gcc/ChangeLog

   1 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
   2
   3         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
   4         BUILTIN_VEC_XST entries for pointer to double and long long.
   5
   6 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
   7
   8         PR target/85990
   9         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
  10         Update comments.
  11         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
  12         Likewise.
  13
  14 2018-06-12  Martin Liska  <mliska@suse.cz>
  15
  16         * doc/options.texi: Document IntegerRange.
  17
  18 2018-06-12  Martin Liska  <mliska@suse.cz>
  19
  20         * config/i386/i386.opt: Make MPX-related options as Deprecated.
  21         * opt-functions.awk: Handle Deprecated flag.
  22         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
  23         and report error.
  24         (read_cmdline_option): Report warning for a deprecated option.
  25         * opts.h (struct cl_option): Add new field cl_deprecated.
  26         (CL_ERR_DEPRECATED): New.
  27
  28 2018-06-12  Martin Liska  <mliska@suse.cz>
  29
  30         * doc/options.texi: Document Deprecated option flag.
  31
  32 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
  33
  34         * config/arc/arc-arch.h (arc_extras): New enum.
  35         (arc_cpu_t):Add field extra.
  36         (arc_cpu_types): Consider the extras.
  37         * config/arc/arc-cpus.def: Add extras info.
  38         * config/arc/arc-opts.h (processor_type): Consider extra field.
  39         * config/arc/arc.c (arc_override_options): Handle extra field.
  40
  41 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
  42
  43         * config/arc/arc-arch.h: Update ARC_OPTX macro.
  44         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
  45         field.
  46         * config/arc/arc.c (arc_init): Update pic warning.
  47         (irq_range): Update irq range parsing warnings.
  48         (arc_override_options): Update various warning messages.
  49         (arc_handle_aux_attribute): Likewise.
  50
  51 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
  52
  53         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
  54
  55 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
  56
  57         * doc/sourcebuild.texi: Document usage of line number 0 in verify
  58         compiler messages directives.
  59
  60 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
  61
  62         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
  63         * config/mips/mips-tables.opt: Regenerate.
  64         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
  65         mips64r6.
  66         * doc/invoke.texi: Document -march=i6500.
  67
  68 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
  69
  70         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
  71         (i6400_gpmul): Add cpu_unit.
  72         (i6400_gpdiv): Likewise.
  73         (i6400_msa_add_d): Update reservations.
  74         (i6400_msa_int_add) Likewise.
  75         (i6400_msa_short_logic3) Likewise.
  76         (i6400_msa_short_logic2) Likewise.
  77         (i6400_msa_short_logic) Likewise.
  78         (i6400_msa_move) Likewise.
  79         (i6400_msa_cmp) Likewise.
  80         (i6400_msa_short_float2) Likewise.
  81         (i6400_msa_div_d) Likewise.
  82         (i6400_msa_long_logic1) Likewise.
  83         (i6400_msa_long_logic2) Likewise.
  84         (i6400_msa_mult) Likewise.
  85         (i6400_msa_long_float2) Likewise.
  86         (i6400_msa_long_float4) Likewise.
  87         (i6400_msa_long_float5) Likewise.
  88         (i6400_msa_long_float8) Likewise.
  89         (i6400_fpu_fadd): Include frint type.
  90         (i6400_fpu_store): New define_insn_reservation.
  91         (i6400_fpu_load): Likewise.
  92         (i6400_fpu_move): Likewise.
  93         (i6400_fpu_fcmp): Likewise.
  94         (i6400_fpu_fmadd): Likewise.
  95         (i6400_int_mult): Include imul3nc type and update reservation.
  96         (i6400_int_div): Include idiv3 type and update reservation.
  97         (i6400_int_load): Update to check type not move_type.
  98         (i6400_int_store): Likewise.
  99         (i6400_int_prefetch): Set zero latency.
 100
 101 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
 102
 103         * gcc.c: Document new %@{...} sequence.
 104         (LINK_COMMAND_SPEC): Use it for the -L switches.
 105         (cpp_unique_options): Use it for the -I switches.
 106         (at_file_argbuf): New global variable.
 107         (in_at_file): Likewise.
 108         (alloc_args): Create at_file_argbuf.
 109         (clear_args): Truncate at_file_argbuf.
 110         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
 111         (open_at_file): New function.
 112         (close_at_file): Likewise.
 113         (create_at_file): Delete.
 114         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
 115         <'o'>: Likewise.
 116         <'@'>: New case.
 117         (validate_switches_from_spec): Deal with %@{...} sequence.
 118         (validate_switches): Likewise.
 119         (driver::finalize): Call clear_args.
 120
 121 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
 122
 123         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
 124
 125 2018-06-11  Martin Sebor  <msebor@redhat.com>
 126
 127         * doc/invoke.texi (-Wall): List -Wc++17-compat.
 128         (Wno-class-memaccess): Add @opindex.
 129         (Wno-templates, Wno-multiple-inheritance): Same.
 130         (Wno-virtual-inheritance, Wno-namespaces): Same.
 131         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
 132         (Wno-format-overflow, Wno-format-truncation): Same.
 133         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
 134         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
 135         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
 136         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
 137         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
 138         (Wno-misspelled-isr): Same.
 139
 140 2018-06-11  Martin Sebor  <msebor@redhat.com>
 141
 142         * PR tree-optimization/86083
 143         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
 144
 145 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
 146
 147         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
 148
 149 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
 150
 151         PR target/85755
 152         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
 153         on the correct operand.
 154         (*movdi_internal64): Ditto.
 155
 156 2018-06-11  Martin Liska  <mliska@suse.cz>
 157
 158         PR tree-optimization/86089
 159         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
 160
 161 2018-06-11  Julia Koval  <julia.koval@intel.com>
 162
 163         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
 164         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
 165         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
 166
 167 2018-06-11  Olivier Hainque  <hainque@adacore.com>
 168
 169         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
 170         for Ada with strict dwarf2.
 171
 172 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
 173
 174         PR target/85755
 175         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
 176         addresses.
 177
 178 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
 179
 180         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
 181
 182 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
 183
 184         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
 185         TARGET_ELF.
 186
 187 2018-06-08  Martin Liska  <mliska@suse.cz>
 188
 189         * tree-cfg.h (debug_function): Fix argument type to match
 190         implementation.
 191
 192 2018-06-08  Martin Liska  <mliska@suse.cz>
 193
 194         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
 195         Remove usage of MPX-related (and removed) fields.
 196         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
 197
 198 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
 199
 200         * cfg.c (debug): Use TDF_NONE rather than 0.
 201         * cfghooks.c (debug): Likewise.
 202         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
 203         (struct dump_option_value_info): Convert to...
 204         (struct kv_pair): ...this template type.
 205         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
 206         rather than 0.
 207         (optinfo_verbosity_options): Likewise.
 208         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
 209         OPTGROUP_NONE.
 210         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
 211         than int for "optgroup_flags" param.
 212         (dump_generic_expr_loc): Use dump_flags_t rather than int for
 213         "dump_kind" param.
 214         (dump_dec): Likewise.
 215         (dump_finish): Use TDF_NONE rather than 0.
 216         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
 217         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
 218         than 0.  Update for change to option_ptr.
 219         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
 220         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
 221         0.  Update for changes to optinfo_verbosity_options and
 222         optgroup_options.
 223         (opt_info_switch_p): Convert optgroup_flags from int to
 224         optgroup_flags_t.
 225         (dump_basic_block): Use dump_flags_t rather than int
 226         for "dump_kind" param.
 227         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
 228         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
 229         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
 230         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
 231         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
 232         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
 233         TDF_NONE): Convert from macros to...
 234         (enum dump_flag): ...this new enum.
 235         (dump_flags_t): Update to use enum.
 236         (operator|, operator&, operator~, operator|=, operator&=):
 237         Implement for dump_flags_t.
 238         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
 239         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
 240         Convert from macros to...
 241         (enum optgroup_flag): ...this new enum.
 242         (optgroup_flags_t): New typedef.
 243         (operator|, operator|=): Implement for optgroup_flags_t.
 244         (struct dump_file_info): Convert field "alt_flags" to
 245         dump_flags_t.  Convert field "optgroup_flags" to
 246         optgroup_flags_t.
 247         (dump_basic_block): Use dump_flags_t rather than int for param.
 248         (dump_generic_expr_loc): Likewise.
 249         (dump_dec): Likewise.
 250         (dump_register): Convert param "optgroup_flags" to
 251         optgroup_flags_t.
 252         (opt_info_enable_passes): Likewise.
 253         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
 254         than 0.
 255         * gimple-pretty-print.c (debug): Likewise.
 256         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
 257         (merged_store_group::apply_stores): Likewise.
 258         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
 259         * gimple.c (verify_gimple_pp): Likewise.
 260         * graphite-poly.c (print_pbb_body): Likewise.
 261         * passes.c (pass_manager::register_one_dump_file): Convert
 262         local "optgroup_flags" to optgroup_flags_t.
 263         * print-tree.c (print_node): Use TDF_NONE rather than 0.
 264         (debug): Likewise.
 265         (debug_body): Likewise.
 266         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
 267         to optgroup_flags_t.
 268         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
 269         than 0.
 270         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
 271         (convert_mult_to_fma): Likewise.
 272         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
 273         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
 274         * tree-vect-data-refs.c (dump_lower_bound): Convert param
 275         "dump_kind" to dump_flags_t.
 276
 277 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
 278
 279         * config/rs6000/rs6000.c (min, max): Delete.
 280
 281 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
 282
 283         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
 284         -mabi=spe and -mabi=no-spe.
 285
 286 2018-06-08  Martin Liska  <mliska@suse.cz>
 287
 288         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
 289         where we expect an existing summary.
 290
 291 2018-06-08  Martin Liska  <mliska@suse.cz>
 292
 293         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
 294         * ipa-inline.h (estimate_edge_growth): Likewise.
 295
 296 2018-06-08  Martin Liska  <mliska@suse.cz>
 297
 298         * cgraph.c (function_version_hasher::hash): Use
 299         cgraph_node::get_uid ().
 300         (function_version_hasher::equal):
 301         * cgraph.h (cgraph_node::get_uid): New method.
 302         * ipa-inline.c (update_caller_keys): Use
 303         cgraph_node::get_uid ().
 304         (update_callee_keys): Likewise.
 305         * ipa-utils.c (searchc): Likewise.
 306         (ipa_reduced_postorder): Likewise.
 307         * lto-cgraph.c (input_node): Likewise.
 308         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
 309         * symbol-summary.h (symtab_insertion): Likewise.
 310         (symtab_removal): Likewise.
 311         (symtab_duplication): Likewise.
 312         * tree-pretty-print.c (dump_function_header): Likewise.
 313         * tree-sra.c (convert_callers_for_node): Likewise.
 314
 315 2018-06-08  Martin Liska  <mliska@suse.cz>
 316
 317         * cgraph.c (symbol_table::create_edge): Always assign a new
 318         unique number.
 319         (symbol_table::free_edge): Do not recycle numbers.
 320         * cgraph.h (cgraph_edge::get): New method.
 321         * symbol-summary.h (symtab_removal): Use it.
 322         (symtab_duplication): Likewise.
 323         (call_summary::hashable_uid): Remove.
 324
 325 2018-06-08  Martin Liska  <mliska@suse.cz>
 326
 327         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
 328         (initialize_growth_caches): Remove.
 329         (free_growth_caches): Likewise.
 330         (do_estimate_edge_time): Use edge_growth_cache.
 331         (do_estimate_edge_size): Likewise.
 332         (do_estimate_edge_hints): Likewise.
 333         * ipa-inline.c (reset_edge_caches): Likewise.
 334         (recursive_inlining): Likewise.
 335         (inline_small_functions): Likewise.
 336         * ipa-inline.h (initialize_growth_caches): Remove.
 337         (estimate_edge_size): Likewise.
 338         (estimate_edge_time): Likewise.
 339         (estimate_edge_hints): Likewise.
 340         (reset_edge_growth_cache): Likewise.
 341         * symbol-summary.h (call_summary::remove): New method.
 342
 343 2018-06-08  Martin Liska  <mliska@suse.cz>
 344
 345         * ipa-cp.c (class edge_clone_summary): New summary.
 346         (grow_edge_clone_vectors): Remove.
 347         (ipcp_edge_duplication_hook): Remove.
 348         (class edge_clone_summary_t): New call_summary class.
 349         (ipcp_edge_removal_hook): Remove.
 350         (edge_clone_summary_t::duplicate): New function.
 351         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
 352         (create_specialized_node): Likewise.
 353         (ipcp_driver): Initialize edge_clone_summaries and do not
 354         register hooks.
 355
 356 2018-06-08  Martin Liska  <mliska@suse.cz>
 357
 358         * symbol-summary.h (get): New function.
 359         (call_summary::m_initialize_when_cloning): New class member.
 360
 361 2018-06-08  Martin Liska  <mliska@suse.cz>
 362
 363         * cgraph.c (cgraph_node::remove): Do not recycle uid.
 364         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
 365         (symbol_table::allocate_cgraph_symbol): Do not set uid.
 366         * passes.c (uid_hash_t): Record removed_nodes by their uids.
 367         (remove_cgraph_node_from_order): Use the removed_nodes set.
 368         (do_per_function_toporder): Likwise.
 369         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
 370         instead of summary_uid.
 371         (symtab_removal): Likewise.
 372         (symtab_duplication): Likewise.
 373
 374 2018-06-08  Martin Liska  <mliska@suse.cz>
 375
 376         * ipa-cp.c (ipcp_store_bits_results): Use
 377         ipcp_transformation_sum.
 378         (ipcp_store_vr_results): Likewise.
 379         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
 380         to ...
 381         (ipcp_transformation_initialize): ... this.
 382         (ipa_set_node_agg_value_chain):
 383         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
 384         (write_ipcp_transformation_info): Likewise.
 385         (read_ipcp_transformation_info): Likewise.
 386         (ipcp_update_bits): Likewise.
 387         (ipcp_update_vr): Likewise.
 388         (ipcp_transform_function): Likewise.
 389         * ipa-prop.h: Rename ipcp_transformation_summary to
 390         ipcp_transformation.
 391         (class ipcp_transformation_t): New function summary.
 392         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
 393         (ipa_get_agg_replacements_for_node): Likewise.
 394
 395 2018-06-08  Martin Liska  <mliska@suse.cz>
 396
 397         * ipa-pure-const.c (struct funct_state_d): Do it class instead
 398         of struct.
 399         (class funct_state_summary_t): New function_summary class.
 400         (has_function_state): Remove.
 401         (get_function_state): Likewise.
