret): New.
1003         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1004         new arguments to interrupt attribute.
1005
1006 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
1007
1008         PR target/63177
1009         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1010         Don't handle -mcpu=power8 if -mpower9-vector is also used.
1011
1012 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1013
1014         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1015         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1016         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1017         several redundant entries.
1018
1019 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
1020
1021         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1022         type from "rtx" to "rtx_insn *".
1023         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1024         for local "call_insn", removing cast.
1025         (ix86_expand_call): Likewise, introducing a "call_insn" local.
1026
1027 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
1028
1029         PR tree-optimization/86066
1030         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1031         for BIT_INSERT_EXPR stores.
1032
1033 2018-06-06  Richard Biener  <rguenther@suse.de>
1034
1035         PR tree-optimization/86062
1036         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1037         component refs ontop
1038         of to be offsetted base.
1039
1040 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1041
1042         * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1043         to be static and remove check on interrupt attribute name.
1044
1045         gcc/testsuite/gcc.target/msp430/
1046         * function-attributes-4.c: New test.
1047         * static-interrupts.c: New test.
1048
1049 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1050
1051         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1052         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1053
1054 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
1055
1056         PR target/79924
1057         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1058         second argument.
1059         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1060         Remove second argument, change how error is called.
1061         (aarch64_layout_arg): Remove second argument from
1062         aarch64_err_no_fpadvsimd call.
1063         (aarch64_init_cumulative_args): Ditto.
1064         (aarch64_gimplify_va_arg_expr): Ditto.
1065         * config/aarch64/aarch64.md (mov<mode>): Ditto.
1066
1067 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
1068
1069         * config/i386/i386.md (simple_return_indirect_internal): New expander.
1070         (*simple_return_indirect_internal<mode>): Rename from
1071         simple_return_indirect_internal.  Use W mode iterator.
1072         (rstorssp): New expander.
1073         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
1074         (clrssbsy): New expander.
1075         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
1076
1077 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1078
1079         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1080         __typeof__.
1081         (cmse_check_pointed_object): Likewise.
1082
1083 2018-06-05  Martin Liska  <mliska@suse.cz>
1084
1085         PR gcov-profile/47618
1086         * doc/invoke.texi: Document how -fprofile-dir format
1087         is extended.
1088
1089 2018-06-05  Richard Biener  <rguenther@suse.de>
1090
1091         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1092         removal pretend DOM info isn't available so we do not update
1093         it and only remove edges, not dominated blocks.  Actually free
1094         DOM info in case we removed something.  Remove unreachable blocks.
1095         (mfb_keep_latches): Work with either DOM info or marked backedges.
1096         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1097         first.  Mark backedges if DOM info isn't available.
1098         (Re-)compute DOM info after cleanup_control_flow_pre.
1099
1100 2018-06-05  Richard Biener  <rguenther@suse.de>
1101
1102         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1103         (locus_discrim_hasher::hash): Adjust.
1104         (locus_discrim_hasher::equal): Likewise.
1105         (next_discriminator_for_locus): Work on line directly.
1106         (same_line_p): Pass in expanded locus1 as well.
1107         (assign_discriminators): Avoid redundant location expansions.
1108
1109 2018-06-05  Richard Biener  <rguenther@suse.de>
1110
1111         PR tree-optimization/86046
1112         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1113         if required after clearing TREE_ADDRESSABLE.
1114
1115 2018-06-05  Richard Biener  <rguenther@suse.de>
1116
1117         PR tree-optimization/86047
1118         * tree-ssa-loop.c (for_each_index): Glob handling of all
1119         decls and constants and really handle all of them.
1120
1121 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1122
1123         PR target/81497
1124         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1125         qualifier_void_pointer and qualifier_const_void_pointer.
1126         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1127         (arm_init_builtins): Handle the above.
1128         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1129         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1130         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1131         void intrinsics.
1132
1133 2018-06-05  Martin Liska  <mliska@suse.cz>
1134
1135         * auto-profile.c (read_autofdo_file): Do not use
1136         gcov_ctr_summary struct.
1137         (afdo_callsite_hot_enough_for_early_inline): Likewise.
1138         * coverage.c (struct counts_entry): Likewise.
1139         (read_counts_file): Read just single summary entry.
1140         (get_coverage_counts): Use gcov_summary struct.
1141         * coverage.h (get_coverage_counts): Likewise.
1142         * gcov-dump.c (dump_working_sets): Likewise.
1143         (tag_summary): Dump just single summary.
1144         * gcov-io.c (gcov_write_summary): Write just histogram
1145         summary.
1146         (gcov_read_summary): Read just single summary.
1147         (compute_working_sets): Use gcov_summary struct.
1148         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1149         of GCOV_COUNTERS_SUMMABLE.
1150         (GCOV_COUNTERS_SUMMABLE): Remove.
1151         (GCOV_FIRST_VALUE_COUNTER): Replace with
1152         GCOV_COUNTER_V_INTERVAL.
1153         (struct gcov_ctr_summary): Remove.
1154         (struct gcov_summary): Directly use fields of former
1155         gcov_ctr_summary.
1156         (compute_working_sets): Use gcov_summary struct.
1157         * gcov.c (read_count_file): Do not use ctrs fields.
1158         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1159         struct.
1160         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1161         struct.
1162         * profile.c: Likewise.
1163         * profile.h: Likewise.
1164
1165 2018-06-05  Martin Liska  <mliska@suse.cz>
1166
1167         PR gcov-profile/84846
1168         * gcov.c (output_lines): Print working directory only
1169         in intermediate format.
1170
1171 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
1172
1173         * config/s390/s390-builtin-types.def: Add void function type.
1174         * config/s390/s390-builtins.def: Use the function type for the
1175         tbeginc builtin.
1176
1177 2018-06-04  Jim Wilson  <jimw@sifive.com>
1178
1179         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1180         to int.
1181         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1182         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
1183         handle EH_RETURN_DATA_REGNO registers properly.
1184         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1185         (riscv_expand_epilogue): Update comment.  Change argument name and
1186         type.  Update code to use new name and type.  Pass new args to
1187         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
1188         EXCEPTION_RETURN.
1189         * config/riscv/riscv.md (NORMAL_RETURN): New.
1190         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1191         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1192         (eh_return): Call gen_eh_return_internal and emit barrier.
1193         (eh_return_internal): Call riscv_expand_epilogue.
1194
1195 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
1196
1197         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1198         bit_insertion field and declare can_be_merged_into method.
1199         (merged_store_group::can_be_merged_into): New method.
1200         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1201         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
1202         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1203
1204 2018-06-04  Richard Biener  <rguenther@suse.de>
1205
1206         PR tree-optimization/85955
1207         * builtins.c (fold_builtin_sincos): Convert pointers to
1208         destination to appropriate type before dereferencing.
1209
1210 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
1211
1212         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1213
1214 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
1215
1216         * expr.c (expand_expr_real_1): Force the operand into memory if
1217         its TYPE_MODE is BLKmode and if there is no integer mode for
1218         the number of bits being extracted.
1219
1220 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
1221
1222         PR target/85832
1223         PR target/86036
1224         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1225         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1226
1227 2018-06-04  Richard Biener  <rguenther@suse.de>
1228
1229         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1230         (cleanup_tree_cfg_noloop): ... single caller.  Do
1231         start_recording_case_labels later.
1232
1233 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
1234
1235         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1236         to _IMMINTRIN_H_INCLUDED.
1237         * config/i386/pconfigintrin.h: Ditto.
1238         * config/i386/waitpkgintrin.h: Ditto.
1239         * config/i386/immintrin.h: Add includes for sgxintrin.h,
1240         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1241         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1242         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1243         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1244         waitpkgintrin.h and cldemoteintrin.h.
1245
1246 2018-06-04  Richard Biener  <rguenther@suse.de>
1247
1248         PR tree-optimization/86038
1249         * tracer.c (find_best_successor): Check probability for
1250         being initialized, bail out if not.
1251
1252 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
1253
1254         PR target/86003
1255         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1256         of bits to ignore when comparing architectures.
1257
1258 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
1259
1260         PR tree-optimization/69615
1261         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1262         maximum or minimum of the type, try to merge it also as if
1263         range1 is + [-, x - 1] or + [x + 1, -].
1264
1265         PR c++/86025
1266         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1267
1268 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
1269
1270         PR tree-optimization/86034
1271         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1272         the unsigned bitfield type in a bit insertion sequence if it does not
1273         have a larger precision than the bitfield size.
1274         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1275
1276 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
1277
1278         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1279
1280 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
1281
1282         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1283         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1284         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1285         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1286
1287 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
1288
1289         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1290         Disable -fdelete-null-pointer-checks for ELF toolchain.
1291
1292 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
1293             Kito Cheng  <kito.cheng@gmail.com>
1294
1295         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1296         (nds32le-*-*, nds32be-*-*): Integrate checking process.
1297         (nds32*-*-*): Add glibc and uclibc conditions.
1298         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1299         (TARGET_EXCEPT_UNWIND_INFO): Define.
1300         * config/nds32/elf.h: New file.
1301         * config/nds32/linux.h: New file.
1302         * config/nds32/nds32-elf.opt: New file.
1303         * config/nds32/nds32-linux.opt: New file.
1304         * config/nds32/nds32-fp-as-gp.c
1305         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1306         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1307         TARGET_LINUX_ABI.
1308         (nds32_asm_file_end): Ditto.
1309         (nds32_print_operand): Ditto.
1310         (nds32_insert_attributes): Ditto.
1311         (nds32_init_libfuncs): New function.
1312         (TARGET_HAVE_TLS): Define.
1313         (TARGET_INIT_LIBFUNCS): Define.
1314         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1315         spec content.
1316         (TARGET_ELF): Apply different mcmodel setting.
1317         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1318         been migrated into elf.h and linux.h files.
1319         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1320         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1321         (mcmodel): The content has been migrated into nds32-elf.opt and
1322         nds32-linux.opt files.
1323         * config/nds32/t-elf: New file.
1324         * config/nds32/t-linux: New file.
1325
1326 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
1327             Shiva Chen  <shiva0217@gmail.com>
1328
1329         * config/nds32/constants.md (unspec_volatile_element): Add
1330         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1331         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1332         optimization.
1333         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1334         (make_pass_nds32_fp_as_gp): Declare.
1335         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1336         optmization pass.
1337         (nds32_asm_function_end_prologue): Remove unused asm output.
1338         (nds32_asm_function_begin_epilogue): Remove unused asm output.
1339         (nds32_asm_file_start): Output necessary fp_as_gp information.
1340         (nds32_option_override): Adjust register usage.
1341         (nds32_expand_prologue): Consider fp_as_gp situation.
1342         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1343         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1344         (epilogue): Ditto.
1345         (return): Ditto.
1346         (simple_return): Ditto.
1347         (omit_fp_begin): Output special directive for fp_as_gp.
1348         (omit_fp_end): Output special directive for fp_as_gp.
1349         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1350         mforbid-fp-as-gp): New options.
1351
1352 2018-06-01  Mark Wielaard  <mark@klomp.org>
1353
1354         * dwarf2out.c (dwarf2out_finish): Remove generation of
1355         DW_AT_loclists_base.
1356
1357 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
1358
1359         * gimple-ssa-store-merging.c: Include gimple-fold.h.
1360         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1361         (struct merged_store_group): Add bit_insertion field.
1362         (dump_char_array): Use standard hexadecimal format.
1363         (merged_store_group::merged_store_group): Set bit_insertion to false.
1364         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
1365         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
1366         also print the mask in the dump file.
1367         (pass_store_merging::gate): Minor tweak.
1368         (imm_store_chain_info::coalesce_immediate): Fix wrong association
1369         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
1370         stores with INTEGER_CST stores.
1371         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1372         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1373         and use it throughout.  Generate bit insertion sequences if need be.
1374         (pass_store_merging::process_store): Remove redundant condition.
1375         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1376
1377 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
1378
1379         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1380         the 128-bit floating point types.  Fix function comment.
1381
1382 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1383
1384         * config/aarch64/aarch64-simd.md
1385         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1386         mnemonics.
1387         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1388         mnemonics.
1389
1390 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
1391
1392         PR tree-optimization/85989
1393         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1394         variable.
1395         (backprop::intersect_uses): Check it when deciding whether this
1396         is a backedge reference.
1397         (backprop::process_block): Add each phi to m_visited_phis
1398         after visiting it, then clear it at the end.
1399
1400 2018-06-01  Richard Biener  <rguenther@suse.de>
1401
1402         * tree-vectorizer.h (vect_dr_stmt): New function.
1403         (vect_get_load_cost): Adjust.
1404         (vect_get_store_cost): Likewise.
1405         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1406         Use vect_dr_stmt instead of DR_SMTT.
1407         (vect_record_base_alignments): Likewise.
1408         (vect_calculate_target_alignment): Likewise.
1409         (vect_compute_data_ref_alignment): Likewise and make static.
1410         (vect_update_misalignment_for_peel): Likewise.
1411         (vect_verify_datarefs_alignment): Likewise.
1412         (vector_alignment_reachable_p): Likewise.
1413         (vect_get_data_access_cost): Likewise.  Pass down
1414         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1415         (vect_get_peeling_costs_all_drs): Likewise.
1416         (vect_peeling_hash_get_lowest_cost): Likewise.
1417         (vect_enhance_data_refs_alignment): Likewise.
1418         (vect_find_same_alignment_drs): Likewise.
1419         (vect_analyze_data_refs_alignment): Likewise.
1420         (vect_analyze_group_access_1): Likewise.
1421         (vect_analyze_group_access): Likewise.
1422         (vect_analyze_data_ref_access): Likewise.
1423         (vect_analyze_data_ref_accesses): Likewise.
1424         (vect_vfa_segment_size): Likewise.
1425         (vect_small_gap_p): Likewise.
1426         (vectorizable_with_step_bound_p): Likewise.
1427         (vect_prune_runtime_alias_test_list): Likewise.
1428         (vect_analyze_data_refs): Likewise.
1429         (vect_supportable_dr_alignment): Likewise.
1430         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1431         (vect_gen_prolog_loop_niters): Likewise.
1432         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1433         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1434         modify DR_STMT.
1435         (vect_recog_mask_conversion_pattern): Likewise.
1436         (vect_try_gather_scatter_pattern): Likewise.
1437         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1438         to vect_get_store_cost.
1439         (vect_get_store_cost): Get stmt_info instead of DR.
1440         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1441         (vect_get_load_cost): Get stmt_info instead of DR.
1442
1443 2018-06-01  Richard Biener  <rguenther@suse.de>
1444
1445         PR middle-end/86017
1446         * gimple-fold.c (var_decl_component_p): Also allow offsetted
1447         vars wrapped in MEM_REFs.
1448
1449 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
1450
1451         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1452         Fix subreg tests so that we only return a choice between
1453         GENERAL_REGS and FP_REGS if the original classes included both.
1454
1455 2018-06-01  Richard Biener  <rguenther@suse.de>
1456
1457         PR ipa/85960
1458         * tree-ssa-structalias.c (get_function_part_constraint):
1459         Handle NULL fi->decl.
1460         (find_func_aliases_for_call): Properly handle indirect
1461         fi from direct call.
1462         (find_func_clobbers): Likewise.
1463         (ipa_pta_execute): Likewise.
1464         (create_variable_info_for): For functions that are ifunc_resolver
1465         resolve to a varinfo that contains the result of the resolver call.
1466         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1467         aliases.
1468
1469 2018-05-31  Michael Collison  <michael.collison@arm.com>
1470
1471         * config/aarch64/aarch64.md:
1472         (*fix_to_zero_extenddfdi2): New pattern.
1473         * gcc.target/aarch64/fix_extend1.c: New testcase.
1474
1475 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
1476
1477         PR middle-end/78809
1478         PR middle-end/83026
1479         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1480         and BUILT_IN_STRNCMP_EQ.
1481         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1482         BUILT_IN_STRNCMP_EQ.
1483         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1484         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1485         (gimple_fold_builtin): Likewise.
1486         * tree-ssa-strlen.c (compute_string_length): New function.
1487         (determine_min_obsize): New function.
1488         (handle_builtin_string_cmp): New function to handle calls to
1489         string compare functions.
1490         (strlen_optimize_stmt): Add handling to builtin string compare
1491         calls.
1492         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1493         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1494         * tree.c (build_common_builtin_nodes): Add new defines of
1495         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1496
1497 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
1498
1499         PR target/85984
1500         * bb-reorder.c (pass_partition_blocks::gate): Return false for
1501         functions with naked attribute.
1502
1503 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
1504
1505         * config/i386/sse.md (avx_vec_concat<mode>):
1506         Substitute concat_tg_mode mode attribute with xtg_mode.
1507         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1508         (concat_tg_mode): Remove mode attribute.
1509
1510 2018-05-31  Martin Sebor  <msebor@redhat.com>
1511
1512         PR c/82063
1513         * calls.c (alloc_max_size): Correct a logic error/typo.
1514         Treat excessive arguments as infinite.  Warn for invalid arguments.
1515         * doc/invoke.texi (-Walloc-size-larger-than): Update.
1516
1517 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
1518
1519         PR target/85829
1520         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1521         and movx for Haswell.
1522
1523 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
1524             Cesar Philippidis  <cesar@codesourcery.com>
1525
1526         PR middle-end/85879
1527         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1528         when emitting error on private/firstprivate reductions.
1529         * omp-low.c (lower_omp_target): Avoid reference-type processing
1530         on pointers for firstprivate clause.
1531
1532 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1533
1534         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1535         (st1x2): Likewise.
1536         (st1x3): Likewise.
1537         * config/aarch64/aarch64-simd.md
1538         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1539         (aarch64_ld1_x3_<mode>): Likewise
1540         (aarch64_st1x2<VALLDIF:mode>): Likewise
1541         (aarch64_st1_x2_<mode>): Likewise
1542         (aarch64_st1x3<VALLDIF:mode>): Likewise
1543         (aarch64_st1_x3_<mode>): Likewise
1544         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1545         (vld1_s8_x3): Likewise.
1546         (vld1_u16_x3): Likewise.
1547         (vld1_s16_x3): Likewise.
1548         (vld1_u32_x3): Likewise.
1549         (vld1_s32_x3): Likewise.
1550         (vld1_u64_x3): Likewise.
1551         (vld1_s64_x3): Likewise.
1552         (vld1_f16_x3): Likewise.
1553         (vld1_f32_x3): Likewise.
1554         (vld1_f64_x3): Likewise.
1555         (vld1_p8_x3): Likewise.
1556         (vld1_p16_x3): Likewise.
1557         (vld1_p64_x3): Likewise.
1558         (vld1q_u8_x3): Likewise.
1559         (vld1q_s8_x3): Likewise.
1560         (vld1q_u16_x3): Likewise.
1561         (vld1q_s16_x3): Likewise.
1562         (vld1q_u32_x3): Likewise.
1563         (vld1q_s32_x3): Likewise.
1564         (vld1q_u64_x3): Likewise.
1565         (vld1q_s64_x3): Likewise.
1566         (vld1q_f16_x3): Likewise.
1567         (vld1q_f32_x3): Likewise.
1568         (vld1q_f64_x3): Likewise.
1569         (vld1q_p8_x3): Likewise.
1570         (vld1q_p16_x3): Likewise.
1571         (vld1q_p64_x3): Likewise.
1572         (vst1_s64_x2): Likewise.
1573         (vst1_u64_x2): Likewise.
1574         (vst1_f64_x2): Likewise.
1575         (vst1_s8_x2): Likewise.
1576         (vst1_p8_x2): Likewise.
1577         (vst1_s16_x2): Likewise.
1578         (vst1_p16_x2): Likewise.
1579         (vst1_s32_x2): Likewise.
1580         (vst1_u8_x2): Likewise.
1581         (vst1_u16_x2): Likewise.
1582         (vst1_u32_x2): Likewise.
1583         (vst1_f16_x2): Likewise.
1584         (vst1_f32_x2): Likewise.
1585         (vst1_p64_x2): Likewise.
1586         (vst1q_s8_x2): Likewise.
1587         (vst1q_p8_x2): Likewise.
1588         (vst1q_s16_x2): Likewise.
1589         (vst1q_p16_x2): Likewise.
1590         (vst1q_s32_x2): Likewise.
1591         (vst1q_s64_x2): Likewise.
1592         (vst1q_u8_x2): Likewise.
1593         (vst1q_u16_x2): Likewise.
1594         (vst1q_u32_x2): Likewise.
1595         (vst1q_u64_x2): Likewise.
1596         (vst1q_f16_x2): Likewise.
1597         (vst1q_f32_x2): Likewise.
1598         (vst1q_f64_x2): Likewise.
1599         (vst1q_p64_x2): Likewise.
1600         (vst1_s64_x3): Likewise.
1601         (vst1_u64_x3): Likewise.
1602         (vst1_f64_x3): Likewise.
1603         (vst1_s8_x3): Likewise.
1604         (vst1_p8_x3): Likewise.
1605         (vst1_s16_x3): Likewise.
1606         (vst1_p16_x3): Likewise.
1607         (vst1_s32_x3): Likewise.
1608         (vst1_u8_x3): Likewise.
1609         (vst1_u16_x3): Likewise.
1610         (vst1_u32_x3): Likewise.
1611         (vst1_f16_x3): Likewise.
1612         (vst1_f32_x3): Likewise.
1613         (vst1_p64_x3): Likewise.
1614         (vst1q_s8_x3): Likewise.
1615         (vst1q_p8_x3): Likewise.
1616         (vst1q_s16_x3): Likewise.
1617         (vst1q_p16_x3): Likewise.
1618         (vst1q_s32_x3): Likewise.
1619         (vst1q_s64_x3): Likewise.
1620         (vst1q_u8_x3): Likewise.
1621         (vst1q_u16_x3): Likewise.
1622         (vst1q_u32_x3): Likewise.
1623         (vst1q_u64_x3): Likewise.
1624         (vst1q_f16_x3): Likewise.
1625         (vst1q_f32_x3): Likewise.
1626         (vst1q_f64_x3): Likewise.
1627         (vst1q_p64_x3): Likewise.
1628
1629 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1630
1631         * config/msp430/msp430.c (msp430_output_labelref): Prepend
1632         user_label_prefix to name.
1633
1634         * tree-core.h: Update comment about the format of NAME string
1635         passed to handler in attribute_spec.
1636
1637         * config/msp430/msp430.md: Remove erroneous subreg expression from
1638         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1639         zero_extend{q,h}isi2.
1640
1641 2018-05-30  Borislav Petkov  <bp@suse.de>
1642
1643         * doc/extend.texi: Document some architecture specific
1644         constraints and sort entries.
1645
1646 2018-05-30  Martin Sebor  <msebor@redhat.com>
1647
1648         PR middle-end/85369
1649         * builtins.c (expand_builtin_stpcpy_1): New function.
1650         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1651         only if the former succeeds.
1652
1653 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1654
1655         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1656         in saphira.
1657
1658 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
1659
1660         * doc/invoke.texi (-flinker-output): Document
1661
1662 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
1663
1664         * passes.c (ipa_write_summaries): Only modify statements if body
1665         is in memory.
1666         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1667         incrementally linking.
1668         (ipa_passes): Likewise.
1669         * lto-cgraph.c (lto_output_node): When incrementally linking do not
1670         pass down resolution info.
1671         * common.opt (flag_incremental_link): Update info.
1672         * gcc.c (plugin specs): Turn flinker-output=* to
1673         -plugin-opt=-linker-output-known
1674         * toplev.c (compile_file): Also cut compilation when doing incremental
1675         link.
1676         * flag-types. (enum lto_partition_model): Add
1677         LTO_LINKER_OUTPUT_NOLTOREL.
1678         (invoke.texi): Add -flinker-output docs.
1679         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1680         link same way as WPA; do not stream in dead initializers.
1681
1682         * dwarf2out.c (dwarf2out_die_ref_for_decl,
1683         darf2out_register_external_decl): Support incremental link.
1684
1685 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
1686
1687         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1688
1689 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
1690
1691         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1692         it down to simple_object_copy_lto_debug_sections.
1693         (run_gcc): Determine incremental LTO link time and configure
1694         lto1 into non-wpa mode, disable renaming of debug sections.
1695
1696 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1697
1698         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1699         descriptions of various incorrectly documented functions.
1700
1701 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1702
1703         Revert:
1704         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1705         address check not strict.
1706
1707 2018-05-30  Richard Biener  <rguenther@suse.de>
1708
1709         PR tree-optimization/85964
1710         * tracer.c (better_p): Drop initialized count check, we only
1711         call the function with initialized counts now.
1712         (find_best_successor): Do find a best edge if one
1713         has uninitialized count.
1714         (find_best_predecessor): Likewise.  Do BB frequency check only
1715         if count is initialized.
1716
1717 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
1718
1719         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1720         (aarch64_ldrstr_offset_compare): New.
1721         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1722         load/store orderings.
1723         (aarch64_gen_adjusted_ldpstp): Likewise.
1724
1725 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
1726
1727         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1728         Check for subset of GENERAL_REGS and FP_REGS.
1729         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1730         r=w alternative.
1731
1732 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
1733
1734         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1735         and wi::to_poly_offset.  Add the current offset and then check
1736         whether the sum fits, rather than using an unchecked addition of
1737         a checked term.  Check for a shwi rather than a uhwi.
1738         * expr.c (get_bit_range): Use tree_to_poly_uint64.
1739         (store_constructor): Use poly_int_tree_p.
1740         (expand_expr_real_1): Likewise.
1741         * function.c (assign_temp): Likewise.
1742         * fold-const.c (const_binop): Use poly_int_tree_p and
1743         wi::to_poly_offset.
1744         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
1745         division.
1746         * ipa-icf-gimple.c (func_checker::compare_operand): Use
1747         to_poly_offset for MEM offsets.
1748         * ipa-icf.c (sem_variable::equals): Likewise.
1749         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1750         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1751         wi::to_poly_offset for BIT_FIELD_REF offsets.
1752         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1753         wi::to_poly_offset.
1754         * var-tracking.c (emit_note_insn_var_location): Use
1755         tree_to_poly_uint64.
1756
1757 2018-05-29  Jim Wilson  <jimw@sifive.com>
1758
1759         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1760
1761 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
1762
1763         PR target/85950
1764         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1765         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1766         sequence.
1767         (sse4_1_round<mode>2): Use nonimmediate_operand
1768         for operand 1 predicate.
1769
1770 2018-05-29  Martin Sebor  <msebor@redhat.com>
1771             Richard Biener  <rguenther@suse.de>
1772
1773         PR testsuite/85888
1774         * calls.c (get_size_range): Call determine_value_range instead
1775         of get_value_range..
1776         * tree-vrp.h (determine_value_range): Declared new function.
1777         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1778
1779 2018-05-29  Richard Biener  <rguenther@suse.de>
1780
1781         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1782         sure to use non-pattern stmts for get_earlier_stmt arguments.
1783         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1784         called on pattern stmts.
1785         (get_later_stmt): Likewise.
1786
1787 2018-05-29  Martin Liska  <mliska@suse.cz>
1788
1789         PR gcov-profile/85759
1790         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1791         env variables.
1792
1793 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
1794
1795         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1796         VEC_UNPACK_*_EXPR.
1797         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1798         VEC_PACK_*_EXPR.
1799
1800         PR target/85918
1801         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1802         VEC_PACK_FLOAT_EXPR): New tree codes.
1803         * tree-pretty-print.c (op_code_prio): Handle
1804         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1805         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1806         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1807         * tree-inline.c (estimate_operator_cost): Likewise.
1808         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1809         * fold-const.c (const_binop): Likewise.
1810         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1811         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1812         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1813         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1814         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1815         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1816         * expr.c (expand_expr_real_2): Likewise.
1817         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1818         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1819         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1820         optabs.
1821         * optabs.c (expand_widen_pattern_expr): For
1822         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1823         sign from result type rather than operand's type.
1824         (expand_binop_directly): For vec_packu_float_optab and
1825         vec_packs_float_optab allow result type to be different from operand's
1826         type.
1827         * optabs-tree.c (optab_for_tree_code): Handle
1828         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1829         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
1830         * tree-vect-generic.c (expand_vector_operations_1):  Handle
1831         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1832         VEC_PACK_FLOAT_EXPR.
1833         * tree-vect-stmts.c (supportable_widening_operation): Handle
1834         FIX_TRUNC_EXPR.
1835         (supportable_narrowing_operation): Handle FLOAT_EXPR.
1836         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1837         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1838         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
1839         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1840         mode attributes.
1841         (vec_pack<floatprefix>_float_<mode>): New expander.
1842         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1843         attributes.
1844         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1845         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1846         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1847         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
1848         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
1849         Document.
1850         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
1851         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
1852         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1853         VEC_PACK_FLOAT_EXPR): Document.
1854
1855 2018-05-29  Richard Biener  <rguenther@suse.de>
1856
1857         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
1858         member.
1859         (stmt_vec_info_vec): Make pointer.
1860         (init_stmt_vec_info_vec): Remove.
1861         (free_stmt_vec_info_vec): Likewise.
1862         (set_stmt_vec_info_vec): New function.
1863         (free_stmt_vec_infos): Likewise.
1864         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
1865         (set_vinfo_for_stmt): Likewise.
1866         (get_earlier_stmt): Likewise.
1867         (get_later_stmt): Likewise.
1868         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
1869         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
1870         (vec_info::~vec_info): Free stmt_vec_infos.
1871         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
1872         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
1873         (pass_slp_vectorize::execute): Likewise.
1874         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
1875         (free_stmt_vec_info_vec): Likewise.
1876         (set_stmt_vec_info_vec): New function.
1877         (free_stmt_vec_infos): Likewise.
1878         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
1879         the global stmt_vec_info_vec.
1880         * tree-parloops.c (gather_scalar_reductions): Use
1881         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
1882         vector.
1883
1884 2018-05-29  Richard Biener  <rguenther@suse.de>
1885
1886         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
1887
1888 2018-05-29  Martin Liska  <mliska@suse.cz>
1889             David Malcolm  <dmalcolm@redhat.com>
1890
1891         * vec.c (test_reverse): New.
1892         (vec_c_tests): Add new test.
1893         * vec.h (vl_ptr>::reverse): New function.
1894
1895 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
1896
1897         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
1898
1899         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
1900         and later.
1901
1902 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1903
1904         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
1905
1906 2018-05-28  Richard Biener  <rguenther@suse.de>
1907
1908         PR tree-optimization/85933
1909         * tree-vect-data-refs.c (vect_record_base_alignments): Only
1910         look at stmts marked as vectorizable.
1911
1912 2018-05-28  Richard Biener  <rguenther@suse.de>
1913
1914         PR tree-optimization/85934
1915         * tree-vect-generic.c (expand_vector_operations_1): Hoist
1916         vector boolean check before scalar optimization.
1917
1918 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
1919
1920         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1921         for armv5te.
1922
1923 2018-05-28  Mark Wielaard  <mark@klomp.org>
1924
1925         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1926         if it is an expression containing a minus sign.
1927
1928 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
1929
1930         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1931
1932 2018-05-27  Paul Koning  <ni1d@arrl.net>
1933
1934         * config/pdp11/pdp11.md (truncsihi2): Remove.
1935
1936 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
1937             Chung-Ju Wu  <jasonwucj@gmail.com>
1938
1939         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1940         implementation.
1941         (unaligned_store_dw): Ditto.
1942         * config/nds32/nds32-memory-manipulation.c
1943         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1944         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1945         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1946         (emit_setmem_word_loop): Rename to ...
1947         (emit_setmem_doubleword_loop): ... this.
1948         (nds32_gen_dup_4_byte_to_word_value): New function.
1949         (nds32_gen_dup_8_byte_to_double_word_value): New function.
1950         (nds32_expand_setmem_loop): Refine implementation.
1951         (nds32_expand_setmem_loop_v3m): Ditto.
1952         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1953         pattern.
1954
1955 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
1956
1957         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1958
1959 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
1960
1961         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1962         (nds32_init_machine_status): Initialize machine->attr_naked_p and
1963         machine->attr_no_prologue_p.
1964         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1965         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1966         (nds32_expand_epilogue): Consider attr_naked_p.
1967         (nds32_expand_epilogue_v3pop): Likewise.
1968         (nds32_can_use_return_insn): Likewise.
1969         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1970         attr_no_prologue_p fields.
1971         * config/nds32/nds32.opt (mret-in-naked-func): New option.
1972
1973 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
1974
1975         PR target/85918
1976         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1977         attributes.
1978         * config/i386/sse.md
1979         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1980         Rename to ...
1981         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1982         ... this.
1983         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1984         Rename to ...
1985         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1986         ... this.
1987         (*<floatsuffix>floatv2div2sf2): Rename to ...
1988         (*float<floatunssuffix>v2div2sf2): ... this.
1989         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1990         (float<floatunssuffix>v2div2sf2_mask): ... this.
1991         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1992         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1993         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1994         to ...
1995         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1996         ... this.
1997         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1998         Rename to ...
1999         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2000         ... this.
2001         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2002         Rename to ...
2003         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2004         ... this.
2005         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2006         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2007         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2008         gen_ufix_truncv8dfv8si2.
2009         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2010         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2011         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2012         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2013         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2014         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2015         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2016         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2017
2018 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
2019
2020         PR target/85900
2021         PR target/85345
2022         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2023
2024 2018-05-25  Jim Wilson  <jimw@sifive.com>
2025
2026         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2027         * config/riscv/riscv.c (struct machine_function): Add
2028         interrupt_handler_p and attribute_checked_p fields.
2029         (riscv_attribute_table): Add interrupt.
2030         (riscv_interrupt_type_p): New.
2031         (riscv_save_reg_p): Save extra regs for interrupt handler.
2032         (riscv_use_save_libcall): Return false  for interrupt handler.
2033         (riscv_first_stack_step): Add forward declaration.
2034         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
2035         for interrupt handler with large frame.  Use it for saved reg list.
2036         (riscv_expand_prologue): Move flag_stack_usage_info support to
2037         eliminate duplication.
2038         (riscv_expand_epilogue): Generate mret for interrupt handler.
2039         (riscv_epilogue_uses): New.
2040         (riscv_can_use_return_insn): Return false for interrupt handler.
2041         (riscv_function_ok_for_sibcall): Likewise.
2042         (riscv_set_current_function): Add interrupt handler support.
2043         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2044         * config/riscv/riscv.md (UNSPECV_MRET): New.
2045         (GP_REGNUM): New.
2046         (riscv_frflags, riscv_fsflags): Use tab after opcode.
2047         (riscv_mret): New.
2048         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2049
2050 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2051
2052         PR tree-optimization/85712
2053         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2054         this candidate has already been replaced in-situ by a copy.
2055
2056 2018-05-25  Jason Merrill  <jason@redhat.com>
2057
2058         PR c++/80485 - inline function non-zero address.
2059         * symtab.c (nonzero_address): Check DECL_COMDAT.
2060
2061 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
2062
2063         PR target/83628
2064         * config/alpha/alpha.md (ashlsi3): New insn pattern.
2065         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
2066         extension of SImode operation.  Use const123_operand predicate.
2067         (*saddsi_1): Remove.
2068         (*saddl_se_1): Ditto.
2069         (*ssubsi_1): Ditto.
2070         (*ssubl_se_1): Ditto.
2071         * config/alpha/predicates.md (const123_operand): New predicate.
2072         * config/alpha/constraints.md (P): Use IN_RANGE.
2073
2074 2018-05-25  Richard Biener  <rguenther@suse.de>
2075
2076         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2077         defaulted to true.
2078         (ref_maybe_used_by_stmt_p): Likewise.
2079         (stmt_may_clobber_ref_p): Likewise.
2080         (stmt_may_clobber_ref_p_1): Likewise.
2081         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2082         and pass it along.
2083         (ref_maybe_used_by_stmt_p): Likewise.
2084         (stmt_may_clobber_ref_p): Likewise.
2085         (stmt_may_clobber_ref_p_1): Likewise.
2086         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2087         the alias oracle to disambiguate DRs with stmts DR analysis
2088         couldn't handle.
2089         (vect_analyze_data_refs): Do not give up on not analyzable
2090         DRs for BB vectorization.  Remove code truncating the dataref
2091         vector.
2092
2093 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
2094
2095         PR target/85832
2096         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2097         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
2098         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2099
2100 2018-05-25  Richard Biener  <rguenther@suse.de>
2101
2102         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2103         function, combining stmt data ref gathering and fatal analysis
2104         parts.
2105         (vect_analyze_data_refs): Remove now redudnant code and simplify.
2106         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2107         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2108         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2109         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2110
2111 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
2112
2113         PR tree-optimization/85720
2114         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2115         SCC if all partitions are builtins.
2116         (version_loop_by_alias_check): New parameter.  Generate cancelable
2117         runtime alias check if all partitions are builtins.
2118         (distribute_loop): Update call to above function.
2119
2120 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
2121
2122         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2123         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2124         (parm_default_def_partition_arg): Ditto.
2125         (set_parm_default_def_partition): Ditto.
2126         (get_parm_default_def_partitions): Ditto and make it static.
2127         (get_undefined_value_partitions): Ditto and make it static.
2128         (remove_ssa_form): Refactor call to init_var_map here.
2129         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2130         computation for loop region.
2131         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2132         (register_default_def): Delete.
2133         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2134         (parm_default_def_partition_arg): Ditto.
2135         (set_parm_default_def_partition): Ditto.
2136         (get_parm_default_def_partitions): Ditto and make it static.
2137         (get_undefined_value_partitions): Ditto and make it static.
2138         (coalesce_with_default, coalesce_with_default): Update comment.
2139         (create_coalesce_list_for_region): New func factored out from
2140         create_outofssa_var_map.
2141         (populate_coalesce_list_for_outofssa): New func factored out from
2142         create_outofssa_var_map and coalesce_ssa_name.
2143         (create_outofssa_var_map): Delete.
2144         (coalesce_ssa_name): Refactor to support live range computation.
2145         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2146         (get_parm_default_def_partitions): Delete.
2147         (get_undefined_value_partitions): Ditto.
2148         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2149         computation for loop region.
2150         (new_tree_live_info, loe_visit_block): Ditto.
2151         (live_worklist, set_var_live_on_entry): Ditto.
2152         (calculate_live_on_exit, verify_live_on_entry): Ditto.
2153         * tree-ssa-live.h (struct _var_map): New fields.
2154         (init_var_map): Change decl.
2155         (region_contains_p): New.
2156
2157 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
2158
2159         * tree-ssa-live.h (live_merge_and_clear): Delete.
2160
2161 2018-05-25  Richard Biener  <rguenther@suse.de>
2162
2163         PR c++/85912
2164         * tree-dump.c (dequeue_and_dump): Remove access to removed
2165         operand 2 of a SWITCH_EXPR.
2166
2167 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2168
2169         * doc/sourcebuild.texi (vect_double_cond_arith): Include
2170         multiplication and division.
2171         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2172         (cond_udiv@var{m}, cond_umod@var{m}): Document.
2173         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2174         (cond_udiv_optab, cond_umod_optab): New optabs.
2175         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2176         (IFN_COND_RDIV): New internal functions.
2177         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2178         TRUNC_MOD_EXPR and RDIV_EXPR.
2179         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2180         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2181         New unspecs.
2182         (SVE_INT_BINARY): Include mult.
2183         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2184         (optab, sve_int_op): Handle mult.
2185         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2186         UNSPEC_COND_DIV.
2187         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2188         for SVE_INT_BINARY_SD.
2189
2190 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2191
2192         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2193         (optab, sve_int_op): Handle div and udiv.
2194         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2195         for SVE_INT_BINARY_SD.
2196         (*<optab><mode>3): New insn for the same.
2197
2198 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2199
2200         * tree-vect-patterns.c: Include predict.h.
2201         (vect_recog_divmod_pattern): Restrict check for division support
2202         to when optimizing for size.
2203
2204 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2205
2206         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2207         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2208         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2209         (gimple_match_op::set_op): Likewise.
2210         (gimple_resimplify4): Declare.
2211         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2212         (expr::gen_transform): Likewise.
2213         (decision_tree::gen): Generate a simplification routine for 4 operands.
2214         * gimple-match-head.c (gimple_simplify): Add an overload for
2215         4 operands.  In the top-level function, handle up to 4 call
2216         arguments and call gimple_resimplify4.
2217         (gimple_resimplify4): New function.
2218         (build_call_internal): Pass a fourth operand.
2219         (maybe_push_to_seq): Likewise.
2220         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2221         Fold VEC_COND_EXPRs of an operation and a default value into
2222         an IFN_COND_* function if possible.
2223         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2224         New unspecs.
2225         (SVE_COND_FP_BINARY): Include them.
2226         (optab, sve_fp_op): Handle them.
2227         (SVE_INT_BINARY_REV): New code iterator.
2228         (SVE_COND_FP_BINARY_REV): New int iterator.
2229         (commutative): New int attribute.
2230         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2231         Declare.
2232         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2233         function.
2234         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2235         (*cond_<optab><mode>): New patterns for reversed operands.
2236
2237 2018-05-25  Richard Biener  <rguenther@suse.de>
2238
2239         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2240         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2241         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2242         (STMT_VINFO_GROUPED_ACCESS): Adjust.
2243         * tree-vect-data-refs.c (everywhere): Adjust users.
2244         * tree-vect-loop.c (everywhere): Likewise.
2245         * tree-vect-slp.c (everywhere): Likewise.
2246         * tree-vect-stmts.c (everywhere): Likewise.
2247         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2248
2249 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2250
2251         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2252         Rename to...
2253         (gcc_cv_as_section_exclude): ... this.
2254         Try Solaris as #exclude syntax.
2255         * configure: Regenerate.
2256         * config.in: Regenerate.
2257         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2258         SECTION_EXCLUDE.
2259         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2260         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2261
2262         * varasm.c (default_elf_asm_named_section): Don't check if
2263         HAVE_GAS_SECTION_EXCLUDE is defined.
2264
2265 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2266
2267         * doc/md.texi: Update the documentation of the cond_* optabs
2268         to mention the new final operand.  Fix GET_MODE_NUNITS call.
2269         Describe the scalar case too.
2270         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2271         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2272         instead of 2.
2273         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2274         (get_conditional_internal_fn): Update comment.
2275         * tree-vect-loop.c (vectorizable_reduction): Pass the original
2276         accumulator value as a final argument to conditional functions.
2277         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2278         a define_expand and add an "else" operand.  Assert for now that
2279         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
2280         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2281         (*cond_<optab><mode>): New patterns.
2282         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2283         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2284         (UNSPEC_COND_EOR): Delete.
2285         (optab): Remove associated mappings.
2286         (SVE_INT_BINARY): New code iterator.
2287         (sve_int_op): Remove int attribute and add "minus" to the code
2288         attribute.
2289         (SVE_COND_INT_OP): Delete.
2290         (SVE_COND_FP_OP): Rename to...
2291         (SVE_COND_FP_BINARY): ...this.
2292
2293 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2294
2295         * optabs.c (can_reuse_operands_p): New function.
2296         (maybe_legitimize_operands): Try to reuse the results for
2297         earlier operands.
2298
2299 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
2300
2301         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2302         Add {q} suffix to insn mnemonic.
2303
2304 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2305
2306         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2307         (msp430_warn_func_return): New.
2308
2309 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
2310
2311         * fold-const.c (tree_nonzero_bits): New function.
2312         * fold-const.h (tree_nonzero_bits): Likewise.
2313         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2314         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2315
2316 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
2317
2318         PR target/85900
2319         PR target/85345
2320         * varasm.c (assemble_alias): Check ifunc_resolver only on
2321         FUNCTION_DECL.
2322
2323 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
2324
2325         PR target/85903
2326         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2327         when memory input operand is handled.
2328
2329 2018-05-24  Luis Machado  <luis.machado@linaro.org>
2330
2331         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2332         global.
2333         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2334
2335 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
2336
2337         * match.pd: Delay FMA folds until after vectorization.
2338
2339 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2340
2341         PR target/83009
2342         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2343         address check not strict.
2344
2345 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
2346
2347         * gimple-match.h (gimple_match_op): New class.
2348         (mprts_hook): Replace parameters with a gimple_match_op *.
2349         (maybe_build_generic_op): Likewise.
2350         (gimple_simplified_result_is_gimple_val): Replace parameters with
2351         a const gimple_match_op *.
2352         (gimple_simplify): Replace code_helper * and tree * parameters with
2353         a gimple_match_op * parameter.
2354         (gimple_resimplify1): Replace code_helper *, tree and tree *
2355         parameters with a gimple_match_op * parameter.
2356         (gimple_resimplify2): Likewise.
2357         (gimple_resimplify3): Likewise.
2358         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2359         parameters with a gimple_match_op * parameter.
2360         * gimple-match-head.c (gimple_simplify): Change prototypes of
2361         auto-generated functions to take a gimple_match_op * instead of
2362         separate code_helper * and tree * parameters.  Make the same
2363         change in the top-level overload and update calls to the
2364         gimple_resimplify routines.  Update calls to the auto-generated
2365         functions and to maybe_push_res_to_seq in the publicly-facing
2366         operation-specific gimple_simplify overloads.
2367         (gimple_match_op::MAX_NUM_OPS): Define.
2368         (gimple_resimplify1): Replace rcode and ops with a single res_op
2369         parameter.  Update call to gimple_simplify.
2370         (gimple_resimplify2): Likewise.
2371         (gimple_resimplify3): Likewise.
2372         (mprts_hook): Replace parameters with a gimple_match_op *.
2373         (maybe_build_generic_op): Likewise.
2374         (build_call_internal): Replace type, nargs and ops with
2375         a gimple_match_op *.
2376         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2377         with a single gimple_match_op *.  Update calls to mprts_hook,
2378         build_call_internal and gimple_simplified_result_is_gimple_val.
2379         Factor out code that is common to the tree_code and combined_fn cases.
2380         * genmatch.c (expr::gen_transform): Replace tem_code and
2381         tem_ops with a gimple_match_op called tem_op.  Update calls
2382         to the gimple_resimplify functions and maybe_push_res_to_seq.
2383         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2384         res_ops.  Update call to the gimple_resimplify functions.
2385         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2386         (decision_tree::gen): Make the functions take a gimple_match_op *
2387         called res_op instead of separate res_code and res_ops parameters.
2388         Update call accordingly.
2389         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2390         and ops with a single res_op parameter.  Update calls to
2391         maybe_build_generic_op and maybe_push_res_to_seq.
2392         (fold_stmt_1): Update calls to gimple_simplify and
2393         replace_stmt_with_simplification.
2394         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2395         and gimple_simplified_result_is_gimple_val.
2396         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2397         gimple_simplify.
2398         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2399         with a gimple_match_op *.
2400         (vn_nary_build_or_lookup): Likewise.  Update call to
2401         vn_nary_build_or_lookup_1.
2402         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2403         gimple_match_op *.  Update calls to the gimple_resimplify routines
2404         and to gimple_simplified_result_is_gimple_val.
2405         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2406         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2407         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2408         (visit_nary_op): Likewise.
2409         (visit_reference_op_load): Likewise.
2410
2411 2018-05-23  Luis Machado  <luis.machado@linaro.org>
2412
2413         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2414         modifier for printing the step amount.
2415
2416 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
2417
2418         PR target/78849
2419         * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2420         types.
2421
2422 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
2423
2424         * doc/sourcebuild.texi (Endianness): New subsubsection.
2425
2426 2018-05-23  Luis Machado  <luis.machado@linaro.org>
2427
2428         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2429         <prefetch_dynamic_strides>: New const bool field.
2430         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2431         prefetch_dynamic_strides.
2432         (exynosm1_prefetch_tune): Likewise.
2433         (thunderxt88_prefetch_tune): Likewise.
2434         (thunderx_prefetch_tune): Likewise.
2435         (thunderx2t99_prefetch_tune): Likewise.
2436         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2437         false.
2438         (aarch64_override_options_internal): Update to set
2439         PARAM_PREFETCH_DYNAMIC_STRIDES.
2440         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2441         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2442         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2443         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2444         prefetch-dynamic-strides setting.
2445
2446 2018-05-23  Luis Machado  <luis.machado@linaro.org>
2447
2448         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2449         <minimum_stride>: New const int field.
2450         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2451         minimum_stride field defaulting to -1.
2452         (exynosm1_prefetch_tune): Likewise.
2453         (thunderxt88_prefetch_tune): Likewise.
2454         (thunderx_prefetch_tune): Likewise.
2455         (thunderx2t99_prefetch_tune): Likewise.
2456         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2457         <default_opt_level>: Set to 3.
2458         (aarch64_override_options_internal): Update to set
2459         PARAM_PREFETCH_MINIMUM_STRIDE.
2460         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2461         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2462         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2463         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2464         stride is constant and is below the minimum stride threshold.
2465
2466 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2467
2468         * config/arm/arm-cpus.in (mode26): Delete.
2469         (armv4): Delete mode26 reference.
2470         * config/arm/arm.c (arm_configure_build_target): Delete use of
2471         isa_bit_mode26.
2472
2473 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
2474
2475         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2476         New insn pattern.
2477         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2478         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
2479         for non-SSE modes.
2480         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
2481         (floatunsdidf2): Ditto.
2482
2483 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
2484
2485         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2486         (fixuns_trunc<mode>si2_avx512f): Ditto.
2487         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2488         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2489         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2490
2491 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
2492
2493         PR rtl-optimization/79985
2494         * df-scan.c (df_insn_refs_collect): Remove special case for
2495         global registers and asm statements.
2496
2497 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
2498
2499         * extend.texi (Global Register Variables): Rewrite the bullet list.
2500         Note that the register is available for allocation. Note that access
2501         via inline asm must use constraints. Add note about async-signal
2502         handlers. Remove paragraph about automagic register selection.
2503
2504 2018-05-23  Richard Biener  <rguenther@suse.de>
2505
2506         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2507         of fixed offset from memset VN.
2508
2509 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
2510
2511         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2512         first_interp field.
2513         (alloc_cand_and_find_basis): Initialize first_interp field.
2514         (slsr_process_mul): Modify first_interp field.
2515         (slsr_process_add): Likewise.
2516         (slsr_process_cast): Modify first_interp field for each new
2517         interpretation.
2518         (slsr_process_copy): Likewise.
2519         (dump_candidate): Dump first_interp field.
2520         (replace_mult_candidate): Process all interpretations, not just
2521         subsequent ones.
2522         (replace_rhs_if_not_dup): Likewise.
2523         (replace_one_candidate): Likewise.
2524
2525 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
2526
2527         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2528         Add new boolean.
2529         (aarch64_needs_frame_chain): New function.
2530         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2531
2532 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
2533
2534         PR target/84882
2535         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2536         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2537         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2538         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2539         as true for strict-align.
2540         (aarch64_can_inline_p): Perform checks even when callee has no
2541         attributes to check for strict alignment.
2542         * doc/extend.texi (AArch64 Function Attributes): Document
2543         no-strict-align.
2544         * doc/invoke.texi: (AArch64 Options): Likewise.
2545
2546 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
2547
2548         PR tree-optimization/85853
2549         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2550         the handling of the root of the node to...
2551         (vect_slp_analyze_node_operations_1): ...this new function,
2552         and run the whole thing with the child nodes' def types
2553         set according to their SLP node's def type.
2554
2555 2018-05-23  Richard Biener  <rguenther@suse.de>
2556
2557         PR middle-end/85874
2558         * tree-data-ref.c (create_runtime_alias_checks): Defer
2559         and ignore overflow warnings.
2560
2561 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
2562
2563         PR tree-optimization/85822
2564         * tree-vrp.c (is_masked_range_test): Fix handling of negative
2565         constants.
2566
2567 2018-05-23  Richard Biener  <rguenther@suse.de>
2568
2569         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2570         memset constants via native_interpret_expr.
2571
2572 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
2573
2574         PR target/85345
2575         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2576         attribute.
2577         (cgraph_node::create_alias): Likewise.
2578         (cgraph_node::get_availability): Check ifunc_resolver instead
2579         of looking up ifunc attribute.
2580         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2581         * varasm.c (do_assemble_alias): Likewise.
2582         (assemble_alias): Likewise.
2583         (default_binds_local_p_3): Likewise.
2584         * cgraph.h (cgraph_node): Add ifunc_resolver.
2585         (cgraph_node::only_called_directly_or_aliased_p): Return false
2586         for IFUNC resolver.
2587         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2588         attribute.
2589         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2590         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2591         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2592         instead of looking up ifunc attribute.
2593
2594 2018-05-22  Luis Machado  <luis.machado@linaro.org>
2595
2596         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2597
2598 2018-05-22  Martin Sebor  <msebor@redhat.com>
2599
2600         PR middle-end/85359
2601         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2602         only when expasion succeeds.
2603         (expand_builtin_strcmp): Same.
2604         (expand_builtin_strncmp): Same.
2605
2606 2018-05-22  Martin Sebor  <msebor@redhat.com>
2607
2608         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2609
2610 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
2611             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2612
2613         * config/aarch64/aarch64-ldpstp.md: Replace uses of
2614         aarch64_mem_pair_operand with memory_operand and delete operand swapping
2615         code.
2616         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2617         Add check for legitimate_address.
2618         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2619         (aarch64_swap_ldrstr_operands): New.
2620         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2621         Define prototype.
2622
2623 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
2624             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2625
2626         * config/aarch64/aarch64.md: New patterns to generate stp
2627         and ldp.
2628         (store_pair_sw, store_pair_dw): New patterns to generate stp for
2629         single words and double words.
2630         (load_pair_sw, load_pair_dw): Likewise.
2631         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2632         Delete.
2633         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2634         Delete.
2635         * config/aarch64/aarch64-ldpstp.md: Modify peephole
2636         for different mode ldpstp and add peephole for merged zero stores.
2637         Likewise for loads.
2638         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2639         Add size check.
2640         (aarch64_gen_store_pair): Rename calls to match new patterns.
2641         (aarch64_gen_load_pair): Rename calls to match new patterns.
2642         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2643         (load_pair<DREG:mode><DREG2:mode>): ... This.
2644         (store_pair<mode>): Rename to...
2645         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2646         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2647         New mode iterators.
2648         (V_INT_EQUIV): Handle SImode.
2649         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2650         New predicate.
2651
2652 2018-05-22  Martin Sebor  <msebor@redhat.com>
2653
2654         PR c/85623
2655         * calls.c (maybe_warn_nonstring_arg): Use string length to set
2656         or ajust the presumed bound on an operation to avoid unnecessary
2657         warnings.
2658
2659 2018-05-22  Martin Sebor  <msebor@redhat.com>
2660
2661         PR tree-optimization/85826
2662         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2663         assuming that a DECL necesarily has a constant size.
2664
2665 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
2666
2667         PR middle-end/85862
2668         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2669
2670 2018-05-22  Richard Biener  <rguenther@suse.de>
2671
2672         PR tree-optimization/85834
2673         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2674         non-constant and non-zero memset arguments.
2675
2676 2018-05-22  Martin Liska  <mliska@suse.cz>
2677
2678         PR ipa/85607
2679         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2680
2681 2018-05-22  Richard Biener  <rguenther@suse.de>
2682
2683         PR tree-optimization/85863
2684         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2685         comparisons when vectype is specified.
2686         (vectorizable_condition): Do not specify vectype for
2687         vect_is_simple_cond when SLP vectorizing.
2688
2689 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
2690
2691         PR target/85657
2692         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2693         define __ibm128 as long double.
2694         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2695         as a distinct type when IEEE 128-bit support is enabled.
2696         (init_float128_ieee): Fix up conversions between IFmode and IEEE
2697         128-bit types to use the correct functions.
2698         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2699         convert between 128-bit floating point types that have different
2700         modes but the same representation, instead of using gen_lowpart to
2701         makean alias.
2702         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2703         KFmode.
2704         (IFKF_reg): New attributes to give the register constraints for
2705         IFmode and KFmode.
2706         (extend<mode>tf2_internal): New insns to mark an explicit
2707         conversion between 128-bit floating point types that have a
2708         different mode but share the same representation.
2709
2710 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
2711
2712         PR tree-optimization/85814
2713         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2714         a null return from get_strinfo when unsharing the next
2715         strinfo in the chain.
2716
2717 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
2718
2719         PR gcc/84923
2720         * varasm.c (weak_finish): Clean up weak_decls.
2721
2722 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2723
2724         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2725         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2726         UNSPEC_UADALP values.
2727         * config/aarch64/iterators.md (ABAL): New int iterator.
2728         (ABDL2): Likewise.
2729         (ADALP): Likewise.
2730         (sur): Add mappings for the above.
2731         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2732         New define_insn.
2733         (aarch64_<sur>abal<mode>_4): Likewise.
2734         (aarch64_<sur>adalp<mode>_3): Likewise.
2735         (<sur>sadv16qi): New define_expand.
2736
2737 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
2738
2739         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2740         (*movdf_internal): Ditto.
2741         (*rcpsf2_sse): Ditto.
2742         (*rsqrtsf2_sse): Ditto.
2743         (*sqrt<mode>2_sse): Ditto.
2744
2745 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
2746
2747         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2748         eor3q<mode>4.
2749         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2750         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2751         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2752         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2753         vbcaxq_s64): New.
2754         * config/aarch64/arm_neon.h: Likewise.
2755         * config/aarch64/iterators.md (VQ_I): New.
2756
2757 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
2758
2759         * config.gcc: Add arc/t-multilib-linux to tmake_file for
2760         arc*-*-linux*.
2761         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2762         MULTILIB_DIRNAMES
2763
2764 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
2765
2766         * config/nds32/constraints.md (S): New constraint.
2767         * config/nds32/nds32.md (call_internal): Use constraint S.
2768         (call_value_internal): Likewise.
2769         (sibcall_internal): Likewise.
2770         (sibcall_value_internal): Likewise.
2771
2772 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
2773             Chung-Ju Wu  <jasonwucj@gmail.com>
2774
2775         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2776         into consideration.
2777
2778 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
2779             Chung-Ju Wu  <jasonwucj@gmail.com>
2780
2781         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2782         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2783         (nds32_rtx_costs_impl): Simplify.
2784         (nds32_address_cost_impl): Simplify.
2785         (nds32_init_rtx_costs): New function.
2786         (nds32_rtx_costs_speed_prefer): Likewise.
2787         (nds32_rtx_costs_size_prefer): Likewise.
2788         (nds32_address_cost_speed_prefer): Likewise.
2789         (nds32_address_cost_speed_fwprop): Likewise.
2790         (nds32_address_cost_size_prefer): Likewise.
2791         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2792         * config/nds32/nds32.c (nds32_option_override): Use
2793         nds32_init_rtx_costs function.
2794
2795 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
2796
2797         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2798         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2799         (TARGET_PIPELINE_N8): Likewise.
2800         (TARGET_PIPELINE_N10): Likewise.
2801         (TARGET_PIPELINE_N13): Likewise.
2802         (TARGET_PIPELINE_GRAYWOLF): Likewise.
2803
2804 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
2805
2806         * config/nds32/nds32-fpu.md: Update copyright year.
2807
2808 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2809
2810         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2811
2812 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2813
2814         * config/nds32/nds32.c
2815         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2816         * config/nds32/nds32.opt (minline-asm-r15): New option.
2817
2818 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2819
2820         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2821         MASK_HW_ABS.
2822         * config/nds32/nds32.md (abssi2): New pattern.
2823
2824 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
2825
2826         * config/i386/i386.md (rex64namesuffix): New mode attribute.
2827         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2828         Merge insn pattern from sse_cvtsi2ss<round_name> and
2829         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2830         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2831         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2832         using SWI48 mode iterator.
2833         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2834         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2835         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2836         pattern from sse_cvttss2si<round_saeonly_name>
2837         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2838         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2839         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2840         using SWI48 mode iterator.
2841         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2842         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2843         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2844         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2845         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2846         using SWI48 mode iterator.
2847         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
2848         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
2849         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
2850         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
2851         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
2852         SWI48 mode iterator.
2853         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
2854         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
2855         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
2856         pattern from sse_cvttsd2si<round_saeonly_name>
2857         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
2858
2859 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2860
2861         * config/nds32/nds32-md-auxiliary.c
2862         (nds32_valid_smw_lwm_base_p): Refine.
2863         (nds32_output_smw_single_word): Refine.
2864         (nds32_output_smw_double_word): New.
2865         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
2866
2867 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2868
2869         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
2870         (nds32_output_stack_pop): Refine.
2871         (nds32_expand_unaligned_load): Refine.
2872         (nds32_expand_unaligned_store): Refine.
2873
2874 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
2875             Chung-Ju Wu  <jasonwucj@gmail.com>
2876
2877         * config/nds32/constants.md: Add TP_REGNUM constant.
2878         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
2879         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
2880         UNSPEC_ADD32.
2881         * config/nds32/nds32-doubleword.md: Consider flag_pic.
2882         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
2883         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
2884         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
2885         and PIC code generation.
2886         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
2887         code generation.
2888         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
2889         optimization.
2890         * config/nds32/nds32.md: Support TLS and PIC.
2891         * config/nds32/nds32.c: Support TLS and PIC.
2892         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
2893         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
2894         predicate.
2895
2896 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2897
2898         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
2899         mode with E_ prefix.