 402         (set_function_state): Likewise.
 403         (add_new_function): Likewise.
 404         (funct_state_summary_t::insert): New function.
 405         (duplicate_node_data): Remove.
 406         (remove_node_data): Remove.
 407         (funct_state_summary_t::duplicate): New function.
 408         (register_hooks): Create new funct_state_summaries.
 409         (pure_const_generate_summary): Use it.
 410         (pure_const_write_summary): Likewise.
 411         (pure_const_read_summary): Likewise.
 412         (propagate_pure_const): Likewise.
 413         (propagate_nothrow): Likewise.
 414         (dump_malloc_lattice): Likewise.
 415         (propagate_malloc): Likewise.
 416         (execute): Do not register hooks, just remove summary
 417         instead.
 418         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
 419         constructor.
 420
 421 2018-06-08  Martin Liska  <mliska@suse.cz>
 422
 423         * ipa-reference.c (remove_node_data): Remove.
 424         (duplicate_node_data): Likewise.
 425         (class ipa_ref_var_info_summary_t): New class.
 426         (class ipa_ref_opt_summary_t): Likewise.
 427         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
 428         (get_reference_optimization_summary): Use
 429         ipa_ref_opt_sum_summaries.
 430         (set_reference_vars_info): Remove.
 431         (set_reference_optimization_summary): Likewise.
 432         (ipa_init): Create summaries.
 433         (init_function_info): Use function summary.
 434         (ipa_ref_opt_summary_t::duplicate): New function.
 435         (ipa_ref_opt_summary_t::remove): New function.
 436         (get_read_write_all_from_node): Fix GNU coding style.
 437         (propagate): Use function summary.
 438         (write_node_summary_p): Fix GNU coding style.
 439         (stream_out_bitmap): Likewise.
 440         (ipa_reference_read_optimization_summary): Use function summary.
 441         (ipa_reference_c_finalize): Do not release hooks.
 442
 443 2018-06-08  Martin Liska  <mliska@suse.cz>
 444
 445         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
 446         (analyze_function_body): Extract multiple calls of get_create.
 447         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
 448         * ipa-inline.c (recursive_inlining): Use ::get method.
 449         * ipa-inline.h (estimate_edge_growth): Likewise.
 450
 451 2018-06-08  Martin Liska  <mliska@suse.cz>
 452
 453         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
 454         HSA_INVALID.
 455         (hsa_function_summary::hsa_function_summary): Use the new enum
 456         value.
 457         (hsa_gpu_implementation_p): Use hsa_summaries::get.
 458         * hsa-gen.c (hsa_get_host_function): Likewise.
 459         (get_brig_function_name): Likewise.
 460         * ipa-hsa.c (process_hsa_functions): Likewise.
 461         (ipa_hsa_write_summary): Likewise.
 462         * symbol-summary.h (symtab_duplication): Use ::get function/
 463         (get): New function.
 464
 465 2018-06-08  Martin Liska  <mliska@suse.cz>
 466
 467         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
 468         of get.
 469         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
 470         (hsa_register_kernel): Likewise.
 471         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
 472         * hsa-gen.c (hsa_get_host_function): Likewise.
 473         (get_brig_function_name): Likewise.
 474         (generate_hsa): Likewise.
 475         (pass_gen_hsail::execute): Likewise.
 476         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
 477         (devirtualization_time_bonus): Likewise.
 478         (ipcp_propagate_stage): Likewise.
 479         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
 480         (edge_set_predicate): Likewise.
 481         (evaluate_conditions_for_known_args): Likewise.
 482         (evaluate_properties_for_edge): Likewise.
 483         (ipa_fn_summary::reset): Likewise.
 484         (ipa_fn_summary_t::duplicate): Likewise.
 485         (dump_ipa_call_summary): Likewise.
 486         (ipa_dump_fn_summary): Likewise.
 487         (analyze_function_body): Likewise.
 488         (compute_fn_summary): Likewise.
 489         (estimate_edge_devirt_benefit): Likewise.
 490         (estimate_edge_size_and_time): Likewise.
 491         (estimate_calls_size_and_time): Likewise.
 492         (estimate_node_size_and_time): Likewise.
 493         (inline_update_callee_summaries): Likewise.
 494         (remap_edge_change_prob): Likewise.
 495         (remap_edge_summaries): Likewise.
 496         (ipa_merge_fn_summary_after_inlining): Likewise.
 497         (ipa_update_overall_fn_summary): Likewise.
 498         (read_ipa_call_summary): Likewise.
 499         (inline_read_section): Likewise.
 500         (write_ipa_call_summary): Likewise.
 501         (ipa_fn_summary_write): Likewise.
 502         (ipa_free_fn_summary): Likewise.
 503         * ipa-hsa.c (process_hsa_functions): Likewise.
 504         (ipa_hsa_write_summary): Likewise.
 505         (ipa_hsa_read_section): Likewise.
 506         * ipa-icf.c (sem_function::merge): Likewise.
 507         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
 508         (do_estimate_edge_time): Likewise.
 509         (estimate_size_after_inlining): Likewise.
 510         (estimate_growth): Likewise.
 511         (growth_likely_positive): Likewise.
 512         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
 513         (inline_call): Likewise.
 514         * ipa-inline.c (caller_growth_limits): Likewise.
 515         (can_inline_edge_p): Likewise.
 516         (can_inline_edge_by_limits_p): Likewise.
 517         (compute_uninlined_call_time): Likewise.
 518         (compute_inlined_call_time): Likewise.
 519         (want_inline_small_function_p): Likewise.
 520         (edge_badness): Likewise.
 521         (update_caller_keys): Likewise.
 522         (update_callee_keys): Likewise.
 523         (recursive_inlining): Likewise.
 524         (inline_small_functions): Likewise.
 525         (inline_to_all_callers_1): Likewise.
 526         (dump_overall_stats): Likewise.
 527         (early_inline_small_functions): Likewise.
 528         (early_inliner): Likewise.
 529         * ipa-inline.h (estimate_edge_growth): Likewise.
 530         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
 531         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
 532         * ipa-prop.h (IPA_NODE_REF): Likewise.
 533         (IPA_EDGE_REF): Likewise.
 534         * ipa-pure-const.c (malloc_candidate_p): Likewise.
 535         (propagate_malloc): Likewise.
 536         * ipa-split.c (execute_split_functions): Likewise.
 537         * symbol-summary.h: Rename get to get_create.
 538         (get): Likewise.
 539         (get_create): Likewise.
 540         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
 541
 542 2018-06-08  Martin Liska  <mliska@suse.cz>
 543
 544         * symbol-summary.h (release): Move definition out of class
 545         declaration.
 546         (symtab_removal): Likewise.
 547         (symtab_duplication): Likewise.
 548
 549 2018-06-08  Martin Liska  <mliska@suse.cz>
 550
 551         * symbol-summary.h (function_summary): Move constructor
 552         implementation out of class declaration.
 553         (release): Likewise.
 554         (symtab_insertion): Likewise.
 555         (symtab_removal): Likewise.
 556         (symtab_duplication): Likewise.
 557         (get): Likewise.
 558
 559 2018-06-08  Martin Liska  <mliska@suse.cz>
 560
 561         * Makefile.in: Remove support for MPX (macros, related functions,
 562         fields in cgraph_node, ...).
 563         * builtin-types.def (BT_BND): Likewise.
 564         (BT_FN_BND_CONST_PTR): Likewise.
 565         (BT_FN_CONST_PTR_BND): Likewise.
 566         (BT_FN_VOID_PTR_BND): Likewise.
 567         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
 568         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
 569         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
 570         (expand_builtin_mempcpy_with_bounds): Likewise.
 571         (expand_builtin_memset_with_bounds): Likewise.
 572         (expand_builtin_memset_args): Likewise.
 573         (std_expand_builtin_va_start): Likewise.
 574         (expand_builtin): Likewise.
 575         (expand_builtin_with_bounds): Likewise.
 576         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
 577         (DEF_LIB_BUILTIN_CHKP): Likewise.
 578         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
 579         (DEF_CHKP_BUILTIN): Likewise.
 580         (BUILT_IN_MEMCPY): Likewise.
 581         (BUILT_IN_MEMMOVE): Likewise.
 582         (BUILT_IN_MEMPCPY): Likewise.
 583         (BUILT_IN_MEMSET): Likewise.
 584         (BUILT_IN_STPCPY): Likewise.
 585         (BUILT_IN_STRCAT): Likewise.
 586         (BUILT_IN_STRCHR): Likewise.
 587         (BUILT_IN_STRCPY): Likewise.
 588         (BUILT_IN_STRLEN): Likewise.
 589         (BUILT_IN_MEMCPY_CHK): Likewise.
 590         (BUILT_IN_MEMMOVE_CHK): Likewise.
 591         (BUILT_IN_MEMPCPY_CHK): Likewise.
 592         (BUILT_IN_MEMSET_CHK): Likewise.
 593         (BUILT_IN_STPCPY_CHK): Likewise.
 594         (BUILT_IN_STRCAT_CHK): Likewise.
 595         (BUILT_IN_STRCPY_CHK): Likewise.
 596         * calls.c (store_bounds): Likewise.
 597         (emit_call_1): Likewise.
 598         (special_function_p): Likewise.
 599         (maybe_warn_nonstring_arg): Likewise.
 600         (initialize_argument_information): Likewise.
 601         (finalize_must_preallocate): Likewise.
 602         (compute_argument_addresses): Likewise.
 603         (expand_call): Likewise.
 604         * cfgexpand.c (expand_call_stmt): Likewise.
 605         (expand_return): Likewise.
 606         (expand_gimple_stmt_1): Likewise.
 607         (pass_expand::execute): Likewise.
 608         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
 609         (cgraph_node::remove): Likewise.
 610         (cgraph_node::dump): Likewise.
 611         (cgraph_node::verify_node): Likewise.
 612         * cgraph.h (chkp_function_instrumented_p): Likewise.
 613         (symtab_node::get_alias_target): Likewise.
 614         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
 615         (cgraph_local_p): Likewise.
 616         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
 617         (cgraph_edge::rebuild_references): Likewise.
 618         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
 619         (walk_polymorphic_call_targets): Likewise.
 620         (cgraph_node::expand_thunk): Likewise.
 621         (symbol_table::output_weakrefs): Likewise.
 622         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
 623         (ix86_handle_option): Likewise.
 624         * config/i386/constraints.md: Likewise.
 625         * config/i386/i386-builtin-types.def (BND): Likewise.
 626         (VOID): Likewise.
 627         (PVOID): Likewise.
 628         (ULONG): Likewise.
 629         * config/i386/i386-builtin.def (BDESC_END): Likewise.
 630         (BDESC_FIRST): Likewise.
 631         (BDESC): Likewise.
 632         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
 633         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
 634         * config/i386/i386.c (enum reg_class): Likewise.
 635         (ix86_target_string): Likewise.
 636         (ix86_option_override_internal): Likewise.
 637         (ix86_conditional_register_usage): Likewise.
 638         (ix86_valid_target_attribute_inner_p): Likewise.
 639         (ix86_set_indirect_branch_type): Likewise.
 640         (ix86_set_current_function): Likewise.
 641         (ix86_function_arg_regno_p): Likewise.
 642         (init_cumulative_args): Likewise.
 643         (ix86_function_arg_advance): Likewise.
 644         (ix86_function_arg): Likewise.
 645         (ix86_pass_by_reference): Likewise.
 646         (ix86_function_value_regno_p): Likewise.
 647         (ix86_function_value_1): Likewise.
 648         (ix86_function_value_bounds): Likewise.
 649         (ix86_return_in_memory): Likewise.
 650         (ix86_setup_incoming_vararg_bounds): Likewise.
 651         (ix86_va_start): Likewise.
 652         (indirect_thunk_need_prefix): Likewise.
 653         (print_reg): Likewise.
 654         (ix86_print_operand): Likewise.
 655         (ix86_expand_call): Likewise.
 656         (ix86_output_function_return): Likewise.
 657         (reg_encoded_number): Likewise.
 658         (BDESC_VERIFYS): Likewise.
 659         (ix86_init_mpx_builtins): Likewise.
 660         (ix86_init_builtins): Likewise.
 661         (ix86_emit_cmove): Likewise.
 662         (ix86_emit_move_max): Likewise.
 663         (ix86_expand_builtin): Likewise.
 664         (ix86_builtin_mpx_function): Likewise.
 665         (ix86_get_arg_address_for_bt): Likewise.
 666         (ix86_load_bounds): Likewise.
 667         (ix86_store_bounds): Likewise.
 668         (ix86_load_returned_bounds): Likewise.
 669         (ix86_store_returned_bounds): Likewise.
 670         (ix86_class_likely_spilled_p): Likewise.
 671         (ix86_hard_regno_mode_ok): Likewise.
 672         (x86_order_regs_for_local_alloc): Likewise.
 673         (ix86_mitigate_rop): Likewise.
 674         (ix86_bnd_prefixed_insn_p): Likewise.
 675         (ix86_mpx_bound_mode): Likewise.
 676         (ix86_make_bounds_constant): Likewise.
 677         (ix86_initialize_bounds): Likewise.
 678         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
 679         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
 680         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
 681         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
 682         (TARGET_CHKP_BOUND_MODE): Likewise.
 683         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
 684         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
 685         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
 686         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
 687         * config/i386/i386.h (TARGET_MPX): Likewise.
 688         (TARGET_MPX_P): Likewise.
 689         (VALID_BND_REG_MODE): Likewise.
 690         (FIRST_BND_REG): Likewise.
 691         (LAST_BND_REG): Likewise.
 692         (enum reg_class): Likewise.
 693         (BND_REG_P): Likewise.
 694         (BND_REGNO_P): Likewise.
 695         (BNDmode): Likewise.
 696         (ADJUST_INSN_LENGTH): Likewise.
 697         * config/i386/i386.md: Likewise.
 698         * config/i386/i386.opt: Likewise.
 699         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
 700         (defined): Likewise.
 701         (LINK_MPX): Likewise.
 702         (MPX_SPEC): Likewise.
 703         (LIBMPX_SPEC): Likewise.
 704         (LIBMPXWRAPPERS_SPEC): Likewise.
 705         (CHKP_SPEC): Likewise.
 706         * config/i386/predicates.md: Likewise.
 707         * dbxout.c (dbxout_type): Likewise.
 708         * doc/extend.texi: Likewise.
 709         * doc/invoke.texi: Likewise.
 710         * doc/md.texi: Likewise.
 711         * doc/tm.texi: Likewise.
 712         * doc/tm.texi.in: Likewise.
 713         * dwarf2out.c (is_base_type): Likewise.
 714         (gen_formal_types_die): Likewise.
 715         (gen_subprogram_die): Likewise.
 716         (gen_type_die_with_usage): Likewise.
 717         (gen_decl_die): Likewise.
 718         (dwarf2out_late_global_decl): Likewise.
 719         * expr.c (expand_assignment): Likewise.
 720         (emit_storent_insn): Likewise.
 721         (store_expr_with_bounds): Likewise.
 722         (store_expr): Likewise.
 723         (expand_expr_real_1): Likewise.
 724         * expr.h (store_expr_with_bounds): Likewise.
 725         * function.c (use_register_for_decl): Likewise.
 726         (struct bounds_parm_data): Likewise.
 727         (assign_parms_augmented_arg_list): Likewise.
 728         (assign_parm_find_entry_rtl): Likewise.
 729         (assign_parm_is_stack_parm): Likewise.
 730         (assign_parm_load_bounds): Likewise.
 731         (assign_bounds): Likewise.
 732         (assign_parms): Likewise.
 733         (expand_function_start): Likewise.
 734         * gcc.c (CHKP_SPEC): Likewise.
 735         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
 736         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
 737         (wrestrict_dom_walker::check_call): Likewise.
 738         * gimple.c (gimple_build_call_from_tree): Likewise.
 739         * gimple.h (enum gf_mask): Likewise.
 740         (gimple_call_with_bounds_p): Likewise.
 741         (gimple_call_set_with_bounds): Likewise.
 742         * gimplify.c (gimplify_init_constructor): Likewise.
 743         * ipa-cp.c (initialize_node_lattices): Likewise.
 744         (propagate_constants_across_call): Likewise.
 745         (find_more_scalar_values_for_callers_subset): Likewise.
 746         * ipa-hsa.c (process_hsa_functions): Likewise.
 747         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
 748         * ipa-icf.c (sem_function::merge): Likewise.
 749         * ipa-inline.c (early_inliner): Likewise.
 750         * ipa-pure-const.c (warn_function_noreturn): Likewise.
 751         (warn_function_cold): Likewise.
 752         (propagate_pure_const): Likewise.
 753         * ipa-ref.h (enum GTY): Likewise.
 754         * ipa-split.c (find_retbnd): Likewise.
 755         (consider_split): Likewise.
 756         (split_function): Likewise.
 757         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
 758         * ipa.c (walk_polymorphic_call_targets): Likewise.
 759         (symbol_table::remove_unreachable_nodes): Likewise.
 760         (process_references): Likewise.
 761         (cgraph_build_static_cdtor_1): Likewise.
 762         * lto-cgraph.c (lto_output_node): Likewise.
 763         (output_refs): Likewise.
 764         (compute_ltrans_boundary): Likewise.
 765         (input_overwrite_node): Likewise.
 766         (input_node): Likewise.
 767         (input_cgraph_1): Likewise.
 768         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
 769         * passes.c (pass_manager::execute_early_local_passes): Likewise.
 770         (class pass_chkp_instrumentation_passes): Likewise.
 771         (make_pass_chkp_instrumentation_passes): Likewise.
 772         * passes.def: Likewise.
 773         * rtl.h (struct GTY): Likewise.
 774         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
 775         * stor-layout.c (layout_type): Likewise.
 776         * symtab.c: Likewise.
 777         * target.def: Likewise.
 778         * targhooks.c (default_chkp_bound_type): Likewise.
 779         (default_chkp_bound_mode): Likewise.
 780         (default_builtin_chkp_function): Likewise.
 781         (default_chkp_function_value_bounds): Likewise.
 782         (default_chkp_make_bounds_constant): Likewise.
 783         (default_chkp_initialize_bounds): Likewise.
 784         * targhooks.h (default_chkp_bound_type): Likewise.
 785         (default_chkp_bound_mode): Likewise.
 786         (default_builtin_chkp_function): Likewise.
 787         (default_chkp_function_value_bounds): Likewise.
 788         (default_chkp_make_bounds_constant): Likewise.
 789         (default_chkp_initialize_bounds): Likewise.
 790         * toplev.c (compile_file): Likewise.
 791         (process_options): Likewise.
 792         * tree-core.h (DEF_BUILTIN): Likewise.
 793         (DEF_BUILTIN_CHKP): Likewise.
 794         * tree-inline.c (declare_return_variable): Likewise.
 795         (remap_gimple_stmt): Likewise.
 796         (copy_bb): Likewise.
 797         (initialize_inlined_parameters): Likewise.
 798         (expand_call_inline): Likewise.
 799         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
 800         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
 801         (make_pass_ipa_chkp_produce_thunks): Likewise.
 802         (make_pass_chkp): Likewise.
 803         (make_pass_chkp_opt): Likewise.
 804         (make_pass_chkp_instrumentation_passes): Likewise.
 805         * tree-pretty-print.c (dump_generic_node): Likewise.
 806         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
 807         * tree-ssa-dce.c (propagate_necessity): Likewise.
 808         (eliminate_unnecessary_stmts): Likewise.
 809         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
 810         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
 811         * tree-ssa-sccvn.h: Likewise.
 812         * tree-ssa-strlen.c (get_string_length): Likewise.
 813         (valid_builtin_call): Likewise.
 814         (adjust_last_stmt): Likewise.
 815         (handle_builtin_strchr): Likewise.
 816         (handle_builtin_strcpy): Likewise.
 817         (handle_builtin_stxncpy): Likewise.
 818         (handle_builtin_memcpy): Likewise.
 819         (handle_builtin_strcat): Likewise.
 820         (strlen_check_and_optimize_stmt): Likewise.
 821         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
 822         * tree-streamer-in.c: Likewise.
 823         * tree-streamer.c (record_common_node): Likewise.
 824         * tree.c (tree_code_size): Likewise.
 825         (wide_int_to_tree_1): Likewise.
 826         (type_contains_placeholder_1): Likewise.
 827         (build_common_tree_nodes): Likewise.
 828         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
 829         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
 830         (POINTER_BOUNDS_P): Likewise.
 831         (BOUNDED_TYPE_P): Likewise.
 832         (BOUNDED_P): Likewise.
 833         (CALL_WITH_BOUNDS_P): Likewise.
 834         (pointer_bounds_type_node): Likewise.
 835         * value-prof.c (gimple_ic): Likewise.
 836         * var-tracking.c (vt_add_function_parameters): Likewise.
 837         * varasm.c (make_decl_rtl): Likewise.
 838         (assemble_start_function): Likewise.
 839         (output_constant): Likewise.
 840         (maybe_assemble_visibility): Likewise.
 841         * varpool.c (ctor_for_folding): Likewise.
 842         * chkp-builtins.def: Remove.
 843         * ipa-chkp.c: Remove.
 844         * ipa-chkp.h: Remove.
 845         * rtl-chkp.c: Remove.
 846         * rtl-chkp.h: Remove.
 847         * tree-chkp-opt.c: Remove.
 848         * tree-chkp.c: Remove.
 849         * tree-chkp.h: Remove.
 850
 851 2018-06-07  Carl Love  <cel@us.ibm.com>
 852
 853         * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
 854         vextract_fp_from_shortl): Add BE support.
 855
 856 2018-06-07  Paul Koning  <ni1d@arrl.net>
 857
 858         * compare-elim.c (try_merge_compare): Don't merge compare if
 859         address contains a side effect.
 860         (try_eliminate_compare): Likewise.
 861
 862 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
 863
 864         * config.gcc: Support "tremont".
 865         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
 866         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
 867         PROCESSOR_TREMONT.
 868         * config/i386/i386.c (m_TREMONT): Define.
 869         (processor_target_table): Add "tremont".
 870         (PTA_TREMONT): Define.
 871         (ix86_lea_outperforms): Add TARGET_TREMONT.
 872         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
 873         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
 874         and M_INTEL_GOLDMONT_PLUS.
 875         (fold_builtin_cpu): Add "tremont".
 876         (ix86_add_stmt_cost): Add TARGET_TREMONT.
 877         (ix86_option_override_internal): Add "tremont".
 878         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
 879         (processor_type): Add PROCESSOR_TREMONT.
 880         * config/i386/x86-tune.def: Add m_TREMONT.
 881         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
 882
 883 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
 884
 885         * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
 886         symbol defined for msp430i* devices to be lower case.
 887
 888         gcc/testsuite/gcc.target/msp430/
 889         * msp430i-device-symbol.c: New test.
 890         * msp430f-device-symbol.c: New test.
 891         * msp430.h: New test header file.
 892
 893 2018-06-07  Richard Biener  <rguenther@suse.de>
 894
 895         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
 896         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
 897         Properly wrap signed arithmetic if overflow wraps.
 898
 899 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
 900
 901         PR tree-optimization/69615
 902         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
 903         of a cast from a same precision integral SSA_NAME in a bb dominated
 904         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
 905         cast to utype if rhs2 has already a compatible type.
 906
 907 2018-06-07  Richard Biener  <rguenther@suse.de>
 908
 909         PR tree-optimization/85935
 910         * graphite-scop-detection.c (find_params_in_bb): Analyze
 911         condition operands with respect to the correct loop.  Assert
 912         the analysis doesn't fail.
 913
 914 2018-06-04  Carl Love  <cel@us.ibm.com>
 915
 916         * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
 917         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
 918         as it is slightly cheaper.
 919         (first_match_or_eos_index_<mode>):
 920         Calculate index using natural element order.
 921         (first_match_index_<mode>):
 922         Calculate index using natural element order.
 923         (first_match_or_eos_index_<mode>):
 924         Calculate index using natural order.
 925         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
 926         for BE and LE modes.
 927         * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
 928         P9V_BUILTIN_VCLZLSBB_V16QI.
 929         * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
 930         specific.
 931
 932 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 933
 934         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
 935         indentation and line wrap for many prototypes.  Add missing
 936         @smallexample directives around block of prototypes for vec_xl and
 937         vec_xst.
 938
 939 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
 940
 941         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
 942         track if we pass or return IEEE 128-bit floating point.
 943         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
 944         C++ mangling that is compatible with GCC 8.1.
 945         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
 946         (init_cumulative_args): Note if we pass or return IEEE 128-bit
 947         floating point types.
 948         (rs6000_function_arg_advance_1): Likewise.
 949         (rs6000_mangle_type): Optionally generate mangled names that match
 950         what GCC 8.1 generated for IEEE 128-bit floating point types.
 951         (rs6000_globalize_decl_name): If we have an external function that
 952         passes or returns IEEE 128-bit types, generate a weak reference
 953         from the mangled name used in GCC 8.1 to the current mangled
 954         name.
 955         (rs6000_init_builtins): Make __ibm128 use the long double type if
 956         long double is IBM extended double.  Make __float128 use the long
 957         double type if long double is IEEE 128-bit.
 958
 959         PR target/85657
 960         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
 961         macro for __ibm128 built-in functions.
 962         (PACK_IF): Add __ibm128 pack/unpack functions.
 963         (UNPACK_IF): Likewise.
 964         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
 965         enable long double built-in functions if long double is IEEE
 966         128-bit floating point.
 967         (rs6000_invalid_builtin): Update long double built-in function
 968         error message.
 969         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
 970         functions, adjust the built-in function to use the long double
 971         built-in function if __ibm128 and long double are the same type.
 972         * doc/extend.texi (PowerPC builtins): Update documention for
 973         __builtin_{,un}pack_longdouble.  Add documentation for
 974         __builtin_{,un}pack_ibm128.
 975
 976 2018-06-06  Jim Wilson  <jimw@sifive.com>
 977
 978         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
 979         (struct machine_function): New field interrupt_mode.
 980         (riscv_handle_type_attribute): New function.  Add forward declaration.
 981         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
 982         (riscv_expand_epilogue): Check interrupt_mode field.
 983         (riscv_set_current_function): Check interrupt attribute args and
 984         set interrupt_mode field.
 985         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
 986         (riscv_sret, riscv_uret): New.
 987         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
 988         new arguments to interrupt attribute.
 989
 990 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
 991
 992         PR target/63177
 993         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
 994         Don't handle -mcpu=power8 if -mpower9-vector is also used.
 995
 996 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 997
 998         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
 999         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1000         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1001         several redundant entries.
1002
1003 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
1004
1005         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1006         type from "rtx" to "rtx_insn *".
1007         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1008         for local "call_insn", removing cast.
1009         (ix86_expand_call): Likewise, introducing a "call_insn" local.
1010
1011 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
1012
1013         PR tree-optimization/86066
1014         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1015         for BIT_INSERT_EXPR stores.
1016
1017 2018-06-06  Richard Biener  <rguenther@suse.de>
1018
1019         PR tree-optimization/86062
1020         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1021         component refs ontop
1022         of to be offsetted base.
1023
1024 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1025
1026         * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1027         to be static and remove check on interrupt attribute name.
1028
1029         gcc/testsuite/gcc.target/msp430/
1030         * function-attributes-4.c: New test.
1031         * static-interrupts.c: New test.
1032
1033 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1034
1035         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1036         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1037
1038 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
1039
1040         PR target/79924
1041         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1042         second argument.
1043         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1044         Remove second argument, change how error is called.
1045         (aarch64_layout_arg): Remove second argument from
1046         aarch64_err_no_fpadvsimd call.
1047         (aarch64_init_cumulative_args): Ditto.
1048         (aarch64_gimplify_va_arg_expr): Ditto.
1049         * config/aarch64/aarch64.md (mov<mode>): Ditto.
1050
1051 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
1052
1053         * config/i386/i386.md (simple_return_indirect_internal): New expander.
1054         (*simple_return_indirect_internal<mode>): Rename from
1055         simple_return_indirect_internal.  Use W mode iterator.
1056         (rstorssp): New expander.
1057         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
1058         (clrssbsy): New expander.
1059         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
1060
1061 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1062
1063         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1064         __typeof__.
1065         (cmse_check_pointed_object): Likewise.
1066
1067 2018-06-05  Martin Liska  <mliska@suse.cz>
1068
1069         PR gcov-profile/47618
1070         * doc/invoke.texi: Document how -fprofile-dir format
1071         is extended.
1072
1073 2018-06-05  Richard Biener  <rguenther@suse.de>
1074
1075         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1076         removal pretend DOM info isn't available so we do not update
1077         it and only remove edges, not dominated blocks.  Actually free
1078         DOM info in case we removed something.  Remove unreachable blocks.
1079         (mfb_keep_latches): Work with either DOM info or marked backedges.
1080         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1081         first.  Mark backedges if DOM info isn't available.