2900
2901 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
2902             Chung-Ju Wu  <jasonwucj@gmail.com>
2903
2904         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
2905         * config/nds32/nds32-md-auxiliary.c
2906         (symbolic_reference_mentioned_p): New.
2907         (nds32_legitimize_ict_address): New.
2908         (nds32_expand_ict_move): New.
2909         (nds32_indirect_call_referenced_p): New.
2910         (nds32_symbol_binds_local_p): Delete.
2911         (nds32_long_call_p): Modify.
2912         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
2913         * config/nds32/nds32-protos.h
2914         (symbolic_reference_mentioned_p): Declare.
2915         (nds32_legitimize_ict_address): Declare.
2916         (nds32_expand_ict_move): Declare.
2917         (nds32_indirect_call_referenced_p): Declare.
2918         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2919         (nds32_relax_group): Use nds32_ict_const_p as condition.
2920         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2921         (nds32_asm_file_start): Output ict_model directive in asm code.
2922         (nds32_legitimate_address_p): Consider indirect call.
2923         (nds32_print_operand): Consider indirect call.
2924         (nds32_print_operand_address): Consider indirect call.
2925         (nds32_insert_attributes): Handle "indirect_call" attribute.
2926         (TARGET_LEGITIMATE_ADDRESS_P): Define.
2927         (TARGET_LEGITIMATE_CONSTANT_P): Define.
2928         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2929         (TARGET_DELEGITIMIZE_ADDRESS): Define.
2930         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2931         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2932         (TARGET_ICT_MODEL_SMALL): Define.
2933         (TARGET_ICT_MODEL_LARGE): Define.
2934         * config/nds32/nds32.md (movsi): Consider ict model.
2935         (call, call_value): Consider ict model.
2936         (sibcall, sibcall_value): Consider ict model.
2937         * config/nds32/nds32.opt (mict-model): New option.
2938         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2939         model.
2940
2941 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
2942             Monk Chiang  <sh.chiang04@gmail.com>
2943             Jim Wilson <jimw@sifive.com>
2944
2945         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2946         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
2947         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2948         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2949         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
2950         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2951         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2952         compute save_libcall_adjustment properly.
2953         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2954         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2955         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2956         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2957         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2958         (ABI_SPEC): Handle mabi=ilp32e.
2959         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2960         (RVE): Add RVE mask.
2961         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2962         <-march>: Add rv32e as an example.
2963
2964 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
2965
2966         PR c++/82899
2967         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2968         (intra_create_variable_infos): Handle C++ constructors.
2969
2970 2018-05-18  Martin Liska  <mliska@suse.cz>
2971
2972         * passes.def: Remove a redundant pass.
2973
2974 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
2975
2976         PR bootstrap/85838
2977         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2978
2979 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2980
2981         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2982         (ARMv4): Update.
2983         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2984         (ARMv6m): Update.
2985         (armv2, armv2a, armv3, armv3m): Delete architectures.
2986         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2987         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2988         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2989         Delete cpus.
2990         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2991         (*mulsidi3adddi): Likewise.
2992         (mulsidi3): Likewise.
2993         (*mulsidi3_nov6): Likewise.
2994         (umulsidi3): Likewise.
2995         (umulsidi3_nov6): Likewise.
2996         (umaddsidi4): Likewise.
2997         (*umulsidi3adddi): Likewise.
2998         (smulsi3_highpart): Likewise.
2999         (*smulsi3_highpart_nov6): Likewise.
3000         (umulsi3_highpart): Likewise.
3001         (*umulsi3_highpart_nov6): Likewise.
3002         * config/arm/arm.h (arm_arch3m): Delete.
3003         * config/arm/arm.c (arm_arch3m): Delete.
3004         (arm_option_override_internal): Update armv3-related comment.
3005         (arm_configure_build_target): Delete use of isa_bit_mode32.
3006         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3007         (arm_rtx_costs_internal): Delete check of arm_arch3m.
3008         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3009         (mulsa3): Likewise.
3010         (mulusa3): Likewise.
3011         * config/arm/arm-protos.h (arm_arch3m): Delete.
3012         * config/arm/arm-tables.opt: Regenerate.
3013         * config/arm/arm-tune.md: Likewise.
3014         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3015         deleted architectures.
3016
3017 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3018
3019         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3020         (armv5t, armv5te): New features.
3021         (ARMv5, ARMv5e): Delete fgroups.
3022         (ARMv5t, ARMv5te): Adjust for above changes.
3023         (ARMv6m): Likewise.
3024         (armv5, armv5e): Delete arches.
3025         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3026         arm_arch5.
3027         (*call_reg_arm): Likewise.
3028         (*call_value_reg_armv5): Likewise.
3029         (*call_value_reg_arm): Likewise.
3030         (*call_symbol): Likewise.
3031         (*call_value_symbol): Likewise.
3032         (*sibcall_insn): Likewise.
3033         (*sibcall_value_insn): Likewise.
3034         (clzsi2): Likewise.
3035         (prefetch): Likewise.
3036         (define_split and define_peephole2 dependent on arm_arch5):
3037         Likewise.
3038         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3039         arm_arch5e.
3040         (TARGET_ARM_QBIT): Likewise.
3041         (TARGET_DSP_MULTIPLY): Likewise.
3042         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3043         (arm_arch5, arm_arch5e): Delete.
3044         (arm_arch5t, arm_arch5te): Declare.
3045         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3046         (arm_arch5t): Declare.
3047         (arm_option_reconfigure_globals): Update for the above.
3048         (arm_options_perform_arch_sanity_checks): Update comment, replace
3049         use of arm_arch5 with arm_arch5t.
3050         (use_return_insn): Likewise.
3051         (arm_emit_call_insn): Likewise.
3052         (output_return_instruction): Likewise.
3053         (arm_final_prescan_insn): Likewise.
3054         (arm_coproc_builtin_available): Likewise.
3055         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3056         arm_arch5e with arm_arch5t and arm_arch5te.
3057         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3058         (arm_arch5t, arm_arch5te): Declare.
3059         * config/arm/arm-tables.opt: Regenerate.
3060         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3061         * config/arm/t-multilib: Likewise.
3062         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3063         instead of arm_arch5.
3064         (*call_reg_thumb1): Likewise.
3065         (*call_value_reg_thumb1_v5): Likewise.
3066         (*call_value_reg_thumb1): Likewise.
3067         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3068         unreachable path.
3069         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3070
3071 2018-05-18  Martin Liska  <mliska@suse.cz>
3072
3073         PR gcov-profile/84846
3074         * doc/gcov.texi: Document -t option of gcov tool.
3075
3076 2018-05-18  Martin Liska  <mliska@suse.cz>
3077
3078         PR gcov-profile/84846
3079         * gcov.c (print_usage): Add new -t option.
3080         (process_args): Handle the option.
3081         (generate_results): Use stdout as output when requested by
3082         the option.
3083
3084 2018-05-18  Martin Liska  <mliska@suse.cz>
3085
3086         PR gcov-profile/84846
3087         * coverage.c (coverage_init): Write PWD to .gcno file.
3088         * doc/gcov.texi: Document how working directory is printed.
3089         * gcov-dump.c (dump_gcov_file): Print PWD.
3090         * gcov.c (output_intermediate_file): Likewise.
3091         (read_graph_file): Read PWD string.
3092         (output_lines): Print PWD.
3093
3094 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3095
3096         PR middle-end/85817
3097         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3098         for retval and return false if all args to phi are zero.
3099
3100 2018-05-18  Richard Biener  <rguenther@suse.de>
3101
3102         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3103         method.
3104         (evrp_dom_walker::before_dom_children): Call it.
3105
3106 2018-05-18  Richard Biener  <rguenther@suse.de>
3107
3108         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3109         results when processing array refs with variable index.
3110
3111 2018-05-18  Toon Moene  <toon@moene.org>
3112
3113         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3114         directly after that of -floop-interchange. Indicate that both
3115         options are enabled by default when specifying -O3.
3116
3117 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3118
3119         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3120         iterator.  Delete separate integer-mode vec_set<mode> expander.
3121         (aarch64_simd_vec_setv2di): Delete.
3122         (vec_setv2di): Delete.
3123         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3124         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
3125         the "w, r" alternative.
3126
3127 2018-05-18  Martin Liska  <mliska@suse.cz>
3128
3129         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3130         * tree-pass.h (make_pass_lower_switch_O0): New function.
3131         * tree-switch-conversion.c (node_has_low_bound): Remove.
3132         (node_has_high_bound): Likewise.
3133         (node_is_bounded): Likewise.
3134         (class pass_lower_switch): Make it a template type and create
3135         two instances.
3136         (pass_lower_switch::execute): Add template argument.
3137         (make_pass_lower_switch): New function.
3138         (make_pass_lower_switch_O0): New function.
3139         (do_jump_if_equal): Remove.
3140         (emit_case_nodes): Simplify to just handle all 3 cases and leave
3141         all the hard work to tree optimization passes.
3142
3143 2018-05-18  Martin Liska  <mliska@suse.cz>
3144
3145         * dbgcnt.c (limit_low): Renamed from limit.
3146         (limit_high): New variable.
3147         (dbg_cnt_is_enabled): Check for upper limit.
3148         (dbg_cnt): Adjust dumping.
3149         (dbg_cnt_set_limit_by_index): Add new argument for high
3150         value.
3151         (dbg_cnt_set_limit_by_name): Likewise.
3152         (dbg_cnt_process_single_pair): Parse new format.
3153         (dbg_cnt_process_opt): Use strtok.
3154         (dbg_cnt_list_all_counters): Remove 'value' and add
3155         'limit_high'.
3156         * doc/invoke.texi: Document changes.
3157
3158 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
3159
3160         * doc/sourcebuild.texi (scalar_all_fma): Document.
3161         * tree.def (FMA_EXPR): Delete.
3162         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3163         * internal-fn.c (ternary_direct): New macro.
3164         (expand_ternary_optab_fn): Likewise.
3165         (direct_ternary_optab_supported_p): Likewise.
3166         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3167         * builtins.c (fold_builtin_fma): Delete.
3168         (fold_builtin_3): Don't call it.
3169         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3170         * expr.c (expand_expr_real_2): Likewise.
3171         * fold-const.c (operand_equal_p): Likewise.
3172         (fold_ternary_loc): Likewise.
3173         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3174         * gimple.c (DEFTREECODE): Likewise.
3175         * gimplify.c (gimplify_expr): Likewise.
3176         * optabs-tree.c (optab_for_tree_code): Likewise.
3177         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3178         * tree-eh.c (operation_could_trap_p): Likewise.
3179         (stmt_could_throw_1_p): Likewise.
3180         * tree-inline.c (estimate_operator_cost): Likewise.
3181         * tree-pretty-print.c (dump_generic_node): Likewise.
3182         (op_code_prio): Likewise.
3183         * tree-ssa-loop-im.c (stmt_cost): Likewise.
3184         * tree-ssa-operands.c (get_expr_operands): Likewise.
3185         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3186         * fold-const-call.h (fold_fma): Delete.
3187         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3188         CFN_FNMA and CFN_FNMS.
3189         (fold_fma): Delete.
3190         * genmatch.c (combined_fn): New enum.
3191         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3192         (commutative_op): New function.
3193         (commutate): Use it.  Handle more than 2 operands.
3194         (dt_operand::gen_gimple_expr): Use commutative_op.
3195         (parser::parse_expr): Allow :c to be used with non-binary
3196         operators if the commutative operand is known.
3197         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3198         CFN_FMS, CFN_FNMA and CFN_FNMS.
3199         (backprop::process_assign_use): Remove FMA_EXPR handling.
3200         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3201         (gen_hsa_fma): New function.
3202         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3203         IFN_FNMA and IFN_FNMS.
3204         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3205         * gimple-fold.h (follow_all_ssa_edges): Declare.
3206         * gimple-fold.c (follow_all_ssa_edges): New function.
3207         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3208         gimple_build interface and use follow_all_ssa_edges to fold the result.
3209         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3210         instead of checking for optabs directly.
3211         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3212         rather than FMA_EXPRs.
3213         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3214         call to IFN_FMA instead of an FMA_EXPR.
3215
3216 2018-05-17  Jim Wilson  <jimw@sifive.com>
3217
3218         * expr.c (do_tablejump): When converting index to Pmode, if we have a
3219         sign extended promoted subreg, and the range does not have the sign bit
3220         set, then do a sign extend.
3221
3222         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3223         test, check for sign extended subreg and/or constant operands, and
3224         do a sign extend in that case.
3225
3226 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
3227
3228         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3229         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3230         Add untyped.
3231         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3232         Change logics_shift_reg to logics_shift_imm.
3233         (thunderx2t99_fp_loadpair_basic): Delete.
3234         (thunderx2t99_fp_storepair_basic): Delete.
3235         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3236         (thunderx2t99_asimd_polynomial): Delete.
3237         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3238         and neon_fp_mul_d_scalar_q.
3239         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3240         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3241         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3242         (thunderx2t99_asimd_lut): Add missing tbl types.
3243         (thunderx2t99_asimd_ext): Delete.
3244         (thunderx2t99_asimd_load1_1_mult): Delete.
3245         (thunderx2t99_asimd_load1_2_mult): Delete.
3246         (thunderx2t99_asimd_load1_ldp): New.
3247         (thunderx2t99_asimd_load1): New.
3248         (thunderx2t99_asimd_load2): Add missing *load2* types.
3249         (thunderx2t99_asimd_load3): New.
3250         (thunderx2t99_asimd_load4): New.
3251         (thunderx2t99_asimd_store1_1_mult): Delete.
3252         (thunderx2t99_asimd_store1_2_mult): Delete.
3253         (thunderx2t99_asimd_store2_mult): Delete.
3254         (thunderx2t99_asimd_store2_onelane): Delete.
3255         (thunderx2t99_asimd_store_stp): New.
3256         (thunderx2t99_asimd_store1): New.
3257         (thunderx2t99_asimd_store2): New.
3258         (thunderx2t99_asimd_store3): New.
3259         (thunderx2t99_asimd_store4): New.
3260
3261 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
3262
3263         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3264         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
3265
3266 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
3267             Segher Boessenkool  <segher@kernel.crashing.org>
3268
3269         PR target/85698
3270         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3271         operand.
3272
3273 2018-05-17  Richard Biener  <rguenther@suse.de>
3274
3275         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3276         for pruning loop and prune defs feeding only already visited PHIs.
3277
3278 2018-05-17  Richard Biener  <rguenther@suse.de>
3279
3280         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3281
3282 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
3283             Richard Biener  <rguenther@suse.de>
3284
3285         PR tree-optimization/85793
3286         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3287         for VMAT_ELEMENTWISE.
3288
3289 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
3290
3291         * internal-fn.h (lookup_internal_fn): Declare
3292         * internal-fn.c (lookup_internal_fn): New function.
3293         * gimple.c (gimple_build_call_from_tree): Handle calls to
3294         internal functions.
3295         * gimple-pretty-print.c (dump_gimple_call): Print "." before
3296         internal function names.
3297         * tree-pretty-print.c (dump_generic_node): Likewise.
3298         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3299
3300 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
3301
3302         * gimple-fold.h (gimple_build): Make the function forms take
3303         combined_fn rather than built_in_function.
3304         (gimple_simplify): Likewise.
3305         * gimple-match-head.c (gimple_simplify): Likewise.
3306         * gimple-fold.c (gimple_build): Likewise.
3307         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3308         rather than gimple_build_call_internal.
3309         (get_initial_defs_for_reduction): Likewise.
3310         (vect_create_epilog_for_reduction): Likewise.
3311         (vectorizable_live_operation): Likewise.
3312
3313 2018-05-17  Martin Liska  <mliska@suse.cz>
3314
3315         * gimple-ssa-sprintf.c (format_directive): Do not use
3316         space in between 'G_' and '('.
3317
3318 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
3319
3320         PR target/85323
3321         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3322         even if the mask is not all ones.
3323
3324         PR target/85323
3325         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3326         vector.
3327         (ix86_gimple_fold_builtin): Likewise.
3328
3329         PR target/85323
3330         * config/i386/i386.c: Include tree-vector-builder.h.
3331         (ix86_vector_shift_count): New function.
3332         (ix86_fold_builtin): Fold shift builtins by scalar count.
3333         (ix86_gimple_fold_builtin): Likewise.
3334
3335         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3336         _mm512_setzero): New intrinsics.
3337
3338 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
3339             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3340
3341         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3342         code generation for cases where splatting a value is not useful.
3343         * simplify-rtx.c (simplify_ternary_operation): Simplify
3344         vec_merge across a vec_duplicate and a paradoxical subreg forming
3345         a vector mode to a vec_concat.
3346
3347 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
3348
3349         * config.gcc: Support "goldmont-plus".
3350         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3351         "goldmont-plus".
3352         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3353         PROCESSOR_GOLDMONT_PLUS.
3354         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3355         (processor_target_table): Add "goldmont-plus".
3356         (PTA_GOLDMONT_PLUS): Define.
3357         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3358         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3359         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3360         (fold_builtin_cpu): Add "goldmont-plus".
3361         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3362         (ix86_option_override_internal): Add "goldmont-plus".
3363         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3364         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3365         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3366         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3367
3368 2018-05-17  Richard Biener  <rguenther@suse.de>
3369
3370         PR tree-optimization/85757
3371         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3372         remove defs that only feed that PHI from further processing.
3373
3374 2018-05-16  Jim Wilson  <jimw@sifive.com>
3375
3376         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3377         asterisk to name.
3378         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3379
3380 2018-05-16  Mark Wielaard  <mark@klomp.org>
3381
3382         * dwarf2out.c (count_index_strings): New function.
3383         (output_indirect_strings): Call count_index_strings and generate
3384         header for dwarf_version >= 5.
3385
3386 2018-05-16  Mark Wielaard  <mark@klomp.org>
3387
3388         * dwarf2out.c (dwarf_FORM): New function.
3389         (set_indirect_string): Use dwarf_FORM.
3390         (reset_indirect_string): Likewise.
3391         (size_of_die): Likewise.
3392         (value_format): Likewise.
3393         (output_die): Likewise.
3394         (add_skeleton_AT_string): Likewise.
3395         (output_macinfo_op): Likewise.
3396         (index_string): Likewise.
3397         (output_index_string_offset): Likewise.
3398         (output_index_string): Likewise.
3399         (count_index_strings): Likewise.
3400
3401 2018-05-16  Carl Love  <cel@us.ibm.com>
3402
3403         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3404         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3405
3406 2018-05-16  Martin Jambor  <mjambor@suse.cz>
3407
3408         * ipa-prop.c (ipa_free_all_edge_args): Remove.
3409         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3410
3411 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
3412
3413         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3414         (fnma<mode>4): Likewise.
3415         (fms<mode>4): Likewise.
3416         (fnms<mode>4): Likewise.
3417         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3418         (aarch64_fnma<mode>4): Likewise.
3419         (aarch64_fms<mode>4): Likewise.
3420         (aarch64_fnms<mode>4): Likewise.
3421         (aarch64_fnmadd<mode>4): Likewise.
3422
3423 2018-05-16  Jason Merrill  <jason@redhat.com>
3424
3425         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
3426
3427 2018-05-16  Richard Biener  <rguenther@suse.de>
3428
3429         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3430         (dump_stmt_cost): Declare.
3431         (add_stmt_cost): Dump cost we add.
3432         (add_stmt_costs): New function.
3433         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3434         No longer exported.
3435         (vect_analyze_stmt): Adjust prototype.
3436         (vectorizable_condition): Likewise.
3437         (vectorizable_live_operation): Likewise.
3438         (vectorizable_reduction): Likewise.
3439         (vectorizable_induction): Likewise.
3440         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3441         cost vector to pass to vectorizable_ and record afterwards.
3442         (vect_model_reduction_cost): Take cost vector argument and adjust.
3443         (vect_model_induction_cost): Likewise.
3444         (vectorizable_reduction): Likewise.
3445         (vectorizable_induction): Likewise.
3446         (vectorizable_live_operation): Likewise.
3447         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3448         SLP_TREE_NUMBER_OF_VEC_STMTS.
3449         (vect_analyze_slp_cost_1): Remove.
3450         (vect_analyze_slp_cost): Likewise.
3451         (vect_slp_analyze_node_operations): Take visited args and
3452         a target cost vector.  Avoid processing already visited stmt sets.
3453         (vect_slp_analyze_operations): Use a local cost vector to gather
3454         costs and register those of non-discarded instances.
3455         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3456         (vect_schedule_slp_instance): Remove copying of
3457         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
3458         zero.
3459         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3460         adding cost.  Record cost entry location.
3461         (vect_prologue_cost_for_slp_op): Function to compute cost of
3462         a constant or invariant generated for SLP vect in the prologue,
3463         split out from vect_analyze_slp_cost_1.
3464         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
3465         (vect_model_promotion_demotion_cost): Likewise.
3466         (vect_model_store_cost): Likewise, make static.
3467         (vect_model_load_cost): Likewise.
3468         (vectorizable_bswap): Add cost vector arg and adjust.
3469         (vectorizable_call): Likewise.
3470         (vectorizable_simd_clone_call): Likewise.
3471         (vectorizable_conversion): Likewise.
3472         (vectorizable_assignment): Likewise.
3473         (vectorizable_shift): Likewise.
3474         (vectorizable_operation): Likewise.
3475         (vectorizable_store): Likewise.
3476         (vectorizable_load): Likewise.
3477         (vectorizable_condition): Likewise.
3478         (vectorizable_comparison): Likewise.
3479         (can_vectorize_live_stmts): Likewise.
3480         (vect_analyze_stmt): Likewise.
3481         (vect_transform_stmt): Adjust calls to vectorizable_*.
3482         * tree-vectorizer.c: Include gimple-pretty-print.h.
3483         (dump_stmt_cost): New function.
3484
3485 2018-05-16  Richard Biener  <rguenther@suse.de>
3486
3487         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3488         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3489         * tree-ssa-dse.c: Include tree-ssa-loop.h.
3490         (check_name): New callback.
3491         (dse_classify_store): Track cycles via a visited bitmap of PHI
3492         defs and simplify handling of in-loop and across loop dead stores
3493         and properly fail for loop-variant refs.  Handle byte-tracking with
3494         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3495         limiting the walk.
3496
3497 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
3498
3499         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3500         (vect_get_mask_type_for_stmt): Likewise.
3501         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3502         split out from...
3503         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
3504         to determine the statement's vector type and the vector type that
3505         should be used for calculating nunits.  Deal with cases in which
3506         the type has to be deferred.
3507         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3508         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3509         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3510         (vect_determine_vf_for_stmt): New functions, split out from...
3511         (vect_determine_vectorization_factor): ...here.
3512         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3513         (vect_get_mask_type_for_stmt): New functions, split out from
3514         vect_determine_vectorization_factor.
3515
3516 2018-05-16  Richard Biener  <rguenther@suse.de>
3517
3518         * tree-cfg.c (verify_gimple_assign_ternary): Properly
3519         verify the [VEC_]COND_EXPR embedded comparison.
3520
3521 2018-05-15  Martin Sebor  <msebor@redhat.com>
3522
3523         PR tree-optimization/85753
3524         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3525         RECORD_TYPE in addition to ARRAY_TYPE.
3526
3527 2018-05-15  Martin Sebor  <msebor@redhat.com>
3528
3529         PR middle-end/85643
3530         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3531
3532 2018-05-15  Richard Biener  <rguenther@suse.de>
3533
3534         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3535         add by_clobber_p one.  Change algorithm to collect all defs
3536         representing uses we need to walk and try reducing them to
3537         a single one before failing.
3538         (dse_dom_walker::dse_optimize_stmt): Adjust.
3539
3540 2018-05-13  Mark Wielaard  <mark@klomp.org>
3541
3542         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3543         (size_of_loc_descr): Likewise.
3544         (output_loc_operands): Likewise.
3545         (output_loc_operands_raw): Likewise.
3546         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3547         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3548         (hash_loc_operands): Likewise.
3549         (compare_loc_operands): Likewise.
3550
3551 2018-05-14  Mark Wielaard  <mark@klomp.org>
3552
3553         * dwarf2out.c (count_index_addrs): New function.
3554         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3555
3556 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3557
3558         PR tree-optimization/83648
3559         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3560         return value as malloc candidate.
3561
3562 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3563
3564         PR ipa/85734
3565         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3566         param as true in call to suggest_attribute.
3567
3568 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
3569
3570         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3571         -mreadonly-in-sdata.
3572
3573 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3574
3575         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3576         New pattern.
3577         (aarch64_crypto_aesd_fused): Likewise.
3578
3579 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
3580
3581         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3582         (movsi_aarch64): Likewise.
3583         (load_pairsi): Likewise.
3584         (load_pairdi): Likewise.
3585         (store_pairsi): Likewise.
3586         (store_pairdi): Likewise.
3587         (load_pairsf): Likewise.
3588         (load_pairdf): Likewise.
3589         (store_pairsf): Likewise.
3590         (store_pairdf): Likewise.
3591         (zero_extend): Likewise.
3592         (trunc): Swap alternatives.
3593         (fcvt_target): Add '?' to prefer w over r.
3594
3595 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
3596
3597         PR target/85756
3598         * config/i386/i386.md: Disallow non-commutative arithmetics in
3599         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3600         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
3601         in the peephole2 before it.
3602
3603 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
3604
3605         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3606         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3607         (ix86_handle_option): Handle -mcldemote.
3608         * config.gcc: New header.
3609         * config/i386/cldemoteintrin.h: New file.
3610         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3611         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3612         -mcldemote.
3613         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3614         OPTION_MASK_ISA_CLDEMOTE.
3615         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3616         (ix86_valid_target_attribute_inner_p): Ditto.
3617         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3618         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3619         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3620         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3621         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3622         (cldemote): New.
3623         * config/i386/i386.opt: Add -mcldemote.
3624         * config/i386/x86intrin.h: New header.
3625         * doc/invoke.texi: Add -mcldemote.
3626
3627 2018-05-14  Richard Biener  <rguenther@suse.de>
3628
3629         * doc/match-and-simplify.texi: Adjust :s documentation.
3630
3631 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
3632
3633         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3634         intended memcpy size.
3635         (REORDER_45): Likewise.
3636
3637 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
3638
3639         * sort.cc: New file.
3640         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3641         * vec.c (qsort_chk): Use gcc_qsort.
3642         * Makefile.in (OBJS-libcommon): Add sort.o.
3643         (build/sort.o): New target.  Use it...
3644         (BUILD_RTL): ... here, and...
3645         (build/gencfn-macros): ... here, and...
3646         (build/genmatch): ... here.
3647
3648 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
3649             Chung-Ju Wu  <jasonwucj@gmail.com>
3650
3651         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3652         * config/nds32/nds32-graywolf.md: New file.
3653         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3654         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3655         pipeline.
3656         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3657         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3658         * config/nds32/nds32.md (pipeline_model): Add graywolf.
3659         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3660         * config/nds32/pipelines.md: Include n15 settings.
3661
3662 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
3663             Chung-Ju Wu  <jasonwucj@gmail.com>
3664
3665         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3666         * config/nds32/nds32-n13.md: New file.
3667         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3668         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3669         pipeline.
3670         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3671         * config/nds32/nds32.md (pipeline_model): Add n13.
3672         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3673         * config/nds32/pipelines.md: Include n13 settings.
3674
3675 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
3676             Chung-Ju Wu  <jasonwucj@gmail.com>
3677
3678         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3679         * config/nds32/nds32-n10.md: New file.
3680         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3681         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3682         pipeline.
3683         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3684         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3685         * config/nds32/nds32.md (pipeline_model): Add n10.
3686         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3687         * config/nds32/pipelines.md: Include n10 settings.
3688
3689 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
3690             Kito Cheng  <kito.cheng@gmail.com>
3691             Chung-Ju Wu  <jasonwucj@gmail.com>
3692
3693         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3694         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3695         Add enum values for DSP extension instructions.
3696         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3697         New constraints.
3698         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3699         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3700         New code iterators.
3701         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3702         * config/nds32/nds32-dspext.md: New file for DSP implementation.
3703         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3704         * config/nds32/nds32-intrinsic.md: Likewise.
3705         * config/nds32/nds32_intrinsic.h: Likewise.
3706         * config/nds32/nds32-md-auxiliary.c: Likewise.
3707         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3708         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3709         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3710         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3711         * config/nds32/nds32-protos.h: New declarations for DSP extension.
3712         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3713         TYPE_DMAC in switch statement.
3714         * config/nds32/nds32.c: New checking and implementation for DSP
3715         extension instructions.
3716         * config/nds32/nds32.h: Likewise.
3717         * config/nds32/nds32.md: Likewise.
3718         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3719         * config/nds32/predicates.md: Implement new predicates for DSP
3720         extension.
3721
3722 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
3723
3724         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3725         Reformat alternatives and attributes so it is easier to identify
3726         which constraints/attributes go with which instruction.
3727         (mov<mode>_hardfloat32, FMOVE64): Likewise.
3728         (mov<mode>_softfloat32, FMOVE64): Likewise.
3729         (mov<mode>_hardfloat64, FMOVE64): Likewise.
3730         (mov<mode>_softfloat64, FMOVE64): Likewise.
3731
3732 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3733
3734         * doc/extend.texi (PowerPC Built-in Functions): Rename this
3735         subsection.
3736         (Basic PowerPC Built-in Functions): The new name of the
3737         subsection previously known as "PowerPC Built-in Functions".
3738         (Basic PowerPC Built-in Functions Available on all Configurations):
3739         New subsubsection.
3740         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3741         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3742         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3743         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3744
3745 2018-05-11  Martin Jambor  <mjambor@suse.cz>
3746
3747         PR ipa/85655
3748         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3749         single const.
3750
3751 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
3752
3753         PR target/85733
3754         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3755
3756 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
3757
3758         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3759         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3760         (ix86_handle_option): Handle -mwaitpkg.
3761         * config.gcc: New header.
3762         * config/i386/cpuid.h (bit_WAITPKG): New bit.
3763         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3764         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3765         function type.
3766         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3767         OPTION_MASK_ISA_WAITPKG.
3768         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3769         (ix86_option_override_internal): Add PTA_WAITPKG.
3770         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3771         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3772         IX86_BUILTIN_TPAUSE.
3773         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3774         __builtin_ia32_umwait and __builtin_ia32_tpause.
3775         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3776         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3777         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3778         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3779         UNSPECV_TPAUSE): New.
3780         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3781         * config/i386/i386.opt: Add -mwaitpkg.
3782         * config/i386/waitpkgintrin.h: New file.
3783         * config/i386/x86intrin.h: New header.
3784         * doc/invoke.texi: Add -mwaitpkg.
3785
3786 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
3787
3788         PR target/85606
3789         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3790         equivalent.
3791         (cortex-m0): Use armv6s-m isa.
3792         (cortex-m0plus): Likewise.
3793         (cortex-m1): Likewise.
3794         (cortex-m0.small-multiply): Likewise.
3795         (cortex-m0plus.small-multiply): Likewise.
3796         (cortex-m1.small-multiply): Likewise.
3797
3798 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
3799             Jakub Jelinek  <jakub@redhat.com>
3800
3801         PR tree-optimization/85692
3802         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3803         source permute as well.
3804
3805 2018-05-11  Martin Liska  <mliska@suse.cz>
3806
3807         PR sanitizer/85556
3808         * doc/extend.texi: Document LLVM style format for no_sanitize
3809         attribute.
3810
3811 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
3812
3813         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3814         mode_supports_vsx_dform_quad to mode_supports_dq_form.
3815         (mode_supports_vsx_dform_quad): Likewise.
3816         (mode_supports_vmx_dform): Move these functions to be next to the
3817         other mode_supports functions.
3818         (mode_supports_dq_form): Likewise.
3819         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3820         mode_supports_dq_form.
3821         (reg_offset_addressing_ok_p): Likewise.
3822         (offsettable_ok_by_alignment): Likewise.
3823         (rs6000_legitimate_offset_address_p): Likewise.
3824         (legitimate_lo_sum_address_p): Likewise.
3825         (rs6000_legitimize_address): Likewise.
3826         (rs6000_legitimize_reload_address): Likewise.
3827         (rs6000_secondary_reload_inner): Likewise.
3828         (rs6000_preferred_reload_class): Likewise.
3829         (rs6000_output_move_128bit): Likewise.
3830
3831 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
3832
3833         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3834         Generate SImode target register for null target.
3835         <case IX86_BUILTIN_XGETBV>: Ditto.
3836         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3837         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3838
3839 2018-05-10  Carl Love  <cel@us.ibm.com>
3840
3841         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3842         dcbtt and dcbtstt if operands[2] is 0.
3843
3844 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
3845
3846         PR target/85693
3847         * config/i386/sse.md (usadv64qi): New expander.
3848
3849 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
3850
3851         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
3852         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
3853         -maltivec=be support.
3854         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
3855         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
3856         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
3857         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
3858         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
3859         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
3860         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
3861         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
3862         altivec_vsumsws): Adjust.
3863         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
3864         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
3865         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
3866         support.
3867         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
3868         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
3869         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
3870         (altivec_lve<VI_char>x): Delete expand.
3871         (*altivec_lve<VI_char>x_internal): Rename to...
3872         (altivec_lve<VI_char>x): ... this.
3873         (altivec_lvxl_<mode>): Delete expand.
3874         (*altivec_lvxl_<mode>_internal): Rename to ...
3875         (altivec_lvxl_<mode>): ... this.
3876         (altivec_stvxl_<mode>): Delete expand.
3877         (*altivec_stvxl_<mode>_internal): Rename to ...
3878         (altivec_stvxl_<mode>): ... this.
3879         (altivec_stve<VI_char>x): Delete expand.
3880         (*altivec_stve<VI_char>x_internal): Rename to ...
3881         (altivec_stve<VI_char>x): ... this.
3882         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
3883         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
3884         reduc_plus_scal_<mode>): Adjust.
3885         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
3886         comment.
3887         (rs6000_cpu_cpp_builtins): Adjust.
3888         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
3889         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
3890         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
3891         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
3892         -maltivec=be support.
3893         (rs6000_split_vec_extract_var): Adjust.
3894         (rs6000_split_v4si_init): Adjust.
3895         (swap_selector_for_mode): Delete.
3896         (altivec_expand_lvx_be, altivec_expand_stvx_be,
3897         altivec_expand_stvex_be): Delete.
3898         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
3899         -maltivec=be support.
3900         (rs6000_gimple_fold_builtin): Ditto.
3901         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
3902         Adjust.
3903         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
3904         (TARGET_DIRECT_MOVE_64BIT): Adjust.
3905         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
3906         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
3907         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
3908         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
3909         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
3910         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
3911         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
3912         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
3913         anonymous split): Adjust.
3914         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3915         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3916
3917 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
3918
3919         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3920         when --with-gxx-include-dir is also specified.
3921         * configure: Regenerate.
3922
3923 2018-05-09  Jim Wilson  <jimw@sifive.com>
3924
3925         PR target/84797
3926         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3927         * config/riscv/t-withmultilib: New.
3928         * config/riscv/withmultilib.h: New.
3929         * doc/install.texi: Document RISC-V --with-multilib-list support.
3930
3931 2018-05-09  Richard Biener  <rguenther@suse.de>
3932
3933         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3934         vector.
3935         (vect_bb_vectorization_profitable_p): Adjust.  Compute
3936         actual scalar cost using the cost vector and the add_stmt_cost
3937         machinery.
3938
3939 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
3940
3941         PR rtl-optimization/85645
3942         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3943         in the REG_CFA_REGISTER note for LR, don't leave it empty.
3944
3945 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
3946
3947         PR rtl-optimization/85645
3948         * shrink-wrap.c (spread_components): Return a boolean saying if
3949         anything was changed.
3950         (try_shrink_wrapping_separate): Iterate spread_components until
3951         nothing changes anymore.
3952
3953 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
3954
3955         PR rtl-optimization/85645
3956         * regrename.c (build_def_use): Also kill the chains that include the
3957         destination of a REG_CFA_REGISTER note.
3958
3959 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
3960
3961         PR rtl-optimization/85645
3962         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3963         insn that has a REG_CFA_REGISTER note.
3964
3965 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
3966
3967         * cfgexpand.c (expand_clobber): New function.
3968         (expand_gimple_stmt_1): Use it.
3969         * tree-vect-stmts.c (vect_clobber_variable): New function,
3970         split out from...
3971         (vectorizable_simd_clone_call): ...here.
3972         (vectorizable_store): Emit a clobber either side of an
3973         IFN_STORE_LANES sequence.
3974         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3975
3976 2018-05-09  Tom de Vries  <tom@codesourcery.com>
3977
3978         PR target/85626
3979         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3980         (define_insn "trap_if_false"): Add exit after trap.
3981
3982 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
3983
3984         PR rtl-optimization/85638
3985         * bb-reorder.c: Include common/common-target.h.
3986         (create_forwarder_block): New function extracted from...
3987         (fix_up_crossing_landing_pad): ...here.  Rename into...
3988         (dw2_fix_up_crossing_landing_pad): ...this.
3989         (sjlj_fix_up_crossing_landing_pad): New function.
3990         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3991         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3992         from both partitions and exit the loop after one iteration.
3993
3994 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3995
3996         Revert:
3997         * doc/extend.texi (PowerPC Built-in Functions): Rename this
3998         subsection.
3999         (Basic PowerPC Built-in Functions): The new name of the
4000         subsection previously known as "PowerPC Built-in Functions".
4001         (Basic PowerPC Built-in Functions Available on all Configurations):
4002         New subsubsection.
4003         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4004         subsubsection.
4005         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4006         subsubsection.
4007         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4008         subsubsection.
4009         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4010         subsubsection.
4011
4012 2018-05-08  Jim Wilson  <jimw@sifive.com>
4013
4014         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4015         (LD_EMUL_SUFFIX): New.
4016         (LINK_SPEC): Use it.
4017
4018 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4019
4020         * doc/extend.texi (PowerPC Built-in Functions): Rename this
4021         subsection.
4022         (Basic PowerPC Built-in Functions): The new name of the
4023         subsection previously known as "PowerPC Built-in Functions".
4024         (Basic PowerPC Built-in Functions Available on all Configurations):
4025         New subsubsection.
4026         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4027         subsubsection.
4028         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4029         subsubsection.
4030         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4031         subsubsection.
4032         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4033         subsubsection.
4034
4035 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
4036
4037         PR target/85683
4038         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4039         after cmpelim optimization.
4040
4041 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
4042
4043         * config.gcc: Support "goldmont".
4044         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4045         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4046         PROCESSOR_GOLDMONT.
4047         * config/i386/i386.c (m_GOLDMONT): Define.
4048         (processor_target_table): Add "goldmont".
4049         (PTA_GOLDMONT): Define.
4050         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4051         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4052         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4053         (fold_builtin_cpu): Add "goldmont".
4054         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4055         (ix86_option_override_internal): Add "goldmont".
4056         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4057         (processor_type): Add PROCESSOR_GOLDMONT.
4058         * config/i386/i386.md: Add CPU "glm".
4059         * config/i386/glm.md: New file.
4060         * config/i386/x86-tune.def: Add m_GOLDMONT.
4061         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4062
4063 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
4064
4065         PR target/85572
4066         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4067         E_V4DImode.
4068         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4069         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
4070         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
4071
4072         PR target/85317
4073         * config/i386/i386.c (ix86_fold_builtin): Handle
4074         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4075
4076         PR target/85480
4077         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4078         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4079
4080 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
4081
4082         PR target/85658
4083         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4084         (check_arch): Likewise.
4085         (check_fpu): Return the result rather than printing it.
4086         (end arch): Fix operator precedence.
4087         (end cpu): Likewise.
4088         (END): Print the result from check_fpu.
4089
4090 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
4091             Alan Hayward  <alan.hayward@arm.com>
4092             David Sherwood  <david.sherwood@arm.com>
4093
4094         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4095         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4096         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4097         (*fcmuo<mode>_and): New patterns.
4098
4099 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
4100
4101         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4102         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4103         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4104         (cmp_op, sve_imm_con): New code attributes.
4105         (SVE_COND_INT_CMP, imm_con): Delete.
4106         (cmp_op): Remove above unspecs from int attribute.
4107         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4108         to...
4109         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
4110         comparison-specific unspecs.
4111         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4112         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4113         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4114         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4115         (*vec_fcm<cmp_op><mode>): Rename to...
4116         (*fcm<cmp_op><mode>): ...this and adjust likewise.
4117         (*vec_fcmuo<mode>): Rename to...
4118         (*fcmuo<mode>): ...this and adjust likewise.
4119         (*pred_fcm<cmp_op><mode>): New pattern.
4120         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4121         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4122         functions.
4123         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4124         and UNORDERED.
4125         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4126         (aarch64_emit_sve_predicated_cond): New function.
4127         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4128         (aarch64_emit_unspec_cond_or): Replace with...
4129         (aarch64_emit_sve_or_conds): ...this new function.  Use
4130         aarch64_emit_sve_ptrue_op for the individual comparisons and
4131         aarch64_emit_binop to OR them together.
4132         (aarch64_emit_inverted_unspec_cond): Replace with...
4133         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
4134         aarch64_emit_sve_ptrue_op for the comparison and
4135         aarch64_emit_unop to invert the result.
4136         (aarch64_expand_sve_vec_cmp_float): Update after the above
4137         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
4138
4139 2018-05-07  Nathan Sidwell  <nathan@acm.org>
4140
4141         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4142         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4143         (Backwards Compatibility): Likewise.
4144
4145 2018-05-07  Luis Machado  <luis.machado@linaro.org>
4146
4147         PR bootstrap/85681
4148         Revert:
4149         2018-05-07  Luis Machado  <luis.machado@linaro.org>
4150
4151         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4152         <prefetch_dynamic_strides>: New const bool field.
4153         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4154         prefetch_dynamic_strides.
4155         (exynosm1_prefetch_tune): Likewise.
4156         (thunderxt88_prefetch_tune): Likewise.
4157         (thunderx_prefetch_tune): Likewise.
4158         (thunderx2t99_prefetch_tune): Likewise.
4159         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4160         to false.
4161         (aarch64_override_options_internal): Update to set
4162         PARAM_PREFETCH_DYNAMIC_STRIDES.
4163         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4164         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4165         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4166         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4167         prefetch-dynamic-strides setting.
4168
4169         2018-05-07  Luis Machado  <luis.machado@linaro.org>
4170
4171         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4172         <minimum_stride>: New const int field.
4173         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4174         minimum_stride field.
4175         (exynosm1_prefetch_tune): Likewise.
4176         (thunderxt88_prefetch_tune): Likewise.
4177         (thunderx_prefetch_tune): Likewise.
4178         (thunderx2t99_prefetch_tune): Likewise.
4179         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4180         (aarch64_override_options_internal): Update to set
4181         PARAM_PREFETCH_MINIMUM_STRIDE.
4182         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4183         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4184         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4185         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4186         stride is constant and is below the minimum stride threshold.
4187
4188 2018-05-07  Luis Machado  <luis.machado@linaro.org>
4189
4190         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4191         to 512.
4192
4193 2018-05-07  Luis Machado  <luis.machado@linaro.org>
4194
4195         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4196         <prefetch_dynamic_strides>: New const bool field.
4197         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4198         prefetch_dynamic_strides.
4199         (exynosm1_prefetch_tune): Likewise.
4200         (thunderxt88_prefetch_tune): Likewise.
4201         (thunderx_prefetch_tune): Likewise.
4202         (thunderx2t99_prefetch_tune): Likewise.
4203         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4204         to false.
4205         (aarch64_override_options_internal): Update to set
4206         PARAM_PREFETCH_DYNAMIC_STRIDES.
4207         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4208         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4209         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4210         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4211         prefetch-dynamic-strides setting.
4212
4213 2018-05-07  Luis Machado  <luis.machado@linaro.org>
4214
4215         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4216         <minimum_stride>: New const int field.
4217         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4218         minimum_stride field.
4219         (exynosm1_prefetch_tune): Likewise.
4220         (thunderxt88_prefetch_tune): Likewise.
4221         (thunderx_prefetch_tune): Likewise.
4222         (thunderx2t99_prefetch_tune): Likewise.
4223         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4224         (aarch64_override_options_internal): Update to set
4225         PARAM_PREFETCH_MINIMUM_STRIDE.
4226         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4227         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4228         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4229         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4230         stride is constant and is below the minimum stride threshold.
4231
4232 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
4233
4234         PR c++/85659
4235         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4236         the type is addressable.  Don't force op into register if it has
4237         BLKmode.
4238
4239 2018-05-05  Roland McGrath  <mcgrathr@google.com>
4240
4241         PR other/77609
4242         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4243         any section for which we don't know a specific type it should have,
4244         regardless of name.  Previously this was done only for the exact
4245         names ".init_array", ".fini_array", and ".preinit_array".
4246         (default_elf_asm_named_section): Add comment about
4247         relationship with default_section_type_flags and SECTION_NOTYPE.
4248         (get_section): Don't consider it a type conflict if one side has
4249         SECTION_NOTYPE and the other doesn't, as long as neither has the
4250         SECTION_BSS et al used in the default_section_type_flags logic.
4251
4252 2018-05-05  Tom de Vries  <tom@codesourcery.com>
4253
4254         PR target/85653
4255         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4256         (workaround_barsyncs): New function.
4257         (nvptx_reorg): Use workaround_barsyncs.
4258         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4259         (define_expand "nvptx_membar_cta"): New define_expand.
4260         (define_insn "*nvptx_membar_cta"): New insn.
4261
4262 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
4263
4264         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4265         To improve optimization opportunities.
4266         * builtin-types.def: The new needed builtin types for the above.
4267
4268 2018-05-04  Richard Biener  <rguenther@suse.de>
4269
4270         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4271         * gimple-ssa-store-merging.c
4272         (imm_store_chain_info::output_merged_store): Remove redundant create,
4273         release split_store vector contents on failure.
4274         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4275         scalar stmt vector on cache hit.
4276
4277 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
4278
4279         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4280         Xilinx FP support.
4281         * config.gcc (powerpc-xilinx-eabi*): Remove.
4282         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4283         support.
4284         (fusion_addis_mem_combo_load): Ditto.
4285         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4286         FP support.
4287         (rs6000_cpu_cpp_builtins): Ditto.
4288         * config/rs6000/rs6000-linux.c
4289         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4290         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4291         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4292         support.
4293         (rs6000_setup_reg_addr_masks): Ditto.
4294         (rs6000_init_hard_regno_mode_ok): Ditto.
4295         (rs6000_option_override_internal): Ditto.
4296         (legitimate_lo_sum_address_p): Ditto.
4297         (rs6000_legitimize_address): Ditto.
4298         (rs6000_legitimize_reload_address): Ditto.
4299         (rs6000_legitimate_address_p): Ditto.
4300         (abi_v4_pass_in_fpr): Ditto.
4301         (setup_incoming_varargs): Ditto.
4302         (rs6000_gimplify_va_arg): Ditto.
4303         (rs6000_split_multireg_move): Ditto.
4304         (rs6000_savres_strategy): Ditto.
4305         (rs6000_emit_prologue_components): Ditto.
4306         (rs6000_emit_epilogue_components): Ditto.
4307         (rs6000_emit_prologue): Ditto.
4308         (rs6000_emit_epilogue): Ditto.
4309         (rs6000_elf_file_end): Ditto.
4310         (rs6000_function_value): Ditto.
4311         (rs6000_libcall_value): Ditto.
4312         * config/rs6000/rs6000.h: Ditto.
4313         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4314         (TARGET_MINMAX): ... this.  New.
4315         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4316         * config/rs6000/rs6000.md: Remove Xilinx FP support.
4317         (*movsi_internal1_single): Delete.
4318         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4319         mfpu=, mxilinx-fpu): Delete.
4320         * config/rs6000/singlefp.h: Delete.
4321         * config/rs6000/sysv4.h: Remove Xilinx FP support.
4322         * config/rs6000/t-rs6000: Ditto.
4323         * config/rs6000/t-xilinx: Delete.
4324         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4325         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4326         (VStype_simple): Delete.
4327         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4328         * config/rs6000/xfpu.h: Delete.
4329         * config/rs6000/xfpu.md: Delete.
4330         * config/rs6000/xilinx.h: Delete.
4331         * config/rs6000/xilinx.opt: Delete.
4332         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4333         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4334
4335 2018-05-04  Tom de Vries  <tom@codesourcery.com>
4336
4337         PR libgomp/85639
4338         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4339         if ignore == 0.
4340
4341 2018-05-04  Richard Biener  <rguenther@suse.de>
4342
4343         PR middle-end/85627
4344         * tree-complex.c (update_complex_assignment): We are always in SSA form.
4345         (expand_complex_div_wide): Likewise.
4346         (expand_complex_operations_1): Likewise.
4347         (expand_complex_libcall): Preserve EH info of the original stmt.
4348         (tree_lower_complex): Handle removed blocks.
4349         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4350         on complex multiplication and division libcall builtins.
4351
4352 2018-05-04  Richard Biener  <rguenther@suse.de>
4353
4354         PR middle-end/85574
4355         * fold-const.c (negate_expr_p): Restrict negation of operand
4356         zero of a division to when we know that can happen without
4357         overflow.
4358         (fold_negate_expr_1): Likewise.
4359
4360 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
4361
4362         PR libstdc++/85466
4363         * real.h (real_nextafter): Declare.
4364         * real.c (real_nextafter): New function.
4365         * fold-const-call.c (fold_const_nextafter): New function.
4366         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4367         CASE_CFN_NEXTTOWARD.
4368         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4369         even when arg1_mode is different from arg0_mode.
4370
4371 2018-05-03  Nathan Sidwell  <nathan@acm.org>
4372
4373         * doc/extend.texi (Deprecated Features): Remove
4374         -ffriend-injection.
4375         (Backwards Compatibility): Likewise.
4376         * doc/invoke.texi (C++ Language Options): Likewise.
4377         (C++ Dialect Options): Likewise.
4378
4379 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
4380
4381         PR target/85530
4382         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4383         _mm512_mask_mullox_epi64): New intrinsics.
4384
4385 2018-05-03  Tom de Vries  <tom@codesourcery.com>
4386
4387         PR testsuite/85106
4388         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4389         dump files): Add offload-tree.
4390
4391 2018-05-03  Richard Biener  <rguenther@suse.de>
4392
4393         PR tree-optimization/85615
4394         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4395         to loops not nested in BBs loop father to avoid creating multi-entry
4396         loops.
4397
4398 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4399
4400         PR tree-optimization/70291
4401         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4402         arguments.  Change return type to tree.  Emit libcall as a new
4403         statement rather than replacing existing one when inplace_p is true.
4404         (expand_complex_multiplication_components): New function.
4405         (expand_complex_multiplication): Expand floating-point complex
4406         multiplication using the above.
4407         (expand_complex_division): Rename inner_type parameter to type.
4408         Update expand_complex_libcall call-site.
4409         (expand_complex_operations_1): Update expand_complex_multiplication
4410         and expand_complex_division call-sites.
4411
4412 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
4413
4414         PR target/85582
4415         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4416         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4417         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4418         the highest significant bit of the shift count mask is clear.  In
4419         check whether and[sq]i3 is needed verify that all significant bits
4420         of the shift count other than the highest are set.
4421
4422 2018-05-02  Tom de Vries  <tom@codesourcery.com>
4423
4424         PR libgomp/82428
4425         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4426         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4427         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4428         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4429         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4430         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4431         __builtin_goacc_parlevel_size.
4432
4433 2018-05-02  Richard Biener  <rguenther@suse.de>
4434
4435         PR tree-optimization/85597
4436         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4437         do not use split vect_get_vec_defs call but call vect_get_slp_defs
4438         directly.
4439
4440 2018-05-02  Tom de Vries  <tom@codesourcery.com>
4441
4442         PR testsuite/85106
4443         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4444         dump files): Add ltrans-tree.
4445
4446 2018-05-02  Tom de Vries  <tom@codesourcery.com>
4447
4448         PR testsuite/85106
4449         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4450         dump files): Add wpa-ipa.
4451
4452 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
4453
4454         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
4455         powerpc*-*-linux*paired* target.
4456         * config/rs6000/750cl.h: Delete.
4457         * config/rs6000/paired.h: Delete.
4458         * config/rs6000/paired.md: Delete.
4459         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4460         float support.
4461         * config/rs6000/rs6000-builtin.def: Remove paired float support.
4462         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4463         comment.  Remove paired float support.
4464         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4465         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4466         VECTOR_PAIRED.
4467         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4468         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4469         declarations.
4470         * config/rs6000/rs6000.c: Remove paired float support.
4471         (paired_expand_vector_init, paired_expand_vector_move,
4472         paired_emit_vector_compare, paired_emit_vector_cond_expr,
4473         (paired_expand_lv_builtin, paired_expand_stv_builtin,
4474         paired_expand_builtin, paired_expand_predicate_builtin,
4475         paired_init_builtins): Delete.
4476         * config/rs6000/rs6000.h: Remove paired float support.
4477         * config/rs6000/rs6000.md: Remove paired float support.
4478         (move_from_CR_ov_bit): Delete.
4479         * config/rs6000/rs6000.opt (mpaired): Delete.
4480         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4481         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4482
4483 2018-05-02  Richard Biener  <rguenther@suse.de>
4484
4485         PR middle-end/85567
4486         * gimplify.c (gimplify_save_expr): When in SSA form allow
4487         SAVE_EXPRs to compute to SSA vars.
4488
4489 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
4490
4491         PR target/85582
4492         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4493         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4494         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4495         clobber operands[2], instead use a new pseudo.  Formatting fixes.
4496
4497 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
4498
4499         PR tree-optimization/85586
4500         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4501         exit early for statements in the same group if the accesses are
4502         not strided.
4503
4504 2018-05-02  Tom de Vries  <tom@codesourcery.com>
4505
4506         PR lto/85451
4507         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4508         error message.
4509
4510 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
4511
4512         PR tree-optimization/85143
4513         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4514
4515 2018-05-01  Tom de Vries  <tom@codesourcery.com>
4516
4517         PR lto/85451
4518         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4519         not found" error message.
4520
4521 2018-05-01  Tom de Vries  <tom@codesourcery.com>
4522
4523         PR other/83786
4524         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4525         * vec.c (test_ordered_remove_if): New function.
4526         (vec_c_tests): Call test_ordered_remove_if.
4527         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4528         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4529         * tree-vect-patterns.c (vect_pattern_recog_1): Use
4530         VEC_ORDERED_REMOVE_IF.
4531
4532 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4533
4534         PR tree-optimization/82665
4535         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4536         pointer subtraction where arguments come from a memchr call.
4537
4538 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
4539
4540         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4541         --push-state --as-needed and --pop-state instead of --as-needed and
4542         --no-as-needed if ld supports it.
4543         * configure: Regenerated.
4544
4545         PR web/85578
4546         * doc/install.texi2html: Replace _002d with - and _002a with * in
4547         generated html files using sed.
4548
4549 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4550
4551         PR c++/85523
4552         * gcc-rich-location.c (blank_line_before_p): New function.
4553         (use_new_line): New function.
4554         (gcc_rich_location::add_fixit_insert_formatted): New function.
4555         * gcc-rich-location.h
4556         (gcc_rich_location::add_fixit_insert_formatted): New function.
4557
4558 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4559
4560         * selftest.c (assert_streq): Rename "expected" and "actual" to
4561         "val1" and "val2".  Extend NULL-handling to cover both inputs
4562         symmetrically, while still requiring both to be non-NULL for a pass.
4563         * selftest.h (assert_streq): Rename "expected" and "actual" to
4564         "val1" and "val2".
4565         (ASSERT_EQ): Likewise.
4566         (ASSERT_EQ_AT): Likewise.
4567         (ASSERT_KNOWN_EQ): Likewise.
4568         (ASSERT_KNOWN_EQ_AT): Likewise.
4569         (ASSERT_NE): Likewise.
4570         (ASSERT_MAYBE_NE): Likewise.
4571         (ASSERT_MAYBE_NE_AT): Likewise.
4572         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
4573         the assertion to pass.
4574         (ASSERT_STREQ_AT): Likewise.
4575
4576 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
4577
4578         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4579         interaction with -pie.
4580
4581 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4582
4583         * selftest.h: Fix alphabetization of per-source-file selftest
4584         declarations.
4585
4586 2018-04-30  Jason Merrill  <jason@redhat.com>
4587
4588         PR c++/61982 - dead stores to destroyed objects.
4589         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4590         of clobber.
4591
4592 2018-04-30  Jason Merrill  <jason@redhat.com>
4593
4594         * tree.c (build_clobber): New.
4595         * tree.h: Declare it.
4596         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4597
4598 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4599
4600         * diagnostic-show-locus.c (layout::layout): Update for
4601         location_get_source_line returning a char_span.
4602         (struct char_span): Move to input.h.
4603         (struct correction): Update for fields in char_span becoming
4604         private.
4605         (struct source_line): Update for location_get_source_line
4606         returning a char_span.
4607         (layout::print_line): Likewise.
4608         * edit-context.c (edited_file::print_content): Likewise.
4609         (edited_file::print_diff_hunk): Likewise.
4610         (edited_file::print_run_of_changed_lines): Likewise.
4611         (edited_file::get_num_lines): Likewise.
4612         (edited_line::edited_line): Likewise.
4613         * final.c (asm_show_source): Likewise.
4614         * input.c (location_get_source_line): Convert return type
4615         from const char * to char_span, losing the final "line_len"
4616         param.
4617         (dump_location_info): Update for the above.
4618         (get_substring_ranges_for_loc): Likewise.  Use a char_span
4619         when handling the literal within the line.
4620         (test_reading_source_line): Update for location_get_source_line
4621         returning a char_span.
4622         * input.h (class char_span): Move here from
4623         diagnostic-show-locus.c, converting from a struct to a class.
4624         Make data members private.
4625         (char_span::operator bool): New.
4626         (char_span::length): New.
4627         (char_span::get_buffer): New.
4628         (char_span::operator[]): New.
4629         (char_span::subspan): Make const.
4630         (char_span::xstrdup): New.
4631         (location_get_source_line): Convert return type from const char *
4632         to char_span, losing the final "line_size" param.
4633
4634 2018-04-30  Jan Hubicka  <jh@suse.cz>
4635
4636         * lto-wrapper.c (ltrans_priorities): New static var.
4637         (cmp_priority): New.
4638         (run_gcc): Read priorities and if doing parallel build order
4639         the Makefile by them.
4640
4641 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4642
4643         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4644
4645 2018-04-30  Richard Biener  <rguenther@suse.de>
4646
4647         * tree-cfg.c (verify_address): Remove base argument, add
4648         flag whether to check TREE_ADDRESSABLE and do that.
4649         (verify_expr): Remove.
4650         (verify_types_in_gimple_reference): Add pieces from verify_expr.
4651         (verify_gimple_assign_single): Likewise.
4652         (verify_gimple_switch): Likewise.
4653         (verify_expr_location_1): Dereference tp once.  Add (disabled)
4654         piece from verify_expr.
4655         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4656
4657 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
4658
4659         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4660
4661 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
4662
4663         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4664         (small_data_pattern): Likewise.
4665         (arc_rewrite_small_data): Likewise.
4666         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4667         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4668         (get_symbol_alignment): New function.
4669         (legitimate_small_data_address_p): Likewise.
4670         (legitimate_scaled_address): Update, call
4671         legitimate_small_data_address_p.
4672         (output_sdata): New static variable.
4673         (arc_print_operand): Update how we handle small data operands.
4674         (arc_print_operand_address): Likewise.
4675         (arc_legitimate_address_p): Update, use
4676         legitimate_small_data_address_p.
4677         (arc_rewrite_small_data_p): Remove.
4678         (arc_rewrite_small_data_1): Likewise.
4679         (arc_rewrite_small_data): Likewise.
4680         (small_data_pattern): Likewise.
4681         (compact_sda_memory_operand): Update to use
4682         legitimate_small_data_address_p and get_symbol_alignment.
4683         (prepare_move_operands): Don't rewite sdata pattern.
4684         (prepare_extend_operands): Remove.
4685         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4686         pattern.
4687         (zero_extendqisi2): Likewise.
4688         (zero_extendhisi2): Likewise.
4689         (extendqihi2): Likewise.
4690         (extendqisi2): Likewise.
4691         (extendhisi2): Likewise.
4692         (addsi3): Likewise.
4693         (subsi3): Likewise.
4694         (andsi3): Likewise.
4695         * config/arc/constraints.md (Usd): Change it to memory constraint.
4696
4697 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
4698
4699         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4700         as source of std instructions.
4701         * config/arc/arc.md (movsi_insn): Update pattern predicate to
4702         allow 6-bit constants as source for store instructions.
4703         (movdi_insn): Update instruction pattern to allow 6-bit constants
4704         as source for store instructions.
4705
4706 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
4707
4708         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4709
4710 2018-04-30  Nathan Sidwell  <nathan@acm.org>
4711             Sandra Loosemore <sandra@codesourcery.com>
4712
4713         * dumpfile.c (dump_open): Allow '-' for stdout.
4714         * doc/invoke.texi (Developer Options): Document dump filename
4715         determination early.  Document stdin/stdout selection.
4716
4717 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
4718
4719         Microblaze Target: PIC data text relative
4720
4721         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4722         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4723         Add declaration.
4724         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4725         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4726         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4727         New addressing mode for data-text relative position indepenedent code.
4728         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4729         'ADDRESS_SYMBOLIC_TXT_REL'.
4730         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4731         (microblaze_legitimate_pic_operand): Exclude function calls from
4732         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4733         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4734         addresses cases.
4735         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4736         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4737         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4738         for 'address + offset'.
4739         (microblaze_expand_prologue): Add new function prologue call for
4740         'r20' assignation.
4741         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4742         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4743         table in case of TARGET_PIC_DATA_TEXT_REL.
4744         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4745         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4746         Add new macros 'UNSPEC_TEXT',
4747         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4748         + exclude function calls from 'UNSPEC_PLT' in case of data text
4749         relative mode.
4750         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4751         new target hook for generating address diff vector tables in case of
4752         flag_pic.
4753         * doc/tm.texi : Regenerate.
4754         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4755         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4756         of addr diff vector generation.
4757         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4758         target hook definition.
4759         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4760         Add default function for generate_pic_addr_diff_vec -> flag_pic.
4761         * doc/invoke.texi (Add new pic option): Add new microblaze pic
4762         option for data text relative.
4763
4764 2018-04-30  Richard Biener  <rguenther@suse.de>
4765
4766         * tree-chrec.h (evolution_function_is_constant_p): Remove
4767         redundant check.
4768         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4769
4770 2018-04-30  Richard Biener  <rguenther@suse.de>
4771
4772         PR bootstrap/85571
4773         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4774
4775 2018-04-30  Richard Biener  <rguenther@suse.de>
4776
4777         PR tree-optimization/28364
4778         PR tree-optimization/85275
4779         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4780         copying first exit test.
4781
4782 2018-04-28  Mark Wielaard  <mark@klomp.org>
4783
4784         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4785         dwarf_version >= 5.
4786         (dwarf_AT): Handle DW_AT_addr_base.
4787         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4788
4789 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
4790
4791         PR target/84431
4792         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4793         (*ashl<dwi>3_doubleword_mask_1): Ditto.
4794         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4795         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4796
4797 2018-04-28  Richard Biener  <rguenther@suse.de>
4798
4799         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4800         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4801         to reflect use.  Only add interesting stmts.
4802
4803 2018-04-27  Martin Jambor  <mjambor@suse.cz>
4804
4805         PR ipa/85549
4806         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4807         the jump function allows for passing through aggregate values.
4808
4809 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
4810
4811         * input.h (in_system_header_at): Convert from macro to inline
4812         function.
4813         (from_macro_expansion_at): Likewise.
4814         (from_macro_definition_at): Likewise.
4815
4816 2018-04-27  Jeff Law  <law@redhat.com>
4817
4818         * config.gcc: Mark tile* targets as deprecated/obsolete.
4819
4820 2018-04-27  Richard Biener  <rguenther@suse.de>
4821
4822         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4823         fix for ILP32.
4824
4825 2018-04-27  Richard Biener  <rguenther@suse.de>
4826
4827         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4828
4829 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
4830
4831         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4832         with Yd constraint. Set "preferred_for_speed" attribute from
4833         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4834         with Yd constraint.
4835         (*movdi_internal): Ditto.
4836         (movti_interunit splitters): Remove
4837         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4838         (movdi_interunit splitters): Ditto.
4839         * config/i386/constraints.md (Ye): Remove.
4840         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4841
4842 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4843
4844         PR target/85512
4845         * config/aarch64/constraints.md (Usg): Limit to 31.
4846         (Usj): Limit to 63.
4847
4848 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
4849
4850         PR tree-optimization/85529
4851         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4852         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4853         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4854         zero extension or masking of the MSB bit.
4855         (optimize_range_tests): Add FIRST_BB argument, pass it through
4856         to optimize_range_tests_var_bound.
4857         (maybe_optimize_range_tests, reassociate_bb): Adjust
4858         optimize_range_tests callers.
4859
4860 2018-04-26  Richard Biener  <rguenther@suse.de>
4861             Jakub Jelinek  <jakub@redhat.com>
4862
4863         * cgraph.h (symbol_table): Just declare debug method here.
4864         * symtab.c (symbol_table::debug): Define.
4865
4866 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
4867
4868         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
4869
4870 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
4871
4872         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4873         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
4874         (*movdi_internal): Substitute Yi and Yj constraint with x
4875         and Ym and Yn constraint with y constraint.  Update "isa"
4876         attribute and set "preferred_for_speed" attribute from
4877         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
4878         (*movsi_internal): Ditto.
4879         (*movdf_internal): Ditto.
4880         (*movsf_internal): Ditto.
4881         (*zero_extendsidi2): Ditto.
4882         * config/i386/sse.md (vec_set<mode>_0): Ditto.
4883         (sse2_loadld): Ditto.
4884         (*vec_extract<ssevecmodelower>_0): Ditto.
4885         (*vec_extractv4si_0_zext_sse4): Ditto.
4886         (vec_concatv2di): Ditto.
4887         (*vec_dup<mode>): Ditto.
4888         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
4889         * config/i386/constraints.md (Yi): Remove.
4890         (Yj): Remove.
4891         (Ym): Remove.
4892         (Yn): Remove.
4893
4894 2018-04-26  Nathan Sidwell  <nathan@acm.org>
4895
4896         * dumpfile.c (dump_open): New.
4897         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
4898         (dump_finish): Detect stdio/stderr by value not name.
4899
4900 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
4901
4902         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4903
4904 2018-04-26  Tom de Vries  <tom@codesourcery.com>
4905
4906         PR target/84952
4907         * config/nvptx/nvptx.c (verify_neutering_jumps)
4908         (verify_neutering_labels): New function
4909         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
4910
4911 2018-04-26  Tom de Vries  <tom@codesourcery.com>
4912
4913         PR target/84025
4914         * config/nvptx/nvptx.c (needs_neutering_p): New function.
4915         (nvptx_single): Use needs_neutering_p to skip over insns that do not
4916         need neutering.
4917
4918 2018-04-26  Richard Biener <rguenther@suse.de>
4919             Tom de Vries  <tom@codesourcery.com>
4920
4921         PR lto/85422
4922         * lto-streamer-out.c (output_function): Fixup loops if required to match
4923         discovery done in the reader.
4924
4925 2018-04-26  Richard Biener  <rguenther@suse.de>
4926
4927         PR tree-optimization/85116
4928         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4929         have a loop exit from the single latch predecessor.  Remove
4930         case of header with just condition.
4931         (ch_base::copy_headers): Exclude infinite loops from any
4932         processing.
4933         (pass_ch::execute): Record exits.
4934
4935 2018-04-26  Richard Biener  <rguenther@suse.de>
4936
4937         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4938         prologue cost vector and pass it to vect_get_load_cost.
4939         (vect_get_peeling_costs_all_drs): Likewise.
4940         (vect_peeling_hash_get_lowest_cost): Likewise.
4941         (vect_enhance_data_refs_alignment): Likewise.
4942
4943 2018-04-26  Richard Biener  <rguenther@suse.de>
4944
4945         PR middle-end/85450
4946         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4947         checking of integer<->pointer conversions.
4948         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4949         sign-/zero-extending pointer types.
4950         (expand_omp_for_static_chunk): Likewise.
4951
4952 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
4953             Jean Lee  <xiaoyur347@gmail.com>
4954
4955         * config/mips/mips.c (mips_asan_shadow_offset): New function.
4956         (TARGET_ASAN_SHADOW_OFFSET): Define.
4957         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4958         true for -fsanitize=address.
4959
4960 2018-04-25  Mark Wielaard  <mark@klomp.org>
4961
4962         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4963         shorter ones.
4964
4965 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
4966
4967         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4968         than "alu", remove explicit "memory" and "imm_disp" attributes.
4969         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4970
4971         PR middle-end/85414
4972         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4973         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4974         gen_lowpart_no_emit.
4975
4976 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
4977
4978         PR target/85473
4979         * config/i386/i386.c (ix86_expand_builtin): Change memory
4980         operand to XI, extend p0 to Pmode.
4981         * config/i386/i386.md: Change unspec volatile and operand
4982         1 mode to XI, change operand 0 mode to P.
4983
4984 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
4985
4986         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4987         GET_MODE_MASK before any checking.
4988         (nds32_can_use_bset_p): Likewise.
4989         (nds32_can_use_btgl_p): Likewise.
4990
4991 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
4992
4993         * config/nds32/nds32-doubleword.md: New define_split pattern for
4994         illegal register number.
4995
4996 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
4997
4998         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4999
5000 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
5001
5002         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5003
5004 2018-04-25  Richard Biener  <rguenther@suse.de>
5005
5006         * lto-streamer.h (LTO_major_version): Bump to 8.
5007
5008 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
5009
5010         * BASE-VER: Set to 9.0.0.
5011
5012 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
5013
5014         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5015         in __abskf2 and __powikf2.
5016
5017 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5018
5019         PR target/85512
5020         * config/aarch64/constraints.md (Usg, Usj): New constraints.
5021         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5022         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5023         Use the above on operand 2.  Reindent.
5024         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5025
5026 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
5027
5028         PR target/85485
5029         * common/config/i386/i386-common.c (ix86_handle_option): Don't
5030         handle OPT_mcet.
5031         * config/i386/i386.opt (mcet): Removed.
5032         * doc/install.texi: Remove -mcet documentation.
5033         * doc/invoke.texi: Likewise.
5034
5035 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
5036
5037         PR target/85485
5038         * doc/install.texi: Remove -mcet from bootstrap-cet.
5039
5040 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
5041
5042         PR target/85511
5043         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5044         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5045         if TARGET_64BIT.
5046
5047         PR target/85503
5048         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5049         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5050         containing a CONST_VECTOR.
5051
5052 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
5053
5054         * doc/install.texi: Update newlib dependency for nvptx.
5055
5056 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
5057
5058         PR target/85508
5059         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5060         instead of INTVAL when shifting x left.
5061
5062 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
5063
5064         PR tree-optimization/85478
5065         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5066         vect_grouped_store_supported for single element vectors.
5067
5068 2018-04-24  Richard Biener  <rguenther@suse.de>
5069
5070         PR target/85491
5071         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5072         load cost increase to the case of non-constant step.
5073
5074 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
5075
5076         PR target/84828
5077         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5078         destination if any_malformed_asm.
5079
5080 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
5081
5082         PR middle-end/85496
5083         * expr.c (store_field): In the bitfield case, if the value comes from
5084         a function call and is returned in registers by means of a PARALLEL,
5085         do not change the mode of the temporary unless BLKmode and VOIDmode.
5086
5087 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
5088
5089         PR rtl-optimization/85423
5090         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5091         dependencies to debug insns when the previous insn is non-debug.
5092
5093 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
5094
5095         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5096         enums into a single definition.
5097         (fls): Fix predicates and printing.
5098         (seti): Likewise.
5099
5100 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
5101
5102         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5103         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5104         and short u6 immediate.
5105         (check_if_valid_sleep_operand): Remove.
5106         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5107
5108 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5109
5110         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5111         flag_always_save_lp condition.
5112         * config/nds32/nds32.opt (malways-save-lp): New option.
5113
5114 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
5115
5116         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5117         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5118         * config/nds32/nds32.h
5119         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5120         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5121
5122 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
5123
5124         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5125         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5126
5127 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
5128             Chung-Ju Wu  <jasonwucj@gmail.com>
5129
5130         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5131         Declare.
5132         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5133         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5134
5135 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5136
5137         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5138
5139 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5140
5141         * config/nds32/nds32-protos.h (nds32_data_alignment,
5142         nds32_local_alignment): Declare.
5143         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5144         nds32_local_alignment): New functions.
5145         (TARGET_CONSTANT_ALIGNMENT): Define.
5146         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5147
5148 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5149
5150         * config/nds32/nds32.c
5151         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5152         (TARGET_MODES_TIEABLE_P): Likewise.
5153
5154 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5155
5156         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5157         level Ofast and Og.
5158
5159 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
5160             Chung-Ju Wu  <jasonwucj@gmail.com>
5161
5162         * config/nds32/constants.md (unspec_volatile_element): Add enum values
5163         for unaligned access.
5164         * config/nds32/nds32-intrinsic.c: Implementation of expanding
5165         unaligned access.
5166         * config/nds32/nds32-intrinsic.md: Likewise.
5167         * config/nds32/nds32_intrinsic.h: Likewise.
5168         * config/nds32/nds32.h (nds32_builtins): Likewise.
5169         * config/nds32/nds32.opt (munaligned-access): New option.
5170         * config/nds32/nds32.c (nds32_asm_file_start): Display
5171         flag_unaligned_access status.
5172
5173 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
5174
5175         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5176         -mno-relax is present.
5177         * config/riscv/linux.h (LINK_SPEC): Ditto.
5178
5179 2018-04-20  Martin Sebor  <msebor@redhat.com>
5180
5181         PR c/85365
5182         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5183         for null pointers.
5184         (gimple_fold_builtin_stxcpy_chk): Same.
5185         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5186
5187 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
5188
5189         PR target/85456
5190         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5191         __powikf2 when long double is IEEE 128-bit.
5192
5193 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
5194
5195         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5196         step to make sure stack always aligned.
5197
5198 2018-04-20  Carl Love  <cel@us.ibm.com>
5199
5200         PR target/83402
5201         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5202         size check for arg0.
5203
5204 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
5205             Tom de Vries  <tom@codesourcery.com>
5206
5207         PR target/85445
5208         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5209         Emit insns for calls too.
5210         (nvptx_find_par): Always look for worker-level predecessor insn.
5211         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
5212         calls.
5213         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5214         (nvptx_process_pars): Propagate frames for calls.
5215
5216 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
5217
5218         PR target/85469
5219         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5220         Removed.
5221         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5222         (ix86_handle_option): Don't handle OPT_mibt.
5223         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5224         __SHSTK__.
5225         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5226         has_ibt and ibt.
5227         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5228         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5229         (ix86_target_macros): Define __CET__ with flag_cf_protection
5230         for -fcf-protection.
5231         * config/i386/i386.c (isa2_opts): Remove -mibt.
5232         * config/i386/i386.h (TARGET_IBT): Removed.
5233         (TARGET_IBT_P): Likewise.
5234         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5235         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5236         * config/i386/i386.opt (mcet): Update help message.
5237         (mshstk): Likewise.
5238         (mibt): Removed.
5239         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
5240         -mcet as an alias for -mshstk.
5241
5242 2018-04-20  Richard Biener <rguenther@suse.de>
5243
5244         PR middle-end/85475
5245         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5246         complexity by forcing a single use of the multiply operand.
5247
5248 2018-04-20  Martin Jambor  <mjambor@suse.cz>
5249
5250         ipa/85449
5251         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5252         recursion dependency to only apply to non-clones.
5253
5254 2018-04-20  Martin Jambor  <mjambor@suse.cz>
5255
5256         ipa/85447
5257         * ipa-cp.c (create_specialized_node): Check that clones of
5258         self-recursive edges exist during IPA-CP.
5259
5260 2018-04-19  Toon Moene  <toon@moene.org>
5261
5262         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5263         by -O3.
5264
5265 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
5266
5267         PR tree-optimization/85467
5268         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5269         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
5270         VECTOR_CST element to type.
5271
5272 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
5273
5274         PR target/85397
5275         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5276         * config/i386/i386.md (builtin_setjmp_setup): Removed.
5277         (builtin_longjmp): Likewise.
5278         (save_stack_nonlocal): New pattern.
5279         (restore_stack_nonlocal): Likewise.
5280
5281 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
5282
5283         PR target/85404
5284         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5285         Replace ASM_OUTPUT_LABEL with fprintf.
5286
5287 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
5288
5289         PR target/85417
5290         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5291         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5292         * config/i386/i386-c.c (ix86_target_macros_internal): Also
5293         define __IBT__ and __SHSTK__ for -fcf-protection.
5294         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5295         TARGET_IBT.
5296         (ix86_trampoline_init): Likewise.
5297         (x86_output_mi_thunk): Likewise.
5298         (ix86_notrack_prefixed_insn_p): Likewise.
5299         (ix86_option_override_internal): Don't disallow -fcf-protection.
5300         * config/i386/i386.md (rdssp<mode>): Also enable for
5301         -fcf-protection.
5302         (incssp<mode>): Likewise.
5303         (nop_endbr): Likewise.
5304         * config/i386/i386.opt (mcet): Change help message to built-in
5305         functions only.
5306         (mibt): Likewise.
5307         (mshstk): Likewise.
5308         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5309         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
5310         enable CET built-in functions.
5311
5312 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
5313
5314         * common/config/i386/i386-common.c
5315         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5316         OPTION_MASK_ISA_MOVDIRI_UNSET,
5317         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5318         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5319         * config.gcc (movdirintrin.h): New header.
5320         * config/i386/cpuid.h (bit_MOVDIRI,
5321         bit_MOVDIR64B): New bits.
5322         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5323         and -mmvodir64b.
5324         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5325         (VOID, PVOID, PCVOID)): New function types.
5326         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5327         __builtin_ia32_directstoreu_u64,
5328         __builtin_ia32_movdir64b): New builtins.
5329         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5330         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5331         and -mmovdiri.
5332         (ix86_valid_target_attribute_inner_p): Ditto.
5333         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5334         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5335         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5336         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5337         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5338         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5339         (movdiri<mode>, movdir64b_<mode>): New.
5340         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5341         * config/i386/immintrin.h: Include movdirintrin.h.
5342         * config/i386/movdirintrin.h: New file.
5343         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5344
5345 2018-04-19  Richard Biener  <rguenther@suse.de>
5346
5347         PR middle-end/85455
5348         * cfg.c (clear_bb_flags): When loop state says we have
5349         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5350
5351 2018-04-19  Richard Biener  <rguenther@suse.de>
5352
5353         PR tree-optimization/84737
5354         * tree-vect-data-refs.c (vect_copy_ref_info): New function
5355         copying restrict info.
5356         (vect_setup_realignment): Use it.
5357         * tree-vectorizer.h (vect_copy_ref_info): Declare.
5358         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5359         the first DR to all generated stores.
5360         (vectorizable_load): Likewise for loads.
5361
5362 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
5363
5364         PR tree-optimization/85446
5365         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5366         the integral and pointer types to have the same precision.
5367
5368         * doc/install.texi: Document --disable-cet being the default and
5369         --enable-cet=auto.
5370
5371 2018-04-18  Martin Liska  <mliska@suse.cz>
5372
5373         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5374         style.
5375
5376 2018-04-18  Martin Liska  <mliska@suse.cz>
5377
5378         Revert
5379         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
5380
5381         PR ipa/83983
5382         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5383         arguments if they are comparable.
5384
5385 2018-04-18  Martin Liska  <mliska@suse.cz>
5386
5387         Revert
5388         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
5389
5390         PR lto/84805
5391         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5392         incomplete types.
5393
5394 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
5395
5396         PR target/85388
5397         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5398         ENDBR after calling __morestack.
5399
5400 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
5401
5402         PR jit/85384
5403         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5404         by using gcc_base_ver to generate a gcc_driver_version, and use
5405         it when generating GCC_DRIVER_NAME.
5406         * configure: Regenerate.
5407
5408 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
5409
5410         PR target/81084
5411         * config.gcc: Obsolete powerpc*-*-*spe*.
5412
5413 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
5414
5415         PR debug/84637
5416         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5417         (stabstr_D): Change type of unum from unsigned int to
5418         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
5419         type.
5420
5421 2018-04-17  Jim Wilson  <jimw@sifive.com>
5422
5423         PR 84856
5424         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5425         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5426         Set arg_pointer_offset after using pretend_args_size.
5427
5428 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
5429
5430         PR rtl-optimization/85431
5431         * dse.c (record_store): Ignore zero width stores.
5432
5433         PR sanitizer/85230
5434         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
5435         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5436         __builtin_stack_restore rather than after it.
5437         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5438         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5439         argument instead of virtual_dynamic_stack_rtx.
5440
5441 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5442
5443         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5444         New prototype.
5445         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5446         Add note to error message to explain internal mapping of overloaded
5447         built-in function name to non-overloaded built-in function name.
5448         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5449         function.
5450
5451 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
5452
5453         PR target/85424
5454         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5455         where the inputs overlap with the output.
5456
5457 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
5458
5459         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5460         (=v, v) alternative and explicit "memory" attribute.
5461         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
5462         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5463         attributes.
5464         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5465         "sselog1" type instead of "sselog".
5466         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5467         "sselog".  Remove explicit "memory" attribute.
5468         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5469         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5470         attributes.
5471         (vec_extract_hi_v32hi): Merge all alternatives into one, use
5472         "sselog1" type instead of "sselog".  Remove explicit "memory"
5473         attribute.
5474         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5475         use "sselog1" type instead of "sselog".  Remove explicit "memory"
5476         attribute.
5477         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5478         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5479         attributes.
5480         (vec_extract_hi_v64qi): Merge all alternatives into one, use
5481         "sselog1" type instead of "sselog".  Remove explicit "memory"
5482         attribute.
5483         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5484         use "sselog1" type instead of "sselog".  Remove explicit "memory"
5485         attribute.
5486
5487         PR target/85430
5488         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5489
5490         PR middle-end/85414
5491         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5492         on a SUBREG.
5493
5494 2018-04-17  Martin Jambor  <mjambor@suse.cz>
5495
5496         PR ipa/85421
5497         * ipa-cp.c (create_specialized_node): Call
5498         expand_all_artificial_thunks if necessary.
5499
5500 2018-04-17  Martin Liska  <mliska@suse.cz>
5501
5502         PR lto/85405
5503         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5504         in message, remote space in between '_G' and '('.
5505
5506 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
5507
5508         PR target/85281
5509         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5510         avx512f_vmcmp<mode>3<round_saeonly_name>,
5511         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5512         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5513         avx512f_rndscale<mode><round_saeonly_name>,
5514         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5515         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5516         Use %<iptr>2 instead of %2 for -masm=intel.
5517         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5518         avx512f_vcvttss2usi<round_saeonly_name>,
5519         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5520         -masm=intel.
5521         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5522         avx512f_vcvttsd2usi<round_saeonly_name>,
5523         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5524         Use %q1 instead of %1 for -masm=intel.
5525         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5526         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5527         of %3 for -masm=intel.
5528         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5529         -masm=intel.
5530         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5531         -masm=intel.
5532         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5533         -masm=intel.
5534         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5535         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5536         %g1.
5537         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5538         -masm=intel.
5539         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5540         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5541         %g1 and one with %0 and %1.
5542         (avx512er_vmrcp28<mode><round_saeonly_name>,
5543         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5544         %1 for -masm=intel.
5545         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5546         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5547         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5548         of %0 and %{%4%} for -masm=intel.
5549         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5550         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5551         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5552         order of %0 and %{%5%}%{z%} for -masm=intel.
5553
5554 2018-04-17  Jan Hubicka  <jh@suse.cz>
5555
5556         PR lto/85405
5557         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5558
5559 2018-04-17  Martin Liska  <mliska@suse.cz>
5560
5561         PR ipa/85329
5562         * multiple_target.c (create_dispatcher_calls): Set apostrophes
5563         for target_clone error message.  Make default implementation
5564         clone to be a local declaration.
5565         (separate_attrs): Add new argument and check for an empty
5566         string.
5567         (expand_target_clones): Handle it.
5568         (ipa_target_clone): Make redirection just for target_clones
5569         functions.
5570
5571 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
5572             Tom de Vries  <tom@codesourcery.com>
5573
5574         PR middle-end/84955
5575         * omp-expand.c (expand_oacc_for): Add dummy false branch for
5576         tiled basic blocks without omp continue statements.
5577
5578 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
5579
5580         PR target/83660
5581         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5582         vec_extract expression as having side effects to make sure it gets
5583         a cleanup point.
5584
5585 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
5586
5587         PR target/85403
5588         * config/i386/i386.c (get_builtin_code_for_version): Check
5589         error_mark_node.
5590
5591 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
5592
5593         PR target/84331
5594         * gcc/config.gcc: Support "skylake".
5595         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5596         PROCESSOR_SKYLAKE.
5597         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5598         (processor_target_table): Add "skylake".
5599         (ix86_option_override_internal): Add "skylake".
5600         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5601         PROCESSOR_CANNONLAKE.
5602         (get_builtin_code_for_version): Fix priority for
5603         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5604         PROCESSOR_SKYLAKE-AVX512.
5605         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5606         (processor_type): Add PROCESSOR_SKYLAKE.
5607
5608 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
5609             Jason Merrill  <jason@redhat.com>
5610
5611         PR c++/85112
5612         * convert.c (convert_to_integer_1): Use direct recursion for
5613         enumeral types and types with a precision less than the number
5614         of bits in their mode.
5615
5616 2018-04-16  Julia Koval  <julia.koval@intel.com>
5617
5618         PR target/84413
5619         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5620         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5621
5622 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
5623
5624         PR target/85293
5625         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5626         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5627         and -mno-direct-move.
5628
5629 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
5630
5631         PR target/83402
5632         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5633         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5634         Ensure negative shifts result in {0}.
5635
5636 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
5637
5638         PR rtl-optimization/79916
5639         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5640         regs (if any) to define how to gnerate SD moves when LRA is in
5641         progress.
5642
5643 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
5644
5645         PR rtl-optimization/85393
5646         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5647         * except.c (expand_dw2_landing_pad_for_region): Make static.
5648         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5649         a label and unconditional jump to old_bb, rather than
5650         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5651         basic block.
5652
5653         PR rtl-optimization/85376
5654         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5655         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5656         instead of a specific value.
5657
5658 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
5659             Bin Cheng  <bin.cheng@arm.com>
5660
5661         PR tree-optimization/82965
5662         PR tree-optimization/83991
5663         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5664         by_profile_only parameter.
5665         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5666         information if the loop was predicted to iterate too many times.
5667         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5668
5669 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
5670
5671         PR lto/71991
5672         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5673         always inline.
5674
5675 2018-04-13  Martin Liska  <mliska@suse.cz>
5676             Jakub Jelinek  <jakub@redhat.com>
5677
5678         PR middle-end/81657
5679         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5680         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5681         * builtins.c (expand_builtin_memory_copy_args): Use
5682         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5683         handle dest_addr == pc_rtx.
5684
5685 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
5686
5687         PR target/85291
5688         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5689         asked to not generate direct moves.
5690         (fix_trunc<mode>si2_stfiwx): Similar.
5691         (fix_trunc<mode>si2_internal): Similar.
5692
5693 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
5694
5695         PR debug/83157
5696         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5697         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5698         lookup if dest in some wider mode is known to be const0_rtx and
5699         if so, record permanent equivalence for it to be ZERO_EXTEND of
5700         the narrower mode destination.
5701
5702 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
5703
5704         * lto-streamer-out.c (output_function): Revert 259346.
5705         * omp-expand.c (expand_oacc_for): Likewise.
5706
5707 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
5708
5709         PR rtl-optimization/85354
5710         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5711         * sel-sched.c (sel_global_init): ... here.
5712
5713 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
5714
5715         PR target/85238
5716         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5717         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5718         mode for PE-COFF targets.
5719         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5720         (i386_pe_asm_lto_end): Likewise.
5721         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5722         (TARGET_ASM_LTO_END): Likewise.
5723         * config/i386/winnt.c (saved_debug_info_level): New static variable.
5724         (i386_pe_asm_lto_start): New function.
5725         (i386_pe_asm_lto_end): Likewise.
5726
5727 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
5728             Richard Biener  <rguenther@suse.de>
5729
5730         PR middle-end/84955
5731         * lto-streamer-out.c (output_function): Fix CFG loop state before
5732         streaming out.
5733         * omp-expand.c (expand_oacc_for): Handle calls to internal
5734         functions like regular functions.
5735
5736 2018-04-12  Richard Biener  <rguenther@suse.de>
5737
5738         PR lto/85371
5739         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5740         for the early LTO debug to properly generate references to it
5741         during DIE emission.  Do not re-use that for the skeleton for
5742         split-dwarf.
5743         (dwarf2out_early_finish): Likewise.
5744
5745 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
5746
5747         PR target/85328
5748         * config/i386/sse.md
5749         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5750         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5751         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5752         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5753         and output is a reg, avoid creating invalid lowpart subreg, but
5754         instead split into a 512-bit move.  Don't split if not AVX512VL,
5755         input is xmm16+ reg and output is a mem.
5756         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5757         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5758         xmm16+ reg and output is a mem.
5759
5760 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5761
5762         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5763         also for flag_dwarf2_cfi_asm.
5764
5765 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
5766
5767         PR rtl-optimization/85342
5768         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5769         a bool scalar var inside of the loop instead.  Don't try to update
5770         recog_data.operand after failed apply_change_group.
5771
5772 2018-04-12  Tom de Vries  <tom@codesourcery.com>
5773
5774         PR target/85296
5775         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5776         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
5777         array with flexible array member as array without given dimension.
5778         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5779         argument for undefined param to true.
5780
5781 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
5782
5783         PR target/85321
5784         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5785         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5786         from PowerPC section.
5787         * config/rs6000/sysv4.opt (mcall-): Improve help text.
5788         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5789         help text that is too long.
5790         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5791         help text that is too long.
5792         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5793         help text that is too long.
5794
5795 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
5796
5797         * config/alpha/alpha.md (stack_probe_internal): Rename
5798         from "probe_stack".  Update all callers.
5799
5800 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
5801
5802         PR rtl-optimization/84566
5803         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5804         sched_macro_fuse_insns.
5805
5806 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
5807
5808         PR target/84301
5809         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5810         (compute_block_dependences): ... from here.
5811
5812 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
5813
5814         PR tree-optimization/85331
5815         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5816         from int to HOST_WIDE_INT.
5817
5818 2018-04-11  Martin Jambor  <mjambor@suse.cz>
5819
5820         PR ipa/84149
5821         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5822         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5823         not the same as the source val.
5824         (cgraph_edge_brings_value_p): New parameter.
5825         (gather_edges_for_value): Pass destination value to
5826         cgraph_edge_brings_value_p.
5827         (perhaps_add_new_callers): Likewise.
5828         (get_info_about_necessary_edges): Likewise and exclude values brought
5829         only by self-recursive edges.
5830         (create_specialized_node): Redirect only clones of self-calling edges.
5831         (+self_recursive_pass_through_p): New function.
5832         (find_more_scalar_values_for_callers_subset): Use it.
5833         (find_aggregate_values_for_callers_subset): Likewise.
5834         (known_aggs_to_agg_replacement_list): Removed.
5835         (decide_whether_version_node): Re-calculate known constants for all
5836         remaining context clones.
5837
5838 2018-04-11  Richard Biener  <rguenther@suse.de>
5839
5840         PR lto/85339
5841         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5842         from early DWARF output.
5843         (dwarf2out_early_finish): Output line info unconditionally into
5844         early DWARF and add reference to it.
5845
5846 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
5847
5848         PR target/85281
5849         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5850         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5851         other than V2DFmode using iptr mode attribute.
5852         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5853
5854 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
5855
5856         PR rtl-optimization/84659
5857         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5858
5859 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
5860
5861         PR debug/85302
5862         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5863         SIZEP is NULL.
5864         (output_loc_list): Pass address of a dummy size variable even in the
5865         locview handling loop.
5866         (index_location_lists): Add comment on why skip_loc_list_entry can't
5867         call size_of_locs.
5868
5869 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
5870
5871         PR target/85261
5872         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5873         into register.
5874
5875 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
5876
5877         PR target/85321
5878         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5879         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5880         and -mstring-compare-inline-limit.
5881
5882 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
5883
5884         PR target/85287
5885         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5886         for stack clash protection in a register whenever we need it to be in
5887         a register.
5888
5889 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
5890
5891         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5892         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5893
5894 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
5895
5896         PR target/85321
5897         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5898         the help text.
5899         (mlong-double-): Ditto.
5900         * config/rs6000/sysv4.opt (msdata=): Ditto.
5901         (mtls-size=): Ditto.
5902
5903 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5904
5905         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5906         erroneous entries for
5907         "vector int vec_ldl (int, long int *)", and
5908         "vector unsigned int vec_ldl (int, unsigned long int *)".
5909         Add comments and entries for
5910         "vector bool char vec_ldl (int, bool char *)",
5911         "vector bool short vec_ldl (int, bool short *)",
5912         "vector bool int vec_ldl (int, bool int *)",
5913         "vector bool long long vec_ldl (int, bool long long *)",
5914         "vector pixel vec_ldl (int, pixel *)",
5915         "vector long long vec_ldl (int, long long *)",
5916         "vector unsigned long long vec_ldl (int, unsigned long long *)".
5917         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5918         type tree bool_long_long_type_node and correct definition of
5919         bool_V2DI_type_node to make reference to this new type tree.
5920         (rs6000_mangle_type): Replace erroneous reference to
5921         bool_long_type_node with bool_long_long_type_node.
5922         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5923         comments to emphasize sign distinctions for char and int types and
5924         replace RS6000_BTI_bool_long constant with
5925         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
5926         use of RS6000_BTI_pixel.
5927         (bool_long_type_node): Remove this macro definition.
5928         (bool_long_long_type_node): New macro definition
5929
5930 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
5931
5932         PR rtl-optimization/85300
5933         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5934         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5935         simplify_unary_operation fails.
5936
5937 2018-04-10  Martin Liska  <mliska@suse.cz>
5938
5939         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5940         cgraph_edge and ipa_ref.
5941
5942 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
5943
5944         PR target/85177
5945         PR target/85255
5946         * config/i386/sse.md
5947         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5948         computation of the VEC_MERGE selector from mask.
5949         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5950         Fix decoding of the VEC_MERGE selector into mask.
5951
5952 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
5953
5954         PR tree-optimization/85286
5955         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5956
5957 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
5958
5959         * final.c (final_1): Set insn_last_address as well as
5960         insn_current_address.
5961
5962 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5963
5964         PR target/85173
5965         * explow.c (emit_stack_probe): Call validize_mem on memory location
5966         before passing it to gen_probe_stack.  Create address operand and
5967         legitimize it for the probe_stack_address case.
5968
5969 2018-04-09  Jan Hubicka  <jh@suse.cz>
5970
5971         PR lto/85078
5972         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5973         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5974         * tree.c (free_lang_data_in_type): Fix handling of binfos;
5975         walk basetypes.
5976         (free_lang_data): Rebuild type inheritance graph.
5977
5978 2018-04-09  Martin Sebor  <msebor@redhat.com>
5979
5980         * invoke.texi (-finline-small-functions): Mention other optimization
5981         options.
5982         (-findirect-inlining, -fpartial-inlining): Same.
5983         (-finline-functions-called-once): Same.
5984         (-freorder-blocks-and-partition): Same.
5985
5986 2018-04-09  Jan Hubicka  <jh@suse.cz>
5987
5988         PR rtl/84058
5989         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5990         jumps; choose last target that matches the criteria (i.e.
5991         no partition changes for non-crossing jumps).
5992         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5993         support for redirecting crossing jumps to non-crossing.
5994
5995 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
5996
5997         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5998         also for naked functions.
5999
6000 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
6001
6002         * config/arc/arc.md (add_shift): New pattern.
6003         (add_shift2): Likewise.
6004         (sub_shift): Likewise.
6005         (sub_shift_cmp0_noout): Likewise.
6006         (compare_si_ashiftsi): Likewise.
6007         (xbfu_cmp0_noout): New combine pattern.
6008         (xbfu_cmp0"): Likewise.
6009         (movsi_set_cc_insn): Place the predicable variant first.
6010         (commutative_binary_cmp0_noout): Remove clobber.
6011         (commutative_binary_cmp0): New pattern.
6012         (noncommutative_binary_cmp0): Likewise.
6013         (noncommutative_binary_cmp0_noout): Likewise.
6014         (noncommutative_binary_comparison_result_used): Removed.
6015         (rsub_cmp0): New pattern.
6016         (rsub_cmp0_noout): Likewise.
6017         (extzvsi): Changed, keep only meaningful variants.
6018         (SQH, SEZ): New iterators.
6019         (SQH_postfix): New mode attribute.
6020         (SEZ_prefix): New code attribute.
6021         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6022         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6023         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6024         of numerical value.
6025         (noncommutative_operator): Check the availability of barrel
6026         shifter option.
6027
6028 2018-04-09  Richard Biener  <rguenther@suse.de>
6029
6030         PR tree-optimization/85284
6031         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6032         Only use the niter constraining form of simple_iv when the exit
6033         is always executed.
6034
6035 2018-04-09  Tom de Vries  <tom@codesourcery.com>
6036
6037         PR target/84041
6038         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6039         (define_expand "*memory_barrier"): New define_expand.
6040         (define_insn "memory_barrier"): New insn.
6041
6042 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
6043
6044         PR rtl-optimization/80463
6045         PR rtl-optimization/83972
6046         PR rtl-optimization/83480
6047
6048         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6049         correct producer for the insn.
6050         (tidy_control_flow): Fixup seqnos in case of debug insns.
6051
6052 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
6053
6054         PR rtl-optimization/83913
6055
6056         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6057         different sched-times when merging exprs.
6058
6059 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
6060
6061         PR rtl-optimization/83962
6062
6063         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6064         tidy_fallthru_edge and tidy_control_flow.
6065
6066 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
6067
6068         PR rtl-optimization/83530
6069
6070         * sel-sched.c (force_next_insn): New global variable.
6071         (remove_insn_for_debug): When force_next_insn is true, also leave only
6072         next insn in the ready list.
6073         (sel_sched_region): When the region wasn't scheduled, make another pass
6074         over it with force_next_insn set to 1.
6075
6076 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
6077
6078         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6079         into tm_file.
6080         * config/nds32/constants.md (unspec_volatile_element): Add enum values
6081         for interrupt control.
6082         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6083         functions for interrupt control.
6084         * config/nds32/nds32-intrinsic.md: Likewise.
6085         * config/nds32/nds32_intrinsic.h: Likewise.
6086         * config/nds32/nds32.h (nds32_builtins): Likewise.
6087
6088 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
6089
6090         * config/nds32/nds32.c (nds32_init_machine_status,
6091         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6092         strict_aligned_p field.
6093         (nds32_expand_to_rtl_hook): New function.
6094         (TARGET_EXPAND_TO_RTL_HOOK): Define.
6095         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6096
6097 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
6098             Chung-Ju Wu  <jasonwucj@gmail.com>
6099
6100         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6101         * config/nds32/nds32-n7.md: New file.
6102         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6103         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6104         pipeline.
6105         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6106         * config/nds32/nds32.md (pipeline_model): Add n7.
6107         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6108         * config/nds32/pipelines.md: Include n7 settings.
6109
6110 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
6111             Chung-Ju Wu  <jasonwucj@gmail.com>
6112
6113         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6114         * config/nds32/nds32-e8.md: New file.
6115         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6116         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6117         pipeline.
6118         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6119         * config/nds32/nds32.md (pipeline_model): Add e8.
6120         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6121         * config/nds32/pipelines.md: Include e8 settings.
6122
6123 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
6124             Chung-Ju Wu  <jasonwucj@gmail.com>
6125
6126         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6127         * config/nds32/nds32-n8.md: New file.
6128         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6129         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6130         pipeline.
6131         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6132         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6133         * config/nds32/nds32.md (pipeline_model): Add n8.
6134         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6135         * config/nds32/pipelines.md: Include n8 settings.
6136
6137 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
6138             Chung-Ju Wu  <jasonwucj@gmail.com>
6139
6140         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6141         * config/nds32/nds32-n9-2r1w.md: New file.
6142         * config/nds32/nds32-n9-3r2w.md: New file.
6143         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6144         nds32_register_ports): New or modify for cpu n9.
6145         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6146         pipeline.
6147         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6148         * config/nds32/nds32-utils.c: New file.
6149         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6150         TARGET_MUL_SLOW): Define.
6151         * config/nds32/nds32.md (pipeline_model): New attribute.
6152         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6153         New options that support cpu n9.
6154         * config/nds32/pipelines.md: Include n9 settings.
6155         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6156
6157 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
6158
6159         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6160         information if necessary.
6161         (output_cond_branch_compare_zero): Likewise.
6162         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6163         (nds32_target_alignment): Refine for alignment.
6164         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6165         (FUNCTION_BOUNDARY): Modify.
6166         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6167         align case.
6168         * config/nds32/nds32.opt (malways-align, malign-functions): New.
6169
6170 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
6171
6172         * config/nds32/constants.md (unspec_volatile_element): Add values for
6173         TLB operation and data prefetch.
6174         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6175         functions for TLB operation and data prefetch.
6176         * config/nds32/nds32-intrinsic.md: Likewise.
6177         * config/nds32/nds32_intrinsic.h: Likewise.
6178         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6179         (nds32_print_operand): Likewise.
6180         * config/nds32/nds32.h (nds32_builtins): Likewise.
6181
6182 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
6183         Andrew Pinski <pinsika@gcc.gnu.org>
6184
6185         PR middle-end/82976
6186         * match.pd: Use constant_boolean_node of correct type instead of
6187         boolean_true_node or boolean_false_node for simplifying
6188         pointer comparisons to zero.
6189
6190 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
6191
6192         PR tree-optimization/80021
6193         * tree.c (verify_type_variant): Make error call in verify_variant_match
6194         translatable and remove final full stop.
6195
6196 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
6197
6198         * config/nds32/constants.md (unspec_volatile_element): Add
6199         UNSPEC_VOLATILE_EH_RETURN.
6200         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6201         nds32_output_stack_pop): Support dwarf exception handling process.
6202         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6203         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6204         exception handling process.
6205         (nds32_compute_stack_frame): Likewise.
6206         (nds32_return_addr_rtx): Likewise.
6207         (nds32_initial_elimination_offset): Likewise.
6208         (nds32_expand_prologue): Likewise.
6209         (nds32_expand_epilogue): Likewise.
6210         (nds32_dynamic_chain_address): New function.
6211         * config/nds32/nds32.h (machine_function): Add fields for dwarf
6212         exception handling.
6213         (DYNAMIC_CHAIN_ADDRESS): Define.
6214         (EH_RETURN_DATA_REGNO): Define.
6215         (EH_RETURN_STACKADJ_RTX): Define.
6216         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6217         patterns for dwarf exception handling.
6218
6219 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
6220
6221         * config/nds32/nds32.h: Clean up obsolete macros.
6222
6223 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6224
6225         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6226         Add enum values for particular instructions.
6227         * config/nds32/nds32-intrinsic.c: Implementation of expanding
6228         particular intrinsic functions.
6229         * config/nds32/nds32-intrinsic.md: Likewise.
6230         * config/nds32/nds32_intrinsic.h: Likewise.
6231         * config/nds32/nds32.h (nds32_builtins): Likewise.
6232         * config/nds32/nds32.md (type): Add pbsad and pbsada.
6233         (btst, ave): New patterns for particular instructions.
6234
6235 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6236
6237         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6238         Add enum values for atomic load/store and memory sync.
6239         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6240         and memory sync.
6241         * config/nds32/nds32-intrinsic.md: Likewise.
6242         * config/nds32/nds32_intrinsic.h: Likewise.
6243         * config/nds32/nds32.h (nds32_builtins): Likewise.
6244
6245 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
6246
6247         PR tree-optimization/85257
6248         * fold-const.c (native_encode_vector): If not all elts could fit
6249         and off is -1, return 0 rather than offset.
6250         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6251         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6252         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
6253         adjust buffer in native_interpret_expr call.
6254
6255 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6256
6257         * config/nds32/constants.md (unspec_volatile_element): Add cache
6258         control enum values.
6259         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6260         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6261         * config/nds32/nds32.c (nds32_cctl_names): New.
6262         (nds32_print_operand): Handle cache control register names.
6263         * config/nds32/nds32.h (nds32_builtins): New enum values.
6264         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6265         macros.
6266         * config/nds32/nds32.md (type): Add mmu.
6267         * config/nds32/pipelines.md (simple_insn): Add mmu.
6268
6269 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
6270
6271         * config/nds32/nds32.md (type): Remove call.
6272         * config/nds32/pipelines.md (simple_insn): Likewise.
6273
6274 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6275
6276         * config/nds32/constants.md (unspec_volatile_element): Add
6277         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6278         UNSPEC_VOLATILE_FMFCFG.
6279         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6280         description for fmfcfg and fmfcsr.
6281         (bdesc_1arg): Add fmtcsr.
6282         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6283         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6284         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6285         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6286         unspec_fmfcfg): New patterns.
6287         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6288         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6289         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6290         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6291         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6292         __nds32__fmfcfg): Define.
6293
6294 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6295
6296         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6297         intrinsic register names.
6298         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6299         intrinsic register enum values and macros.
6300
6301 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
6302
6303         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6304         for load/store addressing form.
6305         (nds32_print_operand_address): Likewise.
6306
6307 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
6308
6309         PR target/85196
6310         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6311         based on LABEL_REF.  Remove useless assertion.
6312         (pic_address_needs_scratch): Fix formatting.
6313         (sparc_legitimize_pic_address): Minor tweaks.
6314         (sparc_delegitimize_address): Adjust assertion accordingly.
6315         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6316         into symbolic_operand.
6317         (movsi_high_pic_label_ref): Likewise.
6318         (movsi_lo_sum_pic_label_ref): Likewise.
6319         (movdi_pic_label_ref): Likewise.
6320         (movdi_high_pic_label_ref): Likewise.
6321         (movdi_lo_sum_pic_label_ref): Likewise.
6322
6323 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
6324
6325         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6326         custom LIB_SPEC setup.
6327
6328 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
6329             Kito Cheng  <kito.cheng@gmail.com>
6330
6331         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6332         * config/riscv/freebsd.h: New.
6333
6334 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
6335
6336         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6337         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6338         file.
6339
6340 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
6341             Kito Cheng  <kito.cheng@gmail.com>
6342
6343         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6344         nds32_output_call, nds32_symbol_binds_local_p): New functions.
6345         * config/nds32/nds32-protos.h (nds32_output_call,
6346         nds32_output_return): Declare.
6347         * config/nds32/nds32.md: Refine all the call and return patterns.
6348
6349 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
6350
6351         PR debug/85252
6352         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6353         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6354
6355         PR rtl-optimization/84872
6356         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6357         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6358         EDGE_CROSSING edge.
6359
6360 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
6361
6362         * expr.c (copy_blkmode_to_reg): Revert 254862.
6363         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6364
6365 2018-04-06  Richard Biener  <rguenther@suse.de>
6366
6367         PR middle-end/85244
6368         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6369         after seeing a component reference with an adjacent field.  Treat
6370         refs to arrays at struct end of external decls similar to
6371         refs to unconstrained commons.
6372
6373 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
6374
6375         PR sanitizer/85213
6376         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6377         look through SAVE_EXPRs with non-side-effects argument.  Adjust
6378         recursive calls.
6379         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6380         save_p here.
6381
6382 2018-04-06  Richard Biener  <rguenther@suse.de>
6383
6384         PR middle-end/85180
6385         * alias.c (find_base_term): New wrapper around find_base_term
6386         unwinding CSELIB_VAL_PTR changes.
6387         (find_base_term): Do not restore CSELIB_VAL_PTR during the
6388         recursion.
6389
6390 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
6391
6392         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6393         instructions.
6394         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6395         constant definitions.
6396         ("nop"): lr 0,0 -> nopr r0
6397         ("nop_lr0", "nop_lr1"): New insn definitions.
6398
6399 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
6400
6401         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6402         NDS32_V3PUSH_AVAILABLE_P macro.
6403
6404 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
6405             Chung-Ju Wu  <jasonwucj@gmail.com>
6406
6407         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6408         (nds32*-*-*): Add float and fpu_config into supported_defaults.
6409         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6410         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6411         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6412         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6413         * config/nds32/constraints.md: New constraints and checking for hard
6414         float configuration.
6415         * config/nds32/iterators.md: New mode iterator and attribute for hard
6416         float configuration.
6417         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6418         patterns.
6419         * config/nds32/nds32-fpu.md: New file.
6420         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6421         deal with hard float code generation.
6422         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6423         ARCH_V3S.
6424         (abi_type, float_reg_number): New enum type.
6425         * config/nds32/nds32-predicates.c: New predicates for hard float.
6426         * config/nds32/nds32-protos.h: Declare functions for hard float.
6427         * config/nds32/nds32.c: Implementation for hard float configuration.
6428         * config/nds32/nds32.h: Definitions for hard float configuration.
6429         * config/nds32/nds32.md: Include hard float machine description and
6430         modify patterns for hard float configuration.
6431         * config/nds32/nds32.opt: New options for hard float configuration.
6432         * config/nds32/predicates.md: New predicates for hard float
6433         configuration.
6434
6435 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
6436
6437         * common/config/nds32/nds32-common.c
6438         (nds32_option_optimization_table): Enable -mreleax-hint by default.
6439
6440 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
6441
6442         PR middle-end/85195
6443         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6444         CONSTRUCTOR_ELT (ctor, ...)->value.
6445
6446 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
6447
6448         PR target/85193
6449         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6450
6451 2018-04-05  Tom de Vries  <tom@codesourcery.com>
6452
6453         PR target/85204
6454         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6455         cond jump.
6456
6457 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
6458             Kito Cheng  <kito.cheng@gmail.com>
6459
6460         * config/nds32/constraints.md (U33): Fine-tune checking condition.
6461         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6462         * config/nds32/nds32.h (nds32_16bit_address_type): Add
6463         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6464
6465 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
6466             Kito Cheng  <kito.cheng@gmail.com>
6467
6468         * config/nds32/constraints.md (Ufe): New memory constraint.
6469         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6470         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6471         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6472         operands.
6473         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6474         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6475
6476 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6477
6478         * config/nds32/nds32.md: Use optimize_size in the condition for
6479         alu-shift instructions.
6480
6481 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6482
6483         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6484
6485 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6486
6487         * config/nds32/nds32.md (negsi2): Refine pattern.
6488
6489 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
6490             Chung-Ju Wu  <jasonwucj@gmail.com>
6491
6492         * config/nds32/iterators.md (shift_rotate): New code iterator.
6493         (shift): New code attribute.
6494         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6495         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6496         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6497         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6498         bit-wise operations.
6499         (andsi3, *andsi3): Ditto.
6500         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6501         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6502         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6503         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6504         nds32_ior_operand, nds32_xor_operand): New predicates.
6505
6506 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6507
6508         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6509         (addsi3, subsi3): ... this.
6510
6511 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6512
6513         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6514
6515 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6516
6517         * config/nds32/nds32.md: Adjust indention.
6518
6519 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
6520
6521         * config/nds32/nds32.md (feature): New attribute.
6522
6523 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6524
6525         * config/nds32/nds32.md (subtype): New attribute.
6526
6527 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6528
6529         PR target/85203
6530         * config/arm/arm-builtins.c (arm_expand_builtin): Change
6531         expansion to perform a bitwise AND of the argument followed by a
6532         boolean negation of the result.
6533
6534 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
6535
6536         PR rtl-optimization/84878
6537         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6538         the basic block.  Assert the use reference is not artificial and that
6539         it has an associated insn.
6540
6541 2018-04-04  Michael Matz  <matz@suse.de>
6542
6543         * builtins.c (compute_objsize): Pass correct operand
6544         to array_at_struct_end_p.
6545
6546 2018-04-04  Richard Biener  <rguenther@suse.de>
6547
6548         PR lto/85176
6549         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6550         from contexts for DINFO_LEVEL_TERSE and below.
6551
6552 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
6553
6554         * config/nds32/nds32-doubleword.md (move_<mode>): Require
6555         resiter_operand condition.
6556         * config/nds32/nds32.md (*move<mode>): Ditto.
6557
6558 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
6559             Monk Chiang  <sh.chiang04@gmail.com>
6560
6561         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6562
6563 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
6564
6565         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6566
6567 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
6568             Kito Cheng  <kito.cheng@gmail.com>
6569
6570         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6571         nds32_cond_code_str, output_cond_branch,
6572         output_cond_branch_compare_zero, nds32_expand_cbranch,
6573         nds32_expand_cstore, nds32_expand_movcc,
6574         nds32_output_cbranchsi4_equality_zero,
6575         nds32_output_cbranchsi4_equality_reg,
6576         nds32_output_cbranchsi4_equality_reg_or_const_int,
6577         nds32_output_cbranchsi4_greater_less_zero: New functions.
6578         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6579         nds32_expand_cstore, nds32_expand_movcc,
6580         nds32_output_cbranchsi4_equality_zero,
6581         nds32_output_cbranchsi4_equality_reg,
6582         nds32_output_cbranchsi4_equality_reg_or_const_int,
6583         nds32_output_cbranchsi4_greater_less_zero): Declare.
6584         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6585         nds32_rimm11s_operand): New predicates.
6586         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6587         * config/nds32/nds32.md: Rewrite all the branch and conditional move
6588         patterns.
6589
6590 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
6591
6592         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6593         * config/nds32/nds32.md: Ditto.
6594         * config/nds32/pipelines.md: Ditto.
6595
6596 2018-04-04  Richard Biener  <rguenther@suse.de>
6597
6598         PR tree-optimization/85168
6599         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6600         propagating abnormals.
6601
6602 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
6603
6604         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6605
6606 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
6607             Kito Cheng  <kito.cheng@gmail.com>
6608
6609         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6610         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6611         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6612         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6613         * config/nds32/nds32.md (sibcall_internal): New.
6614         (sibcall_register): Remove.
6615         (sibcall_immediate): Remove.
6616         (sibcall_value_internal): New.
6617         (sibcall_value_register): Remove.
6618         (sibcall_value_immediate): Remove.
6619         * config/nds32/predicates.md (nds32_general_register_operand): New.
6620         (nds32_call_address_operand): New.
6621
6622 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
6623
6624         PR rtl-optimization/85167
6625         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6626         bb_defs if *split_p, instead preinitialize it to NULL.
6627
6628         PR tree-optimization/85156
6629         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6630         evaluating the argument multiple times.
6631
6632 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
6633
6634         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6635         than vector.
6636         (_mm_cvtpd_ps): Likewise.
6637         (_mm_cvttpd_epi32): Likewise.
6638         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6639         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6640         vector, pixel, and bool following altivec.h include.
6641
6642 2018-04-03  Martin Sebor  <msebor@redhat.com>
6643
6644         * doc/extend.texi (Common Function Attributes): Clarify.
6645         (const attribute): Likewise.
6646         (pure attribute): Likewise.
6647
6648 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
6649
6650         PR target/85169
6651         * config/i386/i386.c (ix86_expand_vector_set): Use
6652         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
6653
6654 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
6655
6656         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6657         instructions when changing rounding bits to preserve precision bits
6658         in the x87 control word.
6659
6660 2018-04-03  Martin Liska  <mliska@suse.cz>
6661
6662         PR tree-optimization/82491
6663         * rtl.h (strip_offset_and_add): Replace += suboffset with
6664         poly_uint64 () + suboffset.
6665
6666 2018-03-29  Martin Liska  <mliska@suse.cz>
6667             Martin Jambor  <mjambor@suse.cz>
6668
6669         PR ipa/84947
6670         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6671         param_type is not an integral or pointer type.
6672
6673 2018-04-03  Richard Biener  <rguenther@suse.de>
6674
6675         * sese.h (recompute_all_dominators): Remove.
6676
6677 2018-04-02  Martin Sebor  <msebor@redhat.com>
6678
6679         * doc/invoke.texi (-Wrestrict): Fix typos.
6680
6681 2018-04-02  Jim Wilson  <jimw@sifive.com>
6682
6683         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6684         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6685         (<optab>di3, <optab>si3_extend): Likewise.
6686         (<optab>si3_mask, <optab>si3_mask_1): New.
6687         (<optab>di3_mask, <optab>di3_mask_1): New.
6688         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6689         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6690         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6691
6692 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
6693
6694         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6695         example.
6696
6697 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
6698
6699         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6700         (nds32_canonicalize_comparison): New function.
6701
6702 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
6703             Kito Cheng  <kito.cheng@gmail.com>
6704             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
6705
6706         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6707         * config/nds32/constants.md (unspec_volatile_element): Add
6708         UNSPEC_VOLATILE_RELAX_GROUP.
6709         * config/nds32/nds32-relax-opt.c: New file.
6710         * config/nds32/nds32-predicates.c
6711         (nds32_symbol_load_store_p): New function.
6712         * config/nds32/nds32-protos.h
6713         (nds32_symbol_load_store_p): Declare function.
6714         (make_pass_nds32_relax_opt): Declare new rtl pass function.
6715         * config/nds32/nds32.c
6716         (nds32_register_pass): New function to register pass.
6717         (nds32_register_passes): New function to register passes.
6718         * config/nds32/nds32.md (relax_group): New pattern.
6719         * config/nds32/nds32.opt (mrelax-hint): New option.
6720         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6721
6722 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
6723
6724         * config/nds32/t-nds32: Modify files dependency.
6725
6726 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
6727
6728         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6729         (PROFILE_HOOK): Define its implementation.
6730
6731 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
6732
6733         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6734         type and 32-bit size.
6735
6736 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
6737
6738         PR middle-end/85090
6739         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6740         (V_128_256): New mode iterator.
6741         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6742         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6743         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6744         of V.
6745         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6746         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6747
6748 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
6749
6750         PR target/83315
6751         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6752         NaN inputs correctly.
6753
6754 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
6755
6756         PR target/80546
6757         * config/rs6000/vsx.md (??r): New mode attribute.
6758         (*vsx_mov<mode>_64bit): Use it.
6759         (*vsx_mov<mode>_32bit): Likewise.
6760
6761 2018-03-30  Martin Sebor  <msebor@redhat.com>
6762
6763         PR tree-optimization/84818
6764         * builtins.c (check_access): Use warning_n.
6765
6766 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
6767
6768         PR target/83822
6769         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6770         condition.
6771         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6772         condition.
6773
6774 2018-03-30  Julia Koval  <julia.koval@intel.com>
6775
6776         PR target/84413
6777         * x86-tune.def (movx, partial_reg_dependency): Enable for
6778         m_SKYLAKE_AVX512.
6779
6780 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
6781
6782         PR inline-asm/84985
6783         * lra-constraints.c (process_alt_operands): Move setting
6784         this_alternative_matches below.
6785
6786 2018-03-29  Martin Liska  <mliska@suse.cz>
6787
6788         PR lto/84995.
6789         * doc/invoke.texi: Document how LTO works with debug info.
6790         Describe auto-load support of binutils.  Mention 'x86-64'
6791         as valid option value of -march option.
6792
6793 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
6794
6795         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6796
6797         PR c/85094
6798         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6799         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6800         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6801         checking.
6802
6803 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
6804
6805         PR target/84912
6806         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6807         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6808         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6809         for RS6000_BTM_POWERPC64.
6810         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6811         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6812         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6813         definition.
6814         (DIVDE): Use it.
6815         (DIVDEU): Likewise.
6816
6817 2018-03-28 Carl Love  <cel@us.ibm.com>
6818
6819         Revert
6820         2017-09-27  Carl Love  <cel@us.ibm.com>
6821
6822         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6823         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6824         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6825         fctiw instruction.
6826
6827 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6828
6829         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6830         instead of __vector bool.
6831         (_mm_max_pu8): Likewise.
6832         (_mm_min_pi16): Likewise.
6833
6834 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
6835
6836         PR target/84912
6837         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6838         (DIVWEUO): Likewise.
6839         (DIVDEO): Likewise.
6840         (DIVDEUO): Likewise.
6841         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6842         DIVWEUO and DIVDEUO.
6843         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6844         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6845         (div_extend): Likewise.
6846         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6847         builtin function.
6848         (__builtin_divweuo): Likewise.
6849         (__builtin_divdeo): Likewise.
6850         (__builtin_divdeuo): Likewise.
6851
6852 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
6853
6854         PR target/85095
6855         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6856         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6857
6858         PR tree-optimization/82004
6859         * gimple-match-head.c (optimize_pow_to_exp): New function.
6860         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6861         Don't fold to exp if optimize_pow_to_exp is false.
6862
6863 2018-03-28  Martin Liska  <mliska@suse.cz>
6864
6865         PR other/84819
6866         * calls.c (initialize_argument_information): Fix trailing space.
6867         * common.opt: Fix typo and provide better explanation for
6868         -fsanitize-coverage option.
6869         * config/i386/i386.opt: Fix typo.
6870
6871 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
6872             Martin Liska  <mliska@suse.cz>
6873
6874         PR sanitizer/85081
6875         * gimplify.c (asan_poison_variable): Don't do the check for
6876         gimplify_omp_ctxp here.
6877         (gimplify_decl_expr): Do it here.
6878         (gimplify_target_expr): Likewise.
6879
6880 2018-03-28  Martin Liska  <mliska@suse.cz>
6881
6882         PR target/84988
6883         * config/i386/i386.c (ix86_function_arg_advance): Do not call
6884         chkp_type_bounds_count if MPX is not enabled.
6885
6886 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
6887
6888         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6889
6890 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
6891
6892         PR target/84914
6893         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6894         function to create the function decl for complex long double
6895         multiply and divide for -mabi=ieeelongdouble.
6896         (init_float128_ieee): Call it.
6897
6898 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
6899
6900         PR target/85044
6901         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6902         -fcf-protection=branch -mibt.