1082         (Re-)compute DOM info after cleanup_control_flow_pre.
1083
1084 2018-06-05  Richard Biener  <rguenther@suse.de>
1085
1086         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1087         (locus_discrim_hasher::hash): Adjust.
1088         (locus_discrim_hasher::equal): Likewise.
1089         (next_discriminator_for_locus): Work on line directly.
1090         (same_line_p): Pass in expanded locus1 as well.
1091         (assign_discriminators): Avoid redundant location expansions.
1092
1093 2018-06-05  Richard Biener  <rguenther@suse.de>
1094
1095         PR tree-optimization/86046
1096         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1097         if required after clearing TREE_ADDRESSABLE.
1098
1099 2018-06-05  Richard Biener  <rguenther@suse.de>
1100
1101         PR tree-optimization/86047
1102         * tree-ssa-loop.c (for_each_index): Glob handling of all
1103         decls and constants and really handle all of them.
1104
1105 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1106
1107         PR target/81497
1108         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1109         qualifier_void_pointer and qualifier_const_void_pointer.
1110         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1111         (arm_init_builtins): Handle the above.
1112         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1113         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1114         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1115         void intrinsics.
1116
1117 2018-06-05  Martin Liska  <mliska@suse.cz>
1118
1119         * auto-profile.c (read_autofdo_file): Do not use
1120         gcov_ctr_summary struct.
1121         (afdo_callsite_hot_enough_for_early_inline): Likewise.
1122         * coverage.c (struct counts_entry): Likewise.
1123         (read_counts_file): Read just single summary entry.
1124         (get_coverage_counts): Use gcov_summary struct.
1125         * coverage.h (get_coverage_counts): Likewise.
1126         * gcov-dump.c (dump_working_sets): Likewise.
1127         (tag_summary): Dump just single summary.
1128         * gcov-io.c (gcov_write_summary): Write just histogram
1129         summary.
1130         (gcov_read_summary): Read just single summary.
1131         (compute_working_sets): Use gcov_summary struct.
1132         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1133         of GCOV_COUNTERS_SUMMABLE.
1134         (GCOV_COUNTERS_SUMMABLE): Remove.
1135         (GCOV_FIRST_VALUE_COUNTER): Replace with
1136         GCOV_COUNTER_V_INTERVAL.
1137         (struct gcov_ctr_summary): Remove.
1138         (struct gcov_summary): Directly use fields of former
1139         gcov_ctr_summary.
1140         (compute_working_sets): Use gcov_summary struct.
1141         * gcov.c (read_count_file): Do not use ctrs fields.
1142         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1143         struct.
1144         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1145         struct.
1146         * profile.c: Likewise.
1147         * profile.h: Likewise.
1148
1149 2018-06-05  Martin Liska  <mliska@suse.cz>
1150
1151         PR gcov-profile/84846
1152         * gcov.c (output_lines): Print working directory only
1153         in intermediate format.
1154
1155 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
1156
1157         * config/s390/s390-builtin-types.def: Add void function type.
1158         * config/s390/s390-builtins.def: Use the function type for the
1159         tbeginc builtin.
1160
1161 2018-06-04  Jim Wilson  <jimw@sifive.com>
1162
1163         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1164         to int.
1165         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1166         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
1167         handle EH_RETURN_DATA_REGNO registers properly.
1168         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1169         (riscv_expand_epilogue): Update comment.  Change argument name and
1170         type.  Update code to use new name and type.  Pass new args to
1171         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
1172         EXCEPTION_RETURN.
1173         * config/riscv/riscv.md (NORMAL_RETURN): New.
1174         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1175         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1176         (eh_return): Call gen_eh_return_internal and emit barrier.
1177         (eh_return_internal): Call riscv_expand_epilogue.
1178
1179 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
1180
1181         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1182         bit_insertion field and declare can_be_merged_into method.
1183         (merged_store_group::can_be_merged_into): New method.
1184         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1185         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
1186         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1187
1188 2018-06-04  Richard Biener  <rguenther@suse.de>
1189
1190         PR tree-optimization/85955
1191         * builtins.c (fold_builtin_sincos): Convert pointers to
1192         destination to appropriate type before dereferencing.
1193
1194 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
1195
1196         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1197
1198 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
1199
1200         * expr.c (expand_expr_real_1): Force the operand into memory if
1201         its TYPE_MODE is BLKmode and if there is no integer mode for
1202         the number of bits being extracted.
1203
1204 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
1205
1206         PR target/85832
1207         PR target/86036
1208         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1209         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1210
1211 2018-06-04  Richard Biener  <rguenther@suse.de>
1212
1213         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1214         (cleanup_tree_cfg_noloop): ... single caller.  Do
1215         start_recording_case_labels later.
1216
1217 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
1218
1219         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1220         to _IMMINTRIN_H_INCLUDED.
1221         * config/i386/pconfigintrin.h: Ditto.
1222         * config/i386/waitpkgintrin.h: Ditto.
1223         * config/i386/immintrin.h: Add includes for sgxintrin.h,
1224         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1225         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1226         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1227         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1228         waitpkgintrin.h and cldemoteintrin.h.
1229
1230 2018-06-04  Richard Biener  <rguenther@suse.de>
1231
1232         PR tree-optimization/86038
1233         * tracer.c (find_best_successor): Check probability for
1234         being initialized, bail out if not.
1235
1236 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
1237
1238         PR target/86003
1239         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1240         of bits to ignore when comparing architectures.
1241
1242 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
1243
1244         PR tree-optimization/69615
1245         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1246         maximum or minimum of the type, try to merge it also as if
1247         range1 is + [-, x - 1] or + [x + 1, -].
1248
1249         PR c++/86025
1250         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1251
1252 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
1253
1254         PR tree-optimization/86034
1255         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1256         the unsigned bitfield type in a bit insertion sequence if it does not
1257         have a larger precision than the bitfield size.
1258         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1259
1260 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
1261
1262         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1263
1264 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
1265
1266         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1267         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1268         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1269         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1270
1271 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
1272
1273         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1274         Disable -fdelete-null-pointer-checks for ELF toolchain.
1275
1276 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
1277             Kito Cheng  <kito.cheng@gmail.com>
1278
1279         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1280         (nds32le-*-*, nds32be-*-*): Integrate checking process.
1281         (nds32*-*-*): Add glibc and uclibc conditions.
1282         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1283         (TARGET_EXCEPT_UNWIND_INFO): Define.
1284         * config/nds32/elf.h: New file.
1285         * config/nds32/linux.h: New file.
1286         * config/nds32/nds32-elf.opt: New file.
1287         * config/nds32/nds32-linux.opt: New file.
1288         * config/nds32/nds32-fp-as-gp.c
1289         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1290         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1291         TARGET_LINUX_ABI.
1292         (nds32_asm_file_end): Ditto.
1293         (nds32_print_operand): Ditto.
1294         (nds32_insert_attributes): Ditto.
1295         (nds32_init_libfuncs): New function.
1296         (TARGET_HAVE_TLS): Define.
1297         (TARGET_INIT_LIBFUNCS): Define.
1298         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1299         spec content.
1300         (TARGET_ELF): Apply different mcmodel setting.
1301         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1302         been migrated into elf.h and linux.h files.
1303         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1304         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1305         (mcmodel): The content has been migrated into nds32-elf.opt and
1306         nds32-linux.opt files.
1307         * config/nds32/t-elf: New file.
1308         * config/nds32/t-linux: New file.
1309
1310 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
1311             Shiva Chen  <shiva0217@gmail.com>
1312
1313         * config/nds32/constants.md (unspec_volatile_element): Add
1314         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1315         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1316         optimization.
1317         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1318         (make_pass_nds32_fp_as_gp): Declare.
1319         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1320         optmization pass.
1321         (nds32_asm_function_end_prologue): Remove unused asm output.
1322         (nds32_asm_function_begin_epilogue): Remove unused asm output.
1323         (nds32_asm_file_start): Output necessary fp_as_gp information.
1324         (nds32_option_override): Adjust register usage.
1325         (nds32_expand_prologue): Consider fp_as_gp situation.
1326         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1327         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1328         (epilogue): Ditto.
1329         (return): Ditto.
1330         (simple_return): Ditto.
1331         (omit_fp_begin): Output special directive for fp_as_gp.
1332         (omit_fp_end): Output special directive for fp_as_gp.
1333         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1334         mforbid-fp-as-gp): New options.
1335
1336 2018-06-01  Mark Wielaard  <mark@klomp.org>
1337
1338         * dwarf2out.c (dwarf2out_finish): Remove generation of
1339         DW_AT_loclists_base.
1340
1341 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
1342
1343         * gimple-ssa-store-merging.c: Include gimple-fold.h.
1344         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1345         (struct merged_store_group): Add bit_insertion field.
1346         (dump_char_array): Use standard hexadecimal format.
1347         (merged_store_group::merged_store_group): Set bit_insertion to false.
1348         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
1349         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
1350         also print the mask in the dump file.
1351         (pass_store_merging::gate): Minor tweak.
1352         (imm_store_chain_info::coalesce_immediate): Fix wrong association
1353         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
1354         stores with INTEGER_CST stores.
1355         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1356         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1357         and use it throughout.  Generate bit insertion sequences if need be.
1358         (pass_store_merging::process_store): Remove redundant condition.
1359         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1360
1361 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
1362
1363         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1364         the 128-bit floating point types.  Fix function comment.
1365
1366 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1367
1368         * config/aarch64/aarch64-simd.md
1369         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1370         mnemonics.
1371         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1372         mnemonics.
1373
1374 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
1375
1376         PR tree-optimization/85989
1377         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1378         variable.
1379         (backprop::intersect_uses): Check it when deciding whether this
1380         is a backedge reference.
1381         (backprop::process_block): Add each phi to m_visited_phis
1382         after visiting it, then clear it at the end.
1383
1384 2018-06-01  Richard Biener  <rguenther@suse.de>
1385
1386         * tree-vectorizer.h (vect_dr_stmt): New function.
1387         (vect_get_load_cost): Adjust.
1388         (vect_get_store_cost): Likewise.
1389         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1390         Use vect_dr_stmt instead of DR_SMTT.
1391         (vect_record_base_alignments): Likewise.
1392         (vect_calculate_target_alignment): Likewise.
1393         (vect_compute_data_ref_alignment): Likewise and make static.
1394         (vect_update_misalignment_for_peel): Likewise.
1395         (vect_verify_datarefs_alignment): Likewise.
1396         (vector_alignment_reachable_p): Likewise.
1397         (vect_get_data_access_cost): Likewise.  Pass down
1398         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1399         (vect_get_peeling_costs_all_drs): Likewise.
1400         (vect_peeling_hash_get_lowest_cost): Likewise.
1401         (vect_enhance_data_refs_alignment): Likewise.
1402         (vect_find_same_alignment_drs): Likewise.
1403         (vect_analyze_data_refs_alignment): Likewise.
1404         (vect_analyze_group_access_1): Likewise.
1405         (vect_analyze_group_access): Likewise.
1406         (vect_analyze_data_ref_access): Likewise.
1407         (vect_analyze_data_ref_accesses): Likewise.
1408         (vect_vfa_segment_size): Likewise.
1409         (vect_small_gap_p): Likewise.
1410         (vectorizable_with_step_bound_p): Likewise.
1411         (vect_prune_runtime_alias_test_list): Likewise.
1412         (vect_analyze_data_refs): Likewise.
1413         (vect_supportable_dr_alignment): Likewise.
1414         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1415         (vect_gen_prolog_loop_niters): Likewise.
1416         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1417         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1418         modify DR_STMT.
1419         (vect_recog_mask_conversion_pattern): Likewise.
1420         (vect_try_gather_scatter_pattern): Likewise.
1421         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1422         to vect_get_store_cost.
1423         (vect_get_store_cost): Get stmt_info instead of DR.
1424         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1425         (vect_get_load_cost): Get stmt_info instead of DR.
1426
1427 2018-06-01  Richard Biener  <rguenther@suse.de>
1428
1429         PR middle-end/86017
1430         * gimple-fold.c (var_decl_component_p): Also allow offsetted
1431         vars wrapped in MEM_REFs.
1432
1433 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
1434
1435         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1436         Fix subreg tests so that we only return a choice between
1437         GENERAL_REGS and FP_REGS if the original classes included both.
1438
1439 2018-06-01  Richard Biener  <rguenther@suse.de>
1440
1441         PR ipa/85960
1442         * tree-ssa-structalias.c (get_function_part_constraint):
1443         Handle NULL fi->decl.
1444         (find_func_aliases_for_call): Properly handle indirect
1445         fi from direct call.
1446         (find_func_clobbers): Likewise.
1447         (ipa_pta_execute): Likewise.
1448         (create_variable_info_for): For functions that are ifunc_resolver
1449         resolve to a varinfo that contains the result of the resolver call.
1450         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1451         aliases.
1452
1453 2018-05-31  Michael Collison  <michael.collison@arm.com>
1454
1455         * config/aarch64/aarch64.md:
1456         (*fix_to_zero_extenddfdi2): New pattern.
1457         * gcc.target/aarch64/fix_extend1.c: New testcase.
1458
1459 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
1460
1461         PR middle-end/78809
1462         PR middle-end/83026
1463         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1464         and BUILT_IN_STRNCMP_EQ.
1465         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1466         BUILT_IN_STRNCMP_EQ.
1467         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1468         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1469         (gimple_fold_builtin): Likewise.
1470         * tree-ssa-strlen.c (compute_string_length): New function.
1471         (determine_min_obsize): New function.
1472         (handle_builtin_string_cmp): New function to handle calls to
1473         string compare functions.
1474         (strlen_optimize_stmt): Add handling to builtin string compare
1475         calls.
1476         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1477         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1478         * tree.c (build_common_builtin_nodes): Add new defines of
1479         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1480
1481 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
1482
1483         PR target/85984
1484         * bb-reorder.c (pass_partition_blocks::gate): Return false for
1485         functions with naked attribute.
1486
1487 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
1488
1489         * config/i386/sse.md (avx_vec_concat<mode>):
1490         Substitute concat_tg_mode mode attribute with xtg_mode.
1491         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1492         (concat_tg_mode): Remove mode attribute.
1493
1494 2018-05-31  Martin Sebor  <msebor@redhat.com>
1495
1496         PR c/82063
1497         * calls.c (alloc_max_size): Correct a logic error/typo.
1498         Treat excessive arguments as infinite.  Warn for invalid arguments.
1499         * doc/invoke.texi (-Walloc-size-larger-than): Update.
1500
1501 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
1502
1503         PR target/85829
1504         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1505         and movx for Haswell.