6903         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6904
6905 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6906
6907         PR target/81863
6908         * config/arm/arm.c (arm_valid_symbolic_address): Handle
6909         arm_word_relocations.
6910
6911 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
6912
6913         PR target/85056
6914         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6915         extern array declarations.
6916
6917 2018-03-27  Richard Biener  <rguenther@suse.de>
6918
6919         PR middle-end/84067
6920         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6921         explicit single_use checks.
6922
6923 2018-03-27  Richard Biener  <rguenther@suse.de>
6924
6925         PR tree-optimization/85082
6926         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6927         Valueize the VUSE.
6928
6929 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6930
6931         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6932         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6933         Turn on fasynchronous-unwind-tables and funwind-tables.
6934
6935 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
6936
6937         PR target/85073
6938         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6939         (*bmi_blsr_<mode>_ccz): Ditto.
6940
6941 2018-03-26  Tom de Vries  <tom@codesourcery.com>
6942
6943         PR tree-optimization/85063
6944         * omp-general.c (offloading_function_p): New function.  Factor out
6945         of ...
6946         * omp-offload.c (pass_omp_target_link::gate): ... here.
6947         * omp-general.h (offloading_function_p): Declare.
6948         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6949         with attribute omp declare target for offloading functions.
6950
6951 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
6952
6953         PR tree-optimization/84005
6954         * tree-data-ref.h (get_base_for_alignment): Declare.
6955         * tree-data-ref.c (get_base_for_alignment_1): New function.
6956         (get_base_for_alignment): Likewise.
6957         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6958         get_base_for_alignment to find a suitable base object, instead
6959         of always using drb->base_address.
6960
6961 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
6962
6963         PR inline-asm/85022
6964         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6965         known size by default.
6966
6967 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
6968
6969         PR inline-asm/85030
6970         * lra-constraints.c (process_alt_operands): Don't match BLKmode
6971         and non BLKmode operands.
6972
6973 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6974
6975         PR target/85026
6976         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6977         Clean up attributes.
6978
6979 2018-03-23  Richard Biener  <rguenther@suse.de>
6980
6981         PR debug/85020
6982         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6983         we are going to emit early debug for LTO.
6984
6985 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
6986
6987         PR inline-asm/85034
6988         * function.c (match_asm_constraints_1): Don't optimize if input
6989         doesn't satisfy general_operand predicate for output's mode.
6990
6991         PR inline-asm/85022
6992         * alias.c (write_dependence_p): Don't require for x_canonicalized
6993         non-VOIDmode if x has VOIDmode.
6994
6995         PR sanitizer/85029
6996         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6997         just don't try to optimize it rather than assert it never happens.
6998
6999 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7000
7001         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7002         macro expansions for definition of ST_INTERNAL_<mode> and
7003         LD_INTERNAL_<mode> builtins.
7004         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7005         Remove prototype.
7006         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7007         function.
7008         (altivec_expand_st_builtin): Likewise.
7009         (altivec_expand_builtin): Remove calls to deleted functions.
7010         (rs6000_address_for_altivec): Delete this function.
7011         * config/rs6000/vector.md: Remove expands for
7012         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7013
7014 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
7015
7016         PR target/84826
7017         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7018         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7019         re-computing once computed.
7020         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7021         (arm_init_machine_status): Initialize
7022         machine->static_chain_stack_bytes.
7023
7024 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7025
7026         PR target/84760
7027         * doc/extend.texi: Add four new prototypes for vec_ld.
7028         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7029         definitions for more logical presentation.
7030         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7031         entries for V1TI variants of __builtin_altivec_ld builtin.
7032         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7033         handling of V1TI variant of LVX icode pattern.
7034         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7035         (rs6000_gimple_fold_builtin): Likewise.
7036         (altivec_init_builtins): Add code to define
7037         __builtin_altivec_lvx_v1ti function.
7038
7039 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
7040
7041         PR inline-asm/84941
7042         * function.c (match_asm_constraints_1): Don't do the optimization
7043         if input isn't a REG, SUBREG, MEM or constant.
7044
7045 2018-03-22  Tom de Vries  <tom@codesourcery.com>
7046
7047         PR tree-optimization/84956
7048         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7049         bb_has_abnormal_pred.
7050
7051 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
7052
7053         PR sanitizer/85018
7054         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7055         DECL_INITIAL (decl) to decl at the end.
7056         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7057         adjust the comment.
7058
7059 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
7060
7061         * doc/extend.texi (__builtin_tgmath): Document when complex
7062         integer types are treated as _Complex _Float64.
7063
7064 2018-03-21  Tom de Vries  <tom@codesourcery.com>
7065
7066         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7067
7068 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
7069
7070         PR tree-optimization/84960
7071         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7072         if it is ENTRY block, move them into single succ of ENTRY in that case.
7073
7074 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
7075
7076         PR tree-optimization/84811
7077         * poly-int.h (poly_span_traits): Remove the T3 parameter and
7078         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7079         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7080         (known_subrange_p): Update accordingly.  Cast each value involved
7081         in the size comparison, rather than casting the result of the
7082         subtraction.
7083
7084 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
7085
7086         PR tree-optimization/84982
7087         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7088         by flipping the least significant bit rather than all bits from
7089         bitpos to bitpos + bitsize - 1.
7090
7091 2018-03-21  Nathan Sidwell  <nathan@acm.org>
7092
7093         * doc/extend.texi (Deprecated Features): Remove mention of
7094         long-deleted deprecations.
7095
7096 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7097
7098         PR jit/84288
7099         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7100         * configure: Regenerate.
7101
7102 2018-03-21  Tom de Vries  <tom@codesourcery.com>
7103
7104         PR tree-optimization/83126
7105         * tree-parloops.c (num_phis): New function.
7106         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7107
7108 2018-03-21  Nathan Sidwell  <nathan@acm.org>
7109
7110         * doc/extend.texi (Deprecated Features): Update deprecated flags,
7111         mention anon-struct/union members and trailing attributes.
7112
7113 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
7114
7115         PR tree-optimization/84969
7116         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7117         builtin memset partitions if they set different rhs values.
7118
7119 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
7120
7121         PR rtl-optimization/84989
7122         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7123         VEC_DUPLICATE with scalar result mode.
7124
7125 2018-03-21  Martin Liska  <mliska@suse.cz>
7126
7127         PR ipa/84963
7128         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7129         not intended return statement.
7130
7131 2018-03-21  Martin Liska  <mliska@suse.cz>
7132
7133         PR target/84988
7134         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7135         (chkp_find_bound_slots_1): Limit number of iterations.
7136
7137 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
7138
7139         PR target/84838
7140         * Minor grammar fixes for x86 options.
7141
7142 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
7143
7144         PR debug/84875
7145         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7146         holding REG_CFA_RESTORE notes, instead turn them into a USE.
7147
7148 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
7149
7150         PR target/83789
7151         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7152         (altivec_lvx_<mode>_1op): Likewise.
7153         (altivec_stvx_<mode>_2op): Likewise.
7154         (altivec_stvx_<mode>_1op): Likewise.
7155         (altivec_lvx_<VM2:mode>): New define_expand.
7156         (altivec_stvx_<VM2:mode>): Likewise.
7157         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7158         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7159         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7160         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7161         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7162         (rs6000_gen_lvx): Likewise.
7163         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7164         (altivec_expand_stv_builtin): Likewise.
7165         (altivec_expand_builtin): Likewise.
7166         * config/rs6000/vector.md: Likewise.
7167
7168 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7169
7170         PR target/82518
7171         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7172         BYTES_BIG_ENDIAN.
7173
7174 2018-03-20  Richard Biener  <rguenther@suse.de>
7175
7176         PR target/84986
7177         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7178         sign-conversions as zero, fall back to standard scalar_stmt
7179         cost for the rest.
7180
7181 2018-03-20  Martin Liska  <mliska@suse.cz>
7182
7183         PR ipa/84825
7184         * predict.c (rebuild_frequencies): Handle case when we have
7185         PROFILE_ABSENT, but flag_guess_branch_prob is false.
7186
7187 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
7188
7189         PR target/84990
7190         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7191         flag_section_anchors.
7192         * varasm.c (use_blocks_for_decl_p): Remove hack for
7193         dw2_force_const_mem.
7194
7195         PR target/84845
7196         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7197         to ...
7198         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
7199         be created, use lowpart_subreg of operands[0] rather than operands[0]
7200         itself.
7201         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7202         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7203         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7204         and n constraint instead of aarch64_shift_imm_di and Usd.
7205         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7206         (*aarch64_<optab>_reg_minus<mode>3): ... this.
7207
7208 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
7209
7210         PR target/82989
7211         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7212         to favor GPR over NEON registers.
7213         (<shift>di3_neon): Likewise.
7214
7215 2018-03-20  Tom de Vries  <tom@codesourcery.com>
7216
7217         PR target/84952
7218         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7219         (nvptx_process_pars): Emit bar.sync asap and alap.
7220
7221 2018-03-20  Tom de Vries  <tom@codesourcery.com>
7222
7223         PR target/84954
7224         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7225         seen_label if seen_label is already set.
7226
7227 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
7228
7229         PR target/84945
7230         * config/i386/i386.c (fold_builtin_cpu): For features above 31
7231         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7232         Use 1U instead of 1.  Formatting fixes.
7233
7234         PR c/84953
7235         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7236         instead of TREE_TYPE (s1) for the return value.
7237
7238 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
7239
7240         PR tree-optimization/84946
7241         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7242         bitsize + bitsize in poly_uint64 rather than poly_int64.
7243
7244         PR sanitizer/78651
7245         * dwarf2asm.c: Include fold-const.c.
7246         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7247         of decl rather than decl itself.
7248
7249         PR rtl-optimization/84643
7250         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7251
7252 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
7253
7254         PR sanitizer/78651
7255         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7256         calling assemble_variable.
7257
7258 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
7259
7260         PR target/81647
7261         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7262         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7263
7264 2018-03-19  Jim Wilson  <jimw@sifive.com>
7265
7266         PR bootstrap/84856
7267         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7268         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7269         (riscv_first_stack_step): Likewise.
7270         (riscv_option_override): Use STACK_BOUNDARY instead of
7271         MIN_STACK_BOUNDARY.
7272         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7273         MIN_STACK_BOUNDARY.
7274         (BIGGEST_ALIGNMENT): Set to 128.
7275         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7276         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7277         STACK_BOUNDARY.
7278
7279 2018-03-19  Richard Biener  <rguenther@suse.de>
7280
7281         PR tree-optimization/84933
7282         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7283         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7284
7285 2018-03-19  Richard Biener  <rguenther@suse.de>
7286
7287         PR tree-optimization/84859
7288         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7289         (cond_if_else_store_replacement): Perform sinking operation on
7290         single-store BBs regardless of MAX_STORES_TO_SINK setting.
7291         Generalize what a BB with a single eligible store is.
7292
7293 2018-03-19  Richard Biener  <rguenther@suse.de>
7294
7295         PR tree-optimization/84929
7296         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7297         chrec_is_positive against non-chrec arg.
7298
7299 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
7300
7301         PR target/84711
7302         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7303
7304 2018-03-18  Martin Liska  <mliska@suse.cz>
7305
7306         PR rtl-optimization/84635
7307         * regrename.c (build_def_use): Use matches_mode only when
7308         matches >= 0.
7309
7310 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
7311
7312         PR tree-optimization/84913
7313         * tree-vect-loop.c (vectorizable_reduction): Don't try to
7314         vectorize chains of COND_EXPRs.
7315
7316 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
7317
7318         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7319
7320 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
7321
7322         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7323
7324 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
7325
7326         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7327
7328 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7329             Kito Cheng  <kito.cheng@gmail.com>
7330
7331         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7332         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7333         (nds32_adjust_reg_alloc_order): New function.
7334         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7335
7336 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
7337
7338         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7339         nds32_print_operand, nds32_print_operand_address): Use
7340         HOST_WIDE_INT_PRINT_DEC instead.
7341
7342 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7343
7344         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7345
7346 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
7347
7348         PR target/84902
7349         * config/i386/i386.c (initial_ix86_tune_features,
7350         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7351         unsigned long long.
7352         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7353         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7354         rather than 1u << ix86_tune.  Formatting fix.
7355         (ix86_option_override_internal): Change ix86_arch_mask from
7356         unsigned int to unsigned HOST_WIDE_INT, initialize to
7357         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7358         (ix86_function_specific_restore): Likewise.
7359
7360 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
7361
7362         PR target/84899
7363         * postreload.c (reload_combine_recognize_pattern): Perform
7364         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7365         truncate_int_for_mode the result for the destination's mode.
7366
7367         PR c/84909
7368         * hsa-gen.c (mem_type_for_type): Fix comment typo.
7369         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7370         Likewise.
7371         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7372         Likewise.
7373
7374 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
7375
7376         PR target/84876
7377         * lra-assigns.c (lra_split_hard_reg_for): Don't use
7378         regno_allocno_class_array and sorted_pseudos.
7379         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7380         insns where regno is used.
7381
7382 2018-03-16  Martin Liska  <mliska@suse.cz>
7383
7384         PR ipa/84833
7385         * multiple_target.c (create_dispatcher_calls): Redirect
7386         reference in the symbol table.
7387
7388 2018-03-16  Martin Liska  <mliska@suse.cz>
7389
7390         PR ipa/84722
7391         * multiple_target.c (create_dispatcher_calls): Redirect also
7392         an alias.
7393
7394 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
7395
7396         PR c++/79937
7397         PR c++/82410
7398         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7399         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7400         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7401
7402 2018-03-16  Julia Koval  <julia.koval@intel.com>
7403
7404         * doc/invoke.texi (Skylake Server): Add CLWB.
7405         Cannonlake): Remove CLWB.
7406
7407 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
7408
7409         PR tree-optimization/84841
7410         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7411         1 << 3.
7412         (FLOAT_ONE_CONST_TYPE): Define.
7413         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7414         (sort_by_operand_rank): Put entries with higher constant_type last
7415         rather than first to match comments.
7416
7417 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
7418
7419         * config/nios2/nios2.md (movsi_internal): Fix thinko in
7420         split predicate.
7421
7422 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
7423
7424         PR c++/79085
7425         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7426         check and use address of target always.
7427
7428 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
7429
7430         PR target/84574
7431         * config/i386/i386.c (indirect_thunk_needed): Update comments.
7432         (indirect_thunk_bnd_needed): Likewise.
7433         (indirect_thunks_used): Likewise.
7434         (indirect_thunks_bnd_used): Likewise.
7435         (indirect_return_needed): New.
7436         (indirect_return_bnd_needed): Likewise.
7437         (output_indirect_thunk_function): Add a bool argument for
7438         function return.
7439         (output_indirect_thunk_function): Don't generate alias for
7440         function return thunk.
7441         (ix86_code_end): Call output_indirect_thunk_function to generate
7442         function return thunks.
7443         (ix86_output_function_return): Set indirect_return_bnd_needed
7444         and indirect_return_needed instead of indirect_thunk_bnd_needed
7445         and indirect_thunk_needed.
7446
7447 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
7448
7449         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7450         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7451         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7452
7453 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
7454             Paul Hua <paul.hua.gm@gmail.com>
7455
7456         PR c/84852
7457         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7458
7459 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
7460
7461         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7462         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7463         resp. SFmode cases.
7464
7465 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
7466
7467         PR target/84711
7468         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7469         instead of GET_MODE_SIZE when comparing Units.
7470
7471 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
7472
7473         PR target/68256
7474         * varasm.c (hash_section): Return an unchangeble hash value
7475         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7476         Return !aarch64_can_use_per_function_literal_pools_p ().
7477
7478 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
7479
7480         PR target/84860
7481         * optabs.c (emit_conditional_move): Pass address of cmode's copy
7482         rather than address of cmode as last argument to prepare_cmp_insn.
7483
7484 2018-03-15  Julia Koval  <julia.koval@intel.com>
7485
7486         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7487         F_AVX512VNNI, F_AVX512BITALG): New.
7488
7489 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
7490
7491         PR target/83451
7492         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
7493         insn for floating-point loads and stores.
7494
7495 2018-03-14  Carl Love  <cel@us.ibm.com>
7496
7497         * config/rs6000/rs6000-c.c: Add macro definitions for
7498         ALTIVEC_BUILTIN_VEC_PERMXOR.
7499         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7500         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7501         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7502         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7503         UNSPEC_VPERMXOR.
7504         * config/doc/extend.texi: Add prototypes for vec_permxor.
7505
7506 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
7507
7508         PR c/84852
7509         * diagnostic-show-locus.c (class layout_point): Convert m_line
7510         from int to linenum_type.
7511         (line_span::comparator): Use linenum "compare" function when
7512         comparing line numbers.
7513         (test_line_span): New function.
7514         (layout_range::contains_point): Convert param "row" from int to
7515         linenum_type.
7516         (layout_range::intersects_line_p): Likewise.
7517         (layout::will_show_line_p): Likewise.
7518         (layout::print_source_line): Likewise.
7519         (layout::should_print_annotation_line_p): Likewise.
7520         (layout::print_annotation_line): Likewise.
7521         (layout::print_leading_fixits): Likewise.
7522         (layout::annotation_line_showed_range_p): Likewise.
7523         (struct line_corrections): Likewise for field m_row.
7524         (line_corrections::line_corrections): Likewise for param "row".
7525         (layout::print_trailing_fixits): Likewise.
7526         (layout::get_state_at_point): Likewise.
7527         (layout::get_x_bound_for_row): Likewise.
7528         (layout::print_line): Likewise.
7529         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7530         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7531         * input.c (selftest::test_linenum_comparisons): New function.
7532         (selftest::input_c_tests): Call it.
7533         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7534         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7535         * selftest.h (ASSERT_GT): New macro.
7536         (ASSERT_GT_AT): New macro.
7537         (ASSERT_LT): New macro.
7538         (ASSERT_LT_AT): New macro.
7539
7540 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
7541
7542         PR rtl-optimization/84780
7543         * combine.c (distribute_links): Don't make a link based on pc_rtx.
7544
7545 2018-03-14  Martin Liska  <mliska@suse.cz>
7546
7547         * tree.c (record_node_allocation_statistics): Use
7548         get_stats_node_kind.
7549         (get_stats_node_kind): New function extracted from
7550         record_node_allocation_statistics.
7551         (free_node): Use get_stats_node_kind.
7552
7553 2018-03-14  Richard Biener  <rguenther@suse.de>
7554
7555         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7556         that the value-set of ANTIC_IN doesn't grow.
7557
7558         Revert
7559         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7560         member.
7561         (BB_VISITED_WITH_VISITED_SUCCS): New define.
7562         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7563
7564 2018-03-14  Julia Koval  <julia.koval@intel.com>
7565
7566         * config.gcc (icelake-client, icelake-server): New.
7567         (icelake): Remove.
7568         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7569         (initial_ix86_arch_features): Ditto.
7570         (PTA_SKYLAKE): Add SGX.
7571         (PTA_ICELAKE): Remove.
7572         (PTA_ICELAKE_CLIENT): New.
7573         (PTA_ICELAKE_SERVER): New.
7574         (ix86_option_override_internal): Split up icelake on icelake client and
7575         icelake server.
7576         (get_builtin_code_for_version): Ditto.
7577         (fold_builtin_cpu): Ditto.
7578         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7579         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7580         * config/i386/i386.h (processor_type): Ditto.
7581         * doc/invoke.texi: Ditto.
7582
7583 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
7584
7585         PR sanitizer/83392
7586         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7587         INTEGER_CST offset, add it together with bitpos / 8 and
7588         sign extend based on POINTER_SIZE.
7589
7590         PR target/84844
7591         Revert
7592         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
7593
7594         PR target/78090
7595         * config/i386/constraints.md (Yc): New register constraint.
7596         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7597         Use Yc constraint for alternative 2 of operand 0.  Remove
7598         preferred_for_speed attribute.
7599
7600 2018-03-14  Richard Biener  <rguenther@suse.de>
7601
7602         PR tree-optimization/84830
7603         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7604         with the old one to avoid oscillations.
7605
7606 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
7607
7608         PR target/83712
7609         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7610         pseudos.
7611         (assign_by_spills): Return a flag of reload assignment failure.
7612         Do not process the reload assignment failures.  Do not spill other
7613         reload pseudos if they has the same reg class.  Update n if
7614         necessary.
7615         (lra_assign): Add a return arg.  Set up from the result of
7616         assign_by_spills call.
7617         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7618         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
7619         usage_insns if it is not NULL.
7620         (spill_hard_reg_in_range): New function.
7621         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7622         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7623         function prototypes.
7624         (lra_assign): Change prototype.
7625         * lra.c (lra): Add code to deal with fails by splitting hard reg
7626         live ranges.
7627
7628 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
7629
7630         * config/riscv/riscv.opt (mrelax): New option.
7631         * config/riscv/riscv.c (riscv_file_start): Emit ".option
7632         "norelax" when riscv_mrelax is disabled.
7633         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7634
7635 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
7636
7637         PR target/84743
7638         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7639         reassociation for int modes.
7640
7641 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
7642
7643         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7644         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7645         for big-endian.
7646         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7647         * config/aarch64/aarch64-sve.md
7648         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7649         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7650         (*extend<mode><Vwide>2): Rename to...
7651         (aarch64_sve_extend<mode><Vwide>2): ...this.
7652         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7653         renaming the old pattern to...
7654         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
7655         unsigned packs.
7656         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7657         define_expand, renaming the old pattern to...
7658         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7659         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7660         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7661         account when deciding which SVE instruction the optab should use.
7662         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7663
7664 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
7665
7666         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7667         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7668         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7669         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7670         (tlsdesc_small_<mode>): Turn a define_expand and use
7671         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
7672         (tlsdesc_small_advsimd_<mode>): ...this.
7673         (tlsdesc_small_sve_<mode>): New pattern.
7674
7675 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
7676
7677         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7678         (UNSPEC_UMUL_HIGHPART): New constants.
7679         (MUL_HIGHPART): New int iteraor.
7680         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7681         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7682         define_expand.
7683         (*<su>mul<mode>3_highpart): New define_insn.
7684
7685 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
7686
7687         PR lto/84805
7688         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7689         incomplete types.
7690
7691 2018-03-13  Martin Liska  <mliska@suse.cz>
7692
7693         PR ipa/84658.
7694         * (sem_item_optimizer::sem_item_optimizer): Initialize new
7695         vector.
7696         (sem_item_optimizer::~sem_item_optimizer): Release it.
7697         (sem_item_optimizer::merge_classes): Register variable aliases.
7698         (sem_item_optimizer::fixup_pt_set): New function.
7699         (sem_item_optimizer::fixup_points_to_sets): Likewise.
7700         * ipa-icf.h: Declare new variables and functions.
7701
7702 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
7703
7704         PR middle-end/84834
7705         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7706         integer_pow2p@2 and test integer_pow2p in condition.
7707         (A < 0 ? C : 0): Similarly for @1.
7708
7709         PR middle-end/84831
7710         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7711         characters starting at p contain '\0' character, don't look beyond
7712         that.
7713
7714         PR target/84827
7715         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7716         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7717
7718         PR target/84828
7719         * reg-stack.c (change_stack): Change update_end var from int to
7720         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7721         also call set_block_for_insn on the newly added insns and rescan.
7722
7723         PR target/84786
7724         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7725         on the last operand.
7726
7727         PR c++/84704
7728         * tree.c (stabilize_reference_1): Return save_expr (e) for
7729         STATEMENT_LIST even if it doesn't have side-effects.
7730
7731 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
7732
7733         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7734
7735 2018-03-12  Renlin Li  <renlin.li@arm.com>
7736
7737         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7738         aarch64_output_scalar_simd_mov_immediate.
7739
7740 2018-03-12  Martin Sebor  <msebor@redhat.com>
7741
7742         PR tree-optimization/83456
7743         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7744         for perfectly overlapping calls to memcpy.
7745         (gimple_fold_builtin_memory_chk): Same.
7746         (gimple_fold_builtin_strcpy): Handle no-warning.
7747         (gimple_fold_builtin_stxcpy_chk): Same.
7748         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7749
7750 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
7751
7752         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7753         parameter.  Use it for SFmode.
7754         (rs6000_function_arg_advance_1): Adjust.
7755         (rs6000_function_arg): Adjust.
7756         (rs6000_gimplify_va_arg): Pass false for that new parameter.
7757
7758 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
7759
7760         PR rtl-optimization/84169
7761         PR rtl-optimization/84780
7762         * combine.c (can_combine_p): Check for a 2-insn combination whether
7763         the destination register is used between the two insns, too.
7764
7765 2018-03-12  Richard Biener  <rguenther@suse.de>
7766
7767         PR tree-optimization/84803
7768         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7769         for refs DR analysis didn't process.
7770
7771 2018-03-12  Richard Biener  <rguenther@suse.de>
7772
7773         PR tree-optimization/84777
7774         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7775         force-vectorize loops ignore whether we are optimizing for size.
7776
7777 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7778
7779         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7780         (TARGET_MD_ASM_ADJUST): Define.
7781
7782 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
7783             Kito Cheng  <kito.cheng@gmail.com>
7784             Chung-Ju Wu  <jasonwucj@gmail.com>
7785
7786         * config/nds32/nds32.c (nds32_compute_stack_frame,
7787         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7788         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7789         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7790         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7791         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7792         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7793         * config/nds32/nds32.md (prologue, epilogue): Use macro
7794         NDS32_V3PUSH_AVAILABLE_P to do checking.
7795
7796 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
7797
7798         PR debug/58150
7799         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7800         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7801         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
7802         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
7803         addition of most attributes on !orig_type_die or the attribute not
7804         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7805
7806 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
7807             Chung-Ju Wu  <jasonwucj@gmail.com>
7808
7809         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7810         __NDS32_VH__ macro.
7811         * config/nds32/nds32.opt (mvh): New option.
7812
7813 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
7814             Chung-Ju Wu  <jasonwucj@gmail.com>
7815
7816         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7817         function.
7818         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7819         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7820         definition.
7821
7822 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
7823             Chung-Ju Wu  <jasonwucj@gmail.com>
7824
7825         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7826         function.
7827         * config/nds32/nds32-multiple.md (strlensi): New pattern.
7828         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7829
7830 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
7831             Kito Cheng  <kito.cheng@gmail.com>
7832             Chung-Ju Wu  <jasonwucj@gmail.com>
7833
7834         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7835         UNSPEC_FFMISM and UNSPEC_FLMISM.
7836         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7837         for ffb, ffmism and flmism.
7838         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7839         (unspec_ffmism): Ditto.
7840         (unspec_flmism): Ditto.
7841         (nds32_expand_builtin_impl): Check if string extension is available.
7842         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7843         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7844
7845 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
7846
7847         Reverting patch:
7848         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
7849
7850         PR target/83712
7851         * lra-assigns.c (assign_by_spills): Return a flag of reload
7852         assignment failure.  Do not process the reload assignment
7853         failures.  Do not spill other reload pseudos if they has the same
7854         reg class.
7855         (lra_assign): Add a return arg.  Set up from the result of
7856         assign_by_spills call.
7857         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7858         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
7859         usage_insns if it is not NULL.
7860         (spill_hard_reg_in_range): New function.
7861         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7862         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7863         function prototypes.
7864         (lra_assign): Change prototype.
7865         * lra.c (lra): Add code to deal with fails by splitting hard reg
7866         live ranges.
7867
7868 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
7869
7870         PR target/84807
7871         * config/i386/i386.opt: Replace Enforcment with Enforcement.
7872
7873 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
7874
7875         PR debug/84620
7876         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7877         (dw_val_node): Add val_symbolic_view.
7878         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7879         (symview_upper_bound): New.
7880         (new_line_info_table): Initialize symviews_since_reset.
7881         (dwarf2out_source_line): Count symviews_since_reset and set
7882         symview_upper_bound.
7883         (dw_val_equal_p): Handle symview.
7884         (add_AT_symview): New.
7885         (print_dw_val): Handle symview.
7886         (attr_checksum, attr_checksum_ordered): Likewise.
7887         (same_dw_val_p, size_of_die): Likewise.
7888         (value_format, output_die): Likewise.
7889         (add_high_low_attributes): Use add_AT_symview for entry_view.
7890         (dwarf2out_finish): Reset symview_upper_bound, clear
7891         zero_view_p.
7892
7893 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
7894
7895         PR target/83969
7896         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7897         Add strict argument and use it.
7898         (rs6000_split_multireg_move): Update for new strict argument.
7899         (mem_operand_gpr): Disallow all non-offsettable addresses.
7900         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7901
7902 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
7903
7904         PR target/84772
7905         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7906         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7907         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7908
7909         PR c++/84767
7910         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7911         decl, use remap_type if we want to use the type.
7912
7913 2018-03-09  Martin Sebor  <msebor@redhat.com>
7914
7915         PR tree-optimization/84526
7916         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7917         Remove dead code.
7918         (builtin_access::generic_overlap): Be prepared to handle non-array
7919         base objects.
7920
7921 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
7922
7923         PR rtl-optimization/84682
7924         * lra-constraints.c (process_address_1): Check is_address flag
7925         for address constraints.
7926         (process_alt_operands): Likewise.
7927         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7928         preprocess_constraints.
7929         * recog.h (preprocess_constraints): Add oploc parameter.
7930         Adjust callers.
7931         * recog.c (preprocess_constraints): Test address_operand for
7932         CT_ADDRESS constraints.
7933
7934 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
7935
7936         PR target/83712
7937         * lra-assigns.c (assign_by_spills): Return a flag of reload
7938         assignment failure.  Do not process the reload assignment
7939         failures.  Do not spill other reload pseudos if they has the same
7940         reg class.
7941         (lra_assign): Add a return arg.  Set up from the result of
7942         assign_by_spills call.
7943         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7944         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
7945         usage_insns if it is not NULL.
7946         (spill_hard_reg_in_range): New function.
7947         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7948         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7949         function prototypes.
7950         (lra_assign): Change prototype.
7951         * lra.c (lra): Add code to deal with fails by splitting hard reg
7952         live ranges.
7953
7954 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7955
7956         PR target/83193
7957         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7958         Accept complain bool parameter.  Only emit errors if it is true.
7959         (arm_parse_cpu_option_name): Likewise.
7960         (arm_target_thumb_only): Adjust callers of the above.
7961         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7962         prototype to take a default true bool parameter.
7963         (arm_parse_arch_option_name): Likewise.
7964
7965 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
7966             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
7967
7968         PR jit/64089
7969         PR jit/84288
7970         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7971         * configure: Regenerate.
7972         * configure.ac ("linker --version-script option"): New.
7973         ("linker soname option"): New.
7974
7975 2018-03-09  Richard Biener  <rguenther@suse.de>
7976
7977         PR tree-optimization/84775
7978         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7979         immediate uses of predicate stmts and mark them modified.
7980
7981         Revert
7982         PR tree-optimization/84178
7983         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7984         to caller.
7985         (version_loop_for_if_conversion): Delay update_ssa call.
7986         (tree_if_conversion): Delay update_ssa until after predicate
7987         insertion.
7988
7989 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
7990
7991         PR target/84763
7992         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7993         when the function accesses prior frames.
7994
7995 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
7996
7997         PR debug/84456
7998         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7999         gen_llsym, otherwise call maybe_gen_llsym.
8000
8001         PR inline-asm/84742
8002         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8003         has ',' character inside of it.
8004
8005 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8006
8007         PR target/84748
8008         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8009         as clobbering CC_REGNUM.
8010
8011 2018-03-08  Richard Biener  <rguenther@suse.de>
8012
8013         PR middle-end/84552
8014         * tree-scalar-evolution.c: Include tree-into-ssa.h.
8015         (follow_copies_to_constant): Do not follow SSA names registered
8016         for update.
8017
8018 2018-03-08  Richard Biener  <rguenther@suse.de>
8019
8020         PR tree-optimization/84178
8021         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8022         to caller.
8023         (version_loop_for_if_conversion): Delay update_ssa call.
8024         (tree_if_conversion): Delay update_ssa until after predicate
8025         insertion.
8026
8027 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
8028
8029         PR tree-optimization/84178
8030         * tree-if-conv.c (release_bb_predicate): Remove the
8031         the assertion that the stmts have NULL use_ops.
8032         Discard the statements, asserting that they haven't
8033         yet been added to a BB.
8034
8035 2018-03-08  Richard Biener  <rguenther@suse.de>
8036
8037         PR tree-optimization/84746
8038         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8039         (phi_translate): Pass in destination ANTIC_OUT set.
8040         (phi_translate_1): Likewise.  For a simplified result lookup
8041         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8042         (phi_translate_set): Adjust.
8043         (do_pre_regular_insertion): Likewise.
8044         (do_pre_partial_partial_insertion): Likewise.
8045
8046 2018-03-08  Martin Liska  <mliska@suse.cz>
8047
8048         PR gcov-profile/84735
8049         * doc/gcov.texi: Document usage of profile files.
8050         * gcov-io.h: Document changes in the format.
8051
8052 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
8053
8054         PR debug/84404
8055         PR debug/84408
8056         * dwarf2out.c (struct dw_line_info_table): Update comments for
8057         view == -1.
8058         (FORCE_RESET_NEXT_VIEW): New.
8059         (FORCE_RESETTING_VIEW_P): New.
8060         (RESETTING_VIEW_P): Check for -1 too.
8061         (ZERO_VIEW_P): Likewise.
8062         (new_line_info_table): Force-reset next view.
8063         (dwarf2out_begin_function): Likewise.
8064         (dwarf2out_source_line): Simplify zero_view_p initialization.
8065         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8066         view directly.  Omit view when omitting .loc at line 0.
8067
8068 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
8069
8070         PR tree-optimization/84740
8071         * tree-switch-conversion.c (process_switch): Call build_constructors
8072         only if info.phi_count is non-zero.
8073
8074         PR tree-optimization/84739
8075         * tree-tailcall.c (find_tail_calls): Check call arguments against
8076         DECL_ARGUMENTS (current_function_decl) rather than
8077         DECL_ARGUMENTS (func) when checking for tail recursion.
8078
8079 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
8080
8081         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8082         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8083         Volker Reichelt's entry and add entries for people that perform
8084         GCC fuzzy testing and report numerous bugs.
8085
8086 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
8087
8088         PR target/82411
8089         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8090         readonly data in sdata, if that is disabled.
8091         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8092         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8093         -mreadonly-in-sdata option.
8094
8095 2018-03-07  Martin Sebor  <msebor@redhat.com>
8096
8097         PR tree-optimization/84468
8098         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8099         basic block when looking for nul assignment.
8100
8101 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
8102
8103         PR target/84277
8104         * except.h (output_function_exception_table): Adjust prototype.
8105         * except.c (output_function_exception_table): Remove FNNAME parameter
8106         and add SECTION parameter.  Ouput one part of the table at a time.
8107         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8108         the first part of the exception table and emit unwind directives.
8109         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8110         (i386_pe_seh_cold_init): Likewise.
8111         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8112         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8113         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8114         (ix86_output_call_insn): Emit a nop in one more case for SEH.
8115         * config/i386/winnt.c: Include except.h.
8116         (struct seh_frame_state): Add reg_offset, after_prologue and
8117         in_cold_section fields.
8118         (i386_pe_seh_end_prologue): Set seh->after_prologue.
8119         (i386_pe_seh_cold_init): New function.
8120         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8121         to seh->in_cold_section.
8122         (seh_emit_push): Record the offset of the push.
8123         (seh_emit_save): Record the offet of the save.
8124         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8125         Test seh->after_prologue to disregard the epilogue.
8126         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8127         (i386_pe_end_cold_function): New function.
8128
8129 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
8130
8131         PR fortran/84565
8132         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8133         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8134
8135         PR c++/84704
8136         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8137         on tmp_var.
8138         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8139         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8140
8141         PR middle-end/84723
8142         * multiple_target.c: Include tree-inline.h and intl.h.
8143         (expand_target_clones): Diagnose and fail if node->definition and
8144         !tree_versionable_function_p (node->decl).
8145
8146 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
8147
8148         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8149         sprint_ul.
8150         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8151         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8152         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8153
8154 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
8155
8156         PR target/84710
8157         * combine.c (try_combine): Use reg_or_subregno instead of handling
8158         just paradoxical SUBREGs and REGs.
8159
8160 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
8161
8162         * config/arc/arc.c (arc_finalize_pic): Remove function.
8163         (arc_must_save_register): We use single base PIC register, remove
8164         checks to save/restore the PIC register.
8165         (arc_expand_prologue): Likewise.
8166         * config/arc/arc-protos.h (arc_set_default_type_attributes):
8167         Remove.
8168         (arc_verify_short): Likewise.
8169         (arc_attr_type): Likewise.
8170         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8171         (walk_stores): Likewise.
8172         (arc_address_cost): Make it static.
8173         (arc_verify_short): Likewise.
8174         (branch_dest): Likewise.
8175         (arc_attr_type): Likewise.
8176         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8177         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8178         (arc_final_prescan_insn): Remove inserting the nops due to
8179         hardware hazards.  It is done in reorg step.
8180         (insn_length_variant_t): Remove.
8181         (insn_length_parameters_t): Likewise.
8182         (arc_insn_length_parameters): Likewise.
8183         (arc_get_insn_variants): Likewise.
8184         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8185
8186 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
8187
8188         PR inline-asm/84683
8189         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8190         assertion failure.
8191
8192         PR tree-optimization/84687
8193         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8194         on new_node->decl.
8195         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8196
8197 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8198
8199         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8200         Rename to ppc_speculation_barrier.
8201         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8202         __builtin_ppc_speculation_barrier.
8203
8204 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
8205
8206         PR target/84700
8207         * combine.c (combine_simplify_rtx): Don't try to simplify if
8208         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8209         are equal to x.
8210
8211 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
8212
8213         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8214         to 32 bytes when compiling for POWER9.
8215
8216 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
8217
8218         PR target/84564
8219         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8220         regparm >= 3 with no arg reg available also for calls with
8221         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
8222
8223         PR target/84524
8224         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8225         orig,vex.
8226         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
8227
8228 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
8229
8230         PR target/84264
8231         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8232
8233 2018-03-05  Richard Biener  <rguenther@suse.de>
8234
8235         PR tree-optimization/84486
8236         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8237         When inserting a __builtin_assume_aligned call set the LHS
8238         SSA name alignment info accordingly.
8239
8240 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
8241
8242         PR tree-optimization/84114
8243         * config/aarch64/aarch64.c (aarch64_reassociation_width)
8244         Avoid reassociation of FLOAT_MODE addition.
8245
8246 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
8247
8248         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8249         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8250         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8251         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8252         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8253         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8254         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8255         and -mwbnoinvd.
8256         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8257         __builtin_ia32_wbinvd): New builtins.
8258         (SPECIAL_ARGS2): New.
8259         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8260         (SPECIAL_ARGS2): New.
8261         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8262         (ix86_valid_target_attribute_inner_p): Ditto.
8263         (ix86_init_mmx_sse_builtins): Add special_args2.
8264         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8265         TARGET_WBNOINVD_P): New.
8266         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8267         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8268         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8269         * config/i386/immintrin.h (_wbinvd): New intrinsic.
8270         * config/i386/pconfigintrin.h: New file.
8271         * config/i386/wbnoinvdintrin.h: Ditto.
8272         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8273         wbnoinvdintrin.h.
8274         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8275
8276 2018-03-05  Richard Biener  <rguenther@suse.de>
8277
8278         PR tree-optimization/84670
8279         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8280         member.
8281         (BB_VISITED_WITH_VISITED_SUCCS): New define.
8282         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8283         (compute_antic_aux): Only assert the number of values in ANTIC_IN
8284         doesn't grow if all successors (recursively) were visited at least
8285         once.
8286
8287 2018-03-05  Richard Biener  <rguenther@suse.de>
8288
8289         PR tree-optimization/84650
8290         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8291         if executed in the loop pipeline.
8292
8293 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
8294
8295         * doc/configfiles.texi (Configuration Files): Move info about
8296         conditionalizing $target-protos.h to...
8297         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
8298         differs from $target-protos.h.
8299
8300 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
8301             Chung-Ju Wu  <jasonwucj@gmail.com>
8302
8303         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8304         * config/nds32/nds32-multiple.md (setmemsi): Define.
8305         * config/nds32/nds32-memory-manipulation.c
8306         (nds32_gen_dup_4_byte_to_word_value): New.
8307         (emit_setmem_word_loop): New.
8308         (emit_setmem_byte_loop): New.
8309         (nds32_expand_setmem_loop): New.
8310         (nds32_expand_setmem_loop_v3m): New.
8311         (nds32_expand_setmem_unroll): New.
8312         (nds32_expand_setmem): New.
8313
8314 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
8315             Chung-Ju Wu  <jasonwucj@gmail.com>
8316
8317         * config/nds32/nds32-memory-manipulation.c
8318         (nds32_emit_load_store): New.
8319         (nds32_emit_post_inc_load_store): New.
8320         (nds32_emit_mem_move): New.
8321         (nds32_emit_mem_move_block): New.
8322         (nds32_expand_movmemsi_loop_unknown_size): New.
8323         (nds32_expand_movmemsi_loop_known_size): New.
8324         (nds32_expand_movmemsi_loop): New.
8325         (nds32_expand_movmemsi_unroll): New.
8326         (nds32_expand_movmemqi): Rename ...
8327         (nds32_expand_movmemsi): ... to this.
8328         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8329         (movmemsi): ... to this.
8330         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8331         (nds32_expand_movmemsi): ... to this.
8332
8333 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
8334             Monk Chiang  <sh.chiang04@gmail.com>
8335             Chung-Ju Wu  <jasonwucj@gmail.com>
8336
8337         * config/nds32/nds32-protos.h
8338         (nds32_expand_load_multiple): New arguments.
8339         (nds32_expand_store_multiple): Ditto.
8340         (nds32_valid_multiple_load_store): Rename ...
8341         (nds32_valid_multiple_load_store_p): ... to this.
8342         * config/nds32/nds32-memory-manipulation.c
8343         (nds32_expand_load_multiple): Refine implementation.
8344         (nds32_expand_store_multiple): Ditto.
8345         * config/nds32/nds32-multiple.md
8346         (load_multiple): Update nds32_expand_load_multiple interface.
8347         (store_multiple): Update nds32_expand_store_multiple interface.
8348         * config/nds32/nds32-predicates.c
8349         (nds32_valid_multiple_load_store): Rename ...
8350         (nds32_valid_multiple_load_store_p): ... to this and refine
8351         implementation.
8352         * config/nds32/predicates.md
8353         (nds32_load_multiple_and_update_address_operation): New predicate.
8354         (nds32_store_multiple_and_update_address_operation): New predicate.
8355
8356 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
8357             Chung-Ju Wu  <jasonwucj@gmail.com>
8358
8359         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8360         (combo): New attribute.
8361         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8362
8363 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
8364
8365         * config/nds32/nds32.opt: Change -mcmodel= default value.
8366
8367 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
8368             Monk Chiang  <sh.chiang04@gmail.com>
8369             Chung-Ju Wu  <jasonwucj@gmail.com>
8370
8371         * config/nds32/constants.md (unspec_element): New enum.
8372         * config/nds32/constraints.md (Umw): New constraint.
8373         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8374         * config/nds32/nds32-intrinsic.md: Likewise.
8375         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8376         (nds32_valid_smw_lwm_base_p): New.
8377         (nds32_output_smw_single_word): New.
8378         (nds32_output_lmw_single_word): New.
8379         (nds32_expand_unaligned_load): New.
8380         (nds32_expand_unaligned_store): New.
8381         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8382         (nds32_output_smw_single_word): Declare.
8383         (nds32_output_lmw_single_word): Declare.
8384         (nds32_expand_unaligned_load): Declare.
8385         (nds32_expand_unaligned_store): Declare.
8386         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8387         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8388         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8389         NDS32_BUILTIN_UASTORE_DW.
8390         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8391         predicate.
8392
8393 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
8394             Kito Cheng  <kito.cheng@gmail.com>
8395             Chung-Ju Wu  <jasonwucj@gmail.com>
8396
8397         * config/nds32/nds32-intrinsic.c
8398         (nds32_expand_builtin_null_ftype_reg): Delete.
8399         (nds32_expand_builtin_reg_ftype_imm): Ditto.
8400         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8401         (nds32_read_argument): New.
8402         (nds32_legitimize_target): Ditto.
8403         (nds32_legitimize_argument): Ditto.
8404         (nds32_check_constant_argument): Ditto.
8405         (nds32_expand_unop_builtin): Ditto.
8406         (nds32_expand_unopimm_builtin): Ditto.
8407         (nds32_expand_binop_builtin): Ditto.
8408         (nds32_builtin_decl_impl): Ditto.
8409         (builtin_description): Ditto.
8410         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8411         (nds32_init_builtins_impl): Ditto.
8412         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8413         (nds32_builtin_decl): New.
8414         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8415         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8416
8417 2018-03-02  Jeff Law  <law@redhat.com>
8418
8419         * reorg.c (stop_search_p): Handle DEBUG_INSN.
8420         (redundant_insn, fill_simple_delay_slots): Likewise.
8421         (fill_slots_from_thread): Likewise.
8422         * resource.c (mark_referenced_resources): Likewise.
8423         (mark_set_resources, find_dead_or_set_registers): Likewise.
8424
8425 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
8426
8427         * substring-locations.h (format_warning_va): Formatting fix for
8428         ATTRIBUTE_GCC_DIAG.
8429         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8430         argument.
8431         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8432         * substring-locations.c: Include intl.h.
8433         (format_warning_va): Turned into small wrapper around
8434         format_warning_n_va, renamed to ...
8435         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8436         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8437         use ngettext.
8438         (format_warning_at_substring_n): New function.
8439         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8440         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
8441         format_warning_at_substring with just a shorter name instead of
8442         const function pointer.
8443         (fmtwarn_n): New function.
8444         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8445         appropriate, get rid of all the fmtstr temporaries, move conditionals
8446         with G_() wrapped string literals directly into fmtwarn arguments,
8447         cast dir.len to (int), formatting fixes.
8448
8449 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
8450
8451         * doc/invoke.texi: Remove "Cilk Plus" references.
8452
8453 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
8454             Richard Biener  <rguenther@suse.de>
8455
8456         PR ipa/84628
8457         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8458         for error or warning attributes if CALL_FROM_THUNK_P is set.
8459         Formatting fixes.
8460
8461 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
8462
8463         PR target/56540
8464         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8465         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8466
8467         PR target/56540
8468         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8469         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8470
8471         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8472         instead of -1U in last predictors element's probability member.
8473
8474 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
8475
8476         PR ipa/83983
8477         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8478         arguments if they are comparable.
8479
8480 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
8481
8482         PR tree-optimization/84634
8483         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8484         masks and masked_loop_p with a single loop_masks, making sure it's
8485         null for bb vectorization.
8486
8487 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
8488
8489         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8490         (vect_analyze_data_ref_access): Use loop->safe_len rather than
8491         loop->force_vectorize to check whether there is no alias.
8492
8493 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
8494
8495         PR target/84614
8496         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8497         prototypes.
8498         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8499         comments.
8500         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8501         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8502         instead of a loop around prev_real_insn.
8503         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8504         prev_real_insn.
8505
8506         PR inline-asm/84625
8507         * config/i386/i386.c (ix86_print_operand): Use conditional
8508         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8509         zero vector.
8510
8511 2018-03-02  Richard Biener  <rguenther@suse.de>
8512
8513         PR tree-optimization/84427
8514         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8515         (bitmap_set_subtract_values): Rewrite to handle multiple
8516         exprs per value.
8517         (clean): Likewise.
8518         (prune_clobbered_mems): Likewise.
8519         (phi_translate): Take edge instead of pred/phiblock.
8520         (phi_translate_1): Likewise.
8521         (phi_translate_set): Likewise.  Insert all translated
8522         exprs for a value into the set, keeping possibly multiple
8523         expressions per value.
8524         (compute_antic_aux): Adjust for phi_translate changes.
8525         When intersecting union the expressions and prune those
8526         not in the final value set, keeping possibly multiple
8527         expressions per value.  Do not use value-insertion
8528         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8529         all expressions.  Add verification that the value-sets
8530         only shrink during iteration.
8531         (compute_partial_antic_aux): Adjust for the phi_translate changes.
8532         (do_pre_regular_insertion): Likewise.
8533         (do_pre_partial_partial_insertion): Likewise.
8534
8535 2018-03-02  Richard Biener  <rguenther@suse.de>
8536
8537         PR target/82005
8538         * config/darwin.c (saved_debug_info_level): New static global.
8539         (darwin_asm_lto_start): Disable debug info generation for LTO out.
8540         (darwin_asm_lto_end): Restore debug info generation settings.
8541
8542 2018-03-01  Martin Liska  <mliska@suse.cz>
8543
8544         PR sanitizer/82484
8545         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8546         volatile arguments.
8547
8548 2018-03-01  Richard Biener  <rguenther@suse.de>
8549
8550         PR debug/84645
8551         * dwarf2out.c (gen_variable_die): Properly handle late VLA
8552         type annotation with LTO when debug was disabled at compile-time.
8553
8554 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
8555
8556         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8557         XINT with INTVAL.
8558         (mips_final_postscan_insn): Likewise.
8559
8560 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
8561
8562         PR rtl-optimization/84528
8563         * alias.c (init_alias_target): Add commentary.
8564         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8565         a unique base value if the frame pointer is not eliminated
8566         to the stack pointer.
8567
8568 2018-03-01  Tom de Vries  <tom@codesourcery.com>
8569
8570         PR rtl-optimization/83327
8571         * lra-int.h (hard_regs_spilled_into): Declare.
8572         * lra.c (hard_regs_spilled_into): Define.
8573         (init_reg_info): Init hard_regs_spilled_into.
8574         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8575         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8576         (process_bb_lives): Handle hard_regs_spilled_into.
8577         (lra_create_live_ranges_1): Before doing liveness propagation, clear
8578         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8579
8580 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
8581
8582         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8583         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8584         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8585         * config/rs6000/aix72.h: New file.
8586
8587 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
8588
8589         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8590         instead of warning_at with conditional singular and plural messages
8591         where possible.
8592
8593         PR target/52991
8594         * stor-layout.c (update_alignment_for_field): For
8595         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8596         && !DECL_PACKED (field), do the alignment update, just use
8597         only desired_align instead of MAX (type_align, desired_align)
8598         as the alignment.
8599         (place_field): Don't do known_align < desired_align handling
8600         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8601         is non-NULL, instead do it after rli->prev_field handling and
8602         only if not within a bitfield word.  For DECL_PACKED (field)
8603         use type_align of BITS_PER_UNIT.
8604
8605 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
8606
8607         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8608         superfluous parentheses and trailing spaces.
8609
8610 2018-02-28  Richard Biener  <rguenther@suse.de>
8611
8612         PR tree-optimization/84584
8613         * graphite-scop-detection.c (scop_detection::add_scop): Discard
8614         SCoPs with fake exit edge.
8615
8616 2018-02-28  Martin Liska  <mliska@suse.cz>
8617
8618         PR testsuite/84597
8619         * timevar.c (timer::print): Fix format to properly print 100%
8620         values.
8621
8622 2018-02-28  Richard Biener  <rguenther@suse.de>
8623
8624         PR middle-end/84607
8625         * genmatch.c (capture_info::walk_match): Do not mark
8626         captured expressions without operands as expr_p given
8627         they act more like predicates and should be subject to
8628         "lost tail" side-effect preserving.
8629
8630 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
8631
8632         PR rtl-optimization/81611
8633         * auto-inc-dec.c (attempt_change): Move dead note from
8634         mem_insn if it's the next use of regno
8635         (find_address): Take address use of reg holding
8636         non-incremented value.  Add parm to limit search to the named
8637         reg only.
8638         (merge_in_block): Attempt to use a mem insn that is the next
8639         use of the original regno.
8640
8641 2018-02-27  Martin Sebor  <msebor@redhat.com>
8642
8643         PR c++/83871
8644         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8645         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8646
8647 2018-02-27  Martin Sebor  <msebor@redhat.com>
8648
8649         PR translation/84207
8650         * diagnostic-core.h (warning_n, error_n, inform_n): Change
8651         n argument to unsigned HOST_WIDE_INT.
8652         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8653         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
8654         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8655         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8656
8657 2018-02-27  Richard Biener  <rguenther@suse.de>
8658
8659         PR tree-optimization/84512
8660         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8661         Do not use the estimate returned from record_stmt_cost for
8662         the scalar iteration cost but sum properly using add_stmt_cost.
8663
8664 2018-02-27  Richard Biener  <rguenther@suse.de>
8665
8666         PR tree-optimization/84466
8667         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8668         Adjust last change to less strictly validate use operands.
8669
8670 2018-02-27  Martin Liska  <mliska@suse.cz>
8671
8672         PR gcov-profile/84548
8673         * gcov.c (process_file): Allow partial overlap and consider it
8674         also as group functions.
8675         (output_lines): Properly calculate range of lines for a group.
8676
8677 2018-02-27  Martin Liska  <mliska@suse.cz>
8678
8679         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8680         'ggc' suffixes.  Change first column width.
8681         (timer::print): Fix formatting of the column.
8682
8683 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
8684
8685         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8686         preserve inline entry blocks for the sake of debug inline
8687         entry point markers alone.
8688         (remove_unused_locals): Suggest in comments a better place to
8689         force the preservation of inline entry blocks that are
8690         otherwise unused, but do not preserve them.
8691
8692 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
8693
8694         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8695
8696 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
8697
8698         PR target/84039
8699         * config/i386/constraints.md (Bs): Replace
8700         ix86_indirect_branch_register with
8701         TARGET_INDIRECT_BRANCH_REGISTER.
8702         (Bw): Likewise.
8703         * config/i386/i386.md (indirect_jump): Likewise.
8704         (tablejump): Likewise.
8705         (*sibcall_memory): Likewise.
8706         (*sibcall_value_memory): Likewise.
8707         Peepholes of indirect call and jump via memory: Likewise.
8708         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8709         (*sibcall_value_GOT_32): Likewise.
8710         * config/i386/predicates.md (indirect_branch_operand): Likewise.
8711         (GOT_memory_operand): Likewise.
8712         (call_insn_operand): Likewise.
8713         (sibcall_insn_operand): Likewise.
8714         (GOT32_symbol_operand): Likewise.
8715         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8716
8717 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
8718
8719         PR rtl-optimization/83496
8720         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8721         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
8722         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8723         redundant insn, if any.
8724         (relax_delay_slots): Likewise.
8725         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8726
8727 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
8728
8729         PR tree-optimization/83965
8730         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8731         that grouped statements are part of a reduction chain.  Return
8732         true if the statement is not marked as a reduction itself but
8733         is part of a group.
8734         (vect_recog_dot_prod_pattern): Don't check whether the statement
8735         is part of a group here.
8736         (vect_recog_sad_pattern): Likewise.
8737         (vect_recog_widen_sum_pattern): Likewise.
8738
8739 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
8740
8741         PR debug/84545
8742         * final.c (rest_of_clean_state): Also look for calls inside sequences.
8743
8744 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
8745
8746         PR target/84530
8747         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8748         the bool argument.
8749         (ix86_output_indirect_function_return): New prototype.
8750         (ix86_split_simple_return_pop_internal): Likewise.
8751         * config/i386/i386.c (indirect_return_via_cx): New.
8752         (indirect_return_via_cx_bnd): Likewise.
8753         (indirect_thunk_name): Handle return va CX_REG.
8754         (output_indirect_thunk_function): Create alias for
8755         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8756         (ix86_output_indirect_jmp): Remove the bool argument.
8757         (ix86_output_indirect_function_return): New function.
8758         (ix86_split_simple_return_pop_internal): Likewise.
8759         * config/i386/i386.md (*indirect_jump): Don't pass false
8760         to ix86_output_indirect_jmp.
8761         (*tablejump_1): Likewise.
8762         (simple_return_pop_internal): Change it to define_insn_and_split.
8763         Call ix86_split_simple_return_pop_internal to split it for
8764         -mfunction-return=.
8765         (simple_return_indirect_internal): Call
8766         ix86_output_indirect_function_return instead of
8767         ix86_output_indirect_jmp.
8768
8769 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
8770
8771         PR bootstrap/84405
8772         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8773         memset and value initialization afterwards.
8774
8775 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
8776
8777         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8778
8779 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
8780
8781         PR target/84521
8782         * common/config/aarch64/aarch64-common.c
8783         (aarch_option_optimization_table[]): Switch
8784         off fomit-frame-pointer
8785
8786 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
8787             Chung-Ju Wu  <jasonwucj@gmail.com>
8788
8789         * config/nds32/nds32-multiple.md (load_multiple): Disallow
8790         volatile memory.
8791         (store_multiple): Ditto.
8792
8793 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
8794
8795         * config.gcc: Add --with-cpu support for nds32 target.
8796         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8797         * config/nds32/nds32.opt: Add -mcpu= option.
8798
8799 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
8800
8801         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8802         isel=yes): Warn for these deprecated options.
8803
8804 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
8805
8806         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8807         ISA_2_5_MASKS_EMBEDDED.
8808
8809 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
8810
8811         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8812         p->max as pointers rather than using iterative_hash_expr.
8813
8814 2018-02-23  Carl Love  <cel@us.ibm.com>
8815
8816         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8817         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8818         BU_P8V_OVERLOAD_2.
8819         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8820         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8821         P8V_BUILTIN_VEC_VUNSIGNED2.
8822
8823 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
8824
8825         PR target/81572
8826         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8827         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8828         LRA_UNKNOWN_ALT.
8829         * lra-constraints.c (curr_insn_transform): Set up
8830         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
8831         LRA_UNKNOWN_ALT.
8832         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8833         * lra-eliminations.c (spill_pseudos): Ditto.
8834         (process_insn_for_elimination): Ditto.
8835         * lra-lives.c (reg_early_clobber_p): Use the new macros.
8836         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8837         LRA_NON_CLOBBERED_ALT.
8838
8839 2018-02-22  Martin Sebor  <msebor@redhat.com>
8840
8841         PR tree-optimization/84480
8842         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8843         to maybe_diag_stxncpy_trunc.  Call it.
8844         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8845         from gimple_fold_builtin_strcpy.  Print inlining stack.
8846         (handle_builtin_stxncpy): Print inlining stack.
8847         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8848
8849 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
8850
8851         PR target/84176
8852         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8853         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8854         and -fcheck-pointer-bounds are used together.
8855         (indirect_thunk_prefix): New enum.
8856         (indirect_thunk_need_prefix): New function.
8857         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
8858         "_nt" instead of "_bnd" for NOTRACK prefix.
8859         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8860         (output_indirect_thunk_function): Likewise.
8861         (): Likewise.
8862         (ix86_code_end): Update output_indirect_thunk_function calls.
8863         (ix86_output_indirect_branch_via_reg): Replace
8864         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8865         (ix86_output_indirect_branch_via_push): Likewise.
8866         (ix86_output_function_return): Likewise.
8867         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8868         incompatible with -fcf-protection=branch and
8869         -fcheck-pointer-bounds.
8870
8871 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
8872
8873         PR target/83335
8874         * config/aarch64/aarch64.c (aarch64_print_address_internal):
8875         Change gcc_assert call to output_operand_lossage.
8876
8877 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
8878
8879         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8880
8881 2018-02-22  DJ Delorie  <dj@redhat.com>
8882             Sebastian Perta  <sebastian.perta@renesas.com>
8883             Oleg Endo  <olegendo@gcc.gnu.org>
8884
8885         * config/rx/rx.c (rx_rtx_costs): New function.
8886         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8887
8888 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8889
8890         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8891
8892 2018-02-22  Martin Liska  <mliska@suse.cz>
8893
8894         PR driver/83193
8895         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8896         Add "native" as a possible value.
8897
8898 2018-02-22  Martin Liska  <mliska@suse.cz>
8899
8900         PR driver/83193
8901         * config/i386/i386.c (ix86_option_override_internal):
8902         Add "native" as a possible value for -march and -mtune.
8903
8904 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
8905
8906         PR target/84502
8907         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8908         to all type variants.
8909
8910         PR tree-optimization/84503
8911         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8912         width as info->bitpos + info->bitsize - start.
8913         (merged_store_group::merge_overlapping): Simplify width computation.
8914         (check_no_overlap): New function.
8915         (imm_store_chain_info::try_coalesce_bswap): Compute expected
8916         start + width and last_order of the group, fail if check_no_overlap
8917         fails.
8918         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8919         to group if check_no_overlap fails.
8920
8921 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
8922
8923         * config/rs6000/altivec.md: Delete contraint arguments to
8924         define_expand, define_split, and define_peephole2, and in
8925         define_insn_and_split if always unused.
8926         * config/rs6000/darwin.md: Ditto.
8927         * config/rs6000/dfp.md: Ditto.
8928         * config/rs6000/rs6000.md: Ditto.
8929         * config/rs6000/sync.md: Ditto.
8930         * config/rs6000/vector.md: Ditto.
8931         * config/rs6000/vsx.md: Ditto.
8932
8933 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
8934
8935         * config/rs6000/altivec.md: Write output control strings as braced
8936         blocks instead of double-quoted strings.
8937         * config/rs6000/darwin.md: Ditto.
8938         * config/rs6000/rs6000.md: Ditto.
8939         * config/rs6000/vector.md: Ditto.
8940         * config/rs6000/vsx.md: Ditto.
8941
8942 2018-02-21  Jason Merrill  <jason@redhat.com>
8943
8944         PR c++/84314 - ICE with templates and fastcall attribute.
8945         * attribs.c (build_type_attribute_qual_variant): Remove assert.
8946
8947 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
8948
8949         * ipa-cp.c (determine_versionability): Fix comment typos.
8950
8951 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
8952
8953         PR c/84229
8954         * ipa-cp.c (determine_versionability): Do not version functions caling
8955         va_arg_pack.
8956
8957 2018-02-21  Martin Liska  <mliska@suse.cz>
8958
8959         PR driver/83193
8960         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8961         Add "native" as a possible value.
8962         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
8963         the macro when native cpu detection is available.
8964
8965 2018-02-21  Martin Liska  <mliska@suse.cz>
8966
8967         PR driver/83193
8968         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8969         Add "native" as a possible value.
8970         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8971         when native cpu detection is available.
8972
8973 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
8974             Martin Sebor  <msebor@redhat.com>
8975
8976         PR tree-optimization/84478
8977         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8978         false.
8979         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8980         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8981         support which is conservatively correct, for 2 only stay conservative
8982         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
8983         argument to the 2 argument get_range_strlen, adjust 6 arg
8984         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8985         false.
8986         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8987         (gimple_fold_builtin_strlen): Pass true as last argument to
8988         get_range_strlen.
8989
8990 2018-02-20  Martin Sebor  <msebor@redhat.com>
8991
8992         PR middle-end/84095
8993         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8994         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
8995         (builtin_memref::builtin_memref): Factor out parts into
8996         set_base_and_offset and call it.
8997
8998 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
8999
9000         PR middle-end/84406
9001         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9002         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9003         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
9004         search at the associated MODE_INT.
9005
9006 2018-02-20  Jeff Law  <law@redhat.com>
9007
9008         PR middle-end/82123
9009         PR tree-optimization/81592
9010         PR middle-end/79257
9011         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9012         for range data rather than using global data.
9013         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9014         range data rather than using global data.
9015         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9016         pass it to children as needed.
9017         (struct directive::fmtresult): Similarly.
9018         (struct directive::set_width): Similarly.
9019         (struct directive::set_precision): Similarly.
9020         (format_integer, format_directive, parse_directive): Similarly.
9021         (format_none): Accept unnamed vr_values parameter.
9022         (format_percent, format_floating, format_character): Similarly.
9023         (format_string, format_plain): Similarly.
9024         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9025         the EVRP range analyzer for range data rather than using global data.
9026         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9027         gimple-ssa-evrp-analyze.h
9028         (class sprintf_dom_walker): Add after_dom_children member function.
9029         Add evrp_range_analyzer member.
9030         (sprintf_dom_walker::before_dom_children): Call into the EVRP
9031         range analyzer as needed.
9032         (sprintf_dom_walker::after_dom_children): New member function.
9033         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9034         if not optimizing.
9035         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9036         (evrp_range_analyzer::pop_to_marker): Likewise.
9037
9038 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
9039
9040         PR tree-optimization/84419
9041         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9042         with the required type if its current type is compatible but
9043         different.
9044
9045 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
9046
9047         PR middle-end/82004
9048         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9049         after vectorization.
9050
9051 2018-02-20  Martin Liska  <mliska@suse.cz>
9052
9053         PR driver/83193
9054         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9055         possible values if we don't have a hint.
9056
9057 2018-02-20  Martin Liska  <mliska@suse.cz>
9058
9059         PR c/84310
9060         PR target/79747
9061         * final.c (shorten_branches): Build align_tab array with one
9062         more element.
9063         * opts.c (finish_options): Add alignment option limit check.
9064         (MAX_CODE_ALIGN): Likewise.
9065         (MAX_CODE_ALIGN_VALUE): Likewise.
9066         * doc/invoke.texi: Document maximum allowed option value for
9067         all -falign-* options.
9068
9069 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
9070
9071         PR target/84146
9072         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9073         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9074         * var-tracking.c (emit_note_insn_var_location): Remove all references
9075         to NOTE_INSN_CALL_ARG_LOCATION.
9076         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9077         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9078         Use copy_rtx_if_shared.
9079         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9080         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9081         (dwarf2out_var_location): Remove handling of
9082         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9083         on call_insn.
9084         * final.c (final_scan_insn): Remove all references to
9085         NOTE_INSN_CALL_ARG_LOCATION.
9086         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
9087         before dumping final insns.
9088         * except.c (emit_note_eh_region_end): Remove all references to
9089         NOTE_INSN_CALL_ARG_LOCATION.
9090         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9091         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9092         * config/arc/arc.c (hwloop_optimize): Likewise.
9093         * config/arm/arm.c (create_fix_barrier): Likewise.
9094         * config/s390/s390.c (s390_chunkify_start): Likewise.
9095         * config/sh/sh.c (find_barrier): Likewise.
9096         * config/i386/i386.c (rest_of_insert_endbranch,
9097         ix86_seh_fixup_eh_fallthru): Likewise.
9098         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9099         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9100         * config/frv/frv.c (frv_function_prologue): Likewise.
9101         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
9102         reg note.
9103         (note_outside_basic_block_p): Remove all references to
9104         NOTE_INSN_CALL_ARG_LOCATION.
9105         * gengtype.c (adjust_field_rtx_def): Likewise.
9106         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9107         Likewise.
9108         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9109         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9110
9111         PR c++/84444
9112         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9113         is ADDR_EXPR.
9114
9115         PR tree-optimization/84452
9116         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9117         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9118         is NULL.
9119
9120 2018-02-19  Martin Liska  <mliska@suse.cz>
9121
9122         PR sanitizer/82183
9123         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9124
9125 2018-02-19  Martin Liska  <mliska@suse.cz>
9126             Richard Sandiford  <richard.sandiford@linaro.org>
9127
9128         PR tree-optimization/82491
9129         * gimple-fold.c (get_base_constructor): Make earlier bail out
9130         to prevent ubsan.
9131
9132 2018-02-19  Carl Love  <cel@us.ibm.com>
9133
9134         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9135         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9136         BU_P8V_OVERLOAD_1.
9137         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9138         P8V_BUILTIN_VEC_NEG.
9139
9140 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
9141
9142         * config/rl78/rl78.md (movdf): New define expand.
9143
9144 2018-02-19  Martin Liska  <mliska@suse.cz>
9145
9146         PR other/80589
9147         * doc/invoke.texi: Fix typo.
9148         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9149
9150 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
9151
9152         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9153         handle rs6000_single_float and rs6000_double_float specially for
9154         e500 family CPUs.
9155
9156 2018-02-16  Jeff Law  <law@redhat.com>
9157
9158         * config/rx/rx.c (add_pop_cfi_notes): New function.;
9159         (pop_regs): Use it.
9160
9161 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
9162
9163         PR ipa/84425
9164         * ipa-inline.c (inline_small_functions): Fix a typo.
9165
9166 2018-02-16  Nathan Sidwell  <nathan@acm.org>
9167
9168         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9169
9170 2018-02-16  Carl Love  <cel@us.ibm.com>
9171
9172         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9173         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9174         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9175         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9176         expansion to P8V_BUILTIN_VEC_FLOAT2.
9177
9178 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
9179
9180         PR rtl-optimization/70023
9181         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9182         src_regno into account.
9183
9184 2018-02-16  Carl Love  <cel@us.ibm.com>
9185
9186         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9187         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9188         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9189         * config/rs6000/rs6000.c: Remove case statements for
9190         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9191         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9192         and P9V_BUILTIN_VEC_VINSERT4B.
9193         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9194         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9195         * config/rs6000/vsx.md:
9196         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9197         vec_insert4b.
9198
9199 2018-02-16  Carl Love  <cel@us.ibm.com>
9200
9201         * config/rs6000/altivec.h: Add builtin names vec_extract4b
9202         vec_insert4b.
9203         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9204         definitions.
9205         * config/rs6000/rs6000-c.c: Add the definitions for
9206         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9207         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9208         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9209         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
9210         definition for insert4b and define insn *insert3b_internal.
9211         * doc/extend.texi: Add documentation for vec_extract4b.
9212
9213 2018-02-16  Nathan Sidwell  <nathan@acm.org>
9214
9215         * doc/extend.texi (Backwards Compatibility): Mention friend
9216         injection.  Note for-scope is deprecated.
9217         * doc/invoke.texi (-ffriend-injection): Deprecate.
9218
9219 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
9220
9221         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9222         that moved to I2, also allow destinations that are a paradoxical
9223         subreg (instead of a normal reg).
9224
9225 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
9226
9227         PR target/83831
9228         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9229         to QImode.
9230
9231 2018-02-16  Richard Biener  <rguenther@suse.de>
9232
9233         PR tree-optimization/84037
9234         PR tree-optimization/84016
9235         PR target/82862
9236         * config/i386/i386.c (ix86_builtin_vectorization_cost):
9237         Adjust vec_construct for the fact we need additional higher latency
9238         128bit inserts for AVX256 and AVX512 vector builds.
9239         (ix86_add_stmt_cost): Scale vector construction cost for
9240         elementwise loads.
9241
9242 2018-02-16  Richard Biener  <rguenther@suse.de>
9243
9244         PR tree-optimization/84417
9245         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9246         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9247         (non_rewritable_lvalue_p): Likewise, use poly-ints.
9248
9249 2018-02-16  Martin Liska  <mliska@suse.cz>
9250
9251         PR sanitizer/84307
9252         * internal-fn.def (ASAN_CHECK): Set proper flags.
9253         (ASAN_MARK): Likewise.
9254
9255 2018-02-16  Julia Koval  <julia.koval@intel.com>
9256
9257         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9258         from PTA_CANNONLAKE.
9259
9260 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
9261
9262         PR target/84272
9263         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9264         Use ++iter rather than iter++ for std::list iterators.
9265         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
9266         defer deleting them until all nodes in the forest are processed.  Do
9267         free even leaf nodes.  Change to_process into auto_vec.
9268
9269         PR bootstrap/84405
9270         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9271         * vec.h (vec_default_construct): Use memset instead of placement new
9272         if BROKEN_VALUE_INITIALIZATION is defined.
9273         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9274         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9275         is defined.
9276
9277         PR rtl-optimization/83723
9278         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9279         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
9280         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
9281         recursive calls.
9282         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9283         callers.
9284         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9285
9286 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
9287
9288         PR rtl-optimization/81443
9289         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9290         from inner REGs to paradoxical SUBREGs.
9291
9292 2018-02-16  Richard Biener  <rguenther@suse.de>
9293
9294         PR tree-optimization/84399
9295         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9296         For operands we can analyze at their definition make sure we can
9297         analyze them at each use as well.
9298
9299 2018-02-16  Richard Biener  <rguenther@suse.de>
9300
9301         PR tree-optimization/84190
9302         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9303         volatile accesses if the decl isn't volatile.
9304
9305 2018-02-15  Jason Merrill  <jason@redhat.com>
9306
9307         PR c++/84314 - ICE with templates and fastcall attribute.
9308         * attribs.c (build_type_attribute_qual_variant): Don't clobber
9309         TYPE_CANONICAL on an existing type.
9310
9311 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
9312
9313         PR tree-optimization/84383
9314         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9315         dstoff nor call operand_equal_p if dstbase is NULL.
9316
9317         PR tree-optimization/84334
9318         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9319         also a CONSTANT_CLASS_P, punt.
9320
9321 2018-02-14  Jim Wilson  <jimw@sifive.com>
9322
9323         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9324         first SMALL_OPERAND check.  New local min_second_step.  Move assert
9325         to where locals are set.  Add TARGET_RVC support.
9326         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9327
9328 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
9329
9330         * doc/invoke.texi: Correct -Wformat-overflow code sample.
9331
9332 2018-02-14  Martin Sebor  <msebor@redhat.com>
9333
9334         PR tree-optimization/83698
9335         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9336         arrays constrain the offset range to their bounds.
9337         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9338         (builtin_access::overlap): Avoid setting the size of overlap if it's
9339         already been set.
9340         (maybe_diag_overlap): Also consider arrays when deciding what values
9341         of offsets to include in diagnostics.
9342
9343 2018-02-14  Martin Sebor  <msebor@redhat.com>
9344
9345         PR c/84108
9346         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9347         that correspond to the kind of a declaration.
9348
9349 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
9350
9351         PR target/83984
9352         * config/pa/pa.md: Load address of PIC label using the linkage table
9353         if the label is nonlocal.
9354
9355 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9356
9357         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9358         warning message if user requests -maltivec=be.
9359         * doc/invoke.texi: Document deprecation of -maltivec=be.
9360
9361 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
9362
9363         PR target/84220
9364         * config/rs6000/rs6000-c.c: Update definitions for
9365         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9366         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9367
9368 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
9369
9370         PR target/84239
9371         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9372         add _get_ssp intrinsics. Remove argument from
9373         __builtin_ia32_rdssp[d|q].
9374         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9375         * config/i386/i386-builtin.def: Remove argument from
9376         __builtin_ia32_rdssp[d|q].
9377         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9378         ix86_expand_special_args_builtin for _rdssp[d|q].
9379         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9380         Clear register before usage.
9381         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9382         Add documentation for new _get_ssp and _inc_ssp intrinsics.
9383
9384 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
9385
9386         PR tree-optimization/84357
9387         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9388         operand 1 of an ARRAY_REF too.
9389
9390 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
9391
9392         PR target/83831
9393         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9394         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9395         declarations.
9396         (set_of_reg): New struct.
9397         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9398         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9399         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9400         functions.
9401         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9402         Split into bitclr, bitset, bitinvert patterns if appropriate.
9403         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9404         use rx_fuse_in_memory_bitop.
9405         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9406         to named insn, correct maximum insn length.
9407
9408 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
9409
9410         PR target/79242
9411         * machmode.def: Define a complex mode for PARTIAL_INT.
9412         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9413         MODE_PARTIAL_INT.
9414         * doc/rtl.texi: Document CSPImode.
9415         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9416         handling.
9417         (msp430_hard_regno_nregs_with_padding): Likewise.
9418
9419 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
9420
9421         PR target/84279
9422         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9423
9424 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
9425
9426         PR rtl-optimization/84169
9427         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
9428         we generated a parallel as new i3 and we split that to new i2 and i3
9429         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
9430         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
9431         those to i2, not i1.  Partially rewrite this scan code.
9432
9433 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
9434
9435         PR c/82210
9436         * stor-layout.c (place_field): For variable length fields, adjust
9437         offset_align afterwards not just based on the field's alignment,
9438         but also on the size.
9439
9440         PR middle-end/84309
9441         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9442         of exps and logs in the use_exp2 case.
9443
9444 2018-02-13  Jeff Law  <law@redhat.com>
9445
9446         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9447         entry for "vector".
9448
9449         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9450         ARGS as unused.
9451
9452 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
9453
9454         PR debug/84342
9455         PR debug/84319
9456         * common.opt (gas-loc-support, gas-locview-support): New.
9457         (ginline-points, ginternal-reset-location-views): New.
9458         * doc/invoke.texi: Document them.  Use @itemx where intended.
9459         (gvariable-location-views): Adjust.
9460         * target.def (reset_location_view): New.
9461         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9462         (TARGET_RESET_LOCATION_VIEW): New.
9463         * doc/tm.texi: Rebuilt.
9464         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9465         (dwarf2out_default_as_locview_support): New.
9466         (output_asm_line_debug_info): Use option variables.
9467         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9468         (output_loc_list): Likewise.
9469         (add_high_low_attributes): Check option variables.
9470         Don't output entry view attribute in strict mode.
9471         (gen_inlined_subroutine_die): Check option variables.
9472         (dwarf2out_inline_entry): Likewise.
9473         (init_sections_and_labels): Likewise.
9474         (dwarf2out_early_finish): Likewise.
9475         (maybe_reset_location_view): New, from...
9476         (dwarf2out_var_location): ... here.  Call it.
9477         * debug.h (dwarf2out_default_as_loc_support): Declare.
9478         (dwarf2out_default_as_locview_support): Declare.
9479         * hooks.c (hook_int_rtx_insn_0): New.
9480         * hooks.h (hook_int_rtx_insn_0): Declare.
9481         * toplev.c (process_options): Take -gas-loc-support and
9482         -gas-locview-support from dwarf2out.  Enable
9483         -gvariable-location-views by default only with locview
9484         assembler support.  Enable -ginternal-reset-location-views by
9485         default only if the target defines the corresponding hook.
9486         Enable -ginline-points by default if location views are
9487         enabled; force it disabled if statement frontiers are
9488         disabled.
9489         * tree-inline.c (expand_call_inline): Check option variables.
9490         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9491
9492 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
9493
9494         PR tree-optimization/84321
9495         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9496         handling.  Also check whether the anti-range contains any values
9497         that satisfy the mask; switch to a VR_RANGE if not.
9498
9499 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
9500
9501         PR sanitizer/84340
9502         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9503
9504 2018-02-13  Martin Jambor  <mjambor@suse.cz>
9505
9506         PR c++/83990
9507         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9508         of call statements, also set location of a load to a temporary.
9509
9510 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
9511
9512         * config/rl78/rl78.c (add_vector_labels): New function.
9513         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9514         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9515         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9516         which checks that no arguments are passed.
9517         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9518         * doc/extend.texi: Documentation for the new attribute.
9519
9520 2018-02-13  Andreas Schwab  <schwab@suse.de>
9521
9522         * config/riscv/linux.h (CPP_SPEC): Define.
9523
9524 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
9525
9526         PR target/84335
9527         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9528         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9529         OPTION_MASK_ISA_AES as first argument to def_builtin_const
9530         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9531         instead of OPTION_MASK_ISA_PCLMUL as first argument to
9532         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9533         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9534         temporarily for AES and PCLMUL builtins.
9535
9536         PR tree-optimization/84339
9537         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9538         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9539         Formatting fixes.
9540
9541         PR middle-end/84309
9542         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9543         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9544         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9545         inline function.
9546         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9547         inline function.
9548         * omp-simd-clone.h: New file.
9549         * omp-simd-clone.c: Include omp-simd-clone.h.
9550         (expand_simd_clones): No longer static.
9551         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9552         cgraph.h and omp-simd-clone.h.
9553         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9554         (vect_recog_widen_shift_pattern): Formatting fix.
9555         (vect_pattern_recog_1): Don't check optab for calls.
9556
9557         PR target/84336
9558         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9559         operands[2] into a REG before using gen_lowpart on it.
9560
9561 2018-02-12  Jeff Law  <law@redhat.com>
9562
9563         PR target/83760
9564         * config/sh/sh.c (find_barrier): Consider a sibling call
9565         a barrier as well.
9566
9567         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9568         successfully back substituting a reg.
9569
9570 2018-02-12  Richard Biener  <rguenther@suse.de>
9571
9572         PR tree-optimization/84037
9573         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9574         parameter, move visited init to caller.
9575         (vect_slp_analyze_operations): Separate cost from validity
9576         check, initialize visited once for all instances.
9577         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9578         for all instances.
9579         * tree-vect-stmts.c (vect_model_simple_cost): Make early
9580         out an assert.
9581         (vect_model_promotion_demotion_cost): Likewise.
9582         (vectorizable_bswap): Guard cost modeling with !slp_node
9583         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9584         SLP stmts.
9585         (vectorizable_call): Likewise.
9586         (vectorizable_conversion): Likewise.
9587         (vectorizable_assignment): Likewise.
9588         (vectorizable_shift): Likewise.
9589         (vectorizable_operation): Likewise.
9590         (vectorizable_store): Likewise.
9591         (vectorizable_load): Likewise.
9592         (vectorizable_condition): Likewise.
9593         (vectorizable_comparison): Likewise.
9594
9595 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
9596
9597         PR sanitizer/84307
9598         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9599         (ASAN_MARK): Fix fnspec to account for return value, change pointer
9600         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9601
9602 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
9603
9604         PR middle-end/83665
9605         * params.def (inline-min-speedup): Increase from 8 to 15.
9606         (max-inline-insns-auto): Decrease from 40 to 30.
9607         * ipa-split.c (consider_split): Add some buffer for function to
9608         be considered inlining candidate.
9609         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9610         default values.
9611
9612 2018-02-12  Richard Biener  <rguenther@suse.de>
9613
9614         PR tree-optimization/84037
9615         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9616         matched stmts if we cannot swap the non-matched ones.
9617
9618 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
9619
9620         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9621         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9622         _mm_maskz_scalef_round_ss): New intrinsics.
9623         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9624         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9625         __builtin_ia32_scalefss_round): Remove.
9626         (__builtin_ia32_scalefsd_mask_round,
9627         __builtin_ia32_scalefss_mask_round): New intrinsics.
9628         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9629         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9630         ((match_operand:VF_128 2 "<round_nimm_predicate>"
9631         "<round_constraint>")): Changed to ...
9632         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9633         "<round_scalar_constraint>")): ... this.
9634         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9635         %0, %1, %2<round_op3>}"): Changed to ...
9636         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9637         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9638         %2<round_scalar_mask_op3>}"): ... this.
9639         * config/i386/subst.md (round_scalar_nimm_predicate): New.
9640
9641 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
9642
9643         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9644         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9645         (_mm_maskz_sqrt_round_ss): New intrinsics.
9646         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9647         (__builtin_ia32_sqrtsd_mask_round)
9648         (__builtin_ia32_sqrtss_mask_round): New builtins.
9649         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9650         (__builtin_ia32_sqrtss_round): Remove.
9651         (__builtin_ia32_sqrtsd_mask_round)
9652         (__builtin_ia32_sqrtss_mask_round): New builtins.
9653         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9654         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9655         ((match_operand:VF_128 1 "vector_operand"
9656         "xBm,<round_constraint>")): Changed to ...
9657         ((match_operand:VF_128 1 "vector_operand"
9658         "xBm,<round_scalar_constraint>")): ... this.
9659         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9660         %0, %2, %<iptr>1<round_op3>}): Changed to ...
9661         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9662         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9663         %<iptr>1<round_scalar_mask_op3>}): ... this.
9664         ((set_attr "prefix" "<round_prefix>")): Changed to ...
9665         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9666
9667 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
9668
9669         PR target/84266
9670         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9671         Cast vec_cmpeq result to correct type.
9672         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9673         Cast vec_cmpgt result to correct type.
9674
9675 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
9676
9677         * final.c (final_scan_insn_1): Renamed from...
9678         (final_scan_insn): ... this.  New wrapper, to recover
9679         seen from the outermost call in recursive ones.
9680         * config/sparc/sparc.c (output_return): Drop seen from call.
9681         (output_sibcall): Likewise.
9682         * config/visium/visium.c (output_branch): Likewise.
9683
9684 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
9685
9686         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9687         function label.
9688
9689 2018-02-10  Alan Modra  <amodra@gmail.com>
9690
9691         PR target/84300
9692         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9693         Specify LR as an input.
9694
9695 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
9696
9697         PR sanitizer/83987
9698         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9699         remove_member_access_dummy_vars): New functions.
9700         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9701         lower_omp_1, execute_lower_omp): Use them.
9702
9703         PR rtl-optimization/84308
9704         * shrink-wrap.c (spread_components): Release todo vector.
9705
9706 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
9707
9708         PR rtl-optimization/57193
9709         * ira-color.c (struct allocno_color_data): Add member
9710         conflict_allocno_hard_prefs.
9711         (update_conflict_allocno_hard_prefs): New.
9712         (bucket_allocno_compare_func): Add a preference based on
9713         conflict_allocno_hard_prefs.
9714         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9715         (color_allocnos): Remove a dead code.  Initiate
9716         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
9717
9718 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
9719
9720         PR target/84226
9721         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9722         constraint from =wa to wa.  Avoid a subreg on the output operand,
9723         instead use a pseudo and subreg it in a move.
9724         (p9_xxbrd_<mode>): Changed to ...
9725         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9726         (p9_xxbrd_v2df): New expander.
9727         (p9_xxbrw_<mode>): Changed to ...
9728         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9729         (p9_xxbrw_v4sf): New expander.
9730
9731 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
9732
9733         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9734
9735 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
9736
9737         PR target/83926
9738         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9739         multiply in 32-bit mode.
9740         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9741         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9742         mode.
9743
9744 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
9745
9746         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9747         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9748         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9749         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9750
9751 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
9752
9753         PR lto/84213
9754         * dwarf2out.c (is_trivial_indirect_ref): New function.
9755         (dwarf2out_late_global_decl): Do not generate a location
9756         attribute for variables that have a non-trivial DECL_VALUE_EXPR
9757         and that are not defined in the current unit.
9758
9759 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
9760
9761         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9762         instead of a libcall for UNORDERED.
9763
9764 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
9765
9766         PR target/82641
9767         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9768         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9769
9770 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
9771
9772         PR target/PR84295
9773         * config/s390/s390.c (s390_set_current_function): Invoke
9774         s390_indirect_branch_settings also if fndecl didn't change.
9775
9776 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
9777
9778         * config/rs6000/rs6000.md (blockage): Set length to zero.
9779
9780 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
9781
9782         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9783
9784 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
9785
9786         PR sanitizer/84285
9787         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9788         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9789         -static-lib*san.
9790
9791         PR debug/84252
9792         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9793         PARALLEL incoming that failed vt_get_decl_and_offset check.
9794
9795         PR middle-end/84237
9796         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9797         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9798         TREE_READONLY bit.
9799         (get_variable_section): For decls in named .bss* sections pass true as
9800         second argument to bss_initializer_p.
9801
9802 2018-02-09  Marek Polacek  <polacek@redhat.com>
9803             Jakub Jelinek  <jakub@redhat.com>
9804
9805         PR c++/83659
9806         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9807         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
9808         Sync some changes from cxx_fold_indirect_ref.
9809
9810 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
9811
9812         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9813         markers.
9814         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9815         (BLOCK_INLINE_ENTRY_LABEL): New.
9816         (dwarf2out_var_location): Disregard inline entry markers.
9817         (inline_entry_data): New struct.
9818         (inline_entry_data_hasher): New hashtable type.
9819         (inline_entry_data_hasher::hash): New.
9820         (inline_entry_data_hasher::equal): New.
9821         (inline_entry_data_table): New variable.
9822         (add_high_low_attributes): Add DW_AT_entry_pc and
9823         DW_AT_GNU_entry_view attributes if a pending entry is found
9824         in inline_entry_data_table.  Add old entry_pc attribute only
9825         if debug nonbinding markers are disabled.
9826         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9827         markers are enabled.
9828         (block_within_block_p, dwarf2out_inline_entry): New.
9829         (dwarf2out_finish): Check that no entries remained in
9830         inline_entry_data_table.
9831         * final.c (reemit_insn_block_notes): Handle inline entry notes.
9832         (final_scan_insn, notice_source_line): Likewise.
9833         (rest_of_clean_state): Skip inline entry markers.
9834         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9835         markers.
9836         * gimple.c (gimple_build_debug_inline_entry): New.
9837         * gimple.h (enum gimple_debug_subcode): Add
9838         GIMPLE_DEBUG_INLINE_ENTRY.
9839         (gimple_build_debug_inline_entry): Declare.
9840         (gimple_debug_inline_entry_p): New.
9841         (gimple_debug_nonbind_marker_p): Adjust.
9842         * insn-notes.def (INLINE_ENTRY): New.
9843         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9844         inline entry marker notes.
9845         (print_insn): Likewise.
9846         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9847         (INSN_DEBUG_MARKER_KIND): Likewise.
9848         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9849         * tree-inline.c (expand_call_inline): Build and insert
9850         debug_inline_entry stmt.
9851         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9852         inline entry blocks early, if nonbind markers are enabled.
9853         (dump_scope_block): Dump fragment info.
9854         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9855         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9856         (gimple_build_debug_inline_entry): New.
9857         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9858         Enable/disable inline entry points too.
9859         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9860         (DEBUG_INSN): Describe inline entry markers.
9861
9862         * common.opt (gvariable-location-views): New.
9863         (gvariable-location-views=incompat5): New.
9864         * config.in: Rebuilt.
9865         * configure: Rebuilt.
9866         * configure.ac: Test assembler for view support.
9867         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9868         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9869         * dwarf2out.c (var_loc_view): New typedef.
9870         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9871         (dwarf2out_locviews_in_attribute): New.
9872         (dwarf2out_locviews_in_loclist): New.
9873         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9874         (enum dw_line_info_opcode): Add LI_adv_address.
9875         (struct dw_line_info_table): Add view.
9876         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9877         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9878         (zero_view_p): New variable.
9879         (ZERO_VIEW_P): New macro.
9880         (output_asm_line_debug_info): New.
9881         (struct var_loc_node): Add view.
9882         (add_AT_view_list, AT_loc_list): New.
9883         (add_var_loc_to_decl): Add view param.  Test it against last.
9884         (new_loc_list): Add view params.  Record them.
9885         (AT_loc_list_ptr): Handle loc and view lists.
9886         (view_list_to_loc_list_val_node): New.
9887         (print_dw_val): Handle dw_val_class_view_list.
9888         (size_of_die): Likewise.
9889         (value_format): Likewise.
9890         (loc_list_has_views): New.
9891         (gen_llsym): Set vl_symbol too.
9892         (maybe_gen_llsym, skip_loc_list_entry): New.
9893         (dwarf2out_maybe_output_loclist_view_pair): New.
9894         (output_loc_list): Output view list or entries too.
9895         (output_view_list_offset): New.
9896         (output_die): Handle dw_val_class_view_list.
9897         (output_dwarf_version): New.
9898         (output_compilation_unit_header): Use it.
9899         (output_skeleton_debug_sections): Likewise.
9900         (output_rnglists, output_line_info): Likewise.
9901         (output_pubnames, output_aranges): Update version comments.
9902         (output_one_line_info_table): Output view numbers in asm comments.
9903         (dw_loc_list): Determine current endview, pass it to new_loc_list.
9904         Call maybe_gen_llsym.
9905         (loc_list_from_tree_1): Adjust.
9906         (add_AT_location_description): Create view list attribute if
9907         needed, check it's absent otherwise.
9908         (convert_cfa_to_fb_loc_list): Adjust.
9909         (maybe_emit_file): Call output_asm_line_debug_info for test.
9910         (dwarf2out_var_location): Reset views as needed.  Precompute
9911         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
9912         attribute.  Set view.
9913         (new_line_info_table): Reset next view.
9914         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9915         (dwarf2out_source_line): Likewise.  Output view resets and labels to
9916         the assembler, or select appropriate line info opcodes.
9917         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9918         (optimize_string_length): Catch it.  Adjust.
9919         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
9920         dw_val_class_view_list, and remove it if no longer needed.
9921         (hash_loc_list): Hash view numbers.
9922         (loc_list_hasher::equal): Compare them.
9923         (optimize_location_lists): Check whether a view list symbol is
9924         needed, and whether the locview attribute is present, and
9925         whether they match.  Remove the locview attribute if no longer
9926         needed.
9927         (index_location_lists): Call skip_loc_list_entry for test.
9928         (dwarf2out_finish): Call output_asm_line_debug_info for test.
9929         Use output_dwarf_version.
9930         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9931         (struct dw_val_node): Add val_view_list.
9932         * final.c (SEEN_NEXT_VIEW): New.
9933         (set_next_view_needed): New.
9934         (clear_next_view_needed): New.
9935         (maybe_output_next_view): New.
9936         (final_start_function): Rename to...
9937         (final_start_function_1): ... this.  Take pointer to FIRST,
9938         add SEEN parameter.  Emit param bindings in the initial view.
9939         (final_start_function): Reintroduce SEEN-less interface.
9940         (final): Rename to...
9941         (final_1): ... this.  Take SEEN parameter.  Output final pending
9942         next view at the end.
9943         (final): Reintroduce seen-less interface.
9944         (final_scan_insn): Output pending next view before switching
9945         sections or ending a block.  Mark the next view as needed when
9946         outputting variable locations.  Notify debug backend of section
9947         changes, and of location view changes.
9948         (rest_of_handle_final): Adjust.
9949         * toplev.c (process_options): Autodetect value for debug variable
9950         location views option.  Warn on incompat5 without -gdwarf-5.
9951         * doc/invoke.texi (gvariable-location-views): New.
9952         (gvariable-location-views=incompat5): New.
9953         (gno-variable-location-views): New.
9954
9955 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
9956
9957         PR tree-optimization/84136
9958         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9959         that the result of find_edge is non-NULL.
9960
9961 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
9962
9963         PR target/83008
9964         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9965         storing integer register in SImode.  Fix cost of 256 and 512
9966         byte aligned SSE register store.
9967
9968 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
9969
9970         * config/i386/i386.c (ix86_multiplication_cost): Fix
9971         multiplication cost for TARGET_AVX512DQ.
9972
9973 2018-02-08  Marek Polacek  <polacek@redhat.com>
9974
9975         PR tree-optimization/84238
9976         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9977         get_range_strlen.
9978
9979 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
9980
9981         PR tree-optimization/84265
9982         * tree-vect-stmts.c (vectorizable_store): Don't treat
9983         VMAT_CONTIGUOUS accesses as grouped.
9984         (vectorizable_load): Likewise.
9985
9986 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
9987
9988         PR tree-optimization/81635
9989         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9990         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9991         (test_round_for_mask): New functions.
9992         (wide_int_cc_tests): Call test_round_for_mask.
9993         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9994         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9995         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9996         range returned by get_range_info.
9997
9998 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
9999
10000         PR ipa/81360
10001         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10002         * symtab.c: Include builtins.h
10003         (symtab_node::output_to_lto_symbol_table_p): Move here
10004         from lto-streamer-out.c:output_symbol_p.
10005         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10006         (output_symbol_p): Move all logic to symtab.c
10007         (produce_symtab): Update.
10008
10009 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
10010
10011         * config/s390/s390-opts.h (enum indirect_branch): Define.
10012         * config/s390/s390-protos.h (s390_return_addr_from_memory)
10013         (s390_indirect_branch_via_thunk)
10014         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10015         (enum s390_indirect_branch_type): Define.
10016         * config/s390/s390.c (struct s390_frame_layout, struct
10017         machine_function): Remove.
10018         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10019         (indirect_branch_table_label_no, indirect_branch_table_name):
10020         Define variables.
10021         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10022         (enum s390_indirect_branch_option): Define.
10023         (s390_return_addr_from_memory): New function.
10024         (s390_handle_string_attribute): New function.
10025         (s390_attribute_table): Add new attribute handler.
10026         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10027         (s390_indirect_branch_via_thunk): New function.
10028         (s390_indirect_branch_via_inline_thunk): New function.
10029         (s390_function_ok_for_sibcall): When jumping via thunk disallow
10030         sibling call optimization for non z10 compiles.
10031         (s390_emit_call): Force indirect branch target to be a single
10032         register.  Add r1 clobber for non-z10 compiles.
10033         (s390_emit_epilogue): Emit return jump via return_use expander.
10034         (s390_reorg): Handle JUMP_INSNs as execute targets.
10035         (s390_option_override_internal): Perform validity checks for the
10036         new command line options.
10037         (s390_indirect_branch_attrvalue): New function.
10038         (s390_indirect_branch_settings): New function.
10039         (s390_set_current_function): Invoke s390_indirect_branch_settings.
10040         (s390_output_indirect_thunk_function):  New function.
10041         (s390_code_end): Implement target hook.
10042         (s390_case_values_threshold): Implement target hook.
10043         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10044         macros.
10045         * config/s390/s390.h (struct s390_frame_layout)
10046         (struct machine_function): Move here from s390.c.
10047         (TARGET_INDIRECT_BRANCH_NOBP_RET)
10048         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10049         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10050         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10051         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10052         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10053         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10054         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10055         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10056         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10057         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10058         (mnemonic attribute): Add values which aren't recognized
10059         automatically.
10060         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10061         pattern for branch conversion.  Fix mnemonic attribute.
10062         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10063         indirect branch via thunk if requested.
10064         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10065         ("*indirect_jump"): Disable for branch conversion using out of
10066         line thunks.
10067         ("indirect_jump_via_thunk<mode>_z10")
10068         ("indirect_jump_via_thunk<mode>")
10069         ("indirect_jump_via_inlinethunk<mode>_z10")
10070         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10071         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10072         ("casesi_jump_via_inlinethunk<mode>_z10")
10073         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10074         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10075         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10076         ("*indirect2_jump"): Disable for branch conversion.
10077         ("casesi_jump"): Turn into expander and expand patterns for branch
10078         conversion.
10079         ("return_use"): New expander.
10080         ("*return"): Emit return via thunk and rename it to ...
10081         ("*return<mode>"): ... this one.
10082         * config/s390/s390.opt: Add new options and and enum for the
10083         option values.
10084
10085 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
10086
10087         * lra-constraints.c (match_reload): Unconditionally use
10088         gen_lowpart_SUBREG, rather than selecting between that
10089         and equivalent gen_rtx_SUBREG code.
10090
10091 2018-02-08  Richard Biener  <rguenther@suse.de>
10092
10093         PR tree-optimization/84233
10094         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10095         changed flag instead of boguously re-using phi_inserted.
10096
10097 2018-02-08  Martin Jambor  <mjambor@suse.cz>
10098
10099         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10100         static local variables.
10101
10102 2018-02-08  Richard Biener  <rguenther@suse.de>
10103
10104         PR tree-optimization/84278
10105         * tree-vect-stmts.c (vectorizable_store): When looking for
10106         smaller vector types to perform grouped strided loads/stores
10107         make sure the mode is supported by the target.
10108         (vectorizable_load): Likewise.
10109
10110 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
10111
10112         * config/aarch64/aarch64.c (aarch64_components_for_bb):
10113         Increase LDP/STP opportunities by adding adjacent callee-saves.
10114
10115 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
10116
10117         PR rtl-optimization/84068
10118         PR rtl-optimization/83459
10119         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10120
10121 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
10122
10123         PR tree-optimization/84224
10124         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10125         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10126         non-zero arguments.
10127
10128 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
10129
10130         PR target/84113
10131         * config/rs6000/altivec.md (*restore_world): Remove LR use.
10132         * config/rs6000/predicates.md (restore_world_operation): Adjust op
10133         count, remove one USE.
10134
10135 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
10136
10137         * doc/install.texi (Configuration): Document the
10138         --with-long-double-format={ibm,ieee} PowerPC configuration
10139         options.
10140
10141         PR target/84154
10142         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10143         Convert from define_expand to be define_insn_and_split.  Rework
10144         float/double/_Float128 conversions to QI/HI/SImode to work with
10145         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
10146         conversions to QI/HImode types did a store and then a load to
10147         truncate the value.  For conversions to VSX registers, don't split
10148         the insn, instead emit the code directly.  Use the code iterator
10149         any_fix to combine signed and unsigned conversions.
10150         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10151         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10152         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10153         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10154         (fix_<mode>di2_hw): Likewise.
10155         (fixuns_<mode>di2_hw): Likewise.
10156         (fix_<mode>si2_hw): Likewise.
10157         (fixuns_<mode>si2_hw): Likewise.
10158         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10159         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10160         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10161         fix<uns>_trunc<SFDF:mode>si2_p8.
10162         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10163         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10164         (fix<uns>_<mode>_mem): Likewise.
10165         (fctiw<u>z_<mode>_mem): Likewise.
10166         (fix<uns>_<mode>_mem): Likewise.
10167         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10168         the register allocator from doing a direct move to the GPRs to do
10169         a store, and instead use the ISA 3.0 store byte/half-word from
10170         vector register instruction.  For IEEE 128-bit floating point,
10171         also optimize stores of 32-bit ints.
10172         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10173
10174 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
10175
10176         * genextract.c (push_pathstr_operand): New function to support
10177         [a-zA-Z].
10178         (walk_rtx): Call push_pathstr_operand.
10179         (print_path): Support [a-zA-Z].
10180
10181 2018-02-07  Richard Biener  <rguenther@suse.de>
10182
10183         PR tree-optimization/84037
10184         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10185         (cse_and_gimplify_to_preheader): Declare.
10186         (vect_get_place_in_interleaving_chain): Likewise.
10187         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10188         ivexpr_map.
10189         (_loop_vec_info::~_loop_vec_info): Delete it.
10190         (cse_and_gimplify_to_preheader): New function.
10191         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10192         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10193         (vectorizable_load): Likewise.  For grouped stores always base
10194         the IV on the first element.
10195         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10196         condition before gimplifying.
10197
10198 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
10199
10200         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10201         *DIV_EXPR and *MOD_EXPR.
10202
10203 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
10204
10205         PR target/84248
10206         * config/i386/i386.c (ix86_option_override_internal): Mask out
10207         the CF_SET bit when checking -fcf-protection.
10208
10209 2018-02-07  Tom de Vries  <tom@codesourcery.com>
10210
10211         PR libgomp/84217
10212         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10213         enough.
10214
10215 2018-02-07  Richard Biener  <rguenther@suse.de>
10216
10217         PR tree-optimization/84204
10218         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10219         this place.
10220
10221         PR tree-optimization/84205
10222         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10223         special-case isl_ast_op_zdiv_r.
10224
10225         PR tree-optimization/84223
10226         * graphite-scop-detection.c (gather_bbs::before_dom_children):
10227         Only add conditions from within the region.
10228         (gather_bbs::after_dom_children): Adjust.
10229
10230 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
10231
10232         PR target/84209
10233         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10234         * config/avr/avr.md: Only post-reload split REG-REG moves if
10235         either register is GENERAL_REG_P.
10236
10237 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
10238
10239         PR tree-optimization/84235
10240         * tree-ssa-scopedtables.c
10241         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10242         if the subtraction is performed in floating point type where NaNs are
10243         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10244         build 1.  Formatting fix.
10245
10246 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
10247
10248         PR target/84146
10249         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10250         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10251         and skip it regardless of bb boundaries.  Use CALL_P macro,
10252         don't test INSN_P (insn) together with CALL_P or JUMP_P check
10253         unnecessarily, formatting fix.
10254
10255 2018-02-06  Michael Collison  <michael.collison@arm.com>
10256
10257         * config/arm/thumb2.md:
10258         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10259         (*thumb_mov_notscc): Ditto.
10260
10261 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
10262
10263         PR target/84154
10264         * config/rs6000/rs6000.md (su code attribute): Use "u" for
10265         unsigned_fix, not "s".
10266
10267 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10268
10269         * configure.ac (gcc_fn_eh_frame_ro): New function.
10270         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10271         correct .eh_frame permissions.
10272         * configure: Regenerate.
10273
10274 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
10275
10276         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10277         irrelevant options.
10278
10279 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10280
10281         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10282         Display warning message for -mno-speculate-indirect-jumps.
10283
10284 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
10285
10286         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10287         Undocumented.
10288         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10289
10290 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
10291
10292         PR tree-optimization/84225
10293         * tree-eh.c (find_trapping_overflow): Only call
10294         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10295
10296 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
10297
10298         PR target/84145
10299         * config/i386/i386.c: Reimplement the check of possible options
10300         -mibt/-mshstk conbination. Change error messages.
10301         * doc/invoke.texi: Fix a typo: remove extra '='.
10302
10303 2018-02-06  Marek Polacek  <polacek@redhat.com>
10304
10305         PR tree-optimization/84228
10306         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10307
10308 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
10309
10310         PR target/82641
10311         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10312         emitted arch directives.
10313         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10314         __ARM_FEATURE_COPROC before changing architectures.
10315
10316 2018-02-06  Richard Biener  <rguenther@suse.de>
10317
10318         * config/i386/i386.c (print_reg): Fix typo.
10319         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10320
10321 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
10322
10323         * configure: Regenerate.
10324
10325 2018-02-05  Martin Sebor  <msebor@redhat.com>
10326
10327         PR tree-optimization/83369
10328         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10329         inlining context.
10330
10331 2018-02-05  Martin Liska  <mliska@suse.cz>
10332
10333         * doc/invoke.texi: Cherry-pick upstream r323995.
10334
10335 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
10336
10337         * ira.c (ira_init_register_move_cost): Adjust comment.
10338
10339 2018-02-05  Martin Liska  <mliska@suse.cz>
10340
10341         PR gcov-profile/84137
10342         * doc/gcov.texi: Fix typo in documentation.
10343
10344 2018-02-05  Martin Liska  <mliska@suse.cz>
10345
10346         PR gcov-profile/83879
10347         * doc/gcov.texi: Document necessity of --dynamic-list-data when
10348         using dlopen functionality.
10349
10350 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
10351
10352         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10353         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10354         _mm_maskz_range_ss, _mm_mask_range_round_ss,
10355         _mm_maskz_range_round_ss): New intrinsics.
10356         (__builtin_ia32_rangesd128_round)
10357         (__builtin_ia32_rangess128_round): Remove.
10358         (__builtin_ia32_rangesd128_mask_round,
10359         __builtin_ia32_rangess128_mask_round): New builtins.
10360         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10361         __builtin_ia32_rangess128_round): Remove.
10362         (__builtin_ia32_rangesd128_mask_round,
10363         __builtin_ia32_rangess128_mask_round): New builtins.
10364         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10365         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10366         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10367         "<round_saeonly_constraint>")): Changed to ...
10368         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10369         "<round_saeonly_scalar_constraint>")): ... this.
10370         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10371         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10372         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10373         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10374         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10375
10376 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
10377
10378         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10379         options.
10380         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10381         Remove all values except native, 8540 and 8548.
10382
10383 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
10384
10385         * config/i386/i386.c (ix86_output_function_return): Pass
10386         INVALID_REGNUM, instead of -1, as invalid register number to
10387         indirect_thunk_name and output_indirect_thunk.
10388
10389 2018-02-02  Julia Koval  <julia.koval@intel.com>
10390
10391         * config.gcc: Add -march=icelake.
10392         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10393         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10394         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10395         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10396         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10397         (processor_target_table): Add icelake.
10398         (ix86_option_override_internal): Handle new PTAs.
10399         (get_builtin_code_for_version): Handle icelake.
10400         (M_INTEL_COREI7_ICELAKE): New.
10401         (fold_builtin_cpu): Handle icelake.
10402         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10403         * doc/invoke.texi: Add -march=icelake.
10404
10405 2018-02-02  Julia Koval  <julia.koval@intel.com>
10406
10407         * config/i386/i386.c (ix86_option_override_internal): Change flags type
10408         to wide_int_bitmask.
10409         * wide-int-bitmask.h: New.
10410
10411 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
10412
10413         PR target/84066
10414         * config/i386/i386.md: Replace Pmode with word_mode in
10415         builtin_setjmp_setup and builtin_longjmp to support x32.
10416
10417 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
10418
10419         PR target/56010
10420         PR target/83743
10421         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10422         #include "opts.h".
10423         (rs6000_supported_cpu_names): New static variable.
10424         (linux_cpu_translation_table): Likewise.
10425         (elf_platform) <cpu>: Define new static variable and use it.
10426         Translate kernel AT_PLATFORM name to canonical name if needed.
10427         Error if platform name is unknown.
10428
10429 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
10430
10431         PR target/84089
10432         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10433
10434 2018-02-01  Jeff Law  <law@redhat.com>
10435
10436         PR target/84128
10437         * config/i386/i386.c (release_scratch_register_on_entry): Add new
10438         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
10439         the scratch if RELEASE_VIA_POP is false.
10440         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10441         If we have to save a temporary register, decrement SIZE appropriately.
10442         Pass new arguments to release_scratch_register_on_entry.
10443         (ix86_adjust_stack_and_probe): Likewise.
10444         (ix86_emit_probe_stack_range): Pass new arguments to
10445         release_scratch_register_on_entry.
10446
10447 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
10448
10449         PR rtl-optimization/84157
10450         * combine.c (change_zero_ext): Use REG_P predicate in
10451         front of HARD_REGISTER_P predicate.
10452
10453 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
10454
10455         * config/avr/avr.c (avr_option_override): Move disabling of
10456         -fdelete-null-pointer-checks to...
10457         * common/config/avr/avr-common.c (avr_option_optimization_table):
10458         ...here.
10459
10460 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10461
10462         PR tree-optimization/81635
10463         * tree-data-ref.c (split_constant_offset_1): For types that
10464         wrap on overflow, try to use range info to prove that wrapping
10465         cannot occur.
10466
10467 2018-02-01  Renlin Li  <renlin.li@arm.com>
10468
10469         PR target/83370
10470         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10471         TAILCALL_ADDR_REGS.
10472         (aarch64_register_move_cost): Likewise.
10473         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10474         TAILCALL_ADDR_REGS.
10475         (REG_CLASS_NAMES): Likewise.
10476         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10477         TAILCALL_ADDR_REGS. Remove IP registers.
10478         * config/aarch64/aarch64.md (Ucs): Update register constraint.
10479
10480 2018-02-01  Richard Biener  <rguenther@suse.de>
10481
10482         * domwalk.h (dom_walker::dom_walker): Add additional constructor
10483         for specifying RPO order and allow NULL for that.
10484         * domwalk.c (dom_walker::dom_walker): Likewise.
10485         (dom_walker::walk): Handle NULL RPO order.
10486         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10487         in RPO order.
10488         (rewrite_update_dom_walker): Likewise.
10489         (mark_def_dom_walker): Likewise.
10490
10491 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10492
10493         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10494         (aarch64_maybe_expand_sve_subreg_move): Declare.
10495         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10496         * config/aarch64/predicates.md (aarch64_any_register_operand): New
10497         predicate.
10498         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10499         that are semantically a reverse operation.
10500         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10501         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10502         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10503         functions.
10504         (aarch64_can_change_mode_class): For big-endian, forbid changes
10505         between two SVE modes if they have different element sizes.
10506
10507 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10508
10509         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10510         the TImode handling for big-endian targets.
10511
10512 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10513
10514         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10515         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
10516         not just bytes.
10517         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10518         Remove BSWAP handing for big-endian targets and use the form of
10519         LD1RQ appropariate for the mode.
10520
10521 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10522
10523         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10524         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10525         duplicated element.
10526
10527 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10528
10529         PR tearget/83845
10530         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10531         check for operands that need to go through aarch64_sve_reload_be.
10532
10533 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
10534
10535         PR tree-optimization/81661
10536         PR tree-optimization/84117
10537         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10538         * tree-eh.c: Include gimplify.h.
10539         (find_trapping_overflow, replace_trapping_overflow,
10540         rewrite_to_non_trapping_overflow): New functions.
10541         * tree-vect-loop.c: Include tree-eh.h.
10542         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10543         * tree-data-ref.c: Include tree-eh.h.
10544         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10545
10546 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
10547
10548         PR rtl-optimization/84123
10549         * combine.c (change_zero_ext): Check if hard register satisfies
10550         can_change_dest_mode before calling gen_lowpart_SUBREG.
10551
10552 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
10553
10554         PR target/82444
10555         * ira.c (ira_init_register_move_cost): Remove assert.
10556
10557 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
10558
10559         PR rtl-optimization/84071
10560         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10561         * doc/tm.texi: Regenerate.
10562
10563 2018-01-31  Richard Biener  <rguenther@suse.de>
10564
10565         PR tree-optimization/84132
10566         * tree-data-ref.c (analyze_miv_subscript): Properly
10567         check whether evolution_function_is_affine_multivariate_p
10568         before calling gcd_of_steps_may_divide_p.
10569
10570 2018-01-31  Julia Koval  <julia.koval@intel.com>
10571
10572         PR target/83618
10573         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10574         * config/i386/i386.md (rdpid_rex64) New.
10575         (rdpid): Make 32bit only.
10576
10577 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
10578
10579         PR lto/84105
10580         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10581         an IDENTIFIER_NODE for FUNCTION_TYPE's.
10582
10583 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
10584
10585         Revert
10586         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
10587
10588         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10589
10590 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
10591
10592         PR rtl-optimization/84071
10593         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10594         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10595
10596 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
10597
10598         * config/arc/arc.c (arc_handle_aux_attribute): New function.
10599         (arc_attribute_table): Add 'aux' attribute.
10600         (arc_in_small_data_p): Consider aux like variables.
10601         (arc_is_aux_reg_p): New function.
10602         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10603         (arc_get_aux_arg): New function.
10604         (prepare_move_operands): Handle aux-register access.
10605         (arc_handle_aux_attribute): New function.
10606         * doc/extend.texi (ARC Variable attributes): Add subsection.
10607
10608 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
10609
10610         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10611         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10612         (arc_attribute_table): Add 'uncached' attribute.
10613         (arc_print_operand): Print '.di' flag for uncached memory
10614         accesses.
10615         (arc_in_small_data_p): Do not consider for small data the uncached
10616         types.
10617         (arc_is_uncached_mem_p): New function.
10618         * config/arc/predicates.md (compact_store_memory_operand): Check
10619         for uncached memory accesses.
10620         (nonvol_nonimm_operand): Likewise.
10621         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10622
10623 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
10624
10625         PR c/84100
10626         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10627         falign-loops=): Add Optimization flag.
10628
10629 2018-01-30  Jeff Law  <law@redhat.com>
10630
10631         PR target/84064
10632         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10633         INT_REGISTERS_SAVED.  Check it prior to calling
10634         get_scratch_register_on_entry.
10635         (ix86_adjust_stack_and_probe): Similarly.
10636         (ix86_emit_probe_stack_range): Similarly.
10637         (ix86_expand_prologue): Corresponding changes.
10638
10639 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10640
10641         PR target/40411
10642         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10643         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10644
10645 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
10646
10647         PR target/84112
10648         * lra-constraints.c (curr_insn_transform): Process AND in the
10649         address.
10650
10651 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
10652
10653         PR rtl-optimization/83986
10654         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10655         dependence against last_pending_memory_flush in addition to
10656         pending_jump_insns.
10657
10658 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
10659
10660         PR tree-optimization/81611
10661         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10662         copies.
10663
10664 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
10665
10666         PR target/83758
10667         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10668         a reg rtx.
10669
10670 2018-01-30  Richard Biener  <rguenther@suse.de>
10671             Jakub Jelinek  <jakub@redhat.com>
10672
10673         PR tree-optimization/84111
10674         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10675         inner loops added during recursion, as they don't have up-to-date
10676         SSA form.
10677
10678 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
10679
10680         PR ipa/81360
10681         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10682         (can_inline_edge_by_limits_p): ... here.
10683         (can_early_inline_edge_p, check_callers,
10684         update_caller_keys, update_callee_keys, recursive_inlining,
10685         add_new_edges_to_heap, speculation_useful_p,
10686         inline_small_functions,
10687         inline_small_functions, flatten_function,
10688         inline_to_all_callers_1): Update.
10689
10690 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
10691
10692         * profile-count.c (profile_count::combine_with_ipa_count): Handle
10693         zeros correctly.
10694
10695 2018-01-30  Richard Biener  <rguenther@suse.de>
10696
10697         PR tree-optimization/83008
10698         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10699         invariant and constant vector uses in stmts when they need
10700         more than one stmt.
10701
10702 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10703
10704         PR bootstrap/84017
10705         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10706         * configure: Regenerate.
10707
10708 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
10709
10710         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10711         pattern.
10712         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10713         Use gen_rtx_REG rather than gen_lowpart.
10714
10715 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
10716
10717         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10718         rather than 0 when creating partial subregs.
10719
10720 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
10721
10722         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10723         of usage.
10724
10725 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
10726
10727         PR target/81550
10728         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10729         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10730         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10731         flags.  This restores the settings used before the 2017-07-24.
10732         Turning off pre increment/decrement/modify allows IVOPTS to
10733         optimize DF/SF loops where the index is an int.
10734
10735 2018-01-29  Richard Biener  <rguenther@suse.de>
10736             Kelvin Nilsen  <kelvin@gcc.gnu.org>
10737
10738         PR bootstrap/80867
10739         * tree-vect-stmts.c (vectorizable_call): Don't call
10740         targetm.vectorize_builtin_md_vectorized_function if callee is
10741         NULL.
10742
10743 2018-01-22  Carl Love  <cel@us.ibm.com>
10744
10745         * doc/extend.tex: Fix typo in second arg in
10746         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10747
10748 2018-01-29  Richard Biener  <rguenther@suse.de>
10749
10750         PR tree-optimization/84086
10751         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10752         (flush_ssaname_freelist): When SSA names were released reset
10753         the SCEV hash table.
10754
10755 2018-01-29  Richard Biener  <rguenther@suse.de>
10756
10757         PR tree-optimization/84057
10758         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10759         removed paths when removing edges.
10760
10761 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
10762
10763         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10764         -mfunction-return=@var{choice}.
10765
10766 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10767
10768         PR diagnostic/84034
10769         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10770         Handle CR like TAB.
10771         (layout::print_source_line): Likewise.
10772         (test_get_line_width_without_trailing_whitespace): Add test cases.
10773
10774 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
10775
10776         PR middle-end/84040
10777         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10778         debug insns.
10779
10780 2018-01-26  Jim Wilson  <jimw@sifive.com>
10781
10782         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10783
10784         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10785         specified.
10786
10787 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10788
10789         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10790         and CMP + SUB-immediate -> SUBS.
10791
10792 2018-01-26  Martin Sebor  <msebor@redhat.com>
10793
10794         PR tree-optimization/83896
10795         * tree-ssa-strlen.c (get_string_len): Rename...
10796         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
10797         Avoid assuming length is constant.
10798         (handle_char_store): Use HOST_WIDE_INT for string length.
10799
10800 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
10801
10802         PR target/81763
10803         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10804         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10805
10806 2018-01-26  Richard Biener  <rguenther@suse.de>
10807
10808         PR rtl-optimization/84003
10809         * dse.c (record_store): Only record redundant stores when
10810         the earlier store aliases at least all accesses the later one does.
10811
10812 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
10813
10814         PR rtl-optimization/83985
10815         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10816         REG_CFA_RESTORE insns.
10817         (delete_unmarked_insns): Don't ignore separate shrink wrapping
10818         REG_CFA_RESTORE insns here.
10819
10820         PR c/83989
10821         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10822         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10823
10824 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10825
10826         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10827         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10828         (arc_init): Likewise.
10829         (arc_override_options): Likewise.
10830         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10831         value.
10832         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10833         support.
10834         * config/arc/arc.h (TARGET_DBNZ): Define.
10835         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10836         properly set the tune attribute.
10837         (dbnz): Use TARGET_DBNZ guard.
10838         * config/arc/arc.opt (mtune): Add core3 option.
10839
10840 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10841
10842         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10843         recognize new pic like addresses.
10844         (arc_delegitimize_address): Clean up.
10845
10846 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10847
10848         * config/arc/arc-arches.def: Option mrf16 valid for all
10849         architectures.
10850         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10851         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10852         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10853         * config/arc/arc-tables.opt: Regenerate.
10854         * config/arc/arc.c (arc_conditional_register_usage): Handle
10855         reduced register file case.
10856         (arc_file_start): Set must have build attributes.
10857         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10858         mrf16 option value.
10859         * config/arc/arc.opt (mrf16): Add new option.
10860         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10861         * config/arc/genmultilib.awk: Handle new mrf16 option.
10862         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10863         * config/arc/t-multilib: Regenerate.
10864         * doc/invoke.texi (ARC Options): Document mrf16 option.
10865
10866 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10867
10868         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10869         * config/arc/arc.c (arc_handle_secure_attribute): New function.
10870         (arc_attribute_table): Add 'secure_call' attribute.
10871         (arc_print_operand): Print secure call operand.
10872         (arc_function_ok_for_sibcall): Don't optimize tail calls when
10873         secure.
10874         (arc_is_secure_call_p): New function.  * config/arc/arc.md
10875         (call_i): Add support for sjli instruction.
10876         (call_value_i): Likewise.
10877         * config/arc/constraints.md (Csc): New constraint.
10878
10879 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10880             John Eric Martin  <John.Martin@emmicro-us.com>
10881
10882         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10883         * config/arc/arc.c (_arc_jli_section): New struct.
10884         (arc_jli_section): New type.
10885         (rc_jli_sections): New static variable.
10886         (arc_handle_jli_attribute): New function.
10887         (arc_attribute_table): Add jli_always and jli_fixed attribute.
10888         (arc_file_end): New function.
10889         (TARGET_ASM_FILE_END): Define.
10890         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10891         (arc_add_jli_section): New function.
10892         (jli_call_scan): Likewise.
10893         (arc_reorg): Call jli_call_scan.
10894         (arc_output_addsi): Remove 'S' from printing asm operand.
10895         (arc_is_jli_call_p): New function.
10896         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10897         operand.
10898         (movhi_insn): Likewise.
10899         (movsi_insn): Likewise.
10900         (movsi_set_cc_insn): Likewise.
10901         (loadqi_update): Likewise.
10902         (load_zeroextendqisi_update): Likewise.
10903         (load_signextendqisi_update): Likewise.
10904         (loadhi_update): Likewise.
10905         (load_zeroextendhisi_update): Likewise.
10906         (load_signextendhisi_update): Likewise.
10907         (loadsi_update): Likewise.
10908         (loadsf_update): Likewise.
10909         (movsicc_insn): Likewise.
10910         (bset_insn): Likewise.
10911         (bxor_insn): Likewise.
10912         (bclr_insn): Likewise.
10913         (bmsk_insn): Likewise.