1506
1507 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
1508             Cesar Philippidis  <cesar@codesourcery.com>
1509
1510         PR middle-end/85879
1511         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1512         when emitting error on private/firstprivate reductions.
1513         * omp-low.c (lower_omp_target): Avoid reference-type processing
1514         on pointers for firstprivate clause.
1515
1516 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1517
1518         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1519         (st1x2): Likewise.
1520         (st1x3): Likewise.
1521         * config/aarch64/aarch64-simd.md
1522         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1523         (aarch64_ld1_x3_<mode>): Likewise
1524         (aarch64_st1x2<VALLDIF:mode>): Likewise
1525         (aarch64_st1_x2_<mode>): Likewise
1526         (aarch64_st1x3<VALLDIF:mode>): Likewise
1527         (aarch64_st1_x3_<mode>): Likewise
1528         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1529         (vld1_s8_x3): Likewise.
1530         (vld1_u16_x3): Likewise.
1531         (vld1_s16_x3): Likewise.
1532         (vld1_u32_x3): Likewise.
1533         (vld1_s32_x3): Likewise.
1534         (vld1_u64_x3): Likewise.
1535         (vld1_s64_x3): Likewise.
1536         (vld1_f16_x3): Likewise.
1537         (vld1_f32_x3): Likewise.
1538         (vld1_f64_x3): Likewise.
1539         (vld1_p8_x3): Likewise.
1540         (vld1_p16_x3): Likewise.
1541         (vld1_p64_x3): Likewise.
1542         (vld1q_u8_x3): Likewise.
1543         (vld1q_s8_x3): Likewise.
1544         (vld1q_u16_x3): Likewise.
1545         (vld1q_s16_x3): Likewise.
1546         (vld1q_u32_x3): Likewise.
1547         (vld1q_s32_x3): Likewise.
1548         (vld1q_u64_x3): Likewise.
1549         (vld1q_s64_x3): Likewise.
1550         (vld1q_f16_x3): Likewise.
1551         (vld1q_f32_x3): Likewise.
1552         (vld1q_f64_x3): Likewise.
1553         (vld1q_p8_x3): Likewise.
1554         (vld1q_p16_x3): Likewise.
1555         (vld1q_p64_x3): Likewise.
1556         (vst1_s64_x2): Likewise.
1557         (vst1_u64_x2): Likewise.
1558         (vst1_f64_x2): Likewise.
1559         (vst1_s8_x2): Likewise.
1560         (vst1_p8_x2): Likewise.
1561         (vst1_s16_x2): Likewise.
1562         (vst1_p16_x2): Likewise.
1563         (vst1_s32_x2): Likewise.
1564         (vst1_u8_x2): Likewise.
1565         (vst1_u16_x2): Likewise.
1566         (vst1_u32_x2): Likewise.
1567         (vst1_f16_x2): Likewise.
1568         (vst1_f32_x2): Likewise.
1569         (vst1_p64_x2): Likewise.
1570         (vst1q_s8_x2): Likewise.
1571         (vst1q_p8_x2): Likewise.
1572         (vst1q_s16_x2): Likewise.
1573         (vst1q_p16_x2): Likewise.
1574         (vst1q_s32_x2): Likewise.
1575         (vst1q_s64_x2): Likewise.
1576         (vst1q_u8_x2): Likewise.
1577         (vst1q_u16_x2): Likewise.
1578         (vst1q_u32_x2): Likewise.
1579         (vst1q_u64_x2): Likewise.
1580         (vst1q_f16_x2): Likewise.
1581         (vst1q_f32_x2): Likewise.
1582         (vst1q_f64_x2): Likewise.
1583         (vst1q_p64_x2): Likewise.
1584         (vst1_s64_x3): Likewise.
1585         (vst1_u64_x3): Likewise.
1586         (vst1_f64_x3): Likewise.
1587         (vst1_s8_x3): Likewise.
1588         (vst1_p8_x3): Likewise.
1589         (vst1_s16_x3): Likewise.
1590         (vst1_p16_x3): Likewise.
1591         (vst1_s32_x3): Likewise.
1592         (vst1_u8_x3): Likewise.
1593         (vst1_u16_x3): Likewise.
1594         (vst1_u32_x3): Likewise.
1595         (vst1_f16_x3): Likewise.
1596         (vst1_f32_x3): Likewise.
1597         (vst1_p64_x3): Likewise.
1598         (vst1q_s8_x3): Likewise.
1599         (vst1q_p8_x3): Likewise.
1600         (vst1q_s16_x3): Likewise.
1601         (vst1q_p16_x3): Likewise.
1602         (vst1q_s32_x3): Likewise.
1603         (vst1q_s64_x3): Likewise.
1604         (vst1q_u8_x3): Likewise.
1605         (vst1q_u16_x3): Likewise.
1606         (vst1q_u32_x3): Likewise.
1607         (vst1q_u64_x3): Likewise.
1608         (vst1q_f16_x3): Likewise.
1609         (vst1q_f32_x3): Likewise.
1610         (vst1q_f64_x3): Likewise.
1611         (vst1q_p64_x3): Likewise.
1612
1613 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1614
1615         * config/msp430/msp430.c (msp430_output_labelref): Prepend
1616         user_label_prefix to name.
1617
1618         * tree-core.h: Update comment about the format of NAME string
1619         passed to handler in attribute_spec.
1620
1621         * config/msp430/msp430.md: Remove erroneous subreg expression from
1622         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1623         zero_extend{q,h}isi2.
1624
1625 2018-05-30  Borislav Petkov  <bp@suse.de>
1626
1627         * doc/extend.texi: Document some architecture specific
1628         constraints and sort entries.
1629
1630 2018-05-30  Martin Sebor  <msebor@redhat.com>
1631
1632         PR middle-end/85369
1633         * builtins.c (expand_builtin_stpcpy_1): New function.
1634         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1635         only if the former succeeds.
1636
1637 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1638
1639         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1640         in saphira.
1641
1642 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
1643
1644         * doc/invoke.texi (-flinker-output): Document
1645
1646 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
1647
1648         * passes.c (ipa_write_summaries): Only modify statements if body
1649         is in memory.
1650         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1651         incrementally linking.
1652         (ipa_passes): Likewise.
1653         * lto-cgraph.c (lto_output_node): When incrementally linking do not
1654         pass down resolution info.
1655         * common.opt (flag_incremental_link): Update info.
1656         * gcc.c (plugin specs): Turn flinker-output=* to
1657         -plugin-opt=-linker-output-known
1658         * toplev.c (compile_file): Also cut compilation when doing incremental
1659         link.
1660         * flag-types. (enum lto_partition_model): Add
1661         LTO_LINKER_OUTPUT_NOLTOREL.
1662         (invoke.texi): Add -flinker-output docs.
1663         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1664         link same way as WPA; do not stream in dead initializers.
1665
1666         * dwarf2out.c (dwarf2out_die_ref_for_decl,
1667         darf2out_register_external_decl): Support incremental link.
1668
1669 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
1670
1671         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1672
1673 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
1674
1675         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1676         it down to simple_object_copy_lto_debug_sections.
1677         (run_gcc): Determine incremental LTO link time and configure
1678         lto1 into non-wpa mode, disable renaming of debug sections.
1679
1680 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1681
1682         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1683         descriptions of various incorrectly documented functions.
1684
1685 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1686
1687         Revert:
1688         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1689         address check not strict.
1690
1691 2018-05-30  Richard Biener  <rguenther@suse.de>
1692
1693         PR tree-optimization/85964
1694         * tracer.c (better_p): Drop initialized count check, we only
1695         call the function with initialized counts now.
1696         (find_best_successor): Do find a best edge if one
1697         has uninitialized count.
1698         (find_best_predecessor): Likewise.  Do BB frequency check only
1699         if count is initialized.
1700
1701 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
1702
1703         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1704         (aarch64_ldrstr_offset_compare): New.
1705         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1706         load/store orderings.
1707         (aarch64_gen_adjusted_ldpstp): Likewise.
1708
1709 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
1710
1711         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1712         Check for subset of GENERAL_REGS and FP_REGS.
1713         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1714         r=w alternative.
1715
1716 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
1717
1718         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1719         and wi::to_poly_offset.  Add the current offset and then check
1720         whether the sum fits, rather than using an unchecked addition of
1721         a checked term.  Check for a shwi rather than a uhwi.
1722         * expr.c (get_bit_range): Use tree_to_poly_uint64.
1723         (store_constructor): Use poly_int_tree_p.
1724         (expand_expr_real_1): Likewise.
1725         * function.c (assign_temp): Likewise.
1726         * fold-const.c (const_binop): Use poly_int_tree_p and
1727         wi::to_poly_offset.
1728         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
1729         division.
1730         * ipa-icf-gimple.c (func_checker::compare_operand): Use
1731         to_poly_offset for MEM offsets.
1732         * ipa-icf.c (sem_variable::equals): Likewise.
1733         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1734         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1735         wi::to_poly_offset for BIT_FIELD_REF offsets.
1736         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1737         wi::to_poly_offset.
1738         * var-tracking.c (emit_note_insn_var_location): Use
1739         tree_to_poly_uint64.
1740
1741 2018-05-29  Jim Wilson  <jimw@sifive.com>
1742
1743         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1744
1745 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
1746
1747         PR target/85950
1748         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1749         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1750         sequence.
1751         (sse4_1_round<mode>2): Use nonimmediate_operand
1752         for operand 1 predicate.
1753
1754 2018-05-29  Martin Sebor  <msebor@redhat.com>
1755             Richard Biener  <rguenther@suse.de>
1756
1757         PR testsuite/85888
1758         * calls.c (get_size_range): Call determine_value_range instead
1759         of get_value_range..
1760         * tree-vrp.h (determine_value_range): Declared new function.
1761         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1762
1763 2018-05-29  Richard Biener  <rguenther@suse.de>
1764
1765         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1766         sure to use non-pattern stmts for get_earlier_stmt arguments.
1767         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1768         called on pattern stmts.
1769         (get_later_stmt): Likewise.
1770
1771 2018-05-29  Martin Liska  <mliska@suse.cz>
1772
1773         PR gcov-profile/85759
1774         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1775         env variables.
1776
1777 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
1778
1779         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1780         VEC_UNPACK_*_EXPR.
1781         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1782         VEC_PACK_*_EXPR.
1783
1784         PR target/85918
1785         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1786         VEC_PACK_FLOAT_EXPR): New tree codes.
1787         * tree-pretty-print.c (op_code_prio): Handle
1788         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1789         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1790         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1791         * tree-inline.c (estimate_operator_cost): Likewise.
1792         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1793         * fold-const.c (const_binop): Likewise.
1794         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1795         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1796         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1797         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1798         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1799         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1800         * expr.c (expand_expr_real_2): Likewise.
1801         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1802         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1803         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1804         optabs.
1805         * optabs.c (expand_widen_pattern_expr): For
1806         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1807         sign from result type rather than operand's type.
1808         (expand_binop_directly): For vec_packu_float_optab and
1809         vec_packs_float_optab allow result type to be different from operand's
1810         type.
1811         * optabs-tree.c (optab_for_tree_code): Handle
1812         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1813         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
1814         * tree-vect-generic.c (expand_vector_operations_1):  Handle
1815         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1816         VEC_PACK_FLOAT_EXPR.
1817         * tree-vect-stmts.c (supportable_widening_operation): Handle
1818         FIX_TRUNC_EXPR.
1819         (supportable_narrowing_operation): Handle FLOAT_EXPR.
1820         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1821         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1822         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
1823         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1824         mode attributes.
1825         (vec_pack<floatprefix>_float_<mode>): New expander.
1826         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1827         attributes.
1828         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1829         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1830         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1831         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
1832         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
1833         Document.
1834         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
1835         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
1836         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1837         VEC_PACK_FLOAT_EXPR): Document.
1838
1839 2018-05-29  Richard Biener  <rguenther@suse.de>
1840
1841         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
1842         member.
1843         (stmt_vec_info_vec): Make pointer.
1844         (init_stmt_vec_info_vec): Remove.
1845         (free_stmt_vec_info_vec): Likewise.
1846         (set_stmt_vec_info_vec): New function.
1847         (free_stmt_vec_infos): Likewise.
1848         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
1849         (set_vinfo_for_stmt): Likewise.
1850         (get_earlier_stmt): Likewise.
1851         (get_later_stmt): Likewise.
1852         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
1853         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
1854         (vec_info::~vec_info): Free stmt_vec_infos.
1855         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
1856         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
1857         (pass_slp_vectorize::execute): Likewise.
1858         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
1859         (free_stmt_vec_info_vec): Likewise.
1860         (set_stmt_vec_info_vec): New function.
1861         (free_stmt_vec_infos): Likewise.
1862         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
1863         the global stmt_vec_info_vec.
1864         * tree-parloops.c (gather_scalar_reductions): Use
1865         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
1866         vector.
1867
1868 2018-05-29  Richard Biener  <rguenther@suse.de>
1869
1870         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
1871
1872 2018-05-29  Martin Liska  <mliska@suse.cz>
1873             David Malcolm  <dmalcolm@redhat.com>
1874
1875         * vec.c (test_reverse): New.
1876         (vec_c_tests): Add new test.
1877         * vec.h (vl_ptr>::reverse): New function.
1878
1879 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
1880
1881         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
1882
1883         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
1884         and later.
1885
1886 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1887
1888         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
1889
1890 2018-05-28  Richard Biener  <rguenther@suse.de>
1891
1892         PR tree-optimization/85933
1893         * tree-vect-data-refs.c (vect_record_base_alignments): Only
1894         look at stmts marked as vectorizable.
1895
1896 2018-05-28  Richard Biener  <rguenther@suse.de>
1897
1898         PR tree-optimization/85934
1899         * tree-vect-generic.c (expand_vector_operations_1): Hoist
1900         vector boolean check before scalar optimization.
1901
1902 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
1903
1904         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1905         for armv5te.
1906
1907 2018-05-28  Mark Wielaard  <mark@klomp.org>
1908
1909         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1910         if it is an expression containing a minus sign.
1911
1912 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
1913
1914         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1915
1916 2018-05-27  Paul Koning  <ni1d@arrl.net>
1917
1918         * config/pdp11/pdp11.md (truncsihi2): Remove.
1919
1920 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
1921             Chung-Ju Wu  <jasonwucj@gmail.com>
1922
1923         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1924         implementation.
1925         (unaligned_store_dw): Ditto.
1926         * config/nds32/nds32-memory-manipulation.c
1927         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1928         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1929         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1930         (emit_setmem_word_loop): Rename to ...
1931         (emit_setmem_doubleword_loop): ... this.
1932         (nds32_gen_dup_4_byte_to_word_value): New function.
1933         (nds32_gen_dup_8_byte_to_double_word_value): New function.
1934         (nds32_expand_setmem_loop): Refine implementation.
1935         (nds32_expand_setmem_loop_v3m): Ditto.
1936         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1937         pattern.
1938
1939 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
1940
1941         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1942
1943 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
1944
1945         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1946         (nds32_init_machine_status): Initialize machine->attr_naked_p and
1947         machine->attr_no_prologue_p.
1948         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1949         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1950         (nds32_expand_epilogue): Consider attr_naked_p.
1951         (nds32_expand_epilogue_v3pop): Likewise.
1952         (nds32_can_use_return_insn): Likewise.
1953         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1954         attr_no_prologue_p fields.
1955         * config/nds32/nds32.opt (mret-in-naked-func): New option.
1956
1957 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
1958
1959         PR target/85918
1960         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1961         attributes.
1962         * config/i386/sse.md
1963         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1964         Rename to ...
1965         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1966         ... this.
1967         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1968         Rename to ...
1969         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1970         ... this.
1971         (*<floatsuffix>floatv2div2sf2): Rename to ...
1972         (*float<floatunssuffix>v2div2sf2): ... this.
1973         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1974         (float<floatunssuffix>v2div2sf2_mask): ... this.
1975         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1976         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1977         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1978         to ...
1979         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1980         ... this.
1981         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1982         Rename to ...
1983         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1984         ... this.
1985         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1986         Rename to ...
1987         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1988         ... this.
1989         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1990         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1991         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1992         gen_ufix_truncv8dfv8si2.
1993         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1994         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1995         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1996         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1997         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1998         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1999         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2000         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2001
2002 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
2003
2004         PR target/85900
2005         PR target/85345
2006         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2007
2008 2018-05-25  Jim Wilson  <jimw@sifive.com>
2009
2010         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2011         * config/riscv/riscv.c (struct machine_function): Add
2012         interrupt_handler_p and attribute_checked_p fields.
2013         (riscv_attribute_table): Add interrupt.
2014         (riscv_interrupt_type_p): New.
2015         (riscv_save_reg_p): Save extra regs for interrupt handler.
2016         (riscv_use_save_libcall): Return false  for interrupt handler.
2017         (riscv_first_stack_step): Add forward declaration.
2018         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
2019         for interrupt handler with large frame.  Use it for saved reg list.
2020         (riscv_expand_prologue): Move flag_stack_usage_info support to
2021         eliminate duplication.
2022         (riscv_expand_epilogue): Generate mret for interrupt handler.
2023         (riscv_epilogue_uses): New.
2024         (riscv_can_use_return_insn): Return false for interrupt handler.
2025         (riscv_function_ok_for_sibcall): Likewise.
2026         (riscv_set_current_function): Add interrupt handler support.
2027         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2028         * config/riscv/riscv.md (UNSPECV_MRET): New.
2029         (GP_REGNUM): New.
2030         (riscv_frflags, riscv_fsflags): Use tab after opcode.
2031         (riscv_mret): New.
2032         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2033
2034 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2035
2036         PR tree-optimization/85712
2037         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2038         this candidate has already been replaced in-situ by a copy.
2039
2040 2018-05-25  Jason Merrill  <jason@redhat.com>
2041
2042         PR c++/80485 - inline function non-zero address.
2043         * symtab.c (nonzero_address): Check DECL_COMDAT.
2044
2045 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
2046
2047         PR target/83628
2048         * config/alpha/alpha.md (ashlsi3): New insn pattern.
2049         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
2050         extension of SImode operation.  Use const123_operand predicate.
2051         (*saddsi_1): Remove.
2052         (*saddl_se_1): Ditto.
2053         (*ssubsi_1): Ditto.
2054         (*ssubl_se_1): Ditto.
2055         * config/alpha/predicates.md (const123_operand): New predicate.
2056         * config/alpha/constraints.md (P): Use IN_RANGE.
2057
2058 2018-05-25  Richard Biener  <rguenther@suse.de>
2059
2060         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2061         defaulted to true.
2062         (ref_maybe_used_by_stmt_p): Likewise.
2063         (stmt_may_clobber_ref_p): Likewise.
2064         (stmt_may_clobber_ref_p_1): Likewise.
2065         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2066         and pass it along.
2067         (ref_maybe_used_by_stmt_p): Likewise.
2068         (stmt_may_clobber_ref_p): Likewise.
2069         (stmt_may_clobber_ref_p_1): Likewise.
2070         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2071         the alias oracle to disambiguate DRs with stmts DR analysis
2072         couldn't handle.
2073         (vect_analyze_data_refs): Do not give up on not analyzable
2074         DRs for BB vectorization.  Remove code truncating the dataref
2075         vector.
2076
2077 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
2078
2079         PR target/85832
2080         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2081         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
2082         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2083
2084 2018-05-25  Richard Biener  <rguenther@suse.de>
2085
2086         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2087         function, combining stmt data ref gathering and fatal analysis
2088         parts.
2089         (vect_analyze_data_refs): Remove now redudnant code and simplify.
2090         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2091         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2092         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2093         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2094
2095 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
2096
2097         PR tree-optimization/85720
2098         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2099         SCC if all partitions are builtins.
2100         (version_loop_by_alias_check): New parameter.  Generate cancelable
2101         runtime alias check if all partitions are builtins.
2102         (distribute_loop): Update call to above function.
2103
2104 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
2105
2106         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2107         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2108         (parm_default_def_partition_arg): Ditto.
2109         (set_parm_default_def_partition): Ditto.
2110         (get_parm_default_def_partitions): Ditto and make it static.
2111         (get_undefined_value_partitions): Ditto and make it static.
2112         (remove_ssa_form): Refactor call to init_var_map here.
2113         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2114         computation for loop region.
2115         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2116         (register_default_def): Delete.
2117         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2118         (parm_default_def_partition_arg): Ditto.
2119         (set_parm_default_def_partition): Ditto.
2120         (get_parm_default_def_partitions): Ditto and make it static.
2121         (get_undefined_value_partitions): Ditto and make it static.
2122         (coalesce_with_default, coalesce_with_default): Update comment.
2123         (create_coalesce_list_for_region): New func factored out from
2124         create_outofssa_var_map.
2125         (populate_coalesce_list_for_outofssa): New func factored out from
2126         create_outofssa_var_map and coalesce_ssa_name.
2127         (create_outofssa_var_map): Delete.
2128         (coalesce_ssa_name): Refactor to support live range computation.
2129         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2130         (get_parm_default_def_partitions): Delete.
2131         (get_undefined_value_partitions): Ditto.
2132         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2133         computation for loop region.
2134         (new_tree_live_info, loe_visit_block): Ditto.
2135         (live_worklist, set_var_live_on_entry): Ditto.
2136         (calculate_live_on_exit, verify_live_on_entry): Ditto.
2137         * tree-ssa-live.h (struct _var_map): New fields.
2138         (init_var_map): Change decl.
2139         (region_contains_p): New.
2140
2141 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
2142
2143         * tree-ssa-live.h (live_merge_and_clear): Delete.
2144
2145 2018-05-25  Richard Biener  <rguenther@suse.de>
2146
2147         PR c++/85912
2148         * tree-dump.c (dequeue_and_dump): Remove access to removed
2149         operand 2 of a SWITCH_EXPR.
2150
2151 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2152
2153         * doc/sourcebuild.texi (vect_double_cond_arith): Include
2154         multiplication and division.
2155         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2156         (cond_udiv@var{m}, cond_umod@var{m}): Document.
2157         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2158         (cond_udiv_optab, cond_umod_optab): New optabs.
2159         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2160         (IFN_COND_RDIV): New internal functions.
2161         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2162         TRUNC_MOD_EXPR and RDIV_EXPR.
2163         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2164         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2165         New unspecs.
2166         (SVE_INT_BINARY): Include mult.
2167         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2168         (optab, sve_int_op): Handle mult.
2169         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2170         UNSPEC_COND_DIV.
2171         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2172         for SVE_INT_BINARY_SD.
2173
2174 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2175
2176         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2177         (optab, sve_int_op): Handle div and udiv.
2178         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2179         for SVE_INT_BINARY_SD.
2180         (*<optab><mode>3): New insn for the same.
2181
2182 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2183
2184         * tree-vect-patterns.c: Include predict.h.
2185         (vect_recog_divmod_pattern): Restrict check for division support
2186         to when optimizing for size.
2187
2188 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2189
2190         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2191         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2192         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2193         (gimple_match_op::set_op): Likewise.
2194         (gimple_resimplify4): Declare.
2195         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2196         (expr::gen_transform): Likewise.
2197         (decision_tree::gen): Generate a simplification routine for 4 operands.
2198         * gimple-match-head.c (gimple_simplify): Add an overload for
2199         4 operands.  In the top-level function, handle up to 4 call
2200         arguments and call gimple_resimplify4.
2201         (gimple_resimplify4): New function.
2202         (build_call_internal): Pass a fourth operand.
2203         (maybe_push_to_seq): Likewise.
2204         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2205         Fold VEC_COND_EXPRs of an operation and a default value into
2206         an IFN_COND_* function if possible.
2207         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2208         New unspecs.
2209         (SVE_COND_FP_BINARY): Include them.
2210         (optab, sve_fp_op): Handle them.
2211         (SVE_INT_BINARY_REV): New code iterator.
2212         (SVE_COND_FP_BINARY_REV): New int iterator.
2213         (commutative): New int attribute.
2214         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2215         Declare.
2216         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2217         function.
2218         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2219         (*cond_<optab><mode>): New patterns for reversed operands.
2220
2221 2018-05-25  Richard Biener  <rguenther@suse.de>
2222
2223         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2224         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2225         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2226         (STMT_VINFO_GROUPED_ACCESS): Adjust.
2227         * tree-vect-data-refs.c (everywhere): Adjust users.
2228         * tree-vect-loop.c (everywhere): Likewise.
2229         * tree-vect-slp.c (everywhere): Likewise.
2230         * tree-vect-stmts.c (everywhere): Likewise.
2231         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2232
2233 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2234
2235         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2236         Rename to...
2237         (gcc_cv_as_section_exclude): ... this.
2238         Try Solaris as #exclude syntax.
2239         * configure: Regenerate.
2240         * config.in: Regenerate.
2241         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2242         SECTION_EXCLUDE.
2243         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2244         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2245
2246         * varasm.c (default_elf_asm_named_section): Don't check if
2247         HAVE_GAS_SECTION_EXCLUDE is defined.
2248
2249 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2250
2251         * doc/md.texi: Update the documentation of the cond_* optabs
2252         to mention the new final operand.  Fix GET_MODE_NUNITS call.
2253         Describe the scalar case too.
2254         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2255         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2256         instead of 2.
2257         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2258         (get_conditional_internal_fn): Update comment.
2259         * tree-vect-loop.c (vectorizable_reduction): Pass the original
2260         accumulator value as a final argument to conditional functions.
2261         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2262         a define_expand and add an "else" operand.  Assert for now that
2263         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
2264         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2265         (*cond_<optab><mode>): New patterns.
2266         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2267         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2268         (UNSPEC_COND_EOR): Delete.
2269         (optab): Remove associated mappings.
2270         (SVE_INT_BINARY): New code iterator.
2271         (sve_int_op): Remove int attribute and add "minus" to the code
2272         attribute.
2273         (SVE_COND_INT_OP): Delete.
2274         (SVE_COND_FP_OP): Rename to...
2275         (SVE_COND_FP_BINARY): ...this.
2276
2277 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
2278
2279         * optabs.c (can_reuse_operands_p): New function.
2280         (maybe_legitimize_operands): Try to reuse the results for
2281         earlier operands.
2282
2283 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
2284
2285         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2286         Add {q} suffix to insn mnemonic.
2287
2288 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2289
2290         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2291         (msp430_warn_func_return): New.
2292
2293 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
2294
2295         * fold-const.c (tree_nonzero_bits): New function.
2296         * fold-const.h (tree_nonzero_bits): Likewise.
2297         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2298         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2299
2300 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
2301
2302         PR target/85900
2303         PR target/85345
2304         * varasm.c (assemble_alias): Check ifunc_resolver only on
2305         FUNCTION_DECL.
2306
2307 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
2308
2309         PR target/85903
2310         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2311         when memory input operand is handled.
2312
2313 2018-05-24  Luis Machado  <luis.machado@linaro.org>
2314
2315         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2316         global.
2317         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2318
2319 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
2320
2321         * match.pd: Delay FMA folds until after vectorization.
2322
2323 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2324
2325         PR target/83009
2326         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2327         address check not strict.
2328
2329 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
2330
2331         * gimple-match.h (gimple_match_op): New class.
2332         (mprts_hook): Replace parameters with a gimple_match_op *.
2333         (maybe_build_generic_op): Likewise.
2334         (gimple_simplified_result_is_gimple_val): Replace parameters with
2335         a const gimple_match_op *.
2336         (gimple_simplify): Replace code_helper * and tree * parameters with
2337         a gimple_match_op * parameter.
2338         (gimple_resimplify1): Replace code_helper *, tree and tree *
2339         parameters with a gimple_match_op * parameter.
2340         (gimple_resimplify2): Likewise.
2341         (gimple_resimplify3): Likewise.
2342         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2343         parameters with a gimple_match_op * parameter.
2344         * gimple-match-head.c (gimple_simplify): Change prototypes of
2345         auto-generated functions to take a gimple_match_op * instead of
2346         separate code_helper * and tree * parameters.  Make the same
2347         change in the top-level overload and update calls to the
2348         gimple_resimplify routines.  Update calls to the auto-generated
2349         functions and to maybe_push_res_to_seq in the publicly-facing
2350         operation-specific gimple_simplify overloads.
2351         (gimple_match_op::MAX_NUM_OPS): Define.
2352         (gimple_resimplify1): Replace rcode and ops with a single res_op
2353         parameter.  Update call to gimple_simplify.
2354         (gimple_resimplify2): Likewise.
2355         (gimple_resimplify3): Likewise.
2356         (mprts_hook): Replace parameters with a gimple_match_op *.
2357         (maybe_build_generic_op): Likewise.
2358         (build_call_internal): Replace type, nargs and ops with
2359         a gimple_match_op *.
2360         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2361         with a single gimple_match_op *.  Update calls to mprts_hook,
2362         build_call_internal and gimple_simplified_result_is_gimple_val.