10914         (bicsi3_insn): Likewise.
10915         (cmpsi_cc_c_insn): Likewise.
10916         (movsi_ne): Likewise.
10917         (movsi_cond_exec): Likewise.
10918         (clrsbsi2): Likewise.
10919         (norm_f): Likewise.
10920         (normw): Likewise.
10921         (swap): Likewise.
10922         (divaw): Likewise.
10923         (flag): Likewise.
10924         (sr): Likewise.
10925         (kflag): Likewise.
10926         (ffs): Likewise.
10927         (ffs_f): Likewise.
10928         (fls): Likewise.
10929         (call_i): Remove 'S' asm letter, add jli instruction.
10930         (call_value_i): Likewise.
10931         * config/arc/arc.op (mjli-always): New option.
10932         * config/arc/constraints.md (Cji): New constraint.
10933         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10934         operand.
10935         (subsf3_fpx): Likewise.
10936         (mulsf3_fpx): Likewise.
10937         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10938         asm operand.
10939         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10940         function attrbutes.
10941         * doc/invoke.texi (ARC): Document mjli-always option.
10942
10943 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
10944
10945         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10946         avoid addition with 0 and use incw and decw where possible.
10947
10948 2018-01-26  Richard Biener  <rguenther@suse.de>
10949
10950         PR tree-optimization/81082
10951         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10952         association if it requires casting to unsigned.
10953         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10954         from fold_plusminus_mult_expr to catch important cases late when
10955         range info is available.
10956
10957 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10958
10959         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10960         * configure.ac (hidden_linkonce): New test.
10961         * configure: Regenerate.
10962         * config.in: Regenerate.
10963
10964 2018-01-26  Julia Koval  <julia.koval@intel.com>
10965
10966         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10967         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10968         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10969         _mm_mask_bitshuffle_epi64_mask): Fix type.
10970         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10971         USI_FTYPE_V4DI_V4DI_USI): Remove.
10972         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10973         __builtin_ia32_vpshufbitqmb256_mask,
10974         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10975         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10976         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10977
10978 2018-01-26  Alan Modra  <amodra@gmail.com>
10979
10980         PR target/84033
10981         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10982         UNSPEC_VBPERMQ.  Sort other unspecs.
10983
10984 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
10985
10986         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10987
10988 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
10989
10990         PR middle-end/83055
10991         * predict.c (drop_profile): Do not push/pop cfun; update also
10992         node->count.
10993         (handle_missing_profiles): Fix logic looking for zero profiles.
10994
10995 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
10996
10997         PR middle-end/83977
10998         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10999         on functions with #pragma omp declare simd or functions with simd
11000         attribute.
11001         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11002         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11003         Remove trailing \n from warning_at calls.
11004
11005 2018-01-25  Tom de Vries  <tom@codesourcery.com>
11006
11007         PR target/84028
11008         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11009         for neutered workers.
11010
11011 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
11012
11013         PR target/68467
11014         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11015         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11016
11017 2018-01-24  Jeff Law  <law@redhat.com>
11018
11019         PR target/83994
11020         * i386.c (get_probe_interval): Move to earlier point.
11021         (ix86_compute_frame_layout): If -fstack-clash-protection and
11022         the frame is larger than the probe interval, then use pushes
11023         to save registers rather than reg->mem moves.
11024         (ix86_expand_prologue): Remove conditional for int_registers_saved
11025         assertion.
11026
11027 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
11028
11029         PR target/84014
11030         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11031         min/max for never referenced object.
11032
11033 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
11034
11035         PR middle-end/83977
11036         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11037         here.
11038         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11039         attributes from DECL_ATTRIBUTES (decl) without affecting
11040         DECL_ATTRIBUTES (current_function_decl).
11041         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11042         functions with non-NULL DECL_ABSTRACT_ORIGIN.
11043
11044 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
11045
11046         PR tree-optimization/83979
11047         * fold-const.c (fold_comparison): Use constant_boolean_node
11048         instead of boolean_{true,false}_node.
11049
11050 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
11051
11052         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11053         with zero counts.
11054
11055 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11056
11057         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11058         Simplify the clause that sets the length attribute.
11059         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11060         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11061         clause that sets the length attribute.
11062         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11063
11064 2018-01-24  Tom de Vries  <tom@codesourcery.com>
11065
11066         PR target/83589
11067         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11068         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11069         Add strict parameter.
11070         (prevent_branch_around_nothing): Insert dummy insn between branch to
11071         label and label with no ptx insn inbetween.
11072         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11073
11074 2018-01-24  Tom de Vries  <tom@codesourcery.com>
11075
11076         PR target/81352
11077         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11078         for neutered threads in warp.
11079         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11080
11081 2018-01-24  Richard Biener  <rguenther@suse.de>
11082
11083         PR tree-optimization/83176
11084         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11085         operands.
11086
11087 2018-01-24  Richard Biener  <rguenther@suse.de>
11088
11089         PR tree-optimization/82819
11090         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11091         code generating pluses that are no-ops in the target precision.
11092
11093 2018-01-24  Richard Biener  <rguenther@suse.de>
11094
11095         PR middle-end/84000
11096         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11097
11098 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
11099
11100         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11101         to merge probabilities.
11102         * predict.c (probably_never_executed): Also mark as cold functions
11103         with global 0 profile and guessed local profile.
11104         * profile-count.c (profile_probability::combine_with_count): New
11105         member function.
11106         * profile-count.h (profile_probability::operator*,
11107         profile_probability::operator*=, profile_probability::operator/,
11108         profile_probability::operator/=): Reduce precision to adjusted
11109         and set value to guessed on contradictory divisions.
11110         (profile_probability::combine_with_freq): Remove.
11111         (profile_probability::combine_wiht_count): Declare.
11112         (profile_count::force_nonzero):: Set to adjusted.
11113         (profile_count::probability_in):: Set quality to adjusted.
11114         * tree-ssa-tail-merge.c (replace_block_by): Use
11115         combine_with_count.
11116
11117 2018-01-23  Andrew Waterman  <andrew@sifive.com>
11118             Jim Wilson  <jimw@sifive.com>
11119
11120         * config/riscv/riscv.c (riscv_stack_boundary): New.
11121         (riscv_option_override): Set riscv_stack_boundary.  Handle
11122         riscv_preferred_stack_boundary_arg.
11123         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11124         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11125         (STACK_BOUNDARY): Set to riscv_stack_boundary.
11126         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11127         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11128         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11129
11130 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
11131
11132         PR target/83905
11133         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11134         of struct ix86_frame.
11135         (ix86_expand_epilogue): Likewise.  Add a local variable for
11136         the reg_save_offset field in struct ix86_frame.
11137
11138 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
11139
11140         PR tree-optimization/82604
11141         * tree-loop-distribution.c (enum partition_kind): New enum item
11142         PKIND_PARTIAL_MEMSET.
11143         (partition_builtin_p): Support above new enum item.
11144         (generate_code_for_partition): Ditto.
11145         (compute_access_range): Differentiate cases that equality can be
11146         proven at all loops, the innermost loops or no loops.
11147         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11148         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
11149         (finalize_partitions, distribute_loop): Don't fuse partition of
11150         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11151         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11152         parloop is enabled.
11153
11154 2018-01-23  Martin Liska  <mliska@suse.cz>
11155
11156         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11157         order to ignore the predictor.
11158         (PRED_POLYMORPHIC_CALL): Likewise.
11159         (PRED_RECURSIVE_CALL): Likewise.
11160
11161 2018-01-23  Martin Liska  <mliska@suse.cz>
11162
11163         * tree-profile.c (tree_profiling): Print function header to
11164         aware reader which function we are working on.
11165         * value-prof.c (gimple_find_values_to_profile): Do not print
11166         not interesting value histograms.
11167
11168 2018-01-23  Martin Liska  <mliska@suse.cz>
11169
11170         * profile-count.h (enum profile_quality): Add
11171         profile_uninitialized as the first value. Do not number values
11172         as they are zero based.
11173         (profile_count::verify): Update sanity check.
11174         (profile_probability::verify): Likewise.
11175
11176 2018-01-23  Nathan Sidwell  <nathan@acm.org>
11177
11178         * doc/invoke.texi (ffor-scope): Deprecate.
11179
11180 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
11181
11182         PR tree-optimization/83510
11183         * domwalk.c (set_all_edges_as_executable): New function.
11184         (dom_walker::dom_walker): Convert bool param
11185         "skip_unreachable_blocks" to enum reachability.  Move setup of
11186         edge flags to set_all_edges_as_executable and only do it when
11187         reachability is REACHABLE_BLOCKS.
11188         * domwalk.h (enum dom_walker::reachability): New enum.
11189         (dom_walker::dom_walker): Convert bool param
11190         "skip_unreachable_blocks" to enum reachability.
11191         (set_all_edges_as_executable): New decl.
11192         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
11193         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11194         "reachability".
11195         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11196         but converting true to REACHABLE_BLOCKS.
11197         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11198         * tree-vrp.c
11199         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11200         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11201         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11202         REACHABLE_BLOCKS.
11203         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11204         if check_all_array_refs will be called.
11205
11206 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
11207
11208         * tree.c (selftest::test_location_wrappers): Add more test
11209         coverage.
11210
11211 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
11212
11213         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11214         (selftest::test_bit_in_range): Likewise.
11215
11216 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
11217
11218         PR testsuite/83888
11219         * doc/sourcebuild.texi (vect_float): Say that the selector
11220         only describes the situation when -funsafe-math-optimizations is on.
11221         (vect_float_strict): Document.
11222
11223 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
11224
11225         PR tree-optimization/83965
11226         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11227         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11228         instead of checking only for a reduction.
11229         (vect_recog_widen_sum_pattern): Likewise.
11230
11231 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
11232
11233         * predict.c (probably_never_executed): Only use precise profile info.
11234         (compute_function_frequency): Skip after inlining hack since we now
11235         have quality checking.
11236
11237 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
11238
11239         * profile-count.h (profile_probability::very_unlikely,
11240         profile_probability::unlikely, profile_probability::even): Set
11241         precision to guessed.
11242
11243 2018-01-23  Richard Biener  <rguenther@suse.de>
11244
11245         PR tree-optimization/83963
11246         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11247         Properly terminate dominator walk when crossing the exit edge not
11248         when visiting its source block.
11249
11250 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
11251
11252         PR c++/83918
11253         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11254         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11255
11256 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
11257
11258         PR tree-optimization/83957
11259         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
11260         semicolon after for body surrounded by braces.
11261
11262         PR tree-optimization/83081
11263         * profile-count.h (profile_probability::split): New method.
11264         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11265         Use profile_probability::split.
11266         (do_compare_rtx_and_jump): Fix adjustment of probabilities
11267         when splitting a single conditional jump into 2.
11268
11269 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
11270
11271         PR tree-optimization/69452
11272         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11273         decl.
11274
11275 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11276
11277         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11278         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11279         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11280
11281 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11282
11283         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11284         declaration.
11285         * config/rl78/rl78.md (movdi): New define_expand.
11286         * config/rl78/rl78.c (rl78_split_movdi): New function.
11287
11288 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
11289
11290         PR target/83862
11291         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11292         no longer used.
11293         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11294         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11295         128-bit to produce an UNSPEC move to get the double word with the
11296         signbit and then a shift directly to do signbit.
11297         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11298         implementation with a new version that just does either a direct
11299         move or a regular move.  Move memory interface to separate insns.
11300         Move insns so they are next to the expander.
11301         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11302         with signbit move.  Split big and little endian case.
11303         (signbit<mode>2_dm_mem_le): Likewise.
11304         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11305         (signbit<mode>2_dm2): Likewise.
11306
11307 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11308
11309         * config/rl78/rl78.md (anddi3): New define_expand.
11310
11311 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11312
11313         * config/rl78/rl78.md (umindi3): New define_expand.
11314
11315 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11316
11317         * config/rl78/rl78.md (smindi3): New define_expand.
11318
11319 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11320
11321         * config/rl78/rl78.md (smaxdi3): New define_expand.
11322
11323 2018-01-22  Carl Love  <cel@us.ibm.com>
11324
11325         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11326         LVX_V1TI): Add macro expansion.
11327         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11328         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11329         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11330         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11331         Change check to determine if the instruction is a byte reversing
11332         entry.  Fix typo in comment.
11333         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11334         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11335         Add def_builtin calls for new builtins.
11336         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11337         Add define_insn expansion.
11338
11339 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11340
11341         * config/rl78/rl78.md (umaxdi3): New define_expand.
11342
11343 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11344
11345         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11346         for non-QImode registers.
11347
11348 2018-01-22  Richard Biener  <rguenther@suse.de>
11349
11350         PR tree-optimization/83963
11351         * graphite-scop-detection.c (scop_detection::get_sese): Delay
11352         including the loop exit block.
11353         (scop_detection::merge_sese): Likewise.
11354         (scop_detection::add_scop): Do it here instead.
11355
11356 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11357
11358         * doc/sourcebuild.texi (arm_softfloat): Document.
11359
11360 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
11361
11362         PR gcc/77734
11363         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11364         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11365         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11366
11367 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11368             David Edelsohn  <dje.gcc@gmail.com>
11369
11370         PR target/83946
11371         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11372         Change "crset eq" to "crset 2".
11373         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11374         (*call_indirect_aix<mode>_nospec): Likewise.
11375         (*call_value_indirect_aix<mode>_nospec): Likewise.
11376         (*call_indirect_elfv2<mode>_nospec): Likewise.
11377         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11378         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11379         change assembly output from . to $.
11380         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11381         (indirect_jump<mode>_nospec): Change assembly output from . to $.
11382         (*tablejump<mode>_internal1_nospec): Likewise.
11383
11384 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
11385
11386         PR target/80870
11387         * config/sh/sh_optimize_sett_clrt.cc:
11388         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11389
11390 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
11391
11392         PR tree-optimization/83940
11393         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11394         offset_dt to vect_constant_def rather than vect_unknown_def_type.
11395         (vect_check_load_store_mask): Add a mask_dt_out parameter and
11396         use it to pass back the definition type.
11397         (vect_check_store_rhs): Likewise rhs_dt_out.
11398         (vect_build_gather_load_calls): Add a mask_dt argument and use
11399         it instead of a call to vect_is_simple_use.
11400         (vectorizable_store): Update calls to vect_check_load_store_mask
11401         and vect_check_store_rhs.  Use the dt returned by the latter instead
11402         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
11403         instead of calls to vect_is_simple_use.  Pass the scalar rather
11404         than the vector operand to vect_is_simple_use when handling
11405         second and subsequent copies of an rhs value.
11406         (vectorizable_load): Update calls to vect_check_load_store_mask
11407         and vect_build_gather_load_calls.  Use the cached mask_dt and
11408         gs_info.offset_dt instead of calls to vect_is_simple_use.
11409
11410 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
11411
11412         PR middle-end/83945
11413         * tree-emutls.c: Include gimplify.h.
11414         (lower_emutls_2): New function.
11415         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11416         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11417         it before further processing.
11418
11419         PR target/83930
11420         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11421         UINTVAL (trueop1) instead of INTVAL (op1).
11422
11423 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
11424
11425         PR debug/81570
11426         PR debug/83728
11427         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11428         INCOMING_FRAME_SP_OFFSET if not defined.
11429         (scan_trace): Add ENTRY argument.  If true and
11430         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11431         emit a note to adjust the CFA offset.
11432         (create_cfi_notes): Adjust scan_trace callers.
11433         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11434         INCOMING_FRAME_SP_OFFSET in the CIE.
11435         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11436         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11437         Likewise.
11438         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11439         * doc/tm.texi: Regenerated.
11440
11441 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
11442
11443         PR rtl-optimization/83147
11444         * lra-constraints.c (remove_inheritance_pseudos): Use
11445         lra_substitute_pseudo_within_insn.
11446
11447 2018-01-19  Tom de Vries  <tom@codesourcery.com>
11448             Cesar Philippidis  <cesar@codesourcery.com>
11449
11450         PR target/83920
11451         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11452
11453 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
11454
11455         PR target/83790
11456         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11457         spaces for function labels.
11458
11459 2018-01-19  Martin Liska  <mliska@suse.cz>
11460
11461         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11462         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11463         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11464         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11465         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11466         (PRED_CONST_RETURN): Change from 69 to 65.
11467         (PRED_NULL_RETURN): Change from 91 to 71.
11468         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11469         (PRED_LOOP_GUARD): Change from 66 to 73.
11470
11471 2018-01-19  Martin Liska  <mliska@suse.cz>
11472
11473         * predict.c (predict_insn_def): Add new assert.
11474         (struct branch_predictor): Change type to signed integer.
11475         (test_prediction_value_range): Amend test to cover
11476         PROB_UNINITIALIZED.
11477         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11478         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11479         (PRED_LOOP_ITERATIONS_MAX): Likewise.
11480         (PRED_LOOP_IV_COMPARE): Likewise.
11481         * predict.h (PROB_UNINITIALIZED): Define new constant.
11482
11483 2018-01-19  Martin Liska  <mliska@suse.cz>
11484
11485         * predict.c (dump_prediction): Add new format for
11486         analyze_brprob.py script which is enabled with -details
11487         suboption.
11488         * profile-count.h (precise_p): New function.
11489
11490 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
11491
11492         PR tree-optimization/83922
11493         * tree-vect-loop.c (vect_verify_full_masking): Return false if
11494         there are no statements that need masking.
11495         (vect_active_double_reduction_p): New function.
11496         (vect_analyze_loop_operations): Use it when handling phis that
11497         are not in the loop header.
11498
11499 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
11500
11501         PR tree-optimization/83914
11502         * tree-vect-loop.c (vectorizable_induction): Don't convert
11503         init_expr or apply the peeling adjustment for inductions
11504         that are nested within the vectorized loop.
11505
11506 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11507
11508         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11509         instead of NEG.
11510
11511 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
11512
11513         PR sanitizer/81715
11514         PR testsuite/83882
11515         * function.h (gimplify_parameters): Add gimple_seq * argument.
11516         * function.c: Include gimple.h and options.h.
11517         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11518         for the added local temporaries if needed.
11519         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11520         if there are any parameter cleanups, wrap whole body into a
11521         try/finally with the cleanups.
11522
11523 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
11524
11525         PR target/82964
11526         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11527         Use GET_MODE_CLASS for scalar floating point.
11528
11529 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
11530
11531         PR ipa/82256
11532         patch by PaX Team
11533         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11534         Fix call of call_cgraph_insertion_hooks.
11535
11536 2018-01-18  Martin Sebor  <msebor@redhat.com>
11537
11538         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11539
11540 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
11541
11542         PR ipa/83619
11543         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11544         frequencies.
11545
11546 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
11547
11548         PR other/70268
11549         * common.opt: (-ffile-prefix-map): New option.
11550         * opts.c (common_handle_option): Defer it.
11551         * opts-global.c (handle_common_deferred_options): Handle it.
11552         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11553         * file-prefix-map.h: New file.
11554         (remap_debug_filename, add_debug_prefix_map): ...here.
11555         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11556         * final.c (debug_prefix_map, add_debug_prefix_map
11557         remap_debug_filename): Move to...
11558         * file-prefix-map.c: New file.
11559         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11560         generalize, get rid of alloca(), use strrchr() instead of strchr().
11561         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11562         Implement in terms of add_prefix_map().
11563         (remap_macro_filename, remap_debug_filename): Implement in term of
11564         remap_filename().
11565         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11566         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11567         * dbxout.c: Include file-prefix-map.h.
11568         * varasm.c: Likewise.
11569         * vmsdbgout.c: Likewise.
11570         * xcoffout.c: Likewise.
11571         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11572         * doc/cppopts.texi (-fmacro-prefix-map): Document.
11573         * doc/invoke.texi (-ffile-prefix-map): Document.
11574         (-fdebug-prefix-map): Update description.
11575
11576 2018-01-18  Martin Liska  <mliska@suse.cz>
11577
11578         * config/i386/i386.c (indirect_thunk_name): Document that also
11579         lfence is emitted.
11580         (output_indirect_thunk): Document why both instructions
11581         (pause and lfence) are generated.
11582
11583 2018-01-18  Richard Biener  <rguenther@suse.de>
11584
11585         PR tree-optimization/83887
11586         * graphite-scop-detection.c
11587         (scop_detection::get_nearest_dom_with_single_entry): Remove.
11588         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11589         (scop_detection::merge_sese): Re-implement with a flood-fill
11590         algorithm that properly finds a SESE region if it exists.
11591
11592 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
11593
11594         PR c/61240
11595         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11596         pointer_diff optimizations use view_convert instead of convert.
11597
11598 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11599
11600         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11601         Generate different code for -mno-speculate-indirect-jumps.
11602         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11603         (*call_indirect_aix<mode>): Disable for
11604         -mno-speculate-indirect-jumps.
11605         (*call_indirect_aix<mode>_nospec): New define_insn.
11606         (*call_value_indirect_aix<mode>): Disable for
11607         -mno-speculate-indirect-jumps.
11608         (*call_value_indirect_aix<mode>_nospec): New define_insn.
11609         (*sibcall_nonlocal_sysv<mode>): Generate different code for
11610         -mno-speculate-indirect-jumps.
11611         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11612
11613 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
11614
11615         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11616         long double type, set the flags for noting the default long double
11617         type, even if we don't pass or return a long double type.
11618
11619 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
11620
11621         PR ipa/83051
11622         * ipa-inline.c (flatten_function): Do not overwrite final inlining
11623         failure.
11624
11625 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
11626
11627         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11628         support for merge[hl].
11629         (fold_mergehl_helper): New helper function.
11630         (tree-vector-builder.h): New #include for tree_vector_builder usage.
11631         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11632         (altivec_vmrglw_direct): Add xxmrglw insn.
11633
11634 2018-01-17  Andrew Waterman  <andrew@sifive.com>
11635
11636         * config/riscv/riscv.c (riscv_conditional_register_usage): If
11637         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11638
11639 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
11640
11641         PR lto/83121
11642         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11643         call the lto_location_cache before reading the
11644         DECL_SOURCE_LOCATION of the types.
11645
11646 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
11647             Richard Sandiford  <richard.sandiford@linaro.org>
11648
11649         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11650         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11651         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11652         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11653         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11654         Add declaration.
11655         * config/aarch64/constraints.md (aarch64_movti_operand):
11656         Limit immediates.
11657         * config/aarch64/predicates.md (Uti): Add new constraint.
11658
11659 2018-01-17 Carl Love  <cel@us.ibm.com>
11660
11661         * config/rs6000/vsx.md (define_expand xl_len_r,
11662         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11663         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11664         lxvll.
11665         (define_expand, define_insn): Move the shift left from  the
11666         define_insn to the define_expand for lxvl and stxvl instructions.
11667         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11668         and XL_LEN_R definitions to PURE.
11669
11670 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
11671
11672         * config/i386/i386.c (indirect_thunk_name): Declare regno
11673         as unsigned int.  Compare regno with INVALID_REGNUM.
11674         (output_indirect_thunk): Ditto.
11675         (output_indirect_thunk_function): Ditto.
11676         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
11677         in the call to output_indirect_thunk_function.
11678
11679 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
11680
11681         PR middle-end/83884
11682         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11683         rather than the size of inner_type to determine the stack slot size
11684         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11685
11686 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
11687
11688         PR target/83546
11689         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11690         to PTA_SILVERMONT.
11691
11692 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
11693
11694         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11695         endian Linux systems to optionally enable multilibs for selecting
11696         the long double type if the user configured an explicit type.
11697         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11698         have no long double multilibs if not defined.
11699         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11700         warn if the user used -mabi={ieee,ibm}longdouble and we built
11701         multilibs for long double.
11702         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11703         appropriate multilib option.
11704         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11705         multilib options.
11706         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11707         for building long double multilibs.
11708         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11709
11710 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
11711
11712         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11713         copies.
11714
11715         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11716         64 bits.
11717         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11718         128 bits.
11719
11720         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11721         variables.
11722
11723         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11724         return value.
11725
11726 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
11727
11728         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11729         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11730
11731 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11732
11733         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11734         different rtl trees depending on TARGET_64BIT.
11735         (rs6000_gen_lvx): Likewise.
11736
11737 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
11738
11739         * config/visium/visium.md (nop): Tweak comment.
11740         (hazard_nop): Likewise.
11741
11742 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11743
11744         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11745         -mspeculate-indirect-jumps.
11746         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11747         for -mno-speculate-indirect-jumps.
11748         (*call_indirect_elfv2<mode>_nospec): New define_insn.
11749         (*call_value_indirect_elfv2<mode>): Disable for
11750         -mno-speculate-indirect-jumps.
11751         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11752         (indirect_jump): Emit different RTL for
11753         -mno-speculate-indirect-jumps.
11754         (*indirect_jump<mode>): Disable for
11755         -mno-speculate-indirect-jumps.
11756         (*indirect_jump<mode>_nospec): New define_insn.
11757         (tablejump): Emit different RTL for
11758         -mno-speculate-indirect-jumps.
11759         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11760         (tablejumpsi_nospec): New define_expand.
11761         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11762         (tablejumpdi_nospec): New define_expand.
11763         (*tablejump<mode>_internal1): Disable for
11764         -mno-speculate-indirect-jumps.
11765         (*tablejump<mode>_internal1_nospec): New define_insn.
11766         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11767         option.
11768
11769 2018-01-16  Artyom Skrobov tyomitch@gmail.com
11770
11771         * caller-save.c (insert_save): Drop unnecessary parameter.  All
11772         callers updated.
11773
11774 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
11775             Richard Biener  <rguenth@suse.de>
11776
11777         PR libgomp/83590
11778         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11779         return early, inline manually is_gimple_sizepos.  Make sure if we
11780         call gimplify_expr we don't end up with a gimple constant.
11781         * tree.c (variably_modified_type_p): Don't return true for
11782         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
11783         * gimplify.h (is_gimple_sizepos): Remove.
11784
11785 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
11786
11787         PR tree-optimization/83857
11788         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11789         vectorizable_live_operation for pure SLP statements.
11790         (vectorizable_live_operation): Handle PHIs.
11791
11792 2018-01-16  Richard Biener  <rguenther@suse.de>
11793
11794         PR tree-optimization/83867
11795         * tree-vect-stmts.c (vect_transform_stmt): Precompute
11796         nested_in_vect_loop_p since the scalar stmt may get invalidated.
11797
11798 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
11799
11800         PR c/83844
11801         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11802         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11803         If off is not INTEGER_CST, issue a may not be aligned warning
11804         rather than isn't aligned.  Use isn%'t rather than isn't.
11805         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11806         into MULT_EXPR.
11807         <case MULT_EXPR>: Improve the case when bottom and one of the
11808         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11809         operand, in that case check if the other operand is multiple of
11810         bottom divided by the INTEGER_CST operand.
11811
11812 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
11813
11814         PR target/83858
11815         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11816         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11817         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11818         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11819         * config/pa/pa.c (pa_function_arg_advance): Likewise.
11820         (pa_function_arg, pa_arg_partial_bytes): Likewise.
11821         (pa_function_arg_size): New function.
11822
11823 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
11824
11825         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11826         in a separate statement.
11827
11828 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
11829
11830         PR tree-optimization/83847
11831         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11832         group gathers and scatters.
11833
11834 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
11835
11836         PR rtl-optimization/86620
11837         * params.def (max-sched-ready-insns): Bump minimum value to 1.
11838
11839         PR rtl-optimization/83213
11840         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11841         to last if both are JUMP_INSNs.
11842
11843         PR tree-optimization/83843
11844         * gimple-ssa-store-merging.c
11845         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11846         store_immediate_info for bswap/nop orig_stores.
11847
11848 2018-01-15  Andrew Waterman  <andrew@sifive.com>
11849
11850         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11851         !TARGET_MUL.
11852         <UDIV>: Increase cost if !TARGET_DIV.
11853
11854 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
11855
11856         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11857         (define_attr "cr_logical_3op"): New.
11858         (cceq_ior_compare): Adjust.
11859         (cceq_ior_compare_complement): Adjust.
11860         (*cceq_rev_compare): Adjust.
11861         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11862         (is_cracked_insn): Adjust.
11863         (insn_must_be_first_in_group): Adjust.
11864         * config/rs6000/40x.md: Adjust.
11865         * config/rs6000/440.md: Adjust.
11866         * config/rs6000/476.md: Adjust.
11867         * config/rs6000/601.md: Adjust.
11868         * config/rs6000/603.md: Adjust.
11869         * config/rs6000/6xx.md: Adjust.
11870         * config/rs6000/7450.md: Adjust.
11871         * config/rs6000/7xx.md: Adjust.
11872         * config/rs6000/8540.md: Adjust.
11873         * config/rs6000/cell.md: Adjust.
11874         * config/rs6000/e300c2c3.md: Adjust.
11875         * config/rs6000/e500mc.md: Adjust.
11876         * config/rs6000/e500mc64.md: Adjust.
11877         * config/rs6000/e5500.md: Adjust.
11878         * config/rs6000/e6500.md: Adjust.
11879         * config/rs6000/mpc.md: Adjust.
11880         * config/rs6000/power4.md: Adjust.
11881         * config/rs6000/power5.md: Adjust.
11882         * config/rs6000/power6.md: Adjust.
11883         * config/rs6000/power7.md: Adjust.
11884         * config/rs6000/power8.md: Adjust.
11885         * config/rs6000/power9.md: Adjust.
11886         * config/rs6000/rs64.md: Adjust.
11887         * config/rs6000/titan.md: Adjust.
11888
11889 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11890
11891         * config/i386/predicates.md (indirect_branch_operand): Rewrite
11892         ix86_indirect_branch_register logic.
11893
11894 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11895
11896         * config/i386/constraints.md (Bs): Update
11897         ix86_indirect_branch_register check.  Don't check
11898         ix86_indirect_branch_register with GOT_memory_operand.
11899         (Bw): Likewise.
11900         * config/i386/predicates.md (GOT_memory_operand): Don't check
11901         ix86_indirect_branch_register here.
11902         (GOT32_symbol_operand): Likewise.
11903
11904 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11905
11906         * config/i386/predicates.md (constant_call_address_operand):
11907         Rewrite ix86_indirect_branch_register logic.
11908         (sibcall_insn_operand): Likewise.
11909
11910 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11911
11912         * config/i386/constraints.md (Bs): Replace
11913         ix86_indirect_branch_thunk_register with
11914         ix86_indirect_branch_register.
11915         (Bw): Likewise.
11916         * config/i386/i386.md (indirect_jump): Likewise.
11917         (tablejump): Likewise.
11918         (*sibcall_memory): Likewise.
11919         (*sibcall_value_memory): Likewise.
11920         Peepholes of indirect call and jump via memory: Likewise.
11921         * config/i386/i386.opt: Likewise.
11922         * config/i386/predicates.md (indirect_branch_operand): Likewise.
11923         (GOT_memory_operand): Likewise.
11924         (call_insn_operand): Likewise.
11925         (sibcall_insn_operand): Likewise.
11926         (GOT32_symbol_operand): Likewise.
11927
11928 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
11929
11930         PR middle-end/83837
11931         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11932         type rather than type addr's type points to.
11933         (expand_omp_atomic_mutex): Likewise.
11934         (expand_omp_atomic): Likewise.
11935
11936 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11937
11938         PR target/83839
11939         * config/i386/i386.c (output_indirect_thunk_function): Use
11940         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11941         for  __x86_return_thunk.
11942
11943 2018-01-15  Richard Biener  <rguenther@suse.de>
11944
11945         PR middle-end/83850
11946         * expmed.c (extract_bit_field_1): Fix typo.
11947
11948 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11949
11950         PR target/83687
11951         * config/arm/iterators.md (VF): New mode iterator.
11952         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11953         Remove integer-related logic from pattern.
11954         (neon_vabd<mode>_3): Likewise.
11955
11956 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
11957
11958         PR middle-end/82694
11959         * common.opt (fstrict-overflow): No longer an alias.
11960         (fwrapv-pointer): New option.
11961         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11962         also for pointer types based on flag_wrapv_pointer.
11963         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11964         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11965         opts->x_flag_wrapv got set.
11966         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11967         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11968         POINTER_TYPE_OVERFLOW_UNDEFINED.
11969         * match.pd: Likewise in address comparison pattern.
11970         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11971
11972 2018-01-15  Richard Biener  <rguenther@suse.de>
11973
11974         PR lto/83804
11975         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11976         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
11977         Reset type names to their identifier if their TYPE_DECL doesn't
11978         have linkage (and thus is used for ODR and devirt).
11979         (save_debug_info_for_decl): Remove.
11980         (save_debug_info_for_type): Likewise.
11981         (add_tree_to_fld_list): Adjust.
11982         * tree-pretty-print.c (dump_generic_node): Make dumping of
11983         type names more robust.
11984
11985 2018-01-15  Richard Biener  <rguenther@suse.de>
11986
11987         * BASE-VER: Bump to 8.0.1.
11988
11989 2018-01-14  Martin Sebor  <msebor@redhat.com>
11990
11991         PR other/83508
11992         * builtins.c (check_access): Avoid warning when the no-warning bit
11993         is set.
11994
11995 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
11996
11997         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11998         * ira-color (allocno_hard_regs_compare): Likewise.
11999
12000 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
12001
12002         PR target/83013
12003         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12004         Use .pushsection/.popsection.
12005
12006 2018-01-14  Martin Sebor  <msebor@redhat.com>
12007
12008         PR c++/81327
12009         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12010
12011 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
12012
12013         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12014         entry from extra_headers.
12015         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12016         extra_headers, make the list bitwise identical to the i?86-*-* one.
12017
12018 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12019
12020         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12021         -mcmodel=large with -mindirect-branch=thunk,
12022         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12023         -mfunction-return=thunk-extern.
12024         * doc/invoke.texi: Document -mcmodel=large is incompatible with
12025         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12026         -mfunction-return=thunk and -mfunction-return=thunk-extern.
12027
12028 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12029
12030         * config/i386/i386.c (print_reg): Print the name of the full
12031         integer register without '%'.
12032         (ix86_print_operand): Handle 'V'.
12033         * doc/extend.texi: Document 'V' modifier.
12034
12035 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12036
12037         * config/i386/constraints.md (Bs): Disallow memory operand for
12038         -mindirect-branch-register.
12039         (Bw): Likewise.
12040         * config/i386/predicates.md (indirect_branch_operand): Likewise.
12041         (GOT_memory_operand): Likewise.
12042         (call_insn_operand): Likewise.
12043         (sibcall_insn_operand): Likewise.
12044         (GOT32_symbol_operand): Likewise.
12045         * config/i386/i386.md (indirect_jump): Call convert_memory_address
12046         for -mindirect-branch-register.
12047         (tablejump): Likewise.
12048         (*sibcall_memory): Likewise.
12049         (*sibcall_value_memory): Likewise.
12050         Disallow peepholes of indirect call and jump via memory for
12051         -mindirect-branch-register.
12052         (*call_pop): Replace m with Bw.
12053         (*call_value_pop): Likewise.
12054         (*sibcall_pop_memory): Replace m with Bs.
12055         * config/i386/i386.opt (mindirect-branch-register): New option.
12056         * doc/invoke.texi: Document -mindirect-branch-register option.
12057
12058 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12059
12060         * config/i386/i386-protos.h (ix86_output_function_return): New.
12061         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12062         set function_return_type.
12063         (indirect_thunk_name): Add ret_p to indicate thunk for function
12064         return.
12065         (output_indirect_thunk_function): Pass false to
12066         indirect_thunk_name.
12067         (ix86_output_indirect_branch_via_reg): Likewise.
12068         (ix86_output_indirect_branch_via_push): Likewise.
12069         (output_indirect_thunk_function): Create alias for function
12070         return thunk if regno < 0.
12071         (ix86_output_function_return): New function.
12072         (ix86_handle_fndecl_attribute): Handle function_return.
12073         (ix86_attribute_table): Add function_return.
12074         * config/i386/i386.h (machine_function): Add
12075         function_return_type.
12076         * config/i386/i386.md (simple_return_internal): Use
12077         ix86_output_function_return.
12078         (simple_return_internal_long): Likewise.
12079         * config/i386/i386.opt (mfunction-return=): New option.
12080         (indirect_branch): Mention -mfunction-return=.
12081         * doc/extend.texi: Document function_return function attribute.
12082         * doc/invoke.texi: Document -mfunction-return= option.
12083
12084 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12085
12086         * config/i386/i386-opts.h (indirect_branch): New.
12087         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12088         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12089         with local indirect jump when converting indirect call and jump.
12090         (ix86_set_indirect_branch_type): New.
12091         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12092         (indirectlabelno): New.
12093         (indirect_thunk_needed): Likewise.
12094         (indirect_thunk_bnd_needed): Likewise.
12095         (indirect_thunks_used): Likewise.
12096         (indirect_thunks_bnd_used): Likewise.
12097         (INDIRECT_LABEL): Likewise.
12098         (indirect_thunk_name): Likewise.
12099         (output_indirect_thunk): Likewise.
12100         (output_indirect_thunk_function): Likewise.
12101         (ix86_output_indirect_branch_via_reg): Likewise.
12102         (ix86_output_indirect_branch_via_push): Likewise.
12103         (ix86_output_indirect_branch): Likewise.
12104         (ix86_output_indirect_jmp): Likewise.
12105         (ix86_code_end): Call output_indirect_thunk_function if needed.
12106         (ix86_output_call_insn): Call ix86_output_indirect_branch if
12107         needed.
12108         (ix86_handle_fndecl_attribute): Handle indirect_branch.
12109         (ix86_attribute_table): Add indirect_branch.
12110         * config/i386/i386.h (machine_function): Add indirect_branch_type
12111         and has_local_indirect_jump.
12112         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12113         to true.
12114         (tablejump): Likewise.
12115         (*indirect_jump): Use ix86_output_indirect_jmp.
12116         (*tablejump_1): Likewise.
12117         (simple_return_indirect_internal): Likewise.
12118         * config/i386/i386.opt (mindirect-branch=): New option.
12119         (indirect_branch): New.
12120         (keep): Likewise.
12121         (thunk): Likewise.
12122         (thunk-inline): Likewise.
12123         (thunk-extern): Likewise.
12124         * doc/extend.texi: Document indirect_branch function attribute.
12125         * doc/invoke.texi: Document -mindirect-branch= option.
12126
12127 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
12128
12129         PR ipa/83051
12130         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12131
12132 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
12133
12134         * ipa-inline.c (want_inline_small_function_p): Return false if
12135         inlining has already failed with CIF_FINAL_ERROR.
12136         (update_caller_keys): Call want_inline_small_function_p before
12137         can_inline_edge_p.
12138         (update_callee_keys): Likewise.
12139
12140 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12141
12142         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12143         New function.
12144         (rs6000_quadword_masked_address_p): Likewise.
12145         (quad_aligned_load_p): Likewise.
12146         (quad_aligned_store_p): Likewise.
12147         (const_load_sequence_p): Add comment to describe the outer-most loop.
12148         (mimic_memory_attributes_and_flags): New function.
12149         (rs6000_gen_stvx): Likewise.
12150         (replace_swapped_aligned_store): Likewise.
12151         (rs6000_gen_lvx): Likewise.
12152         (replace_swapped_aligned_load): Likewise.
12153         (replace_swapped_load_constant): Capitalize argument name in
12154         comment describing this function.
12155         (rs6000_analyze_swaps): Add a third pass to search for vector loads
12156         and stores that access quad-word aligned addresses and replace
12157         with stvx or lvx instructions when appropriate.
12158         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12159         New function prototype.
12160         (rs6000_quadword_masked_address_p): Likewise.
12161         (rs6000_gen_lvx): Likewise.
12162         (rs6000_gen_stvx): Likewise.
12163         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12164         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12165         when memory address is aligned.
12166         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12167         this split to select lvx instruction when memory address is aligned.
12168         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12169         instruction when memory address is aligned.
12170         (*vsx_le_perm_load_v16qi): Likewise.
12171         (four unnamed splitters): Modify to select the stvx instruction
12172         when memory is aligned.
12173
12174 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
12175
12176         * predict.c (determine_unlikely_bbs): Handle correctly BBs
12177         which appears in the queue multiple times.
12178
12179 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12180             Alan Hayward  <alan.hayward@arm.com>
12181             David Sherwood  <david.sherwood@arm.com>
12182
12183         * tree-vectorizer.h (vec_lower_bound): New structure.
12184         (_loop_vec_info): Add check_nonzero and lower_bounds.
12185         (LOOP_VINFO_CHECK_NONZERO): New macro.
12186         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12187         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12188         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12189         fields.  Make seg_len the distance travelled, not including the
12190         access size.
12191         (dr_direction_indicator): Declare.
12192         (dr_zero_step_indicator): Likewise.
12193         (dr_known_forward_stride_p): Likewise.
12194         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12195         tree-ssanames.h.
12196         (runtime_alias_check_p): Allow runtime alias checks with
12197         variable strides.
12198         (operator ==): Compare access_size and align.
12199         (prune_runtime_alias_test_list): Rework for new distinction between
12200         the access_size and seg_len.
12201         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
12202         segment lengths.
12203         (get_segment_min_max): New function.
12204         (create_intersect_range_checks): Use it.
12205         (dr_step_indicator): New function.
12206         (dr_direction_indicator): Likewise.
12207         (dr_zero_step_indicator): Likewise.
12208         (dr_known_forward_stride_p): Likewise.
12209         * tree-loop-distribution.c (data_ref_segment_size): Return
12210         DR_STEP * (niters - 1).
12211         (compute_alias_check_pairs): Update call to the dr_with_seg_len
12212         constructor.
12213         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12214         (vect_preserves_scalar_order_p): New function, split out from...
12215         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
12216         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12217         (vect_vfa_access_size): New function.
12218         (vect_vfa_align): Likewise.
12219         (vect_compile_time_alias): Take access_size_a and access_b arguments.
12220         (dump_lower_bound): New function.
12221         (vect_check_lower_bound): Likewise.
12222         (vect_small_gap_p): Likewise.
12223         (vectorizable_with_step_bound_p): Likewise.
12224         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12225         depencies if the vectorization factor is 1.  Convert the checks
12226         for nonzero steps into checks on the bounds of DR_STEP.  Try using
12227         a bunds check for variable steps if the minimum required step is
12228         relatively small. Update calls to the dr_with_seg_len
12229         constructor and to vect_compile_time_alias.
12230         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12231         function.
12232         (vect_loop_versioning): Call it.
12233         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12234         when retrying.
12235         (vect_estimate_min_profitable_iters): Account for any bounds checks.
12236
12237 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12238             Alan Hayward  <alan.hayward@arm.com>
12239             David Sherwood  <david.sherwood@arm.com>
12240
12241         * doc/sourcebuild.texi (vect_scatter_store): Document.
12242         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12243         optabs.
12244         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12245         Document.
12246         * genopinit.c (main): Add supports_vec_scatter_store and
12247         supports_vec_scatter_store_cached to target_optabs.
12248         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12249         IFN_MASK_SCATTER_STORE.
12250         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12251         functions.
12252         * internal-fn.h (internal_store_fn_p): Declare.
12253         (internal_fn_stored_value_index): Likewise.
12254         * internal-fn.c (scatter_store_direct): New macro.
12255         (expand_scatter_store_optab_fn): New function.
12256         (direct_scatter_store_optab_supported_p): New macro.
12257         (internal_store_fn_p): New function.
12258         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12259         IFN_MASK_SCATTER_STORE.
12260         (internal_fn_mask_index): Likewise.
12261         (internal_fn_stored_value_index): New function.
12262         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12263         for scatter stores.
12264         * optabs-query.h (supports_vec_scatter_store_p): Declare.
12265         * optabs-query.c (supports_vec_scatter_store_p): New function.
12266         * tree-vectorizer.h (vect_get_store_rhs): Declare.
12267         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12268         true for scatter stores.
12269         (vect_gather_scatter_fn_p): Handle scatter stores too.
12270         (vect_check_gather_scatter): Consider using scatter stores if
12271         supports_vec_scatter_store_p.
12272         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12273         scatter stores too.
12274         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12275         internal_fn_stored_value_index.
12276         (check_load_store_masking): Handle scatter stores too.
12277         (vect_get_store_rhs): Make public.
12278         (vectorizable_call): Use internal_store_fn_p.
12279         (vectorizable_store): Handle scatter store internal functions.
12280         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12281         when deciding whether the end of the group has been reached.
12282         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12283         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12284         (mask_scatter_store<mode>): New insns.
12285
12286 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12287             Alan Hayward  <alan.hayward@arm.com>
12288             David Sherwood  <david.sherwood@arm.com>
12289
12290         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12291         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12292         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12293         function.
12294         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12295         Use vect_truncate_gather_scatter_offset if we can't treat the
12296         operation as a normal gather load or scatter store.
12297         (get_group_load_store_type): Take the gather_scatter_info
12298         as argument.  Try using a gather load or scatter store for
12299         single-element groups.
12300         (get_load_store_type): Update calls to get_group_load_store_type
12301         and vect_use_strided_gather_scatters_p.
12302
12303 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12304             Alan Hayward  <alan.hayward@arm.com>
12305             David Sherwood  <david.sherwood@arm.com>
12306
12307         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12308         optional tree argument.
12309         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12310         null target hooks.
12311         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12312         but continue to use the current value as a fallback.
12313         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12314         to compare the updates.
12315         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12316         (get_load_store_type): Use it when handling a strided access.
12317         (vect_get_strided_load_store_ops): New function.
12318         (vect_get_data_ptr_increment): Likewise.
12319         (vectorizable_load): Handle strided gather loads.  Always pass
12320         a step to vect_create_data_ref_ptr and bump_vector_ptr.
12321
12322 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12323             Alan Hayward  <alan.hayward@arm.com>
12324             David Sherwood  <david.sherwood@arm.com>
12325
12326         * doc/md.texi (gather_load@var{m}): Document.
12327         (mask_gather_load@var{m}): Likewise.
12328         * genopinit.c (main): Add supports_vec_gather_load and
12329         supports_vec_gather_load_cached to target_optabs.
12330         * optabs-tree.c (init_tree_optimization_optabs): Use
12331         ggc_cleared_alloc to allocate target_optabs.
12332         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12333         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12334         functions.
12335         * internal-fn.h (internal_load_fn_p): Declare.
12336         (internal_gather_scatter_fn_p): Likewise.
12337         (internal_fn_mask_index): Likewise.
12338         (internal_gather_scatter_fn_supported_p): Likewise.
12339         * internal-fn.c (gather_load_direct): New macro.
12340         (expand_gather_load_optab_fn): New function.
12341         (direct_gather_load_optab_supported_p): New macro.
12342         (direct_internal_fn_optab): New function.
12343         (internal_load_fn_p): Likewise.
12344         (internal_gather_scatter_fn_p): Likewise.
12345         (internal_fn_mask_index): Likewise.
12346         (internal_gather_scatter_fn_supported_p): Likewise.
12347         * optabs-query.c (supports_at_least_one_mode_p): New function.
12348         (supports_vec_gather_load_p): Likewise.
12349         * optabs-query.h (supports_vec_gather_load_p): Declare.
12350         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12351         and memory_type field.
12352         (NUM_PATTERNS): Bump to 15.
12353         * tree-vect-data-refs.c: Include internal-fn.h.
12354         (vect_gather_scatter_fn_p): New function.
12355         (vect_describe_gather_scatter_call): Likewise.
12356         (vect_check_gather_scatter): Try using internal functions for
12357         gather loads.  Recognize existing calls to a gather load function.
12358         (vect_analyze_data_refs): Consider using gather loads if
12359         supports_vec_gather_load_p.
12360         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12361         (vect_get_gather_scatter_offset_type): Likewise.
12362         (vect_convert_mask_for_vectype): Likewise.
12363         (vect_add_conversion_to_patterm): Likewise.
12364         (vect_try_gather_scatter_pattern): Likewise.
12365         (vect_recog_gather_scatter_pattern): New pattern recognizer.
12366         (vect_vect_recog_func_ptrs): Add it.
12367         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12368         internal_fn_mask_index and internal_gather_scatter_fn_p.
12369         (check_load_store_masking): Take the gather_scatter_info as an
12370         argument and handle gather loads.
12371         (vect_get_gather_scatter_ops): New function.
12372         (vectorizable_call): Check internal_load_fn_p.
12373         (vectorizable_load): Likewise.  Handle gather load internal
12374         functions.
12375         (vectorizable_store): Update call to check_load_store_masking.
12376         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12377         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12378         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12379         (aarch64_gather_scale_operand_d): New predicates.
12380         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12381         (mask_gather_load<mode>): New insns.
12382
12383 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12384             Alan Hayward  <alan.hayward@arm.com>
12385             David Sherwood  <david.sherwood@arm.com>
12386
12387         * optabs.def (fold_left_plus_optab): New optab.
12388         * doc/md.texi (fold_left_plus_@var{m}): Document.
12389         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12390         * internal-fn.c (fold_left_direct): Define.
12391         (expand_fold_left_optab_fn): Likewise.
12392         (direct_fold_left_optab_supported_p): Likewise.
12393         * fold-const-call.c (fold_const_fold_left): New function.
12394         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12395         * tree-parloops.c (valid_reduction_p): New function.
12396         (gather_scalar_reductions): Use it.
12397         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12398         (vect_finish_replace_stmt): Declare.
12399         * tree-vect-loop.c (fold_left_reduction_fn): New function.
12400         (needs_fold_left_reduction_p): New function, split out from...
12401         (vect_is_simple_reduction): ...here.  Accept reductions that
12402         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12403         (vect_force_simple_reduction): Also store the reduction type in
12404         the assignment's STMT_VINFO_REDUC_TYPE.
12405         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12406         (merge_with_identity): New function.
12407         (vect_expand_fold_left): Likewise.
12408         (vectorize_fold_left_reduction): Likewise.
12409         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
12410         scalar phi in place for it.  Check for target support and reject
12411         cases that would reassociate the operation.  Defer the transform
12412         phase to vectorize_fold_left_reduction.
12413         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12414         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12415         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12416
12417 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12418
12419         * tree-if-conv.c (predicate_mem_writes): Remove redundant
12420         call to ifc_temp_var.
12421
12422 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12423             Alan Hayward  <alan.hayward@arm.com>
12424             David Sherwood  <david.sherwood@arm.com>
12425
12426         * target.def (legitimize_address_displacement): Take the original
12427         offset as a poly_int.
12428         * targhooks.h (default_legitimize_address_displacement): Update
12429         accordingly.
12430         * targhooks.c (default_legitimize_address_displacement): Likewise.
12431         * doc/tm.texi: Regenerate.
12432         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12433         as an argument, moving assert of ad->disp == ad->disp_term to...
12434         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
12435         Try calling targetm.legitimize_address_displacement before expanding
12436         the address rather than afterwards, and adjust for the new interface.
12437         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12438         Match the new hook interface.  Handle SVE addresses.
12439         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12440         new hook interface.
12441
12442 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12443
12444         * Makefile.in (OBJS): Add early-remat.o.
12445         * target.def (select_early_remat_modes): New hook.
12446         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12447         * doc/tm.texi: Regenerate.
12448         * targhooks.h (default_select_early_remat_modes): Declare.
12449         * targhooks.c (default_select_early_remat_modes): New function.
12450         * timevar.def (TV_EARLY_REMAT): New timevar.
12451         * passes.def (pass_early_remat): New pass.
12452         * tree-pass.h (make_pass_early_remat): Declare.
12453         * early-remat.c: New file.
12454         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12455         function.
12456         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12457
12458 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12459             Alan Hayward  <alan.hayward@arm.com>
12460             David Sherwood  <david.sherwood@arm.com>
12461
12462         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12463         vfm1 with a bound_epilog parameter.
12464         (vect_do_peeling): Update calls accordingly, and move the prologue
12465         call earlier in the function.  Treat the base bound_epilog as 0 for
12466         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
12467         this base when peeling for gaps.
12468         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12469         with fully-masked loops.
12470         (vect_estimate_min_profitable_iters): Handle the single peeled
12471         iteration in that case.
12472
12473 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12474             Alan Hayward  <alan.hayward@arm.com>
12475             David Sherwood  <david.sherwood@arm.com>
12476
12477         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12478         single-element interleaving even if the size is not a power of 2.
12479         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12480         accesses for single-element interleaving if the group size is
12481         not a power of 2.
12482
12483 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12484             Alan Hayward  <alan.hayward@arm.com>
12485             David Sherwood  <david.sherwood@arm.com>
12486
12487         * doc/md.texi (fold_extract_last_@var{m}): Document.
12488         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12489         * optabs.def (fold_extract_last_optab): New optab.
12490         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12491         * internal-fn.c (fold_extract_direct): New macro.
12492         (expand_fold_extract_optab_fn): Likewise.
12493         (direct_fold_extract_optab_supported_p): Likewise.
12494         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12495         * tree-vect-loop.c (vect_model_reduction_cost): Handle
12496         EXTRACT_LAST_REDUCTION.
12497         (get_initial_def_for_reduction): Do not create an initial vector
12498         for EXTRACT_LAST_REDUCTION reductions.
12499         (vectorizable_reduction): Leave the scalar phi in place for
12500         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
12501         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
12502         epilogue code for EXTRACT_LAST_REDUCTION and defer the
12503         transform phase to vectorizable_condition.
12504         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12505         split out from...
12506         (vect_finish_stmt_generation): ...here.
12507         (vect_finish_replace_stmt): New function.
12508         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12509         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12510         pattern.
12511         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12512
12513 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12514             Alan Hayward  <alan.hayward@arm.com>
12515             David Sherwood  <david.sherwood@arm.com>
12516
12517         * doc/md.texi (extract_last_@var{m}): Document.
12518         * optabs.def (extract_last_optab): New optab.
12519         * internal-fn.def (EXTRACT_LAST): New internal function.
12520         * internal-fn.c (cond_unary_direct): New macro.
12521         (expand_cond_unary_optab_fn): Likewise.
12522         (direct_cond_unary_optab_supported_p): Likewise.
12523         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12524         loops using EXTRACT_LAST.
12525         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12526         (extract_last_<mode>): ...this optab.
12527         (vec_extract<mode><Vel>): Update accordingly.
12528
12529 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12530             Alan Hayward  <alan.hayward@arm.com>
12531             David Sherwood  <david.sherwood@arm.com>
12532
12533         * target.def (empty_mask_is_expensive): New hook.
12534         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12535         * doc/tm.texi: Regenerate.
12536         * targhooks.h (default_empty_mask_is_expensive): Declare.
12537         * targhooks.c (default_empty_mask_is_expensive): New function.
12538         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12539         if the target says that empty masks are expensive.
12540         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12541         New function.
12542         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12543
12544 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12545             Alan Hayward  <alan.hayward@arm.com>
12546             David Sherwood  <david.sherwood@arm.com>
12547
12548         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12549         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12550         (vect_use_loop_mask_for_alignment_p): New function.
12551         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12552         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12553         niters_skip argument.  Make sure that the first niters_skip elements
12554         of the first iteration are inactive.
12555         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12556         Update call to vect_set_loop_masks_directly.
12557         (get_misalign_in_elems): New function, split out from...
12558         (vect_gen_prolog_loop_niters): ...here.
12559         (vect_update_init_of_dr): Take a code argument that specifies whether
12560         the adjustment should be added or subtracted.
12561         (vect_update_init_of_drs): Likewise.
12562         (vect_prepare_for_masked_peels): New function.
12563         (vect_do_peeling): Skip prologue peeling if we're using a mask
12564         instead.  Update call to vect_update_inits_of_drs.
12565         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12566         mask_skip_niters.
12567         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12568         alignment.  Do not include the number of peeled iterations in
12569         the minimum threshold in that case.
12570         (vectorizable_induction): Adjust the start value down by
12571         LOOP_VINFO_MASK_SKIP_NITERS iterations.
12572         (vect_transform_loop): Call vect_prepare_for_masked_peels.
12573         Take the number of skipped iterations into account when calculating
12574         the loop bounds.
12575         * tree-vect-stmts.c (vect_gen_while_not): New function.
12576
12577 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12578             Alan Hayward  <alan.hayward@arm.com>
12579             David Sherwood  <david.sherwood@arm.com>
12580
12581         * doc/sourcebuild.texi (vect_fully_masked): Document.
12582         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12583         default value to 0.
12584         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12585         split out from...
12586         (vect_analyze_loop_2): ...here. Don't check the vectorization
12587         factor against the number of loop iterations if the loop is
12588         fully-masked.
12589
12590 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12591             Alan Hayward  <alan.hayward@arm.com>
12592             David Sherwood  <david.sherwood@arm.com>
12593
12594         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12595         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12596         (dump_groups): Update accordingly.
12597         (iv_use::mem_type): New member variable.
12598         (address_p): New function.
12599         (record_use): Add a mem_type argument and initialize the new
12600         mem_type field.
12601         (record_group_use): Add a mem_type argument.  Use address_p.
12602         Remove obsolete null checks of base_object.  Update call to record_use.
12603         (find_interesting_uses_op): Update call to record_group_use.
12604         (find_interesting_uses_cond): Likewise.
12605         (find_interesting_uses_address): Likewise.
12606         (get_mem_type_for_internal_fn): New function.
12607         (find_address_like_use): Likewise.
12608         (find_interesting_uses_stmt): Try find_address_like_use before
12609         calling find_interesting_uses_op.
12610         (addr_offset_valid_p): Use the iv mem_type field as the type
12611         of the addressed memory.
12612         (add_autoinc_candidates): Likewise.
12613         (get_address_cost): Likewise.
12614         (split_small_address_groups_p): Use address_p.
12615         (split_address_groups): Likewise.
12616         (add_iv_candidate_for_use): Likewise.
12617         (autoinc_possible_for_pair): Likewise.
12618         (rewrite_groups): Likewise.
12619         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12620         (determine_group_iv_cost): Update after split of USE_ADDRESS.
12621         (get_alias_ptr_type_for_ptr_address): New function.
12622         (rewrite_use_address): Rewrite address uses in calls that were
12623         identified by find_address_like_use.
12624
12625 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12626             Alan Hayward  <alan.hayward@arm.com>
12627             David Sherwood  <david.sherwood@arm.com>
12628
12629         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12630         TARGET_MEM_REFs.
12631         * gimple-expr.h (is_gimple_addressable: Likewise.
12632         * gimple-expr.c (is_gimple_address): Likewise.
12633         * internal-fn.c (expand_call_mem_ref): New function.
12634         (expand_mask_load_optab_fn): Use it.
12635         (expand_mask_store_optab_fn): Likewise.
12636
12637 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12638             Alan Hayward  <alan.hayward@arm.com>
12639             David Sherwood  <david.sherwood@arm.com>
12640
12641         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12642         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12643         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12644         (cond_umax@var{mode}): Document.
12645         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12646         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12647         (cond_umin_optab, cond_umax_optab): New optabs.
12648         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12649         (COND_IOR, COND_XOR): New internal functions.
12650         * internal-fn.h (get_conditional_internal_fn): Declare.
12651         * internal-fn.c (cond_binary_direct): New macro.
12652         (expand_cond_binary_optab_fn): Likewise.
12653         (direct_cond_binary_optab_supported_p): Likewise.
12654         (get_conditional_internal_fn): New function.
12655         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12656         Cope with reduction statements that are vectorized as calls rather
12657         than assignments.
12658         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12659         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12660         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12661         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12662         (UNSPEC_COND_EOR): New unspecs.
12663         (optab): Add mappings for them.
12664         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12665         (sve_int_op, sve_fp_op): New int attributes.
12666
12667 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12668             Alan Hayward  <alan.hayward@arm.com>
12669             David Sherwood  <david.sherwood@arm.com>
12670
12671         * optabs.def (while_ult_optab): New optab.
12672         * doc/md.texi (while_ult@var{m}@var{n}): Document.
12673         * internal-fn.def (WHILE_ULT): New internal function.
12674         * internal-fn.h (direct_internal_fn_supported_p): New override
12675         that takes two types as argument.
12676         * internal-fn.c (while_direct): New macro.
12677         (expand_while_optab_fn): New function.
12678         (convert_optab_supported_p): Likewise.
12679         (direct_while_optab_supported_p): New macro.
12680         * wide-int.h (wi::udiv_ceil): New function.
12681         * tree-vectorizer.h (rgroup_masks): New structure.
12682         (vec_loop_masks): New typedef.
12683         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12684         and fully_masked_p.
12685         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12686         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12687         (vect_max_vf): New function.
12688         (slpeel_make_loop_iterate_ntimes): Delete.
12689         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12690         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12691         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12692         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12693         internal-fn.h, stor-layout.h and optabs-query.h.
12694         (vect_set_loop_mask): New function.
12695         (add_preheader_seq): Likewise.
12696         (add_header_seq): Likewise.
12697         (interleave_supported_p): Likewise.
12698         (vect_maybe_permute_loop_masks): Likewise.
12699         (vect_set_loop_masks_directly): Likewise.
12700         (vect_set_loop_condition_masked): Likewise.
12701         (vect_set_loop_condition_unmasked): New function, split out from
12702         slpeel_make_loop_iterate_ntimes.
12703         (slpeel_make_loop_iterate_ntimes): Rename to..
12704         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
12705         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12706         (vect_do_peeling): Update call accordingly.
12707         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12708         loops.