2363         Factor out code that is common to the tree_code and combined_fn cases.
2364         * genmatch.c (expr::gen_transform): Replace tem_code and
2365         tem_ops with a gimple_match_op called tem_op.  Update calls
2366         to the gimple_resimplify functions and maybe_push_res_to_seq.
2367         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2368         res_ops.  Update call to the gimple_resimplify functions.
2369         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2370         (decision_tree::gen): Make the functions take a gimple_match_op *
2371         called res_op instead of separate res_code and res_ops parameters.
2372         Update call accordingly.
2373         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2374         and ops with a single res_op parameter.  Update calls to
2375         maybe_build_generic_op and maybe_push_res_to_seq.
2376         (fold_stmt_1): Update calls to gimple_simplify and
2377         replace_stmt_with_simplification.
2378         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2379         and gimple_simplified_result_is_gimple_val.
2380         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2381         gimple_simplify.
2382         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2383         with a gimple_match_op *.
2384         (vn_nary_build_or_lookup): Likewise.  Update call to
2385         vn_nary_build_or_lookup_1.
2386         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2387         gimple_match_op *.  Update calls to the gimple_resimplify routines
2388         and to gimple_simplified_result_is_gimple_val.
2389         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2390         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2391         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2392         (visit_nary_op): Likewise.
2393         (visit_reference_op_load): Likewise.
2394
2395 2018-05-23  Luis Machado  <luis.machado@linaro.org>
2396
2397         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2398         modifier for printing the step amount.
2399
2400 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
2401
2402         PR target/78849
2403         * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2404         types.
2405
2406 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
2407
2408         * doc/sourcebuild.texi (Endianness): New subsubsection.
2409
2410 2018-05-23  Luis Machado  <luis.machado@linaro.org>
2411
2412         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2413         <prefetch_dynamic_strides>: New const bool field.
2414         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2415         prefetch_dynamic_strides.
2416         (exynosm1_prefetch_tune): Likewise.
2417         (thunderxt88_prefetch_tune): Likewise.
2418         (thunderx_prefetch_tune): Likewise.
2419         (thunderx2t99_prefetch_tune): Likewise.
2420         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2421         false.
2422         (aarch64_override_options_internal): Update to set
2423         PARAM_PREFETCH_DYNAMIC_STRIDES.
2424         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2425         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2426         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2427         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2428         prefetch-dynamic-strides setting.
2429
2430 2018-05-23  Luis Machado  <luis.machado@linaro.org>
2431
2432         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2433         <minimum_stride>: New const int field.
2434         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2435         minimum_stride field defaulting to -1.
2436         (exynosm1_prefetch_tune): Likewise.
2437         (thunderxt88_prefetch_tune): Likewise.
2438         (thunderx_prefetch_tune): Likewise.
2439         (thunderx2t99_prefetch_tune): Likewise.
2440         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2441         <default_opt_level>: Set to 3.
2442         (aarch64_override_options_internal): Update to set
2443         PARAM_PREFETCH_MINIMUM_STRIDE.
2444         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2445         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2446         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2447         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2448         stride is constant and is below the minimum stride threshold.
2449
2450 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2451
2452         * config/arm/arm-cpus.in (mode26): Delete.
2453         (armv4): Delete mode26 reference.
2454         * config/arm/arm.c (arm_configure_build_target): Delete use of
2455         isa_bit_mode26.
2456
2457 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
2458
2459         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2460         New insn pattern.
2461         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2462         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
2463         for non-SSE modes.
2464         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
2465         (floatunsdidf2): Ditto.
2466
2467 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
2468
2469         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2470         (fixuns_trunc<mode>si2_avx512f): Ditto.
2471         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2472         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2473         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2474
2475 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
2476
2477         PR rtl-optimization/79985
2478         * df-scan.c (df_insn_refs_collect): Remove special case for
2479         global registers and asm statements.
2480
2481 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
2482
2483         * extend.texi (Global Register Variables): Rewrite the bullet list.
2484         Note that the register is available for allocation. Note that access
2485         via inline asm must use constraints. Add note about async-signal
2486         handlers. Remove paragraph about automagic register selection.
2487
2488 2018-05-23  Richard Biener  <rguenther@suse.de>
2489
2490         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2491         of fixed offset from memset VN.
2492
2493 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
2494
2495         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2496         first_interp field.
2497         (alloc_cand_and_find_basis): Initialize first_interp field.
2498         (slsr_process_mul): Modify first_interp field.
2499         (slsr_process_add): Likewise.
2500         (slsr_process_cast): Modify first_interp field for each new
2501         interpretation.
2502         (slsr_process_copy): Likewise.
2503         (dump_candidate): Dump first_interp field.
2504         (replace_mult_candidate): Process all interpretations, not just
2505         subsequent ones.
2506         (replace_rhs_if_not_dup): Likewise.
2507         (replace_one_candidate): Likewise.
2508
2509 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
2510
2511         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2512         Add new boolean.
2513         (aarch64_needs_frame_chain): New function.
2514         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2515
2516 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
2517
2518         PR target/84882
2519         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2520         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2521         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2522         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2523         as true for strict-align.
2524         (aarch64_can_inline_p): Perform checks even when callee has no
2525         attributes to check for strict alignment.
2526         * doc/extend.texi (AArch64 Function Attributes): Document
2527         no-strict-align.
2528         * doc/invoke.texi: (AArch64 Options): Likewise.
2529
2530 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
2531
2532         PR tree-optimization/85853
2533         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2534         the handling of the root of the node to...
2535         (vect_slp_analyze_node_operations_1): ...this new function,
2536         and run the whole thing with the child nodes' def types
2537         set according to their SLP node's def type.
2538
2539 2018-05-23  Richard Biener  <rguenther@suse.de>
2540
2541         PR middle-end/85874
2542         * tree-data-ref.c (create_runtime_alias_checks): Defer
2543         and ignore overflow warnings.
2544
2545 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
2546
2547         PR tree-optimization/85822
2548         * tree-vrp.c (is_masked_range_test): Fix handling of negative
2549         constants.
2550
2551 2018-05-23  Richard Biener  <rguenther@suse.de>
2552
2553         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2554         memset constants via native_interpret_expr.
2555
2556 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
2557
2558         PR target/85345
2559         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2560         attribute.
2561         (cgraph_node::create_alias): Likewise.
2562         (cgraph_node::get_availability): Check ifunc_resolver instead
2563         of looking up ifunc attribute.
2564         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2565         * varasm.c (do_assemble_alias): Likewise.
2566         (assemble_alias): Likewise.
2567         (default_binds_local_p_3): Likewise.
2568         * cgraph.h (cgraph_node): Add ifunc_resolver.
2569         (cgraph_node::only_called_directly_or_aliased_p): Return false
2570         for IFUNC resolver.
2571         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2572         attribute.
2573         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2574         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2575         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2576         instead of looking up ifunc attribute.
2577
2578 2018-05-22  Luis Machado  <luis.machado@linaro.org>
2579
2580         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2581
2582 2018-05-22  Martin Sebor  <msebor@redhat.com>
2583
2584         PR middle-end/85359
2585         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2586         only when expasion succeeds.
2587         (expand_builtin_strcmp): Same.
2588         (expand_builtin_strncmp): Same.
2589
2590 2018-05-22  Martin Sebor  <msebor@redhat.com>
2591
2592         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2593
2594 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
2595             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2596
2597         * config/aarch64/aarch64-ldpstp.md: Replace uses of
2598         aarch64_mem_pair_operand with memory_operand and delete operand swapping
2599         code.
2600         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2601         Add check for legitimate_address.
2602         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2603         (aarch64_swap_ldrstr_operands): New.
2604         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2605         Define prototype.
2606
2607 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
2608             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2609
2610         * config/aarch64/aarch64.md: New patterns to generate stp
2611         and ldp.
2612         (store_pair_sw, store_pair_dw): New patterns to generate stp for
2613         single words and double words.
2614         (load_pair_sw, load_pair_dw): Likewise.
2615         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2616         Delete.
2617         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2618         Delete.
2619         * config/aarch64/aarch64-ldpstp.md: Modify peephole
2620         for different mode ldpstp and add peephole for merged zero stores.
2621         Likewise for loads.
2622         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2623         Add size check.
2624         (aarch64_gen_store_pair): Rename calls to match new patterns.
2625         (aarch64_gen_load_pair): Rename calls to match new patterns.
2626         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2627         (load_pair<DREG:mode><DREG2:mode>): ... This.
2628         (store_pair<mode>): Rename to...
2629         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2630         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2631         New mode iterators.
2632         (V_INT_EQUIV): Handle SImode.
2633         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2634         New predicate.
2635
2636 2018-05-22  Martin Sebor  <msebor@redhat.com>
2637
2638         PR c/85623
2639         * calls.c (maybe_warn_nonstring_arg): Use string length to set
2640         or ajust the presumed bound on an operation to avoid unnecessary
2641         warnings.
2642
2643 2018-05-22  Martin Sebor  <msebor@redhat.com>
2644
2645         PR tree-optimization/85826
2646         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2647         assuming that a DECL necesarily has a constant size.
2648
2649 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
2650
2651         PR middle-end/85862
2652         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2653
2654 2018-05-22  Richard Biener  <rguenther@suse.de>
2655
2656         PR tree-optimization/85834
2657         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2658         non-constant and non-zero memset arguments.
2659
2660 2018-05-22  Martin Liska  <mliska@suse.cz>
2661
2662         PR ipa/85607
2663         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2664
2665 2018-05-22  Richard Biener  <rguenther@suse.de>
2666
2667         PR tree-optimization/85863
2668         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2669         comparisons when vectype is specified.
2670         (vectorizable_condition): Do not specify vectype for
2671         vect_is_simple_cond when SLP vectorizing.
2672
2673 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
2674
2675         PR target/85657
2676         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2677         define __ibm128 as long double.
2678         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2679         as a distinct type when IEEE 128-bit support is enabled.
2680         (init_float128_ieee): Fix up conversions between IFmode and IEEE
2681         128-bit types to use the correct functions.
2682         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2683         convert between 128-bit floating point types that have different
2684         modes but the same representation, instead of using gen_lowpart to
2685         makean alias.
2686         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2687         KFmode.
2688         (IFKF_reg): New attributes to give the register constraints for
2689         IFmode and KFmode.
2690         (extend<mode>tf2_internal): New insns to mark an explicit
2691         conversion between 128-bit floating point types that have a
2692         different mode but share the same representation.
2693
2694 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
2695
2696         PR tree-optimization/85814
2697         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2698         a null return from get_strinfo when unsharing the next
2699         strinfo in the chain.
2700
2701 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
2702
2703         PR gcc/84923
2704         * varasm.c (weak_finish): Clean up weak_decls.
2705
2706 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2707
2708         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2709         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2710         UNSPEC_UADALP values.
2711         * config/aarch64/iterators.md (ABAL): New int iterator.
2712         (ABDL2): Likewise.
2713         (ADALP): Likewise.
2714         (sur): Add mappings for the above.
2715         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2716         New define_insn.
2717         (aarch64_<sur>abal<mode>_4): Likewise.
2718         (aarch64_<sur>adalp<mode>_3): Likewise.
2719         (<sur>sadv16qi): New define_expand.
2720
2721 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
2722
2723         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2724         (*movdf_internal): Ditto.
2725         (*rcpsf2_sse): Ditto.
2726         (*rsqrtsf2_sse): Ditto.
2727         (*sqrt<mode>2_sse): Ditto.
2728
2729 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
2730
2731         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2732         eor3q<mode>4.
2733         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2734         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2735         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2736         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2737         vbcaxq_s64): New.
2738         * config/aarch64/arm_neon.h: Likewise.
2739         * config/aarch64/iterators.md (VQ_I): New.
2740
2741 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
2742
2743         * config.gcc: Add arc/t-multilib-linux to tmake_file for
2744         arc*-*-linux*.
2745         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2746         MULTILIB_DIRNAMES
2747
2748 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
2749
2750         * config/nds32/constraints.md (S): New constraint.
2751         * config/nds32/nds32.md (call_internal): Use constraint S.
2752         (call_value_internal): Likewise.
2753         (sibcall_internal): Likewise.
2754         (sibcall_value_internal): Likewise.
2755
2756 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
2757             Chung-Ju Wu  <jasonwucj@gmail.com>
2758
2759         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2760         into consideration.
2761
2762 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
2763             Chung-Ju Wu  <jasonwucj@gmail.com>
2764
2765         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2766         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2767         (nds32_rtx_costs_impl): Simplify.
2768         (nds32_address_cost_impl): Simplify.
2769         (nds32_init_rtx_costs): New function.
2770         (nds32_rtx_costs_speed_prefer): Likewise.
2771         (nds32_rtx_costs_size_prefer): Likewise.
2772         (nds32_address_cost_speed_prefer): Likewise.
2773         (nds32_address_cost_speed_fwprop): Likewise.
2774         (nds32_address_cost_size_prefer): Likewise.
2775         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2776         * config/nds32/nds32.c (nds32_option_override): Use
2777         nds32_init_rtx_costs function.
2778
2779 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
2780
2781         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2782         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2783         (TARGET_PIPELINE_N8): Likewise.
2784         (TARGET_PIPELINE_N10): Likewise.
2785         (TARGET_PIPELINE_N13): Likewise.
2786         (TARGET_PIPELINE_GRAYWOLF): Likewise.
2787
2788 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
2789
2790         * config/nds32/nds32-fpu.md: Update copyright year.
2791
2792 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2793
2794         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2795
2796 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2797
2798         * config/nds32/nds32.c
2799         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2800         * config/nds32/nds32.opt (minline-asm-r15): New option.
2801
2802 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2803
2804         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2805         MASK_HW_ABS.
2806         * config/nds32/nds32.md (abssi2): New pattern.
2807
2808 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
2809
2810         * config/i386/i386.md (rex64namesuffix): New mode attribute.
2811         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2812         Merge insn pattern from sse_cvtsi2ss<round_name> and
2813         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2814         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2815         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2816         using SWI48 mode iterator.
2817         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2818         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2819         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2820         pattern from sse_cvttss2si<round_saeonly_name>
2821         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2822         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2823         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2824         using SWI48 mode iterator.
2825         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2826         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2827         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2828         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2829         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2830         using SWI48 mode iterator.
2831         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
2832         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
2833         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
2834         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
2835         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
2836         SWI48 mode iterator.
2837         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
2838         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
2839         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
2840         pattern from sse_cvttsd2si<round_saeonly_name>
2841         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
2842
2843 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2844
2845         * config/nds32/nds32-md-auxiliary.c
2846         (nds32_valid_smw_lwm_base_p): Refine.
2847         (nds32_output_smw_single_word): Refine.