12709         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12710         mask_compare_type, can_fully_mask_p and fully_masked_p.
12711         (release_vec_loop_masks): New function.
12712         (_loop_vec_info): Use it to free the loop masks.
12713         (can_produce_all_loop_masks_p): New function.
12714         (vect_get_max_nscalars_per_iter): Likewise.
12715         (vect_verify_full_masking): Likewise.
12716         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12717         retries, and free the mask rgroups before retrying.  Check loop-wide
12718         reasons for disallowing fully-masked loops.  Make the final decision
12719         about whether use a fully-masked loop or not.
12720         (vect_estimate_min_profitable_iters): Do not assume that peeling
12721         for the number of iterations will be needed for fully-masked loops.
12722         (vectorizable_reduction): Disable fully-masked loops.
12723         (vectorizable_live_operation): Likewise.
12724         (vect_halve_mask_nunits): New function.
12725         (vect_double_mask_nunits): Likewise.
12726         (vect_record_loop_mask): Likewise.
12727         (vect_get_loop_mask): Likewise.
12728         (vect_transform_loop): Handle the case in which the final loop
12729         iteration might handle a partial vector.  Call vect_set_loop_condition
12730         instead of slpeel_make_loop_iterate_ntimes.
12731         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12732         (check_load_store_masking): New function.
12733         (prepare_load_store_mask): Likewise.
12734         (vectorizable_store): Handle fully-masked loops.
12735         (vectorizable_load): Likewise.
12736         (supportable_widening_operation): Use vect_halve_mask_nunits for
12737         booleans.
12738         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12739         (vect_gen_while): New function.
12740         * config/aarch64/aarch64.md (umax<mode>3): New expander.
12741         (aarch64_uqdec<mode>): New insn.
12742
12743 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12744             Alan Hayward  <alan.hayward@arm.com>
12745             David Sherwood  <david.sherwood@arm.com>
12746
12747         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12748         (reduc_xor_scal_optab): New optabs.
12749         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12750         (reduc_xor_scal_@var{m}): Document.
12751         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12752         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12753         internal functions.
12754         * fold-const-call.c (fold_const_call): Handle them.
12755         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12756         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12757         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12758         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12759         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12760         (UNSPEC_XORV): New unspecs.
12761         (optab): Add entries for them.
12762         (BITWISEV): New int iterator.
12763         (bit_reduc_op): New int attributes.
12764
12765 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12766             Alan Hayward  <alan.hayward@arm.com>
12767             David Sherwood  <david.sherwood@arm.com>
12768
12769         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12770         * internal-fn.def (VEC_SHL_INSERT): New internal function.
12771         * optabs.def (vec_shl_insert_optab): New optab.
12772         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12773         (duplicate_and_interleave): Likewise.
12774         * tree-vect-loop.c: Include internal-fn.h.
12775         (neutral_op_for_slp_reduction): New function, split out from
12776         get_initial_defs_for_reduction.
12777         (get_initial_def_for_reduction): Handle option 2 for variable-length
12778         vectors by loading the neutral value into a vector and then shifting
12779         the initial value into element 0.
12780         (get_initial_defs_for_reduction): Replace the code argument with
12781         the neutral value calculated by neutral_op_for_slp_reduction.
12782         Use gimple_build_vector for constant-length vectors.
12783         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12784         but the first group_size elements have a neutral value.
12785         Use duplicate_and_interleave otherwise.
12786         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12787         Update call to get_initial_defs_for_reduction.  Handle SLP
12788         reductions for variable-length vectors by creating one vector
12789         result for each scalar result, with the elements associated
12790         with other scalar results stubbed out with the neutral value.
12791         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12792         Require IFN_VEC_SHL_INSERT for double reductions on
12793         variable-length vectors, or SLP reductions that have
12794         a neutral value.  Require can_duplicate_and_interleave_p
12795         support for variable-length unchained SLP reductions if there
12796         is no neutral value, such as for MIN/MAX reductions.  Also require
12797         the number of vector elements to be a multiple of the number of
12798         SLP statements when doing variable-length unchained SLP reductions.
12799         Update call to vect_create_epilog_for_reduction.
12800         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12801         and remove initial values.
12802         (duplicate_and_interleave): Make public.
12803         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12804         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12805
12806 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12807             Alan Hayward  <alan.hayward@arm.com>
12808             David Sherwood  <david.sherwood@arm.com>
12809
12810         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12811         (can_duplicate_and_interleave_p): New function.
12812         (vect_get_and_check_slp_defs): Take the vector of statements
12813         rather than just the current one.  Remove excess parentheses.
12814         Restriction rejectinon of vect_constant_def and vect_external_def
12815         for variable-length vectors to boolean types, or types for which
12816         can_duplicate_and_interleave_p is false.
12817         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12818         (duplicate_and_interleave): New function.
12819         (vect_get_constant_vectors): Use gimple_build_vector for
12820         constant-length vectors and suitable variable-length constant
12821         vectors.  Use duplicate_and_interleave for other variable-length
12822         vectors.  Don't defer the update when inserting new statements.
12823
12824 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12825             Alan Hayward  <alan.hayward@arm.com>
12826             David Sherwood  <david.sherwood@arm.com>
12827
12828         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12829         min_profitable_iters doesn't go negative.
12830
12831 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12832             Alan Hayward  <alan.hayward@arm.com>
12833             David Sherwood  <david.sherwood@arm.com>
12834
12835         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12836         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12837         * optabs.def (vec_mask_load_lanes_optab): New optab.
12838         (vec_mask_store_lanes_optab): Likewise.
12839         * internal-fn.def (MASK_LOAD_LANES): New internal function.
12840         (MASK_STORE_LANES): Likewise.
12841         * internal-fn.c (mask_load_lanes_direct): New macro.
12842         (mask_store_lanes_direct): Likewise.
12843         (expand_mask_load_optab_fn): Handle masked operations.
12844         (expand_mask_load_lanes_optab_fn): New macro.
12845         (expand_mask_store_optab_fn): Handle masked operations.
12846         (expand_mask_store_lanes_optab_fn): New macro.
12847         (direct_mask_load_lanes_optab_supported_p): Likewise.
12848         (direct_mask_store_lanes_optab_supported_p): Likewise.
12849         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12850         parameter.
12851         (vect_load_lanes_supported): Likewise.
12852         * tree-vect-data-refs.c (strip_conversion): New function.
12853         (can_group_stmts_p): Likewise.
12854         (vect_analyze_data_ref_accesses): Use it instead of checking
12855         for a pair of assignments.
12856         (vect_store_lanes_supported): Take a masked_p parameter.
12857         (vect_load_lanes_supported): Likewise.
12858         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12859         vect_store_lanes_supported and vect_load_lanes_supported.
12860         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12861         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12862         parameter.  Don't allow gaps for masked accesses.
12863         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
12864         and vect_load_lanes_supported.
12865         (get_load_store_type): Take a masked_p parameter and update
12866         call to get_group_load_store_type.
12867         (vectorizable_store): Update call to get_load_store_type.
12868         Handle IFN_MASK_STORE_LANES.
12869         (vectorizable_load): Update call to get_load_store_type.
12870         Handle IFN_MASK_LOAD_LANES.
12871
12872 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12873             Alan Hayward  <alan.hayward@arm.com>
12874             David Sherwood  <david.sherwood@arm.com>
12875
12876         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12877         modes for SVE.
12878         * config/aarch64/aarch64-protos.h
12879         (aarch64_sve_struct_memory_operand_p): Declare.
12880         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12881         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12882         (VPRED, vpred): Handle SVE structure modes.
12883         * config/aarch64/constraints.md (Utx): New constraint.
12884         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12885         (aarch64_sve_struct_nonimmediate_operand): New predicates.
12886         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12887         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12888         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12889         structure modes.  Split into pieces after RA.
12890         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12891         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12892         New patterns.
12893         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12894         SVE structure modes.
12895         (aarch64_classify_address): Likewise.
12896         (sizetochar): Move earlier in file.
12897         (aarch64_print_operand): Handle SVE register lists.
12898         (aarch64_array_mode): New function.
12899         (aarch64_sve_struct_memory_operand_p): Likewise.
12900         (TARGET_ARRAY_MODE): Redefine.
12901
12902 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12903             Alan Hayward  <alan.hayward@arm.com>
12904             David Sherwood  <david.sherwood@arm.com>
12905
12906         * target.def (array_mode): New target hook.
12907         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12908         * doc/tm.texi: Regenerate.
12909         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12910         * hooks.c (hook_optmode_mode_uhwi_none): New function.
12911         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12912         targetm.array_mode.
12913         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
12914         type sizes.
12915
12916 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12917             Alan Hayward  <alan.hayward@arm.com>
12918             David Sherwood  <david.sherwood@arm.com>
12919
12920         * fold-const.c (fold_binary_loc): Check the argument types
12921         rather than the result type when testing for a vector operation.
12922
12923 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12924
12925         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12926         * doc/tm.texi: Regenerate.
12927
12928 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12929             Alan Hayward  <alan.hayward@arm.com>
12930             David Sherwood  <david.sherwood@arm.com>
12931
12932         * doc/invoke.texi (-msve-vector-bits=): Document new option.
12933         (sve): Document new AArch64 extension.
12934         * doc/md.texi (w): Extend the description of the AArch64
12935         constraint to include SVE vectors.
12936         (Upl, Upa): Document new AArch64 predicate constraints.
12937         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12938         enum.
12939         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12940         (msve-vector-bits=): New option.
12941         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12942         SVE when these are disabled.
12943         (sve): New extension.
12944         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12945         modes.  Adjust their number of units based on aarch64_sve_vg.
12946         (MAX_BITSIZE_MODE_ANY_MODE): Define.
12947         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12948         aarch64_addr_query_type.
12949         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12950         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12951         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12952         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12953         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12954         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12955         (aarch64_simd_imm_zero_p): Delete.
12956         (aarch64_check_zero_based_sve_index_immediate): Declare.
12957         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12958         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12959         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12960         (aarch64_sve_float_mul_immediate_p): Likewise.
12961         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12962         rather than an rtx.
12963         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12964         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12965         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12966         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12967         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12968         (aarch64_regmode_natural_size): Likewise.
12969         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12970         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12971         left one place.
12972         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12973         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12974         for VG and the SVE predicate registers.
12975         (V_ALIASES): Add a "z"-prefixed alias.
12976         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12977         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12978         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12979         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12980         (REG_CLASS_NAMES): Add entries for them.
12981         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
12982         and the predicate registers.
12983         (aarch64_sve_vg): Declare.
12984         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12985         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12986         (REGMODE_NATURAL_SIZE): Define.
12987         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12988         SVE macros.
12989         * config/aarch64/aarch64.c: Include cfgrtl.h.
12990         (simd_immediate_info): Add a constructor for series vectors,
12991         and an associated step field.
12992         (aarch64_sve_vg): New variable.
12993         (aarch64_dbx_register_number): Handle VG and the predicate registers.
12994         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12995         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12996         (VEC_ANY_DATA, VEC_STRUCT): New constants.
12997         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12998         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12999         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13000         (aarch64_get_mask_mode): New functions.
13001         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13002         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13003         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
13004         predicate modes and predicate registers.  Explicitly restrict
13005         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
13006         to store a vector mode if it is recognized by
13007         aarch64_classify_vector_mode.
13008         (aarch64_regmode_natural_size): New function.
13009         (aarch64_hard_regno_caller_save_mode): Return the original mode
13010         for predicates.
13011         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13012         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13013         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13014         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13015         functions.
13016         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
13017         does not overlap dest if the function is frame-related.  Handle
13018         SVE constants.
13019         (aarch64_split_add_offset): New function.
13020         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13021         them aarch64_add_offset.
13022         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13023         and update call to aarch64_sub_sp.
13024         (aarch64_add_cfa_expression): New function.
13025         (aarch64_expand_prologue): Pass extra temporary registers to the
13026         functions above.  Handle the case in which we need to emit new
13027         DW_CFA_expressions for registers that were originally saved
13028         relative to the stack pointer, but now have to be expressed
13029         relative to the frame pointer.
13030         (aarch64_output_mi_thunk): Pass extra temporary registers to the
13031         functions above.
13032         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
13033         IP0 and IP1 values for SVE frames.
13034         (aarch64_expand_vec_series): New function.
13035         (aarch64_expand_sve_widened_duplicate): Likewise.
13036         (aarch64_expand_sve_const_vector): Likewise.
13037         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13038         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
13039         into the register, rather than emitting a SET directly.
13040         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13041         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13042         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13043         (offset_9bit_signed_scaled_p): New functions.
13044         (aarch64_replicate_bitmask_imm): New function.
13045         (aarch64_bitmask_imm): Use it.
13046         (aarch64_cannot_force_const_mem): Reject expressions involving
13047         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
13048         (aarch64_classify_index): Handle SVE indices, by requiring
13049         a plain register index with a scale that matches the element size.
13050         (aarch64_classify_address): Handle SVE addresses.  Assert that
13051         the mode of the address is VOIDmode or an integer mode.
13052         Update call to aarch64_classify_symbol.
13053         (aarch64_classify_symbolic_expression): Update call to
13054         aarch64_classify_symbol.
13055         (aarch64_const_vec_all_in_range_p): New function.
13056         (aarch64_print_vector_float_operand): Likewise.
13057         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
13058         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
13059         and the FP immediates 1.0 and 0.5.
13060         (aarch64_print_address_internal): Handle SVE addresses.
13061         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13062         (aarch64_regno_regclass): Handle predicate registers.
13063         (aarch64_secondary_reload): Handle big-endian reloads of SVE
13064         data modes.
13065         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13066         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13067         (aarch64_convert_sve_vector_bits): New function.
13068         (aarch64_override_options): Use it to handle -msve-vector-bits=.
13069         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13070         rather than an rtx.
13071         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13072         Handle SVE vector and predicate modes.  Accept VL-based constants
13073         that need only one temporary register, and VL offsets that require
13074         no temporary registers.
13075         (aarch64_conditional_register_usage): Mark the predicate registers
13076         as fixed if SVE isn't available.
13077         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13078         Return true for SVE vector and predicate modes.
13079         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13080         rather than an unsigned int.  Handle SVE modes.
13081         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
13082         SVE modes.
13083         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13084         if SVE is enabled.
13085         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13086         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13087         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13088         (aarch64_sve_float_mul_immediate_p): New functions.
13089         (aarch64_sve_valid_immediate): New function.
13090         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13091         Explicitly reject structure modes.  Check for INDEX constants.
13092         Handle PTRUE and PFALSE constants.
13093         (aarch64_check_zero_based_sve_index_immediate): New function.
13094         (aarch64_simd_imm_zero_p): Delete.
13095         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13096         vector modes.  Accept constants in the range of CNT[BHWD].
13097         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13098         ask for an Advanced SIMD mode.
13099         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13100         (aarch64_simd_vector_alignment): Handle SVE predicates.
13101         (aarch64_vectorize_preferred_vector_alignment): New function.
13102         (aarch64_simd_vector_alignment_reachable): Use it instead of
13103         the vector size.
13104         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13105         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13106         functions.
13107         (MAX_VECT_LEN): Delete.
13108         (expand_vec_perm_d): Add a vec_flags field.
13109         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13110         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13111         (aarch64_evpc_ext): Don't apply a big-endian lane correction
13112         for SVE modes.
13113         (aarch64_evpc_rev): Rename to...
13114         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
13115         (aarch64_evpc_rev_global): New function.
13116         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13117         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13118         MAX_VECT_LEN.
13119         (aarch64_evpc_sve_tbl): New function.
13120         (aarch64_expand_vec_perm_const_1): Update after rename of
13121         aarch64_evpc_rev.  Handle SVE permutes too, trying
13122         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13123         than aarch64_evpc_tbl.
13124         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13125         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13126         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13127         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13128         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13129         (aarch64_expand_sve_vcond): New functions.
13130         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13131         of aarch64_vector_mode_p.
13132         (aarch64_dwarf_poly_indeterminate_value): New function.
13133         (aarch64_compute_pressure_classes): Likewise.
13134         (aarch64_can_change_mode_class): Likewise.
13135         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13136         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13137         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13138         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13139         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13140         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13141         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13142         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13143         constraints.
13144         (Dn, Dl, Dr): Accept const as well as const_vector.
13145         (Dz): Likewise.  Compare against CONST0_RTX.
13146         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13147         of "vector" where appropriate.
13148         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13149         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13150         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13151         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13152         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13153         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13154         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13155         (v_int_equiv): Extend to SVE modes.
13156         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13157         mode attributes.
13158         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13159         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13160         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13161         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13162         (SVE_COND_FP_CMP): New int iterators.
13163         (perm_hilo): Handle the new unpack unspecs.
13164         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13165         attributes.
13166         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13167         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13168         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13169         (aarch64_equality_operator, aarch64_constant_vector_operand)
13170         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13171         (aarch64_sve_nonimmediate_operand): Likewise.
13172         (aarch64_sve_general_operand): Likewise.
13173         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13174         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13175         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13176         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13177         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13178         (aarch64_sve_float_arith_immediate): Likewise.
13179         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13180         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13181         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13182         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13183         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13184         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13185         (aarch64_sve_float_arith_operand): Likewise.
13186         (aarch64_sve_float_arith_with_sub_operand): Likewise.
13187         (aarch64_sve_float_mul_operand): Likewise.
13188         (aarch64_sve_vec_perm_operand): Likewise.
13189         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13190         (aarch64_mov_operand): Accept const_poly_int and const_vector.
13191         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13192         as well as const_vector.
13193         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13194         in file.  Use CONST0_RTX and CONSTM1_RTX.
13195         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
13196         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13197         Use aarch64_simd_imm_zero.
13198         * config/aarch64/aarch64-sve.md: New file.
13199         * config/aarch64/aarch64.md: Include it.
13200         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13201         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13202         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13203         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13204         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13205         (sve): New attribute.
13206         (enabled): Disable instructions with the sve attribute unless
13207         TARGET_SVE.
13208         (movqi, movhi): Pass CONST_POLY_INT operaneds through
13209         aarch64_expand_mov_immediate.
13210         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13211         CNT[BHSD] immediates.
13212         (movti): Split CONST_POLY_INT moves into two halves.
13213         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13214         Split additions that need a temporary here if the destination
13215         is the stack pointer.
13216         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13217         (*add<mode>3_poly_1): New instruction.
13218         (set_clobber_cc): New expander.
13219
13220 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
13221
13222         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13223         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
13224         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13225         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13226         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
13227         Change innermode from fixed_mode_size to machine_mode.
13228         (simplify_subreg): Update call accordingly.  Handle a constant-sized
13229         subreg of a variable-length CONST_VECTOR.
13230
13231 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
13232             Alan Hayward  <alan.hayward@arm.com>
13233             David Sherwood  <david.sherwood@arm.com>
13234
13235         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13236         (add_offset_to_base): New function, split out from...
13237         (create_mem_ref): ...here.  When handling a scale other than 1,
13238         check first whether the address is valid without the offset.
13239         Add it into the base if so, leaving the index and scale as-is.
13240
13241 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
13242
13243         PR c++/83778
13244         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13245         fold_for_warn before checking if arg2 is INTEGER_CST.
13246
13247 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
13248
13249         * config/rs6000/predicates.md (load_multiple_operation): Delete.
13250         (store_multiple_operation): Delete.
13251         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13252         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13253         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13254         guarded by TARGET_STRING.
13255         (rs6000_output_load_multiple): Delete.
13256         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13257         OPTION_MASK_STRING / TARGET_STRING handling.
13258         (print_operand) <'N', 'O'>: Add comment that these are unused now.
13259         (const rs6000_opt_masks) <"string">: Change mask to 0.
13260         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13261         (MASK_STRING): Delete.
13262         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13263         parts.  Simplify.
13264         (load_multiple): Delete.
13265         (*ldmsi8): Delete.
13266         (*ldmsi7): Delete.
13267         (*ldmsi6): Delete.
13268         (*ldmsi5): Delete.
13269         (*ldmsi4): Delete.
13270         (*ldmsi3): Delete.
13271         (store_multiple): Delete.
13272         (*stmsi8): Delete.
13273         (*stmsi7): Delete.
13274         (*stmsi6): Delete.
13275         (*stmsi5): Delete.
13276         (*stmsi4): Delete.
13277         (*stmsi3): Delete.
13278         (movmemsi_8reg): Delete.
13279         (corresponding unnamed define_insn): Delete.
13280         (movmemsi_6reg): Delete.
13281         (corresponding unnamed define_insn): Delete.
13282         (movmemsi_4reg): Delete.
13283         (corresponding unnamed define_insn): Delete.
13284         (movmemsi_2reg): Delete.
13285         (corresponding unnamed define_insn): Delete.
13286         (movmemsi_1reg): Delete.
13287         (corresponding unnamed define_insn): Delete.
13288         * config/rs6000/rs6000.opt (mno-string): New.
13289         (mstring): Replace by deprecation warning stub.
13290         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13291
13292 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
13293
13294         * regrename.c (regrename_do_replace): If replacing the same
13295         reg multiple times, try to reuse last created gen_raw_REG.
13296
13297         PR debug/81155
13298         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13299         main to workaround a bug in GDB.
13300
13301 2018-01-12  Tom de Vries  <tom@codesourcery.com>
13302
13303         PR target/83737
13304         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13305
13306 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
13307
13308         PR rtl-optimization/80481
13309         * ira-color.c (get_cap_member): New function.
13310         (allocnos_conflict_by_live_ranges_p): Use it.
13311         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13312         (setup_slot_coalesced_allocno_live_ranges): Ditto.
13313
13314 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
13315
13316         PR target/83628
13317         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13318         (*saddl_se_1): Ditto.
13319         (*ssubsi_1): Ditto.
13320         (*ssubl_se_1): Ditto.
13321
13322 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
13323
13324         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13325         rather than wi::to_widest for DR_INITs.
13326         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13327         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13328         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13329         INTEGER_CSTs.
13330         (vect_analyze_group_access_1): Note that here.
13331
13332 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
13333
13334         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13335         polynomial type sizes.
13336
13337 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
13338
13339         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13340         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13341         (gimple_add_tmp_var): Likewise.
13342
13343 2018-01-12  Martin Liska  <mliska@suse.cz>
13344
13345         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13346         (gimple_alloc_sizes): Likewise.
13347         (dump_gimple_statistics): Use PRIu64 in printf format.
13348         * gimple.h: Change uint64_t to int.
13349
13350 2018-01-12  Martin Liska  <mliska@suse.cz>
13351
13352         * tree-core.h: Use uint64_t instead of int.
13353         * tree.c (tree_node_counts): Likewise.
13354         (tree_node_sizes): Likewise.
13355         (dump_tree_statistics): Use PRIu64 in printf format.
13356
13357 2018-01-12  Martin Liska  <mliska@suse.cz>
13358
13359         * Makefile.in: As qsort_chk is implemented in vec.c, add
13360         vec.o to linkage of gencfn-macros.
13361         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13362         passing the info to record_node_allocation_statistics.
13363         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13364         and pass the info.
13365         * ggc-common.c (struct ggc_usage): Add operator== and use
13366         it in operator< and compare function.
13367         * mem-stats.h (struct mem_usage): Likewise.
13368         * vec.c (struct vec_usage): Remove operator< and compare
13369         function. Can be simply inherited.
13370
13371 2018-01-12  Martin Jambor  <mjambor@suse.cz>
13372
13373         PR target/81616
13374         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13375         * tree-ssa-math-opts.c: Include domwalk.h.
13376         (convert_mult_to_fma_1): New function.
13377         (fma_transformation_info): New type.
13378         (fma_deferring_state): Likewise.
13379         (cancel_fma_deferring): New function.
13380         (result_of_phi): Likewise.
13381         (last_fma_candidate_feeds_initial_phi): Likewise.
13382         (convert_mult_to_fma): Added deferring logic, split actual
13383         transformation to convert_mult_to_fma_1.
13384         (math_opts_dom_walker): New type.
13385         (math_opts_dom_walker::after_dom_children): New method, body moved
13386         here from pass_optimize_widening_mul::execute, added deferring logic
13387         bits.
13388         (pass_optimize_widening_mul::execute): Moved most of code to
13389         math_opts_dom_walker::after_dom_children.
13390         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13391         * config/i386/i386.c (ix86_option_override_internal): Added
13392         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13393
13394 2018-01-12  Richard Biener  <rguenther@suse.de>
13395
13396         PR debug/83157
13397         * dwarf2out.c (gen_variable_die): Do not reset old_die for
13398         inline instance vars.
13399
13400 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
13401
13402         PR target/81819
13403         * config/rx/rx.c (rx_is_restricted_memory_address):
13404         Handle SUBREG case.
13405
13406 2018-01-12  Richard Biener  <rguenther@suse.de>
13407
13408         PR tree-optimization/80846
13409         * target.def (split_reduction): New target hook.
13410         * targhooks.c (default_split_reduction): New function.
13411         * targhooks.h (default_split_reduction): Declare.
13412         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13413         target requests first reduce vectors by combining low and high
13414         parts.
13415         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13416         (get_vectype_for_scalar_type_and_size): Export.
13417         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13418         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13419         * doc/tm.texi: Regenerate.
13420         * config/i386/i386.c (ix86_split_reduction): Implement
13421         TARGET_VECTORIZE_SPLIT_REDUCTION.
13422
13423 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
13424
13425         PR target/83368
13426         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13427         in PIC mode except for TARGET_VXWORKS_RTP.
13428         * config/sparc/sparc.c: Include cfgrtl.h.
13429         (TARGET_INIT_PIC_REG): Define.
13430         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13431         (sparc_pic_register_p): New predicate.
13432         (sparc_legitimate_address_p): Use it.
13433         (sparc_legitimize_pic_address): Likewise.
13434         (sparc_delegitimize_address): Likewise.
13435         (sparc_mode_dependent_address_p): Likewise.
13436         (gen_load_pcrel_sym): Remove 4th parameter.
13437         (load_got_register): Adjust call to above.  Remove obsolete stuff.
13438         (sparc_expand_prologue): Do not call load_got_register here.
13439         (sparc_flat_expand_prologue): Likewise.
13440         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13441         (sparc_use_pseudo_pic_reg): New function.
13442         (sparc_init_pic_reg): Likewise.
13443         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13444         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13445
13446 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
13447
13448         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13449         Add item for branch_cost.
13450
13451 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
13452
13453         PR rtl-optimization/83565
13454         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13455         not extend the result to a larger mode for rotate operations.
13456         (num_sign_bit_copies1): Likewise.
13457
13458 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13459
13460         PR target/40411
13461         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13462         -symbolic.
13463         Use values-Xc.o for -pedantic.
13464         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13465
13466 2018-01-12  Martin Liska  <mliska@suse.cz>
13467
13468         PR ipa/83054
13469         * ipa-devirt.c (final_warning_record::grow_type_warnings):
13470         New function.
13471         (possible_polymorphic_call_targets): Use it.
13472         (ipa_devirt): Likewise.
13473
13474 2018-01-12  Martin Liska  <mliska@suse.cz>
13475
13476         * profile-count.h (enum profile_quality): Use 0 as invalid
13477         enum value of profile_quality.
13478
13479 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
13480
13481         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13482         -mext-string options.
13483
13484 2018-01-12  Richard Biener  <rguenther@suse.de>
13485
13486         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13487         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13488         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13489         Likewise.
13490         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13491
13492 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
13493
13494         * configure.ac (--with-long-double-format): Add support for the
13495         configuration option to change the default long double format on
13496         PowerPC systems.
13497         * config.gcc (powerpc*-linux*-*): Likewise.
13498         * configure: Regenerate.
13499         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13500         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13501         used without modification.
13502
13503 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13504
13505         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13506         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13507         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13508         MISC_BUILTIN_SPEC_BARRIER.
13509         (rs6000_init_builtins): Likewise.
13510         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13511         enum value.
13512         (speculation_barrier): New define_insn.
13513         * doc/extend.texi: Document __builtin_speculation_barrier.
13514
13515 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
13516
13517         PR target/83203
13518         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13519         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13520         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13521         iterators.
13522         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
13523         integral modes instead of "ss" and "sd".
13524         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13525         vectors with 32-bit and 64-bit elements.
13526         (vecdupssescalarmodesuffix): New mode attribute.
13527         (vec_dup<mode>): Use it.
13528
13529 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
13530
13531         PR target/83330
13532         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13533         frame if argument is passed on stack.
13534
13535 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
13536
13537         PR target/82682
13538         * ree.c (combine_reaching_defs): Optimize also
13539         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13540         reg2=any_extend(exp); reg1=reg2;, formatting fix.
13541
13542 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
13543
13544         PR middle-end/83189
13545         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13546
13547 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
13548
13549         PR middle-end/83718
13550         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13551         after they are computed.
13552
13553 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
13554
13555         PR tree-optimization/83695
13556         * gimple-loop-linterchange.cc
13557         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13558         reset cached scev information after interchange.
13559         (pass_linterchange::execute): Remove call to scev_reset_htab.
13560
13561 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13562
13563         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13564         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13565         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13566         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13567         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13568         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13569         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13570         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13571         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13572         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13573         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13574         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13575         (V_lane_reg): Likewise.
13576         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13577         New define_expand.
13578         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13579         (vfmal_lane_low<mode>_intrinsic,
13580         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13581         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13582         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13583         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13584         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13585         vfmsl_lane_high<mode>_intrinsic): New define_insns.
13586
13587 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13588
13589         * config/arm/arm-cpus.in (fp16fml): New feature.
13590         (ALL_SIMD): Add fp16fml.
13591         (armv8.2-a): Add fp16fml as an option.
13592         (armv8.3-a): Likewise.
13593         (armv8.4-a): Add fp16fml as part of fp16.
13594         * config/arm/arm.h (TARGET_FP16FML): Define.
13595         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13596         when appropriate.
13597         * config/arm/arm-modes.def (V2HF): Define.
13598         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13599         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13600         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13601         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13602         vfmsl_low, vfmsl_high): New set of builtins.
13603         * config/arm/iterators.md (PLUSMINUS): New code iterator.
13604         (vfml_op): New code attribute.
13605         (VFMLHALVES): New int iterator.
13606         (VFML, VFMLSEL): New mode attributes.
13607         (V_reg): Define mapping for V2HF.
13608         (V_hi, V_lo): New mode attributes.
13609         (VF_constraint): Likewise.
13610         (vfml_half, vfml_half_selector): New int attributes.
13611         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13612         define_expand.
13613         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13614         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13615         New define_insn.
13616         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13617         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13618         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13619         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
13620         documentation.
13621         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13622         Document new effective target and option set.
13623
13624 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13625
13626         * config/arm/arm-cpus.in (armv8_4): New feature.
13627         (ARMv8_4a): New fgroup.
13628         (armv8.4-a): New arch.
13629         * config/arm/arm-tables.opt: Regenerate.
13630         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13631         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13632         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13633         Add matching rules for -march=armv8.4-a and extensions.
13634         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13635
13636 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
13637
13638         PR target/81821
13639         * config/rx/rx.md (BW): New mode attribute.
13640         (sync_lock_test_and_setsi): Add mode suffix to insn output.
13641
13642 2018-01-11  Richard Biener  <rguenther@suse.de>
13643
13644         PR tree-optimization/83435
13645         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13646         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13647         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13648
13649 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13650             Alan Hayward  <alan.hayward@arm.com>
13651             David Sherwood  <david.sherwood@arm.com>
13652
13653         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13654         field.
13655         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
13656         (aarch64_print_address_internal): Use it to check for a zero offset.
13657
13658 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13659             Alan Hayward  <alan.hayward@arm.com>
13660             David Sherwood  <david.sherwood@arm.com>
13661
13662         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13663         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13664         Return a poly_int64 rather than a HOST_WIDE_INT.
13665         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13666         rather than a HOST_WIDE_INT.
13667         * config/aarch64/aarch64.h (aarch64_frame): Protect with
13668         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
13669         hard_fp_offset, frame_size, initial_adjust, callee_offset and
13670         final_offset from HOST_WIDE_INT to poly_int64.
13671         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13672         to_constant when getting the number of units in an Advanced SIMD
13673         mode.
13674         (aarch64_builtin_vectorized_function): Check for a constant number
13675         of units.
13676         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13677         GET_MODE_SIZE.
13678         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13679         attribute instead of GET_MODE_NUNITS.
13680         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13681         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13682         GET_MODE_SIZE for fixed-size registers.
13683         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13684         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13685         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13686         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13687         (aarch64_print_operand, aarch64_print_address_internal)
13688         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13689         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13690         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13691         Handle polynomial GET_MODE_SIZE.
13692         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
13693         wider than SImode without modification.
13694         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13695         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13696         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13697         passing and returning SVE modes.
13698         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13699         rather than GEN_INT.
13700         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13701         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13702         (aarch64_allocate_and_probe_stack_space): Likewise.
13703         (aarch64_layout_frame): Cope with polynomial offsets.
13704         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13705         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
13706         polynomial offsets.
13707         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13708         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13709         poly_int64 rather than a HOST_WIDE_INT.
13710         (aarch64_get_separate_components, aarch64_process_components)
13711         (aarch64_expand_prologue, aarch64_expand_epilogue)
13712         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13713         (aarch64_anchor_offset): New function, split out from...
13714         (aarch64_legitimize_address): ...here.
13715         (aarch64_builtin_vectorization_cost): Handle polynomial
13716         TYPE_VECTOR_SUBPARTS.
13717         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13718         GET_MODE_NUNITS.
13719         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13720         number of elements from the PARALLEL rather than the mode.
13721         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13722         rather than GET_MODE_BITSIZE.
13723         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13724         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13725         (aarch64_expand_vec_perm_const_1): Handle polynomial
13726         d->perm.length () and d->perm elements.
13727         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
13728         Apply to_constant to d->perm elements.
13729         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13730         polynomial CONST_VECTOR_NUNITS.
13731         (aarch64_move_pointer): Take amount as a poly_int64 rather
13732         than an int.
13733         (aarch64_progress_pointer): Avoid temporary variable.
13734         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13735         the mode attribute instead of GET_MODE.
13736
13737 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13738             Alan Hayward  <alan.hayward@arm.com>
13739             David Sherwood  <david.sherwood@arm.com>
13740
13741         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13742         x exists before using it.
13743         (aarch64_add_constant_internal): Rename to...
13744         (aarch64_add_offset_1): ...this.  Replace regnum with separate
13745         src and dest rtxes.  Handle the case in which they're different,
13746         including when the offset is zero.  Replace scratchreg with an rtx.
13747         Use 2 additions if there is no spare register into which we can
13748         move a 16-bit constant.
13749         (aarch64_add_constant): Delete.
13750         (aarch64_add_offset): Replace reg with separate src and dest
13751         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
13752         Use aarch64_add_offset_1.
13753         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13754         an rtx rather than an int.  Take the delta as a poly_int64
13755         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
13756         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13757         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13758         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13759         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13760         and aarch64_add_sp.
13761         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13762         aarch64_add_constant.
13763
13764 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13765
13766         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13767         Use scalar_float_mode.
13768
13769 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13770
13771         * config/aarch64/aarch64-simd.md
13772         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13773         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13774         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13775         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13776         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13777         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13778         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13779         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13780         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13781         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13782
13783 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13784
13785         PR target/83514
13786         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13787         targ_options->x_arm_arch_string is non NULL.
13788
13789 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
13790
13791         * config/aarch64/aarch64.h
13792         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
13793
13794 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
13795
13796         PR target/82096
13797         * expmed.c (emit_store_flag_force): Swap if const op0
13798         and change VOIDmode to mode of op0.
13799
13800 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13801
13802         PR rtl-optimization/83761
13803         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13804         than bytes to mode_for_size.
13805
13806 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
13807
13808         PR middle-end/83189
13809         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13810         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13811         profile.
13812
13813 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
13814
13815         PR middle-end/83575
13816         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13817         when in layout mode.
13818         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13819         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13820         partition fixup.
13821
13822 2018-01-10  Michael Collison  <michael.collison@arm.com>
13823
13824         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13825         * config/aarch64/aarch64-option-extension.def: Add
13826         AARCH64_OPT_EXTENSION of 'fp16fml'.
13827         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13828         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13829         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13830         * config/aarch64/constraints.md (Ui7): New constraint.
13831         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13832         (VFMLA_SEL_W): Ditto.
13833         (f16quad): Ditto.
13834         (f16mac1): Ditto.
13835         (VFMLA16_LOW): New int iterator.
13836         (VFMLA16_HIGH): Ditto.
13837         (UNSPEC_FMLAL): New unspec.
13838         (UNSPEC_FMLSL): Ditto.
13839         (UNSPEC_FMLAL2): Ditto.
13840         (UNSPEC_FMLSL2): Ditto.
13841         (f16mac): New code attribute.
13842         * config/aarch64/aarch64-simd-builtins.def
13843         (aarch64_fmlal_lowv2sf): Ditto.
13844         (aarch64_fmlsl_lowv2sf): Ditto.
13845         (aarch64_fmlalq_lowv4sf): Ditto.
13846         (aarch64_fmlslq_lowv4sf): Ditto.
13847         (aarch64_fmlal_highv2sf): Ditto.
13848         (aarch64_fmlsl_highv2sf): Ditto.
13849         (aarch64_fmlalq_highv4sf): Ditto.
13850         (aarch64_fmlslq_highv4sf): Ditto.
13851         (aarch64_fmlal_lane_lowv2sf): Ditto.
13852         (aarch64_fmlsl_lane_lowv2sf): Ditto.
13853         (aarch64_fmlal_laneq_lowv2sf): Ditto.
13854         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13855         (aarch64_fmlalq_lane_lowv4sf): Ditto.
13856         (aarch64_fmlsl_lane_lowv4sf): Ditto.
13857         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13858         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13859         (aarch64_fmlal_lane_highv2sf): Ditto.
13860         (aarch64_fmlsl_lane_highv2sf): Ditto.
13861         (aarch64_fmlal_laneq_highv2sf): Ditto.
13862         (aarch64_fmlsl_laneq_highv2sf): Ditto.
13863         (aarch64_fmlalq_lane_highv4sf): Ditto.
13864         (aarch64_fmlsl_lane_highv4sf): Ditto.
13865         (aarch64_fmlalq_laneq_highv4sf): Ditto.
13866         (aarch64_fmlsl_laneq_highv4sf): Ditto.
13867         * config/aarch64/aarch64-simd.md:
13868         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13869         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13870         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13871         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13872         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13873         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13874         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13875         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13876         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13877         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13878         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13879         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13880         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13881         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13882         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13883         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13884         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13885         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13886         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13887         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13888         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13889         (vfmlsl_low_u32): Ditto.
13890         (vfmlalq_low_u32): Ditto.
13891         (vfmlslq_low_u32): Ditto.
13892         (vfmlal_high_u32): Ditto.
13893         (vfmlsl_high_u32): Ditto.
13894         (vfmlalq_high_u32): Ditto.
13895         (vfmlslq_high_u32): Ditto.
13896         (vfmlal_lane_low_u32): Ditto.
13897         (vfmlsl_lane_low_u32): Ditto.
13898         (vfmlal_laneq_low_u32): Ditto.
13899         (vfmlsl_laneq_low_u32): Ditto.
13900         (vfmlalq_lane_low_u32): Ditto.
13901         (vfmlslq_lane_low_u32): Ditto.
13902         (vfmlalq_laneq_low_u32): Ditto.
13903         (vfmlslq_laneq_low_u32): Ditto.
13904         (vfmlal_lane_high_u32): Ditto.
13905         (vfmlsl_lane_high_u32): Ditto.
13906         (vfmlal_laneq_high_u32): Ditto.
13907         (vfmlsl_laneq_high_u32): Ditto.
13908         (vfmlalq_lane_high_u32): Ditto.
13909         (vfmlslq_lane_high_u32): Ditto.
13910         (vfmlalq_laneq_high_u32): Ditto.
13911         (vfmlslq_laneq_high_u32): Ditto.
13912         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13913         (AARCH64_FL_FOR_ARCH8_4): New.
13914         (AARCH64_ISA_F16FML): New ISA flag.
13915         (TARGET_F16FML): New feature flag for fp16fml.
13916         (doc/invoke.texi): Document new fp16fml option.
13917
13918 2018-01-10  Michael Collison  <michael.collison@arm.com>
13919
13920         * config/aarch64/aarch64-builtins.c:
13921         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13922         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13923         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13924         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13925         (AARCH64_ISA_SHA3): New ISA flag.
13926         (TARGET_SHA3): New feature flag for sha3.
13927         * config/aarch64/iterators.md (sha512_op): New int attribute.
13928         (CRYPTO_SHA512): New int iterator.
13929         (UNSPEC_SHA512H): New unspec.
13930         (UNSPEC_SHA512H2): Ditto.
13931         (UNSPEC_SHA512SU0): Ditto.
13932         (UNSPEC_SHA512SU1): Ditto.
13933         * config/aarch64/aarch64-simd-builtins.def
13934         (aarch64_crypto_sha512hqv2di): New builtin.
13935         (aarch64_crypto_sha512h2qv2di): Ditto.
13936         (aarch64_crypto_sha512su0qv2di): Ditto.
13937         (aarch64_crypto_sha512su1qv2di): Ditto.
13938         (aarch64_eor3qv8hi): Ditto.
13939         (aarch64_rax1qv2di): Ditto.
13940         (aarch64_xarqv2di): Ditto.
13941         (aarch64_bcaxqv8hi): Ditto.
13942         * config/aarch64/aarch64-simd.md:
13943         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13944         (aarch64_crypto_sha512su0qv2di): Ditto.
13945         (aarch64_crypto_sha512su1qv2di): Ditto.
13946         (aarch64_eor3qv8hi): Ditto.
13947         (aarch64_rax1qv2di): Ditto.
13948         (aarch64_xarqv2di): Ditto.
13949         (aarch64_bcaxqv8hi): Ditto.
13950         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13951         (vsha512h2q_u64): Ditto.
13952         (vsha512su0q_u64): Ditto.
13953         (vsha512su1q_u64): Ditto.
13954         (veor3q_u16): Ditto.
13955         (vrax1q_u64): Ditto.
13956         (vxarq_u64): Ditto.
13957         (vbcaxq_u16): Ditto.
13958         * config/arm/types.md (crypto_sha512): New type attribute.
13959         (crypto_sha3): Ditto.
13960         (doc/invoke.texi): Document new sha3 option.
13961
13962 2018-01-10  Michael Collison  <michael.collison@arm.com>
13963
13964         * config/aarch64/aarch64-builtins.c:
13965         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13966         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13967         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13968         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13969         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13970         (AARCH64_ISA_SM4): New ISA flag.
13971         (TARGET_SM4): New feature flag for sm4.
13972         * config/aarch64/aarch64-simd-builtins.def
13973         (aarch64_sm3ss1qv4si): Ditto.
13974         (aarch64_sm3tt1aq4si): Ditto.
13975         (aarch64_sm3tt1bq4si): Ditto.
13976         (aarch64_sm3tt2aq4si): Ditto.
13977         (aarch64_sm3tt2bq4si): Ditto.
13978         (aarch64_sm3partw1qv4si): Ditto.
13979         (aarch64_sm3partw2qv4si): Ditto.
13980         (aarch64_sm4eqv4si): Ditto.
13981         (aarch64_sm4ekeyqv4si): Ditto.
13982         * config/aarch64/aarch64-simd.md:
13983         (aarch64_sm3ss1qv4si): Ditto.
13984         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13985         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13986         (aarch64_sm4eqv4si): Ditto.
13987         (aarch64_sm4ekeyqv4si): Ditto.
13988         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13989         (sm3part_op): Ditto.
13990         (CRYPTO_SM3TT): Ditto.
13991         (CRYPTO_SM3PART): Ditto.
13992         (UNSPEC_SM3SS1): New unspec.
13993         (UNSPEC_SM3TT1A): Ditto.
13994         (UNSPEC_SM3TT1B): Ditto.
13995         (UNSPEC_SM3TT2A): Ditto.
13996         (UNSPEC_SM3TT2B): Ditto.
13997         (UNSPEC_SM3PARTW1): Ditto.
13998         (UNSPEC_SM3PARTW2): Ditto.
13999         (UNSPEC_SM4E): Ditto.
14000         (UNSPEC_SM4EKEY): Ditto.
14001         * config/aarch64/constraints.md (Ui2): New constraint.
14002         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14003         * config/arm/types.md (crypto_sm3): New type attribute.
14004         (crypto_sm4): Ditto.
14005         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14006         (vsm3tt1aq_u32): Ditto.
14007         (vsm3tt1bq_u32): Ditto.
14008         (vsm3tt2aq_u32): Ditto.
14009         (vsm3tt2bq_u32): Ditto.
14010         (vsm3partw1q_u32): Ditto.
14011         (vsm3partw2q_u32): Ditto.
14012         (vsm4eq_u32): Ditto.
14013         (vsm4ekeyq_u32): Ditto.
14014         (doc/invoke.texi): Document new sm4 option.
14015
14016 2018-01-10  Michael Collison  <michael.collison@arm.com>
14017
14018         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14019         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14020         (AARCH64_FL_FOR_ARCH8_4): New.
14021         (AARCH64_FL_V8_4): New flag.
14022         (doc/invoke.texi): Document new armv8.4-a option.
14023
14024 2018-01-10  Michael Collison  <michael.collison@arm.com>
14025
14026         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14027         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14028         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14029         * config/aarch64/aarch64-option-extension.def: Add
14030         AARCH64_OPT_EXTENSION of 'sha2'.
14031         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14032         (crypto): Disable sha2 and aes if crypto disabled.
14033         (crypto): Enable aes and sha2 if enabled.
14034         (simd): Disable sha2 and aes if simd disabled.
14035         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14036         New flags.
14037         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14038         (TARGET_SHA2): New feature flag for sha2.
14039         (TARGET_AES): New feature flag for aes.
14040         * config/aarch64/aarch64-simd.md:
14041         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14042         conditional on TARGET_AES.
14043         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14044         (aarch64_crypto_sha1hsi): Make pattern conditional
14045         on TARGET_SHA2.
14046         (aarch64_crypto_sha1hv4si): Ditto.
14047         (aarch64_be_crypto_sha1hv4si): Ditto.
14048         (aarch64_crypto_sha1su1v4si): Ditto.
14049         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14050         (aarch64_crypto_sha1su0v4si): Ditto.
14051         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14052         (aarch64_crypto_sha256su0v4si): Ditto.
14053         (aarch64_crypto_sha256su1v4si): Ditto.
14054         (doc/invoke.texi): Document new aes and sha2 options.
14055
14056 2018-01-10  Martin Sebor  <msebor@redhat.com>
14057
14058         PR tree-optimization/83781
14059         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14060         as string arrays.
14061
14062 2018-01-11  Martin Sebor  <msebor@gmail.com>
14063             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14064
14065         PR tree-optimization/83501
14066         PR tree-optimization/81703
14067
14068         * tree-ssa-strlen.c (get_string_cst): Rename...
14069         (get_string_len): ...to this.  Handle global constants.
14070         (handle_char_store): Adjust.
14071
14072 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
14073             Jim Wilson  <jimw@sifive.com>
14074
14075         * config/riscv/riscv-protos.h (riscv_output_return): New.
14076         * config/riscv/riscv.c (struct machine_function): New naked_p field.
14077         (riscv_attribute_table, riscv_output_return),
14078         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14079         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14080         (riscv_compute_frame_info): Only compute frame->mask if not a naked
14081         function.
14082         (riscv_expand_prologue): Add early return for naked function.
14083         (riscv_expand_epilogue): Likewise.
14084         (riscv_function_ok_for_sibcall): Return false for naked function.
14085         (riscv_set_current_function): New.
14086         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14087         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14088         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14089         * doc/extend.texi (RISC-V Function Attributes): New.
14090
14091 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
14092
14093         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14094         check for 128-bit long double before checking TCmode.
14095         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14096         128-bit long doubles before checking TFmode or TCmode.
14097         (FLOAT128_IBM_P): Likewise.
14098
14099 2018-01-10  Martin Sebor  <msebor@redhat.com>
14100
14101         PR tree-optimization/83671
14102         * builtins.c (c_strlen): Unconditionally return zero for the empty
14103         string.
14104         Use -Warray-bounds for warnings.
14105         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14106         for non-constant array indices with COMPONENT_REF, arrays of
14107         arrays, and pointers to arrays.
14108         (gimple_fold_builtin_strlen): Determine and set length range for
14109         non-constant character arrays.
14110
14111 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
14112
14113         PR middle-end/81897
14114         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14115         empty blocks.
14116
14117 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
14118
14119         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14120
14121 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
14122
14123         PR target/83399
14124         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14125         VECTOR_MEM_ALTIVEC_OR_VSX_P.
14126         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14127         indexed_or_indirect_operand predicate.
14128         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14129         (*vsx_le_perm_load_v8hi): Likewise.
14130         (*vsx_le_perm_load_v16qi): Likewise.
14131         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14132         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14133         (*vsx_le_perm_store_v8hi): Likewise.
14134         (*vsx_le_perm_store_v16qi): Likewise.
14135         (eight unnamed splitters): Likewise.
14136
14137 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
14138
14139         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14140         * config/rs6000/emmintrin.h: Likewise.
14141         * config/rs6000/mmintrin.h: Likewise.
14142         * config/rs6000/xmmintrin.h: Likewise.
14143
14144 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
14145
14146         PR c++/43486
14147         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14148         "public_flag".
14149         * tree.c (tree_nop_conversion): Return true for location wrapper
14150         nodes.
14151         (maybe_wrap_with_location): New function.
14152         (selftest::check_strip_nops): New function.
14153         (selftest::test_location_wrappers): New function.
14154         (selftest::tree_c_tests): Call it.
14155         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14156         (maybe_wrap_with_location): New decl.
14157         (EXPR_LOCATION_WRAPPER_P): New macro.
14158         (location_wrapper_p): New inline function.
14159         (tree_strip_any_location_wrapper): New inline function.
14160
14161 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
14162
14163         PR target/83735
14164         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14165         stack_realign_offset for the largest alignment of stack slot
14166         actually used.
14167         (ix86_find_max_used_stack_alignment): New function.
14168         (ix86_finalize_stack_frame_flags): Use it.  Set
14169         max_used_stack_alignment if we don't realign stack.
14170         * config/i386/i386.h (machine_function): Add
14171         max_used_stack_alignment.
14172
14173 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
14174
14175         * config/arm/arm.opt (-mbranch-cost): New option.
14176         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14177         account.
14178
14179 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
14180
14181         PR target/83629
14182         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14183         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14184
14185 2018-01-10  Richard Biener  <rguenther@suse.de>
14186
14187         PR debug/83765
14188         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14189         early out so it also covers the case where we have a non-NULL
14190         origin.
14191
14192 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
14193
14194         PR tree-optimization/83753
14195         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14196         for non-strided grouped accesses if the number of elements is 1.
14197
14198 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
14199
14200         PR target/81616
14201         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14202         * i386.h (TARGET_USE_GATHER): Define.
14203         * x86-tune.def (X86_TUNE_USE_GATHER): New.
14204
14205 2018-01-10  Martin Liska  <mliska@suse.cz>
14206
14207         PR bootstrap/82831
14208         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14209         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14210         partitioning.
14211         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14212         CLEANUP_NO_PARTITIONING is not set.
14213
14214 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
14215
14216         * doc/rtl.texi: Remove documentation of (const ...) wrappers
14217         for vectors, as a partial revert of r254296.
14218         * rtl.h (const_vec_p): Delete.
14219         (const_vec_duplicate_p): Don't test for vector CONSTs.
14220         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14221         * expmed.c (make_tree): Likewise.
14222
14223         Revert:
14224         * common.md (E, F): Use CONSTANT_P instead of checking for
14225         CONST_VECTOR.
14226         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14227         checking for CONST_VECTOR.
14228
14229 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
14230
14231         PR middle-end/83575
14232         * predict.c (force_edge_cold): Handle in more sane way edges
14233         with no prediction.
14234
14235 2018-01-09  Carl Love  <cel@us.ibm.com>
14236
14237         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14238         V4SI, V4SF types.
14239         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14240         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14241         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14242         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
14243         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14244         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
14245         * config/rs6000/rs6000-protos.h: Add extern defition for
14246         rs6000_generate_float2_double_code.
14247         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14248         function.
14249         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14250         (float2_v2df): Add define_expand.
14251
14252 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
14253
14254         PR target/83628
14255         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14256         op_mode in the force_to_mode call.
14257
14258 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
14259
14260         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14261         instead of checking each element individually.
14262         (aarch64_evpc_uzp): Likewise.
14263         (aarch64_evpc_zip): Likewise.
14264         (aarch64_evpc_ext): Likewise.
14265         (aarch64_evpc_rev): Likewise.
14266         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14267         instead of checking each element individually.  Return true without
14268         generating rtl if
14269         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14270         whether all selected elements come from the same input, instead of
14271         checking each element individually.  Remove calls to gen_rtx_REG,
14272         start_sequence and end_sequence and instead assert that no rtl is
14273         generated.
14274
14275 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
14276
14277         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14278         order of HIGH and CONST checks.
14279
14280 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
14281
14282         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14283         if the destination isn't an SSA_NAME.
14284
14285 2018-01-09  Richard Biener  <rguenther@suse.de>
14286
14287         PR tree-optimization/83668
14288         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14289         move prologue...
14290         (canonicalize_loop_form): ... here, renamed from ...
14291         (canonicalize_loop_closed_ssa_form): ... this and amended to
14292         swap successor edges for loop exit blocks to make us use
14293         the RPO order we need for initial schedule generation.
14294
14295 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
14296
14297         PR tree-optimization/64811
14298         * match.pd: When optimizing comparisons with Inf, avoid
14299         introducing or losing exceptions from comparisons with NaN.
14300
14301 2018-01-09  Martin Liska  <mliska@suse.cz>
14302
14303         PR sanitizer/82517
14304         * asan.c (shadow_mem_size): Add gcc_assert.
14305
14306 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
14307
14308         Don't save registers in main().
14309
14310         PR target/83738
14311         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14312         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14313         * config/avr/avr.c (avr_set_current_function): Don't error if
14314         naked, OS_task or OS_main are specified at the same time.
14315         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14316         OS_main.
14317         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14318         attribute.
14319         * common/config/avr/avr-common.c (avr_option_optimization_table):
14320         Switch on -mmain-is-OS_task for optimizing compilations.
14321
14322 2018-01-09  Richard Biener  <rguenther@suse.de>
14323
14324         PR tree-optimization/83572
14325         * graphite.c: Include cfganal.h.
14326         (graphite_transform_loops): Connect infinite loops to exit
14327         and remove fake edges at the end.
14328
14329 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
14330
14331         * ipa-inline.c (edge_badness): Revert accidental checkin.
14332
14333 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
14334
14335         PR ipa/80763
14336         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14337         symbols; not inline clones.
14338
14339 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
14340
14341         PR target/83507
14342         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14343         hard registers.  Formatting fixes.
14344
14345         PR preprocessor/83722
14346         * gcc.c (try_generate_repro): Pass
14347         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14348         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14349         do_report_bug.
14350
14351 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
14352             Kito Cheng  <kito.cheng@gmail.com>
14353
14354         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14355         (riscv_leaf_function_p): Delete.
14356         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14357
14358 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14359
14360         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14361         function.
14362         (do_ifelse): New function.
14363         (do_isel): New function.
14364         (do_sub3): New function.
14365         (do_add3): New function.
14366         (do_load_mask_compare): New function.
14367         (do_overlap_load_compare): New function.
14368         (expand_compare_loop): New function.
14369         (expand_block_compare): Call expand_compare_loop() when appropriate.
14370         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14371         option description.
14372         (-mblock-compare-inline-loop-limit): New option.
14373
14374 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14375
14376         PR target/83677
14377         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14378         Reverse order of second and third operands in first alternative.
14379         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14380         of first and second elements in UNSPEC_VPERMR vector.
14381         (altivec_expand_vec_perm_le): Likewise.
14382
14383 2018-01-08  Jeff Law  <law@redhat.com>
14384
14385         PR rtl-optimizatin/81308
14386         * tree-switch-conversion.c (cfg_altered): New file scoped static.
14387         (process_switch): If group_case_labels makes a change, then set
14388         cfg_altered.
14389         (pass_convert_switch::execute): If a switch is converted, then
14390         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
14391
14392         PR rtl-optimization/81308
14393         * recog.c (split_all_insns): Conditionally cleanup the CFG after
14394         splitting insns.
14395
14396 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
14397
14398         PR target/83663 - Revert r255946
14399         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14400         generation for cases where splatting a value is not useful.
14401         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14402         across a vec_duplicate and a paradoxical subreg forming a vector
14403         mode to a vec_concat.
14404
14405 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14406
14407         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14408         -march=armv8.3-a variants.
14409         * config/arm/t-multilib: Likewise.
14410         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
14411
14412 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14413
14414         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14415         to generate rtl.
14416         (cceq_ior_compare_complement): Give it a name so I can use it, and
14417         change boolean_or_operator predicate to boolean_operator so it can
14418         be used to generate a crand.
14419         (eqne): New code iterator.
14420         (bd/bd_neg): New code_attrs.
14421         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14422         a single define_insn.
14423         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14424         decrement (bdnzt/bdnzf/bdzt/bdzf).
14425         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14426         with the new names of the branch decrement patterns, and added the
14427         names of the branch decrement conditional patterns.
14428
14429 2018-01-08  Richard Biener  <rguenther@suse.de>
14430
14431         PR tree-optimization/83563
14432         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14433         cache.
14434
14435 2018-01-08  Richard Biener  <rguenther@suse.de>
14436
14437         PR middle-end/83713
14438         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14439
14440 2018-01-08  Richard Biener  <rguenther@suse.de>
14441
14442         PR tree-optimization/83685
14443         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14444         references to abnormals.
14445
14446 2018-01-08  Richard Biener  <rguenther@suse.de>
14447
14448         PR lto/83719
14449         * dwarf2out.c (output_indirect_strings): Handle empty
14450         skeleton_debug_str_hash.
14451         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14452
14453 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
14454
14455         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14456         (emit_store_direct): Likewise.
14457         (arc_trampoline_adjust_address): Likewise.
14458         (arc_asm_trampoline_template): New function.
14459         (arc_initialize_trampoline): Use asm_trampoline_template.
14460         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14461         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14462         * config/arc/arc.md (flush_icache): Delete pattern.
14463
14464 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
14465
14466         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14467         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14468         munaligned-access.
14469
14470 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14471
14472         PR target/83681
14473         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14474         by not USED_FOR_TARGET.
14475         (make_pass_resolve_sw_modes): Likewise.
14476
14477 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14478
14479         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14480         USED_FOR_TARGET.
14481
14482 2018-01-08  Richard Biener  <rguenther@suse.de>
14483
14484         PR middle-end/83580
14485         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14486
14487 2018-01-08  Richard Biener  <rguenther@suse.de>
14488
14489         PR middle-end/83517
14490         * match.pd ((t * 2) / 2) -> t): Add missing :c.
14491
14492 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
14493
14494         PR middle-end/81897
14495         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14496         basic blocks with a small number of successors.
14497         (convert_control_dep_chain_into_preds): Improve handling of
14498         forwarder blocks.
14499         (dump_predicates): Split apart into...
14500         (dump_pred_chain): ...here...
14501         (dump_pred_info): ...and here.
14502         (can_one_predicate_be_invalidated_p): Add debugging printfs.
14503         (can_chain_union_be_invalidated_p): Improve check for invalidation
14504         of paths.
14505         (uninit_uses_cannot_happen): Avoid unnecessary if
14506         convert_control_dep_chain_into_preds yielded nothing.
14507
14508 2018-01-06  Martin Sebor  <msebor@redhat.com>
14509
14510         PR tree-optimization/83640
14511         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14512         subtracting negative offset from size.
14513         (builtin_access::overlap): Adjust offset bounds of the access to fall
14514         within the size of the object if possible.
14515
14516 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
14517
14518         PR rtl-optimization/83699
14519         * expmed.c (extract_bit_field_1): Restrict the vector usage of
14520         extract_bit_field_as_subreg to cases in which the extracted
14521         value is also a vector.
14522
14523         * lra-constraints.c (process_alt_operands): Test for the equivalence
14524         substitutions when detecting a possible reload cycle.
14525
14526 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
14527
14528         PR debug/83480
14529         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14530         by default if flag_selective_schedling{,2}.  Formatting fixes.
14531
14532         PR rtl-optimization/83682
14533         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14534         if it has non-VECTOR_MODE element mode.
14535         (vec_duplicate_p): Likewise.
14536
14537         PR middle-end/83694
14538         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14539         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14540
14541 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
14542
14543         PR target/83604
14544         * config/i386/i386-builtin.def
14545         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14546         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14547         Require also OPTION_MASK_ISA_AVX512F in addition to
14548         OPTION_MASK_ISA_GFNI.
14549         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14550         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14551         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14552         to OPTION_MASK_ISA_GFNI.
14553         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14554         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14555         OPTION_MASK_ISA_AVX512BW.
14556         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14557         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14558         addition to OPTION_MASK_ISA_GFNI.
14559         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14560         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14561         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14562         to OPTION_MASK_ISA_GFNI.
14563         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14564         a requirement for all ISAs rather than any of them with a few
14565         exceptions.
14566         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14567         processing.
14568         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14569         bitmasks to be enabled with 3 exceptions, instead of requiring any
14570         enabled ISA with lots of exceptions.
14571         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14572         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14573         Change avx512bw in isa attribute to avx512f.
14574         * config/i386/sgxintrin.h: Add license boilerplate.
14575         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
14576         to __AVX512F__ and __AVX512VL to __AVX512VL__.
14577         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14578         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14579         defined.
14580         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14581         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14582         temporarily sse2 rather than sse if not enabled already.
14583
14584         PR target/83604
14585         * config/i386/sse.md (VI248_VLBW): Rename to ...
14586         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
14587         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14588         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14589         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14590         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14591         mode iterator instead of VI248_VLBW.
14592
14593 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
14594
14595         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14596         (record_modified): Skip clobbers; add debug output.
14597         (param_change_prob): Use sreal frequencies.
14598
14599 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
14600
14601         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14602         punt for user-aligned variables.
14603
14604 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
14605
14606         * tree-chrec.c (chrec_contains_symbols): Return true for
14607         POLY_INT_CST.
14608
14609 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
14610
14611         PR target/82439
14612         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14613         of (x|y) == x for BICS pattern.
14614
14615 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
14616
14617         PR tree-optimization/83605
14618         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14619         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14620         can throw.
14621
14622 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14623
14624         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14625         * config/epiphany/rtems.h: New file.
14626
14627 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
14628             Uros Bizjak  <ubizjak@gmail.com>
14629
14630         PR target/83554
14631         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14632         QIreg_operand instead of register_operand predicate.
14633         * config/i386/i386.c (ix86_rop_should_change_byte_p,
14634         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14635         comments instead of -fmitigate[-_]rop.
14636
14637 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14638
14639         PR bootstrap/81926
14640         * cgraphunit.c (symbol_table::compile): Switch to text_section
14641         before calling assembly_start debug hook.
14642         * run-rtl-passes.c (run_rtl_passes): Likewise.
14643         Include output.h.
14644
14645 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14646
14647         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14648         range_int_cst_p rather than !symbolic_range_p before calling
14649         extract_range_from_multiplicative_op_1.
14650
14651 2018-01-04  Jeff Law  <law@redhat.com>
14652
14653         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14654         redundant test in assertion.
14655
14656 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14657
14658         * doc/rtl.texi: Document machine_mode wrapper classes.
14659
14660 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14661
14662         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14663         using tree_to_uhwi.
14664
14665 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14666
14667         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14668         the VEC_PERM_EXPR fold to fail.
14669
14670 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
14671
14672         PR debug/83585
14673         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14674         to switched_sections.
14675
14676 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14677
14678         PR target/83680
14679         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14680         test for d.testing.
14681
14682 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
14683
14684         PR target/83387
14685         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14686         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14687
14688 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
14689
14690         PR debug/83666
14691         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14692         is BLKmode and bitpos not zero or mode change is needed.
14693
14694 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14695
14696         PR target/83675
14697         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14698         TARGET_VIS2.
14699
14700 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
14701
14702         PR target/83628
14703         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14704         instead of MULT rtx.  Update all corresponding splitters.
14705         (*saddl_se): Ditto.
14706         (*ssub<modesuffix>): Ditto.
14707         (*ssubl_se): Ditto.
14708         (*cmp_sadd_di): Update split patterns.
14709         (*cmp_sadd_si): Ditto.
14710         (*cmp_sadd_sidi): Ditto.
14711         (*cmp_ssub_di): Ditto.
14712         (*cmp_ssub_si): Ditto.
14713         (*cmp_ssub_sidi): Ditto.
14714         * config/alpha/predicates.md (const23_operand): New predicate.
14715         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14716         Look for ASHIFT, not MULT inner operand.
14717         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14718
14719 2018-01-04  Martin Liska  <mliska@suse.cz>
14720
14721         PR gcov-profile/83669
14722         * gcov.c (output_intermediate_file): Add version to intermediate
14723         gcov file.
14724         * doc/gcov.texi: Document new field 'version' in intermediate
14725         file format. Fix location of '-k' option of gcov command.
14726
14727 2018-01-04  Martin Liska  <mliska@suse.cz>
14728
14729         PR ipa/82352
14730         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14731
14732 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
14733
14734         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14735
14736 2018-01-03  Martin Sebor  <msebor@redhat.com>
14737
14738         PR tree-optimization/83655
14739         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14740         checking calls with invalid arguments.
14741
14742 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14743
14744         * tree-vect-stmts.c (vect_get_store_rhs): New function.
14745         (vectorizable_mask_load_store): Delete.
14746         (vectorizable_call): Return false for masked loads and stores.
14747         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
14748         instead of gimple_assign_rhs1.
14749         (vectorizable_load): Handle IFN_MASK_LOAD.
14750         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14751
14752 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14753
14754         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14755         split out from..,
14756         (vectorizable_mask_load_store): ...here.
14757         (vectorizable_load): ...and here.
14758
14759 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14760
14761         * tree-vect-stmts.c (vect_build_all_ones_mask)
14762         (vect_build_zero_merge_argument): New functions, split out from...
14763         (vectorizable_load): ...here.
14764
14765 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14766
14767         * tree-vect-stmts.c (vect_check_store_rhs): New function,
14768         split out from...
14769         (vectorizable_mask_load_store): ...here.
14770         (vectorizable_store): ...and here.
14771
14772 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14773
14774         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14775         split out from...
14776         (vectorizable_mask_load_store): ...here.
14777
14778 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14779
14780         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14781         (vect_model_store_cost): Take a vec_load_store_type instead of a
14782         vect_def_type.
14783         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14784         (vect_model_store_cost): Take a vec_load_store_type instead of a
14785         vect_def_type.
14786         (vectorizable_mask_load_store): Update accordingly.
14787         (vectorizable_store): Likewise.
14788         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14789
14790 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14791
14792         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14793         IFN_MASK_LOAD calls here rather than...
14794         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14795
14796 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14797             Alan Hayward  <alan.hayward@arm.com>
14798             David Sherwood  <david.sherwood@arm.com>
14799
14800         * expmed.c (extract_bit_field_1): For vector extracts,
14801         fall back to extract_bit_field_as_subreg if vec_extract
14802         isn't available.
14803
14804 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14805             Alan Hayward  <alan.hayward@arm.com>
14806             David Sherwood  <david.sherwood@arm.com>
14807
14808         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14809         they are variable or constant sized.
14810         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14811         slots for constant-sized data.
14812
14813 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14814             Alan Hayward  <alan.hayward@arm.com>
14815             David Sherwood  <david.sherwood@arm.com>
14816
14817         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14818         handling COND_EXPRs with boolean comparisons, try to find a better
14819         basis for the mask type than the boolean itself.
14820
14821 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14822
14823         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14824         is calculated and how it can be overridden.
14825         * genmodes.c (max_bitsize_mode_any_mode): New variable.
14826         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14827         if defined.
14828         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14829         if nonzero.
14830
14831 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14832             Alan Hayward  <alan.hayward@arm.com>
14833             David Sherwood  <david.sherwood@arm.com>
14834
14835         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14836         Remove the mode argument.
14837         (aarch64_simd_valid_immediate): Remove the mode and inverse
14838         arguments.
14839         * config/aarch64/iterators.md (bitsize): New iterator.
14840         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14841         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14842         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14843         aarch64_simd_valid_immediate.
14844         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14845         (aarch64_reg_or_bic_imm): Likewise.
14846         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14847         with an insn_type enum and msl with a modifier_type enum.
14848         Replace element_width with a scalar_mode.  Change the shift
14849         to unsigned int.  Add constructors for scalar_float_mode and
14850         scalar_int_mode elements.
14851         (aarch64_vect_float_const_representable_p): Delete.
14852         (aarch64_can_const_movi_rtx_p)
14853         (aarch64_simd_scalar_immediate_valid_for_move)
14854         (aarch64_simd_make_constant): Update call to
14855         aarch64_simd_valid_immediate.
14856         (aarch64_advsimd_valid_immediate_hs): New function.
14857         (aarch64_advsimd_valid_immediate): Likewise.
14858         (aarch64_simd_valid_immediate): Remove mode and inverse
14859         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
14860         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14861         and aarch64_float_const_representable_p on the result.
14862         (aarch64_output_simd_mov_immediate): Remove mode argument.
14863         Update call to aarch64_simd_valid_immediate and use of
14864         simd_immediate_info.
14865         (aarch64_output_scalar_simd_mov_immediate): Update call
14866         accordingly.
14867
14868 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14869             Alan Hayward  <alan.hayward@arm.com>
14870             David Sherwood  <david.sherwood@arm.com>
14871
14872         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14873         (mode_nunits): Likewise CONST_MODE_NUNITS.
14874         * machmode.def (ADJUST_NUNITS): Document.
14875         * genmodes.c (mode_data::need_nunits_adj): New field.
14876         (blank_mode): Update accordingly.
14877         (adj_nunits): New variable.
14878         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14879         parameter.
14880         (emit_mode_size_inline): Set need_bytesize_adj for all modes
14881         listed in adj_nunits.
14882         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14883         listed in adj_nunits.  Don't emit case statements for such modes.
14884         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14885         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
14886         nothing if adj_nunits is nonnull.
14887         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14888         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14889         (emit_mode_fbit): Update use of print_maybe_const_decl.
14890         (emit_move_size): Likewise.  Treat the array as non-const
14891         if adj_nunits.
14892         (emit_mode_adjustments): Handle adj_nunits.
14893
14894 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14895
14896         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14897         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14898         (VECTOR_MODES): Use it.
14899         (make_vector_modes): Take the prefix as an argument.
14900
14901 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14902             Alan Hayward  <alan.hayward@arm.com>
14903             David Sherwood  <david.sherwood@arm.com>
14904
14905         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14906         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14907         for MODE_VECTOR_BOOL.
14908         * machmode.def (VECTOR_BOOL_MODE): Document.
14909         * genmodes.c (VECTOR_BOOL_MODE): New macro.
14910         (make_vector_bool_mode): New function.
14911         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14912         MODE_VECTOR_BOOL.
14913         * lto-streamer-in.c (lto_input_mode_table): Likewise.
14914         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14915         Likewise.
14916         * stor-layout.c (int_mode_for_mode): Likewise.
14917         * tree.c (build_vector_type_for_mode): Likewise.
14918         * varasm.c (output_constant_pool_2): Likewise.
14919         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14920         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
14921         for MODE_VECTOR_BOOL.
14922         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14923         of mode class checks.
14924         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14925         instead of a list of mode class checks.
14926         (expand_vector_scalar_condition): Likewise.
14927         (type_for_widest_vector_mode): Handle BImode as an inner mode.
14928
14929 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14930             Alan Hayward  <alan.hayward@arm.com>
14931             David Sherwood  <david.sherwood@arm.com>
14932
14933         * machmode.h (mode_size): Change from unsigned short to
14934         poly_uint16_pod.
14935         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14936         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14937         or if measurement_type is not polynomial.
14938         (fixed_size_mode::includes_p): Check for constant-sized modes.
14939         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14940         return a poly_uint16 rather than an unsigned short.
14941         (emit_mode_size): Change the type of mode_size from unsigned short
14942         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
14943         (emit_mode_adjustments): Cope with polynomial vector sizes.
14944         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14945         for GET_MODE_SIZE.
14946         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14947         for GET_MODE_SIZE.
14948         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14949         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14950         * caller-save.c (setup_save_areas): Likewise.
14951         (replace_reg_with_saved_mem): Likewise.
14952         * calls.c (emit_library_call_value_1): Likewise.
14953         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14954         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14955         (gen_lowpart_for_combine): Likewise.
14956         * convert.c (convert_to_integer_1): Likewise.
14957         * cse.c (equiv_constant, cse_insn): Likewise.
14958         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14959         (cselib_subst_to_values): Likewise.
14960         * dce.c (word_dce_process_block): Likewise.
14961         * df-problems.c (df_word_lr_mark_ref): Likewise.
14962         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14963         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14964         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14965         (rtl_for_decl_location): Likewise.
14966         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14967         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14968         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14969         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14970         (expand_expr_real_1): Likewise.
14971         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14972         (pad_below): Likewise.
14973         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14974         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14975         * ira.c (get_subreg_tracking_sizes): Likewise.
14976         * ira-build.c (ira_create_allocno_objects): Likewise.
14977         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14978         (ira_sort_regnos_for_alter_reg): Likewise.
14979         * ira-costs.c (record_operand_costs): Likewise.
14980         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14981         (resolve_simple_move): Likewise.
14982         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14983         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14984         (lra_constraints): Likewise.
14985         (CONST_POOL_OK_P): Reject variable-sized modes.
14986         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14987         (add_pseudo_to_slot, lra_spill): Likewise.
14988         * omp-low.c (omp_clause_aligned_alignment): Likewise.
14989         * optabs-query.c (get_best_extraction_insn): Likewise.
14990         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14991         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14992         (expand_mult_highpart, valid_multiword_target_p): Likewise.
14993         * recog.c (offsettable_address_addr_space_p): Likewise.
14994         * regcprop.c (maybe_mode_change): Likewise.
14995         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14996         * regrename.c (build_def_use): Likewise.
14997         * regstat.c (dump_reg_info): Likewise.
14998         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14999         (find_reloads, find_reloads_subreg_address): Likewise.
15000         * reload1.c (eliminate_regs_1): Likewise.
15001         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15002         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15003         (simplify_binary_operation_1, simplify_subreg): Likewise.
15004         * targhooks.c (default_function_arg_padding): Likewise.
15005         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15006         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15007         (verify_gimple_assign_ternary): Likewise.
15008         * tree-inline.c (estimate_move_cost): Likewise.
15009         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15010         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15011         (get_address_cost_ainc): Likewise.
15012         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15013         (vect_supportable_dr_alignment): Likewise.
15014         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15015         (vectorizable_reduction): Likewise.
15016         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15017         (vectorizable_operation, vectorizable_load): Likewise.
15018         * tree.c (build_same_sized_truth_vector_type): Likewise.
15019         * valtrack.c (cleanup_auto_inc_dec): Likewise.
15020         * var-tracking.c (emit_note_insn_var_location): Likewise.
15021         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15022         (ADDR_VEC_ALIGN): Likewise.
15023
15024 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15025             Alan Hayward  <alan.hayward@arm.com>
15026             David Sherwood  <david.sherwood@arm.com>
15027
15028         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15029         unsigned short.
15030         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15031         or if measurement_type is polynomial.
15032         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15033         * combine.c (make_extraction): Likewise.
15034         * dse.c (find_shift_sequence): Likewise.
15035         * dwarf2out.c (mem_loc_descriptor): Likewise.
15036         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15037         (extract_bit_field, extract_low_bits): Likewise.
15038         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15039         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15040         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15041         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15042         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15043         * reload.c (find_reloads): Likewise.
15044         * reload1.c (alter_reg): Likewise.
15045         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15046         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15047         * tree-if-conv.c (predicate_mem_writes): Likewise.
15048         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15049         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15050         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15051         * valtrack.c (dead_debug_insert_temp): Likewise.
15052         * varasm.c (mergeable_constant_section): Likewise.
15053         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15054
15055 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15056             Alan Hayward  <alan.hayward@arm.com>
15057             David Sherwood  <david.sherwood@arm.com>
15058
15059         * expr.c (expand_assignment): Cope with polynomial mode sizes
15060         when assigning to a CONCAT.
15061
15062 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15063             Alan Hayward  <alan.hayward@arm.com>
15064             David Sherwood  <david.sherwood@arm.com>
15065
15066         * machmode.h (mode_precision): Change from unsigned short to
15067         poly_uint16_pod.
15068         (mode_to_precision): Return a poly_uint16 rather than an unsigned
15069         short.
15070         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15071         or if measurement_type is not polynomial.
15072         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
15073         in which the mode is already known to be a scalar_int_mode.
15074         * genmodes.c (emit_mode_precision): Change the type of mode_precision
15075         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
15076         initializer.
15077         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15078         for GET_MODE_PRECISION.
15079         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15080         for GET_MODE_PRECISION.
15081         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15082         as polynomial.
15083         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15084         (expand_field_assignment, make_extraction): Likewise.
15085         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15086         (get_last_value): Likewise.
15087         * convert.c (convert_to_integer_1): Likewise.
15088         * cse.c (cse_insn): Likewise.
15089         * expr.c (expand_expr_real_1): Likewise.
15090         * lra-constraints.c (simplify_operand_subreg): Likewise.
15091         * optabs-query.c (can_atomic_load_p): Likewise.
15092         * optabs.c (expand_atomic_load): Likewise.
15093         (expand_atomic_store): Likewise.
15094         * ree.c (combine_reaching_defs): Likewise.
15095         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15096         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15097         * tree.h (type_has_mode_precision_p): Likewise.
15098         * ubsan.c (instrument_si_overflow): Likewise.
15099
15100 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15101             Alan Hayward  <alan.hayward@arm.com>
15102             David Sherwood  <david.sherwood@arm.com>
15103
15104         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15105         polynomial numbers of units.
15106         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15107         (valid_vector_subparts_p): New function.
15108         (build_vector_type): Remove temporary shim and take the number
15109         of units as a poly_uint64 rather than an int.
15110         (build_opaque_vector_type): Take the number of units as a
15111         poly_uint64 rather than an int.
15112         * tree.c (build_vector_from_ctor): Handle polynomial
15113         TYPE_VECTOR_SUBPARTS.
15114         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15115         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15116         (build_vector_from_val): If the number of units is variable,
15117         use build_vec_duplicate_cst for constant operands and
15118         VEC_DUPLICATE_EXPR otherwise.
15119         (make_vector_type): Remove temporary is_constant ().
15120         (build_vector_type, build_opaque_vector_type): Take the number of
15121         units as a poly_uint64 rather than an int.
15122         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15123         VECTOR_CST_NELTS.
15124         * cfgexpand.c (expand_debug_expr): Likewise.
15125         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15126         (store_constructor, expand_expr_real_1): Likewise.
15127         (const_scalar_mask_from_tree): Likewise.
15128         * fold-const-call.c (fold_const_reduction): Likewise.
15129         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15130         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15131         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15132         (fold_relational_const): Likewise.
15133         (native_interpret_vector): Likewise.  Change the size from an
15134         int to an unsigned int.
15135         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15136         TYPE_VECTOR_SUBPARTS.
15137         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15138         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15139         duplicating a non-constant operand into a variable-length vector.
15140         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15141         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15142         * ipa-icf.c (sem_variable::equals): Likewise.
15143         * match.pd: Likewise.
15144         * omp-simd-clone.c (simd_clone_subparts): Likewise.
15145         * print-tree.c (print_node): Likewise.
15146         * stor-layout.c (layout_type): Likewise.
15147         * targhooks.c (default_builtin_vectorization_cost): Likewise.
15148         * tree-cfg.c (verify_gimple_comparison): Likewise.
15149         (verify_gimple_assign_binary): Likewise.
15150         (verify_gimple_assign_ternary): Likewise.
15151         (verify_gimple_assign_single): Likewise.
15152         * tree-pretty-print.c (dump_generic_node): Likewise.
15153         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15154         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15155         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15156         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15157         (vect_shift_permute_load_chain): Likewise.
15158         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15159         (expand_vector_condition, optimize_vector_constructor): Likewise.
15160         (lower_vec_perm, get_compute_type): Likewise.
15161         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15162         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15163         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15164         (vect_recog_mask_conversion_pattern): Likewise.
15165         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15166         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15167         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15168         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15169         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15170         (vectorizable_shift, vectorizable_operation, vectorizable_store)
15171         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15172         (supportable_widening_operation): Likewise.
15173         (supportable_narrowing_operation): Likewise.
15174         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15175         Likewise.
15176         * varasm.c (output_constant): Likewise.
15177
15178 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15179             Alan Hayward  <alan.hayward@arm.com>
15180             David Sherwood  <david.sherwood@arm.com>
15181
15182         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15183         so that both the length == 3 and length != 3 cases set up their
15184         own permute vectors.  Add comments explaining why we know the
15185         number of elements is constant.
15186         (vect_permute_load_chain): Likewise.
15187
15188 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15189             Alan Hayward  <alan.hayward@arm.com>
15190             David Sherwood  <david.sherwood@arm.com>
15191
15192         * machmode.h (mode_nunits): Change from unsigned char to
15193         poly_uint16_pod.
15194         (ONLY_FIXED_SIZE_MODES): New macro.
15195         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15196         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15197         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15198         New typedefs.
15199         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15200         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15201         or if measurement_type is not polynomial.
15202         * genmodes.c (ZERO_COEFFS): New macro.
15203         (emit_mode_nunits_inline): Make mode_nunits_inline return a
15204         poly_uint16.
15205         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15206         Use ZERO_COEFFS when emitting initializers.
15207         * data-streamer.h (bp_pack_poly_value): New function.
15208         (bp_unpack_poly_value): Likewise.
15209         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15210         for GET_MODE_NUNITS.
15211         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15212         for GET_MODE_NUNITS.
15213         * tree.c (make_vector_type): Remove temporary shim and make
15214         the real function take the number of units as a poly_uint64
15215         rather than an int.
15216         (build_vector_type_for_mode): Handle polynomial nunits.
15217         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15218         * emit-rtl.c (const_vec_series_p_1): Likewise.
15219         (gen_rtx_CONST_VECTOR): Likewise.
15220         * fold-const.c (test_vec_duplicate_folding): Likewise.
15221         * genrecog.c (validate_pattern): Likewise.
15222         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15223         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15224         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15225         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15226         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15227         * rtlanal.c (subreg_get_info): Likewise.
15228         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15229         (vect_grouped_load_supported): Likewise.
15230         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15231         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15232         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15233         (simplify_const_unary_operation, simplify_binary_operation_1)
15234         (simplify_const_binary_operation, simplify_ternary_operation)
15235         (test_vector_ops_duplicate, test_vector_ops): Likewise.
15236         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15237         instead of CONST_VECTOR_NUNITS.
15238         * varasm.c (output_constant_pool_2): Likewise.
15239         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15240         explicit-encoded elements in the XVEC for variable-length vectors.
15241
15242 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15243
15244         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15245
15246 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15247             Alan Hayward  <alan.hayward@arm.com>
15248             David Sherwood  <david.sherwood@arm.com>
15249
15250         * coretypes.h (fixed_size_mode): Declare.
15251         (fixed_size_mode_pod): New typedef.
15252         * builtins.h (target_builtins::x_apply_args_mode)
15253         (target_builtins::x_apply_result_mode): Change type to
15254         fixed_size_mode_pod.
15255         * builtins.c (apply_args_size, apply_result_size, result_vector)
15256         (expand_builtin_apply_args_1, expand_builtin_apply)
15257         (expand_builtin_return): Update accordingly.
15258
15259 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15260
15261         * cse.c (hash_rtx_cb): Hash only the encoded elements.
15262         * cselib.c (cselib_hash_rtx): Likewise.
15263         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15264         CONST_VECTOR encoding.
15265
15266 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
15267             Jeff Law  <law@redhat.com>
15268
15269         PR target/83641
15270         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15271         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15272         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15273         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15274
15275         PR target/83641
15276         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15277         explicitly probe *sp in a noreturn function if there were any callee
15278         register saves or frame pointer is needed.
15279
15280 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
15281
15282         PR debug/83621
15283         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15284         BLKmode for ternary, binary or unary expressions.
15285
15286         PR debug/83645
15287         * var-tracking.c (delete_vta_debug_insn): New inline function.
15288         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15289         insns from get_insns () to NULL instead of each bb separately.
15290         Use delete_vta_debug_insn.  No longer static.
15291         (vt_debug_insns_local, variable_tracking_main_1): Adjust
15292         delete_vta_debug_insns callers.
15293         * rtl.h (delete_vta_debug_insns): Declare.
15294         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15295         instead of variable_tracking_main.
15296
15297 2018-01-03  Martin Sebor  <msebor@redhat.com>
15298
15299         PR tree-optimization/83603
15300         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15301         arguments past the endof the argument list in functions declared
15302         without a prototype.
15303         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15304         Avoid checking when arguments are null.
15305
15306 2018-01-03  Martin Sebor  <msebor@redhat.com>
15307
15308         PR c/83559
15309         * doc/extend.texi (attribute const): Fix a typo.
15310         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15311         issuing -Wsuggest-attribute for void functions.
15312
15313 2018-01-03  Martin Sebor  <msebor@redhat.com>
15314
15315         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15316         offset_int::from instead of wide_int::to_shwi.
15317         (maybe_diag_overlap): Remove assertion.
15318         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15319         * gimple-ssa-sprintf.c (format_directive): Same.
15320         (parse_directive): Same.
15321         (sprintf_dom_walker::compute_format_length): Same.
15322         (try_substitute_return_value): Same.
15323
15324 2018-01-03  Jeff Law  <law@redhat.com>
15325
15326         PR middle-end/83654
15327         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15328         non-constant residual for zero at runtime and avoid probing in
15329         that case.  Reorganize code for trailing problem to mirror handling
15330         of the residual.
15331
15332 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15333
15334         PR tree-optimization/83501
15335         * tree-ssa-strlen.c (get_string_cst): New.
15336         (handle_char_store): Call get_string_cst.
15337
15338 2018-01-03  Martin Liska  <mliska@suse.cz>
15339
15340         PR tree-optimization/83593
15341         * tree-ssa-strlen.c: Include tree-cfg.h.
15342         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15343         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15344         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15345         to false.
15346         (strlen_dom_walker::before_dom_children): Call
15347         gimple_purge_dead_eh_edges. Dump tranformation with details
15348         dump flags.
15349         (strlen_dom_walker::before_dom_children): Update call by adding
15350         new argument cleanup_eh.
15351         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15352
15353 2018-01-03  Martin Liska  <mliska@suse.cz>
15354
15355         PR ipa/83549
15356         * cif-code.def (VARIADIC_THUNK): New enum value.
15357         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15358         thunks.
15359
15360 2018-01-03  Jan Beulich  <jbeulich@suse.com>
15361
15362         * sse.md (mov<mode>_internal): Tighten condition for when to use
15363         vmovdqu<ssescalarsize> for TI and OI modes.
15364
15365 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
15366
15367         Update copyright years.
15368
15369 2018-01-03  Martin Liska  <mliska@suse.cz>
15370
15371         PR ipa/83594
15372         * ipa-visibility.c (function_and_variable_visibility): Skip
15373         functions with noipa attribure.
15374
15375 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
15376
15377         * gcc.c (process_command): Update copyright notice dates.
15378         * gcov-dump.c (print_version): Ditto.
15379         * gcov.c (print_version): Ditto.
15380         * gcov-tool.c (print_version): Ditto.
15381         * gengtype.c (create_file): Ditto.
15382         * doc/cpp.texi: Bump @copying's copyright year.
15383         * doc/cppinternals.texi: Ditto.
15384         * doc/gcc.texi: Ditto.
15385         * doc/gccint.texi: Ditto.
15386         * doc/gcov.texi: Ditto.
15387         * doc/install.texi: Ditto.
15388         * doc/invoke.texi: Ditto.
15389
15390 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15391
15392         * vector-builder.h (vector_builder::m_full_nelts): Change from
15393         unsigned int to poly_uint64.
15394         (vector_builder::full_nelts): Update prototype accordingly.
15395         (vector_builder::new_vector): Likewise.
15396         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15397         (vector_builder::operator ==): Likewise.
15398         (vector_builder::finalize): Likewise.
15399         * int-vector-builder.h (int_vector_builder::int_vector_builder):
15400         Take the number of elements as a poly_uint64 rather than an
15401         unsigned int.
15402         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15403         from unsigned int to poly_uint64.
15404         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15405         (vec_perm_indices::new_vector): Likewise.
15406         (vec_perm_indices::length): Likewise.
15407         (vec_perm_indices::nelts_per_input): Likewise.
15408         (vec_perm_indices::input_nelts): Likewise.
15409         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15410         number of elements per input as a poly_uint64 rather than an
15411         unsigned int.  Use the original encoding for variable-length
15412         vectors, rather than clamping each individual element.
15413         For the second and subsequent elements in each pattern,
15414         clamp the step and base before clamping their sum.
15415         (vec_perm_indices::series_p): Handle polynomial element counts.
15416         (vec_perm_indices::all_in_range_p): Likewise.
15417         (vec_perm_indices_to_tree): Likewise.
15418         (vec_perm_indices_to_rtx): Likewise.
15419         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15420         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15421         (tree_vector_builder::new_binary_operation): Handle polynomial
15422         element counts.  Return false if we need to know the number
15423         of elements at compile time.
15424         * fold-const.c (fold_vec_perm): Punt if the number of elements
15425         isn't known at compile time.
15426
15427 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15428
15429         * vec-perm-indices.h (vec_perm_builder): Change element type
15430         from HOST_WIDE_INT to poly_int64.
15431         (vec_perm_indices::element_type): Update accordingly.
15432         (vec_perm_indices::clamp): Handle polynomial element_types.
15433         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15434         (vec_perm_indices::all_in_range_p): Likewise.
15435         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15436         than shwi trees.
15437         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15438         polynomial vec_perm_indices element types.
15439         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15440         * fold-const.c (fold_vec_perm): Likewise.
15441         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15442         * tree-vect-generic.c (lower_vec_perm): Likewise.
15443         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15444         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15445         element type to HOST_WIDE_INT.
15446
15447 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15448             Alan Hayward  <alan.hayward@arm.com>
15449             David Sherwood  <david.sherwood@arm.com>
15450
15451         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15452         rather than an int.  Use plus_constant.
15453         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15454         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15455
15456 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15457             Alan Hayward  <alan.hayward@arm.com>
15458             David Sherwood  <david.sherwood@arm.com>
15459
15460         * calls.c (emit_call_1, expand_call): Change struct_value_size from
15461         a HOST_WIDE_INT to a poly_int64.
15462
15463 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15464             Alan Hayward  <alan.hayward@arm.com>
15465             David Sherwood  <david.sherwood@arm.com>
15466
15467         * calls.c (load_register_parameters): Cope with polynomial
15468         mode sizes.  Require a constant size for BLKmode parameters
15469         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
15470         forces a parameter to be padded at the lsb end in order to
15471         fill a complete number of words, require the parameter size
15472         to be ordered wrt UNITS_PER_WORD.
15473
15474 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15475             Alan Hayward  <alan.hayward@arm.com>
15476             David Sherwood  <david.sherwood@arm.com>
15477
15478         * reload1.c (spill_stack_slot_width): Change element type
15479         from unsigned int to poly_uint64_pod.
15480         (alter_reg): Treat mode sizes as polynomial.
15481
15482 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15483             Alan Hayward  <alan.hayward@arm.com>
15484             David Sherwood  <david.sherwood@arm.com>
15485
15486         * reload.c (complex_word_subreg_p): New function.
15487         (reload_inner_reg_of_subreg, push_reload): Use it.
15488
15489 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15490             Alan Hayward  <alan.hayward@arm.com>
15491             David Sherwood  <david.sherwood@arm.com>
15492
15493         * lra-constraints.c (process_alt_operands): Reject matched
15494         operands whose sizes aren't ordered.
15495         (match_reload): Refer to this check here.
15496
15497 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15498             Alan Hayward  <alan.hayward@arm.com>
15499             David Sherwood  <david.sherwood@arm.com>
15500
15501         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15502         that the mode size is in the set {1, 2, 4, 8, 16}.
15503
15504 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15505             Alan Hayward  <alan.hayward@arm.com>
15506             David Sherwood  <david.sherwood@arm.com>
15507
15508         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15509         Use plus_constant instead of gen_rtx_PLUS.
15510
15511 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15512             Alan Hayward  <alan.hayward@arm.com>
15513             David Sherwood  <david.sherwood@arm.com>
15514
15515         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15516         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15517         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15518         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15519         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15520         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15521         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15522         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15523         * config/i386/i386.c (ix86_push_rounding): ...this new function.
15524         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15525         a poly_int64.
15526         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15527         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15528         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15529         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15530         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15531         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15532         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15533         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15534         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15535         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15536         function.
15537         * expr.c (emit_move_resolve_push): Treat the input and result
15538         of PUSH_ROUNDING as a poly_int64.
15539         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15540         (emit_push_insn): Likewise.
15541         * lra-eliminations.c (mark_not_eliminable): Likewise.
15542         * recog.c (push_operand): Likewise.
15543         * reload1.c (elimination_effects): Likewise.
15544         * rtlanal.c (nonzero_bits1): Likewise.
15545         * calls.c (store_one_arg): Likewise.  Require the padding to be
15546         known at compile time.
15547
15548 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15549             Alan Hayward  <alan.hayward@arm.com>
15550             David Sherwood  <david.sherwood@arm.com>
15551
15552         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15553         Use plus_constant instead of gen_rtx_PLUS.
15554
15555 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15556             Alan Hayward  <alan.hayward@arm.com>
15557             David Sherwood  <david.sherwood@arm.com>
15558
15559         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15560         rather than an int.
15561
15562 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15563             Alan Hayward  <alan.hayward@arm.com>
15564             David Sherwood  <david.sherwood@arm.com>
15565
15566         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15567         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15568         via stack temporaries.  Treat the mode size as polynomial too.
15569
15570 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15571             Alan Hayward  <alan.hayward@arm.com>
15572             David Sherwood  <david.sherwood@arm.com>
15573
15574         * expr.c (expand_expr_real_2): When handling conversions involving
15575         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15576         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
15577         as a poly_uint64 too.
15578
15579 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15580             Alan Hayward  <alan.hayward@arm.com>
15581             David Sherwood  <david.sherwood@arm.com>
15582
15583         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15584
15585 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15586             Alan Hayward  <alan.hayward@arm.com>
15587             David Sherwood  <david.sherwood@arm.com>
15588
15589         * combine.c (can_change_dest_mode): Handle polynomial
15590         REGMODE_NATURAL_SIZE.
15591         * expmed.c (store_bit_field_1): Likewise.
15592         * expr.c (store_constructor): Likewise.
15593         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15594         and polynomial REGMODE_NATURAL_SIZE.
15595         (gen_lowpart_common): Likewise.
15596         * reginfo.c (record_subregs_of_mode): Likewise.
15597         * rtlanal.c (read_modify_subreg_p): Likewise.
15598
15599 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15600             Alan Hayward  <alan.hayward@arm.com>
15601             David Sherwood  <david.sherwood@arm.com>
15602
15603         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15604         numbers of elements.
15605
15606 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15607             Alan Hayward  <alan.hayward@arm.com>
15608             David Sherwood  <david.sherwood@arm.com>
15609
15610         * match.pd: Cope with polynomial numbers of vector elements.
15611
15612 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15613             Alan Hayward  <alan.hayward@arm.com>
15614             David Sherwood  <david.sherwood@arm.com>
15615
15616         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15617         in a POINTER_PLUS_EXPR.
15618
15619 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15620             Alan Hayward  <alan.hayward@arm.com>
15621             David Sherwood  <david.sherwood@arm.com>
15622
15623         * omp-simd-clone.c (simd_clone_subparts): New function.
15624         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15625         (ipa_simd_modify_function_body): Likewise.
15626
15627 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15628             Alan Hayward  <alan.hayward@arm.com>
15629             David Sherwood  <david.sherwood@arm.com>
15630
15631         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15632         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15633         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15634         (expand_vector_condition, vector_element): Likewise.
15635         (subparts_gt): New function.
15636         (get_compute_type): Use subparts_gt.
15637         (count_type_subparts): Delete.
15638         (expand_vector_operations_1): Use subparts_gt instead of
15639         count_type_subparts.
15640
15641 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15642             Alan Hayward  <alan.hayward@arm.com>
15643             David Sherwood  <david.sherwood@arm.com>
15644
15645         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15646         (vect_compile_time_alias): ...this new function.  Do the calculation
15647         on poly_ints rather than trees.
15648         (vect_prune_runtime_alias_test_list): Update call accordingly.
15649
15650 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15651             Alan Hayward  <alan.hayward@arm.com>
15652             David Sherwood  <david.sherwood@arm.com>
15653
15654         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15655         numbers of units.
15656         (vect_schedule_slp_instance): Likewise.
15657
15658 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15659             Alan Hayward  <alan.hayward@arm.com>
15660             David Sherwood  <david.sherwood@arm.com>
15661
15662         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15663         constant and extern definitions for variable-length vectors.
15664         (vect_get_constant_vectors): Note that the number of units
15665         is known to be constant.
15666
15667 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15668             Alan Hayward  <alan.hayward@arm.com>
15669             David Sherwood  <david.sherwood@arm.com>
15670
15671         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15672         of units as polynomial.  Choose between WIDE and NARROW based
15673         on multiple_p.
15674
15675 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15676             Alan Hayward  <alan.hayward@arm.com>
15677             David Sherwood  <david.sherwood@arm.com>
15678
15679         * tree-vect-stmts.c (simd_clone_subparts): New function.
15680         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15681
15682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15683             Alan Hayward  <alan.hayward@arm.com>
15684             David Sherwood  <david.sherwood@arm.com>
15685
15686         * tree-vect-stmts.c (vectorizable_call): Treat the number of
15687         vectors as polynomial.  Use build_index_vector for
15688         IFN_GOMP_SIMD_LANE.
15689
15690 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15691             Alan Hayward  <alan.hayward@arm.com>
15692             David Sherwood  <david.sherwood@arm.com>
15693
15694         * tree-vect-stmts.c (get_load_store_type): Treat the number of
15695         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15696         for variable-length vectors.
15697         (vectorizable_mask_load_store): Treat the number of units as
15698         polynomial, asserting that it is constant if the condition has
15699         already been enforced.
15700         (vectorizable_store, vectorizable_load): Likewise.
15701
15702 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15703             Alan Hayward  <alan.hayward@arm.com>
15704             David Sherwood  <david.sherwood@arm.com>
15705
15706         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15707         of units as polynomial.  Punt if we can't tell at compile time
15708         which vector contains the final result.
15709
15710 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15711             Alan Hayward  <alan.hayward@arm.com>
15712             David Sherwood  <david.sherwood@arm.com>
15713
15714         * tree-vect-loop.c (vectorizable_induction): Treat the number
15715         of units as polynomial.  Punt on SLP inductions.  Use an integer
15716         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
15717         cast of such a series for variable-length floating-point
15718         reductions.
15719
15720 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15721             Alan Hayward  <alan.hayward@arm.com>
15722             David Sherwood  <david.sherwood@arm.com>
15723
15724         * tree.h (build_index_vector): Declare.
15725         * tree.c (build_index_vector): New function.
15726         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15727         of units as polynomial, forcibly converting it to a constant if
15728         vectorizable_reduction has already enforced the condition.
15729         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
15730         to create a {1,2,3,...} vector.
15731         (vectorizable_reduction): Treat the number of units as polynomial.
15732         Choose vectype_in based on the largest scalar element size rather
15733         than the smallest number of units.  Enforce the restrictions
15734         relied on above.
15735
15736 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15737             Alan Hayward  <alan.hayward@arm.com>
15738             David Sherwood  <david.sherwood@arm.com>
15739
15740         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15741         number of units as polynomial.
15742
15743 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15744             Alan Hayward  <alan.hayward@arm.com>
15745             David Sherwood  <david.sherwood@arm.com>
15746
15747         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15748         * target.def (autovectorize_vector_sizes): Return the vector sizes
15749         by pointer, using vector_sizes rather than a bitmask.
15750         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15751         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15752         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15753         Likewise.
15754         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15755         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15756         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15757         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15758         * omp-general.c (omp_max_vf): Likewise.
15759         * omp-low.c (omp_clause_aligned_alignment): Likewise.
15760         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15761         * tree-vect-loop.c (vect_analyze_loop): Likewise.
15762         * tree-vect-slp.c (vect_slp_bb): Likewise.
15763         * doc/tm.texi: Regenerate.
15764         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15765         to a poly_uint64.
15766         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15767         the vector size as a poly_uint64 rather than an unsigned int.
15768         (current_vector_size): Change from an unsigned int to a poly_uint64.
15769         (get_vectype_for_scalar_type): Update accordingly.
15770         * tree.h (build_truth_vector_type): Take the size and number of
15771         units as a poly_uint64 rather than an unsigned int.
15772         (build_vector_type): Add a temporary overload that takes
15773         the number of units as a poly_uint64 rather than an unsigned int.
15774         * tree.c (make_vector_type): Likewise.
15775         (build_truth_vector_type): Take the number of units as a poly_uint64
15776         rather than an unsigned int.
15777
15778 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15779             Alan Hayward  <alan.hayward@arm.com>
15780             David Sherwood  <david.sherwood@arm.com>
15781
15782         * target.def (get_mask_mode): Take the number of units and length
15783         as poly_uint64s rather than unsigned ints.
15784         * targhooks.h (default_get_mask_mode): Update accordingly.
15785         * targhooks.c (default_get_mask_mode): Likewise.
15786         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15787         * doc/tm.texi: Regenerate.
15788
15789 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15790             Alan Hayward  <alan.hayward@arm.com>
15791             David Sherwood  <david.sherwood@arm.com>
15792
15793         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15794         * omp-general.c (omp_max_vf): Likewise.
15795         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15796         (expand_omp_simd): Handle polynomial safelen.
15797         * omp-low.c (omplow_simd_context): Add a default constructor.
15798         (omplow_simd_context::max_vf): Change from int to poly_uint64.
15799         (lower_rec_simd_input_clauses): Update accordingly.
15800         (lower_rec_input_clauses): Likewise.
15801
15802 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15803             Alan Hayward  <alan.hayward@arm.com>
15804             David Sherwood  <david.sherwood@arm.com>
15805
15806         * tree-vectorizer.h (vect_nunits_for_cost): New function.
15807         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15808         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15809         (vect_analyze_slp_cost): Likewise.
15810         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15811         (vect_model_load_cost): Likewise.
15812
15813 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15814             Alan Hayward  <alan.hayward@arm.com>
15815             David Sherwood  <david.sherwood@arm.com>
15816
15817         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15818         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15819         from an unsigned int * to a poly_uint64_pod *.
15820         (calculate_unrolling_factor): New function.
15821         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
15822
15823 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15824             Alan Hayward  <alan.hayward@arm.com>
15825             David Sherwood  <david.sherwood@arm.com>
15826
15827         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15828         from an unsigned int to a poly_uint64.
15829         (_loop_vec_info::slp_unrolling_factor): Likewise.
15830         (_loop_vec_info::vectorization_factor): Change from an int
15831         to a poly_uint64.
15832         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15833         (vect_get_num_vectors): New function.
15834         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15835         (vect_get_num_copies): Use vect_get_num_vectors.
15836         (vect_analyze_data_ref_dependences): Change max_vf from an int *
15837         to an unsigned int *.
15838         (vect_analyze_data_refs): Change min_vf from an int * to a
15839         poly_uint64 *.
15840         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15841         than an unsigned HOST_WIDE_INT.
15842         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15843         (vect_analyze_data_ref_dependence): Change max_vf from an int *
15844         to an unsigned int *.
15845         (vect_analyze_data_ref_dependences): Likewise.
15846         (vect_compute_data_ref_alignment): Handle polynomial vf.
15847         (vect_enhance_data_refs_alignment): Likewise.
15848         (vect_prune_runtime_alias_test_list): Likewise.
15849         (vect_shift_permute_load_chain): Likewise.
15850         (vect_supportable_dr_alignment): Likewise.
15851         (dependence_distance_ge_vf): Take the vectorization factor as a
15852         poly_uint64 rather than an unsigned HOST_WIDE_INT.
15853         (vect_analyze_data_refs): Change min_vf from an int * to a
15854         poly_uint64 *.
15855         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15856         vfm1 as a poly_uint64 rather than an int.  Make the same change
15857         for the returned bound_scalar.
15858         (vect_gen_vector_loop_niters): Handle polynomial vf.
15859         (vect_do_peeling): Likewise.  Update call to
15860         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15861         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15862         be constant.
15863         * tree-vect-loop.c (vect_determine_vectorization_factor)
15864         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15865         (vect_get_known_peeling_cost): Likewise.
15866         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15867         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15868         (vect_transform_loop): Likewise.  Use the lowest possible VF when
15869         updating the upper bounds of the loop.
15870         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
15871         rather than an int.
15872         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15873         polynomial unroll factors.
15874         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15875         (vect_make_slp_decision): Likewise.
15876         (vect_supported_load_permutation_p): Likewise, and polynomial
15877         vf too.
15878         (vect_analyze_slp_cost): Handle polynomial vf.
15879         (vect_slp_analyze_node_operations): Likewise.
15880         (vect_slp_analyze_bb_1): Likewise.
15881         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15882         than an unsigned HOST_WIDE_INT.
15883         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15884         (vectorizable_load): Handle polynomial vf.
15885         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15886         a poly_uint64.
15887         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15888
15889 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15890             Alan Hayward  <alan.hayward@arm.com>
15891             David Sherwood  <david.sherwood@arm.com>
15892
15893         * match.pd: Handle bit operations involving three constants
15894         and try to fold one pair.
15895
15896 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15897
15898         * tree-vect-loop-manip.c: Include gimple-fold.h.
15899         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15900         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
15901         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15902         Add a path that uses a step of VF instead of 1, but disable it
15903         for now.
15904         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15905         and niters_no_overflow parameters.  Update calls to
15906         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15907         Create a new SSA name if the latter choses to use a ste other
15908         than zero, and return it via niters_vector_mult_vf_var.
15909         * tree-vect-loop.c (vect_transform_loop): Update calls to
15910         vect_do_peeling, vect_gen_vector_loop_niters and
15911         slpeel_make_loop_iterate_ntimes.
15912         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15913         (vect_gen_vector_loop_niters): Update declarations after above changes.
15914
15915 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
15916
15917         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15918         128-bit round to integer instructions.
15919         (ceil<mode>2): Likewise.
15920         (btrunc<mode>2): Likewise.
15921         (round<mode>2): Likewise.
15922
15923 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
15924
15925         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15926         unaligned VSX load/store on P8/P9.
15927         (expand_block_clear): Allow the use of unaligned VSX
15928         load/store on P8/P9.
15929
15930 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
15931
15932         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15933         New function.
15934         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15935         swap associated with both a load and a store.
15936
15937 2018-01-02  Andrew Waterman  <andrew@sifive.com>
15938
15939         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15940         * config/riscv/riscv.md (clear_cache): Use it.
15941
15942 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
15943
15944         * web.c: Remove out-of-date comment.
15945
15946 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15947
15948         * expr.c (fixup_args_size_notes): Check that any existing
15949         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15950         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15951         (emit_single_push_insn): ...here.
15952
15953 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15954
15955         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15956         (const_vector_encoded_nelts): New function.
15957         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15958         (const_vector_int_elt, const_vector_elt): Declare.
15959         * emit-rtl.c (const_vector_int_elt_1): New function.
15960         (const_vector_elt): Likewise.
15961         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15962         of CONST_VECTOR_ELT.
15963
15964 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15965
15966         * expr.c: Include rtx-vector-builder.h.
15967         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15968         directly on the tree encoding.
15969         (const_vector_from_tree): Likewise.
15970         * optabs.c: Include rtx-vector-builder.h.
15971         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15972         sequence of "u" values.
15973         * vec-perm-indices.c: Include rtx-vector-builder.h.
15974         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15975         directly on the vec_perm_indices encoding.
15976
15977 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15978
15979         * doc/rtl.texi (const_vector): Describe new encoding scheme.
15980         * Makefile.in (OBJS): Add rtx-vector-builder.o.
15981         * rtx-vector-builder.h: New file.
15982         * rtx-vector-builder.c: Likewise.
15983         * rtl.h (rtx_def::u2): Add a const_vector field.
15984         (CONST_VECTOR_NPATTERNS): New macro.
15985         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15986         (CONST_VECTOR_DUPLICATE_P): Likewise.
15987         (CONST_VECTOR_STEPPED_P): Likewise.
15988         (CONST_VECTOR_ENCODED_ELT): Likewise.
15989         (const_vec_duplicate_p): Check for a duplicated vector encoding.
15990         (unwrap_const_vec_duplicate): Likewise.
15991         (const_vec_series_p): Check for a non-duplicated vector encoding.
15992         Say that the function only returns true for integer vectors.
15993         * emit-rtl.c: Include rtx-vector-builder.h.
15994         (gen_const_vec_duplicate_1): Delete.
15995         (gen_const_vector): Call gen_const_vec_duplicate instead of
15996         gen_const_vec_duplicate_1.
15997         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15998         (gen_const_vec_duplicate): Use rtx_vector_builder.
15999         (gen_const_vec_series): Likewise.
16000         (gen_rtx_CONST_VECTOR): Likewise.
16001         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16002         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16003         Build a new vector rather than modifying a CONST_VECTOR in-place.
16004         (handle_special_swappables): Update call accordingly.
16005         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16006         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16007         Build a new vector rather than modifying a CONST_VECTOR in-place.
16008         (handle_special_swappables): Update call accordingly.
16009
16010 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16011
16012         * simplify-rtx.c (simplify_const_binary_operation): Use
16013         CONST_VECTOR_ELT instead of XVECEXP.
16014
16015 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16016
16017         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16018         the selector elements to be different from the data elements
16019         if the selector is a VECTOR_CST.
16020         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16021         ssizetype for the selector.
16022
16023 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16024
16025         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16026         before testing each element individually.
16027         * tree-vect-generic.c (lower_vec_perm): Likewise.
16028
16029 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16030
16031         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16032         * selftest-run-tests.c (selftest::run_tests): Call it.
16033         * vector-builder.h (vector_builder::operator ==): New function.
16034         (vector_builder::operator !=): Likewise.
16035         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16036         (vec_perm_indices::all_from_input_p): New function.
16037         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16038         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16039         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16040         instead of reading the VECTOR_CST directly.  Detect whether both
16041         vector inputs are the same before constructing the vec_perm_indices,
16042         and update the number of inputs argument accordingly.  Use the
16043         utility functions added above.  Only construct sel2 if we need to.
16044
16045 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16046
16047         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16048         the broadcast of the low byte.
16049         (expand_mult_highpart): Use an explicit encoding for the permutes.
16050         * optabs-query.c (can_mult_highpart_p): Likewise.
16051         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16052         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16053         (vectorizable_bswap): Likewise.
16054         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16055         explicit encoding for the power-of-2 permutes.
16056         (vect_permute_store_chain): Likewise.
16057         (vect_grouped_load_supported): Likewise.
16058         (vect_permute_load_chain): Likewise.
16059
16060 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16061
16062         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16063         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16064         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16065         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16066         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16067         (vect_gen_perm_mask_any): Likewise.
16068
16069 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16070
16071         * int-vector-builder.h: New file.
16072         * vec-perm-indices.h: Include int-vector-builder.h.
16073         (vec_perm_indices): Redefine as an int_vector_builder.
16074         (auto_vec_perm_indices): Delete.
16075         (vec_perm_builder): Redefine as a stand-alone class.
16076         (vec_perm_indices::vec_perm_indices): New function.
16077         (vec_perm_indices::clamp): Likewise.
16078         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16079         (vec_perm_indices::new_vector): New function.
16080         (vec_perm_indices::new_expanded_vector): Update for new
16081         vec_perm_indices class.
16082         (vec_perm_indices::rotate_inputs): New function.
16083         (vec_perm_indices::all_in_range_p): Operate directly on the
16084         encoded form, without computing elided elements.
16085         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16086         encoding.  Update for new vec_perm_indices class.
16087         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16088         the given vec_perm_builder.
16089         (expand_vec_perm_var): Update vec_perm_builder constructor.
16090         (expand_mult_highpart): Use vec_perm_builder instead of
16091         auto_vec_perm_indices.
16092         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16093         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
16094         or double series encoding as appropriate.
16095         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16096         vec_perm_indices instead of auto_vec_perm_indices.
16097         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16098         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16099         (vect_permute_store_chain): Likewise.
16100         (vect_grouped_load_supported): Likewise.
16101         (vect_permute_load_chain): Likewise.
16102         (vect_shift_permute_load_chain): Likewise.
16103         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16104         (vect_transform_slp_perm_load): Likewise.
16105         (vect_schedule_slp_instance): Likewise.
16106         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16107         (vectorizable_mask_load_store): Likewise.
16108         (vectorizable_bswap): Likewise.
16109         (vectorizable_store): Likewise.
16110         (vectorizable_load): Likewise.
16111         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16112         vec_perm_indices instead of auto_vec_perm_indices.  Use
16113         tree_to_vec_perm_builder to read the vector from a tree.
16114         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16115         vec_perm_builder instead of a vec_perm_indices.
16116         (have_whole_vector_shift): Use vec_perm_builder and
16117         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
16118         truncation to calc_vec_perm_mask_for_shift.
16119         (vect_create_epilog_for_reduction): Likewise.
16120         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16121         from auto_vec_perm_indices to vec_perm_indices.
16122         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16123         instead of changing individual elements.
16124         (aarch64_vectorize_vec_perm_const): Use new_vector to install
16125         the vector in d.perm.
16126         * config/arm/arm.c (expand_vec_perm_d::perm): Change
16127         from auto_vec_perm_indices to vec_perm_indices.
16128         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16129         instead of changing individual elements.
16130         (arm_vectorize_vec_perm_const): Use new_vector to install
16131         the vector in d.perm.
16132         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16133         Update vec_perm_builder constructor.
16134         (rs6000_expand_interleave): Likewise.
16135         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16136         (rs6000_expand_interleave): Likewise.
16137
16138 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16139
16140         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16141         to qimode could truncate the indices.
16142         * optabs.c (expand_vec_perm_var): Likewise.
16143
16144 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16145
16146         * Makefile.in (OBJS): Add vec-perm-indices.o.
16147         * vec-perm-indices.h: New file.
16148         * vec-perm-indices.c: Likewise.
16149         * target.h (vec_perm_indices): Replace with a forward class
16150         declaration.
16151         (auto_vec_perm_indices): Move to vec-perm-indices.h.
16152         * optabs.h: Include vec-perm-indices.h.
16153         (expand_vec_perm): Delete.
16154         (selector_fits_mode_p, expand_vec_perm_var): Declare.
16155         (expand_vec_perm_const): Declare.
16156         * target.def (vec_perm_const_ok): Replace with...
16157         (vec_perm_const): ...this new hook.
16158         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16159         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16160         * doc/tm.texi: Regenerate.
16161         * optabs.def (vec_perm_const): Delete.
16162         * doc/md.texi (vec_perm_const): Likewise.
16163         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16164         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16165         expand_vec_perm for constant permutation vectors.  Assert that
16166         the mode of variable permutation vectors is the integer equivalent
16167         of the mode that is being permuted.
16168         * optabs-query.h (selector_fits_mode_p): Declare.
16169         * optabs-query.c: Include vec-perm-indices.h.
16170         (selector_fits_mode_p): New function.
16171         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16172         is defined, instead of checking whether the vec_perm_const_optab
16173         exists.  Use targetm.vectorize.vec_perm_const instead of
16174         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
16175         fit in the vector mode before using a variable permute.
16176         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16177         vec_perm_indices instead of an rtx.
16178         (expand_vec_perm): Replace with...
16179         (expand_vec_perm_const): ...this new function.  Take the selector
16180         as a vec_perm_indices rather than an rtx.  Also take the mode of
16181         the selector.  Update call to shift_amt_for_vec_perm_mask.
16182         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16183         Use vec_perm_indices::new_expanded_vector to expand the original
16184         selector into bytes.  Check whether the indices fit in the vector
16185         mode before using a variable permute.
16186         (expand_vec_perm_var): Make global.
16187         (expand_mult_highpart): Use expand_vec_perm_const.
16188         * fold-const.c: Includes vec-perm-indices.h.
16189         * tree-ssa-forwprop.c: Likewise.
16190         * tree-vect-data-refs.c: Likewise.
16191         * tree-vect-generic.c: Likewise.
16192         * tree-vect-loop.c: Likewise.
16193         * tree-vect-slp.c: Likewise.
16194         * tree-vect-stmts.c: Likewise.
16195         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16196         Delete.
16197         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16198         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16199         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16200         (aarch64_vectorize_vec_perm_const): ...this new function.
16201         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16202         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16203         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16204         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16205         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16206         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16207         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16208         into...
16209         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
16210         check for NEON modes.
16211         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16212         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16213         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16214         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16215         into...
16216         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
16217         the old VEC_PERM_CONST conditions.
16218         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16219         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16220         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16221         (ia64_vectorize_vec_perm_const_ok): Merge into...
16222         (ia64_vectorize_vec_perm_const): ...this new function.
16223         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16224         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16225         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16226         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16227         * config/mips/mips.c (mips_expand_vec_perm_const)
16228         (mips_vectorize_vec_perm_const_ok): Merge into...
16229         (mips_vectorize_vec_perm_const): ...this new function.
16230         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16231         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16232         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16233         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16234         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16235         (rs6000_expand_vec_perm_const): Delete.
16236         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16237         Delete.
16238         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16239         (altivec_expand_vec_perm_const_le): Take each operand individually.
16240         Operate on constant selectors rather than rtxes.
16241         (altivec_expand_vec_perm_const): Likewise.  Update call to
16242         altivec_expand_vec_perm_const_le.
16243         (rs6000_expand_vec_perm_const): Delete.
16244         (rs6000_vectorize_vec_perm_const_ok): Delete.
16245         (rs6000_vectorize_vec_perm_const): New function.
16246         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16247         an element count and rtx array.
16248         (rs6000_expand_extract_even): Update call accordingly.
16249         (rs6000_expand_interleave): Likewise.
16250         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16251         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16252         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16253         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16254         (rs6000_expand_vec_perm_const): Delete.
16255         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16256         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16257         (altivec_expand_vec_perm_const_le): Take each operand individually.
16258         Operate on constant selectors rather than rtxes.
16259         (altivec_expand_vec_perm_const): Likewise.  Update call to
16260         altivec_expand_vec_perm_const_le.
16261         (rs6000_expand_vec_perm_const): Delete.
16262         (rs6000_vectorize_vec_perm_const_ok): Delete.
16263         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
16264         reference to the SPE evmerge intructions.
16265         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16266         an element count and rtx array.
16267         (rs6000_expand_extract_even): Update call accordingly.
16268         (rs6000_expand_interleave): Likewise.
16269         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16270         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16271         new function.
16272         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16273
16274 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16275
16276         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16277         vector mode and that that mode matches the mode of the data
16278         being permuted.
16279         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16280         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
16281         directly using expand_vec_perm_1 when forcing selectors into
16282         registers.
16283         (expand_vec_perm_var): New function, split out from expand_vec_perm.
16284
16285 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16286
16287         * optabs-query.h (can_vec_perm_p): Delete.
16288         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16289         * optabs-query.c (can_vec_perm_p): Split into...
16290         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16291         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16292         particular selector is valid.
16293         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16294         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16295         (vect_grouped_load_supported): Likewise.
16296         (vect_shift_permute_load_chain): Likewise.
16297         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16298         (vect_transform_slp_perm_load): Likewise.
16299         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16300         (vectorizable_bswap): Likewise.
16301         (vect_gen_perm_mask_checked): Likewise.
16302         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
16303         implementations of variable permutation vectors into account
16304         when deciding which selector to use.
16305         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16306         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16307         with a false third argument.
16308         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16309         to test whether the constant selector is valid and can_vec_perm_var_p
16310         to test whether a variable selector is valid.
16311
16312 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16313
16314         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16315         * optabs-query.c (can_vec_perm_p): Likewise.
16316         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16317         instead of vec_perm_indices.
16318         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16319         (vect_gen_perm_mask_checked): Likewise,
16320         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16321         (vect_gen_perm_mask_checked): Likewise,
16322
16323 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16324
16325         * optabs-query.h (qimode_for_vec_perm): Declare.
16326         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16327         (qimode_for_vec_perm): ...this new function.
16328         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16329
16330 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
16331
16332         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16333         does not have a conditional at the top.
16334
16335 2018-01-02  Richard Biener  <rguenther@suse.de>
16336
16337         * ipa-inline.c (big_speedup_p): Fix expression.
16338
16339 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
16340
16341         PR target/81616
16342         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16343         for generic 4->6.
16344
16345 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
16346
16347         PR target/81616
16348         Generic tuning.
16349         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16350         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16351         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16352         cond_taken_branch_cost 3->4.
16353
16354 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
16355
16356         PR tree-optimization/83581
16357         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16358         TODO_cleanup_cfg if any changes have been made.
16359
16360         PR middle-end/83608
16361         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16362         convert_modes if target mode has the right side, but different mode
16363         class.
16364
16365         PR middle-end/83609
16366         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16367         last argument when extracting from CONCAT.  If either from_real or
16368         from_imag is NULL, use expansion through memory.  If result is not
16369         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16370         the parts directly to inner mode, if even that fails, use expansion
16371         through memory.
16372
16373         PR middle-end/83623
16374         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16375         check for bswap in mode rather than HImode and use that in expand_unop
16376         too.
16377 \f
16378 Copyright (C) 2018 Free Software Foundation, Inc.
16379
16380 Copying and distribution of this file, with or without modification,
16381 are permitted in any medium without royalty provided the copyright
16382 notice and this notice are preserved.