2848         (nds32_output_smw_double_word): New.
2849         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
2850
2851 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2852
2853         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
2854         (nds32_output_stack_pop): Refine.
2855         (nds32_expand_unaligned_load): Refine.
2856         (nds32_expand_unaligned_store): Refine.
2857
2858 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
2859             Chung-Ju Wu  <jasonwucj@gmail.com>
2860
2861         * config/nds32/constants.md: Add TP_REGNUM constant.
2862         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
2863         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
2864         UNSPEC_ADD32.
2865         * config/nds32/nds32-doubleword.md: Consider flag_pic.
2866         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
2867         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
2868         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
2869         and PIC code generation.
2870         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
2871         code generation.
2872         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
2873         optimization.
2874         * config/nds32/nds32.md: Support TLS and PIC.
2875         * config/nds32/nds32.c: Support TLS and PIC.
2876         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
2877         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
2878         predicate.
2879
2880 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
2881
2882         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
2883         mode with E_ prefix.
2884
2885 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
2886             Chung-Ju Wu  <jasonwucj@gmail.com>
2887
2888         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
2889         * config/nds32/nds32-md-auxiliary.c
2890         (symbolic_reference_mentioned_p): New.
2891         (nds32_legitimize_ict_address): New.
2892         (nds32_expand_ict_move): New.
2893         (nds32_indirect_call_referenced_p): New.
2894         (nds32_symbol_binds_local_p): Delete.
2895         (nds32_long_call_p): Modify.
2896         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
2897         * config/nds32/nds32-protos.h
2898         (symbolic_reference_mentioned_p): Declare.
2899         (nds32_legitimize_ict_address): Declare.
2900         (nds32_expand_ict_move): Declare.
2901         (nds32_indirect_call_referenced_p): Declare.
2902         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2903         (nds32_relax_group): Use nds32_ict_const_p as condition.
2904         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2905         (nds32_asm_file_start): Output ict_model directive in asm code.
2906         (nds32_legitimate_address_p): Consider indirect call.
2907         (nds32_print_operand): Consider indirect call.
2908         (nds32_print_operand_address): Consider indirect call.
2909         (nds32_insert_attributes): Handle "indirect_call" attribute.
2910         (TARGET_LEGITIMATE_ADDRESS_P): Define.
2911         (TARGET_LEGITIMATE_CONSTANT_P): Define.
2912         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2913         (TARGET_DELEGITIMIZE_ADDRESS): Define.
2914         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2915         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2916         (TARGET_ICT_MODEL_SMALL): Define.
2917         (TARGET_ICT_MODEL_LARGE): Define.
2918         * config/nds32/nds32.md (movsi): Consider ict model.
2919         (call, call_value): Consider ict model.
2920         (sibcall, sibcall_value): Consider ict model.
2921         * config/nds32/nds32.opt (mict-model): New option.
2922         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2923         model.
2924
2925 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
2926             Monk Chiang  <sh.chiang04@gmail.com>
2927             Jim Wilson <jimw@sifive.com>
2928
2929         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2930         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
2931         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2932         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2933         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
2934         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2935         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2936         compute save_libcall_adjustment properly.
2937         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2938         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2939         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2940         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2941         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2942         (ABI_SPEC): Handle mabi=ilp32e.
2943         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2944         (RVE): Add RVE mask.
2945         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2946         <-march>: Add rv32e as an example.
2947
2948 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
2949
2950         PR c++/82899
2951         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2952         (intra_create_variable_infos): Handle C++ constructors.
2953
2954 2018-05-18  Martin Liska  <mliska@suse.cz>
2955
2956         * passes.def: Remove a redundant pass.
2957
2958 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
2959
2960         PR bootstrap/85838
2961         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2962
2963 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2964
2965         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2966         (ARMv4): Update.
2967         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2968         (ARMv6m): Update.
2969         (armv2, armv2a, armv3, armv3m): Delete architectures.
2970         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2971         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2972         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2973         Delete cpus.
2974         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2975         (*mulsidi3adddi): Likewise.
2976         (mulsidi3): Likewise.
2977         (*mulsidi3_nov6): Likewise.
2978         (umulsidi3): Likewise.
2979         (umulsidi3_nov6): Likewise.
2980         (umaddsidi4): Likewise.
2981         (*umulsidi3adddi): Likewise.
2982         (smulsi3_highpart): Likewise.
2983         (*smulsi3_highpart_nov6): Likewise.
2984         (umulsi3_highpart): Likewise.
2985         (*umulsi3_highpart_nov6): Likewise.
2986         * config/arm/arm.h (arm_arch3m): Delete.
2987         * config/arm/arm.c (arm_arch3m): Delete.
2988         (arm_option_override_internal): Update armv3-related comment.
2989         (arm_configure_build_target): Delete use of isa_bit_mode32.
2990         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2991         (arm_rtx_costs_internal): Delete check of arm_arch3m.
2992         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2993         (mulsa3): Likewise.
2994         (mulusa3): Likewise.
2995         * config/arm/arm-protos.h (arm_arch3m): Delete.
2996         * config/arm/arm-tables.opt: Regenerate.
2997         * config/arm/arm-tune.md: Likewise.
2998         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2999         deleted architectures.
3000
3001 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3002
3003         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3004         (armv5t, armv5te): New features.
3005         (ARMv5, ARMv5e): Delete fgroups.
3006         (ARMv5t, ARMv5te): Adjust for above changes.
3007         (ARMv6m): Likewise.
3008         (armv5, armv5e): Delete arches.
3009         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3010         arm_arch5.
3011         (*call_reg_arm): Likewise.
3012         (*call_value_reg_armv5): Likewise.
3013         (*call_value_reg_arm): Likewise.
3014         (*call_symbol): Likewise.
3015         (*call_value_symbol): Likewise.
3016         (*sibcall_insn): Likewise.
3017         (*sibcall_value_insn): Likewise.
3018         (clzsi2): Likewise.
3019         (prefetch): Likewise.
3020         (define_split and define_peephole2 dependent on arm_arch5):
3021         Likewise.
3022         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3023         arm_arch5e.
3024         (TARGET_ARM_QBIT): Likewise.
3025         (TARGET_DSP_MULTIPLY): Likewise.
3026         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3027         (arm_arch5, arm_arch5e): Delete.
3028         (arm_arch5t, arm_arch5te): Declare.
3029         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3030         (arm_arch5t): Declare.
3031         (arm_option_reconfigure_globals): Update for the above.
3032         (arm_options_perform_arch_sanity_checks): Update comment, replace
3033         use of arm_arch5 with arm_arch5t.
3034         (use_return_insn): Likewise.
3035         (arm_emit_call_insn): Likewise.
3036         (output_return_instruction): Likewise.
3037         (arm_final_prescan_insn): Likewise.
3038         (arm_coproc_builtin_available): Likewise.
3039         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3040         arm_arch5e with arm_arch5t and arm_arch5te.
3041         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3042         (arm_arch5t, arm_arch5te): Declare.
3043         * config/arm/arm-tables.opt: Regenerate.
3044         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3045         * config/arm/t-multilib: Likewise.
3046         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3047         instead of arm_arch5.
3048         (*call_reg_thumb1): Likewise.
3049         (*call_value_reg_thumb1_v5): Likewise.
3050         (*call_value_reg_thumb1): Likewise.
3051         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3052         unreachable path.
3053         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3054
3055 2018-05-18  Martin Liska  <mliska@suse.cz>
3056
3057         PR gcov-profile/84846
3058         * doc/gcov.texi: Document -t option of gcov tool.
3059
3060 2018-05-18  Martin Liska  <mliska@suse.cz>
3061
3062         PR gcov-profile/84846
3063         * gcov.c (print_usage): Add new -t option.
3064         (process_args): Handle the option.
3065         (generate_results): Use stdout as output when requested by
3066         the option.
3067
3068 2018-05-18  Martin Liska  <mliska@suse.cz>
3069
3070         PR gcov-profile/84846
3071         * coverage.c (coverage_init): Write PWD to .gcno file.
3072         * doc/gcov.texi: Document how working directory is printed.
3073         * gcov-dump.c (dump_gcov_file): Print PWD.
3074         * gcov.c (output_intermediate_file): Likewise.
3075         (read_graph_file): Read PWD string.
3076         (output_lines): Print PWD.
3077
3078 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3079
3080         PR middle-end/85817
3081         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3082         for retval and return false if all args to phi are zero.
3083
3084 2018-05-18  Richard Biener  <rguenther@suse.de>
3085
3086         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3087         method.
3088         (evrp_dom_walker::before_dom_children): Call it.
3089
3090 2018-05-18  Richard Biener  <rguenther@suse.de>
3091
3092         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3093         results when processing array refs with variable index.
3094
3095 2018-05-18  Toon Moene  <toon@moene.org>
3096
3097         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3098         directly after that of -floop-interchange. Indicate that both
3099         options are enabled by default when specifying -O3.
3100
3101 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3102
3103         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3104         iterator.  Delete separate integer-mode vec_set<mode> expander.
3105         (aarch64_simd_vec_setv2di): Delete.
3106         (vec_setv2di): Delete.
3107         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3108         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
3109         the "w, r" alternative.
3110
3111 2018-05-18  Martin Liska  <mliska@suse.cz>
3112
3113         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3114         * tree-pass.h (make_pass_lower_switch_O0): New function.
3115         * tree-switch-conversion.c (node_has_low_bound): Remove.
3116         (node_has_high_bound): Likewise.
3117         (node_is_bounded): Likewise.
3118         (class pass_lower_switch): Make it a template type and create
3119         two instances.
3120         (pass_lower_switch::execute): Add template argument.
3121         (make_pass_lower_switch): New function.
3122         (make_pass_lower_switch_O0): New function.
3123         (do_jump_if_equal): Remove.
3124         (emit_case_nodes): Simplify to just handle all 3 cases and leave
3125         all the hard work to tree optimization passes.
3126
3127 2018-05-18  Martin Liska  <mliska@suse.cz>
3128
3129         * dbgcnt.c (limit_low): Renamed from limit.
3130         (limit_high): New variable.
3131         (dbg_cnt_is_enabled): Check for upper limit.
3132         (dbg_cnt): Adjust dumping.
3133         (dbg_cnt_set_limit_by_index): Add new argument for high
3134         value.
3135         (dbg_cnt_set_limit_by_name): Likewise.
3136         (dbg_cnt_process_single_pair): Parse new format.
3137         (dbg_cnt_process_opt): Use strtok.
3138         (dbg_cnt_list_all_counters): Remove 'value' and add
3139         'limit_high'.
3140         * doc/invoke.texi: Document changes.
3141
3142 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
3143
3144         * doc/sourcebuild.texi (scalar_all_fma): Document.
3145         * tree.def (FMA_EXPR): Delete.
3146         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3147         * internal-fn.c (ternary_direct): New macro.
3148         (expand_ternary_optab_fn): Likewise.
3149         (direct_ternary_optab_supported_p): Likewise.
3150         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3151         * builtins.c (fold_builtin_fma): Delete.
3152         (fold_builtin_3): Don't call it.
3153         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3154         * expr.c (expand_expr_real_2): Likewise.
3155         * fold-const.c (operand_equal_p): Likewise.
3156         (fold_ternary_loc): Likewise.
3157         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3158         * gimple.c (DEFTREECODE): Likewise.
3159         * gimplify.c (gimplify_expr): Likewise.
3160         * optabs-tree.c (optab_for_tree_code): Likewise.
3161         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3162         * tree-eh.c (operation_could_trap_p): Likewise.
3163         (stmt_could_throw_1_p): Likewise.
3164         * tree-inline.c (estimate_operator_cost): Likewise.
3165         * tree-pretty-print.c (dump_generic_node): Likewise.
3166         (op_code_prio): Likewise.
3167         * tree-ssa-loop-im.c (stmt_cost): Likewise.
3168         * tree-ssa-operands.c (get_expr_operands): Likewise.
3169         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3170         * fold-const-call.h (fold_fma): Delete.
3171         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3172         CFN_FNMA and CFN_FNMS.
3173         (fold_fma): Delete.
3174         * genmatch.c (combined_fn): New enum.
3175         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3176         (commutative_op): New function.
3177         (commutate): Use it.  Handle more than 2 operands.
3178         (dt_operand::gen_gimple_expr): Use commutative_op.
3179         (parser::parse_expr): Allow :c to be used with non-binary
3180         operators if the commutative operand is known.
3181         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3182         CFN_FMS, CFN_FNMA and CFN_FNMS.
3183         (backprop::process_assign_use): Remove FMA_EXPR handling.
3184         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3185         (gen_hsa_fma): New function.
3186         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3187         IFN_FNMA and IFN_FNMS.
3188         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3189         * gimple-fold.h (follow_all_ssa_edges): Declare.
3190         * gimple-fold.c (follow_all_ssa_edges): New function.
3191         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3192         gimple_build interface and use follow_all_ssa_edges to fold the result.
3193         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3194         instead of checking for optabs directly.
3195         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3196         rather than FMA_EXPRs.
3197         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3198         call to IFN_FMA instead of an FMA_EXPR.
3199
3200 2018-05-17  Jim Wilson  <jimw@sifive.com>
3201
3202         * expr.c (do_tablejump): When converting index to Pmode, if we have a
3203         sign extended promoted subreg, and the range does not have the sign bit
3204         set, then do a sign extend.
3205
3206         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3207         test, check for sign extended subreg and/or constant operands, and
3208         do a sign extend in that case.
3209
3210 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
3211
3212         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3213         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3214         Add untyped.
3215         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3216         Change logics_shift_reg to logics_shift_imm.
3217         (thunderx2t99_fp_loadpair_basic): Delete.
3218         (thunderx2t99_fp_storepair_basic): Delete.
3219         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3220         (thunderx2t99_asimd_polynomial): Delete.
3221         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3222         and neon_fp_mul_d_scalar_q.
3223         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3224         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3225         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3226         (thunderx2t99_asimd_lut): Add missing tbl types.
3227         (thunderx2t99_asimd_ext): Delete.
3228         (thunderx2t99_asimd_load1_1_mult): Delete.
3229         (thunderx2t99_asimd_load1_2_mult): Delete.
3230         (thunderx2t99_asimd_load1_ldp): New.
3231         (thunderx2t99_asimd_load1): New.
3232         (thunderx2t99_asimd_load2): Add missing *load2* types.
3233         (thunderx2t99_asimd_load3): New.
3234         (thunderx2t99_asimd_load4): New.
3235         (thunderx2t99_asimd_store1_1_mult): Delete.
3236         (thunderx2t99_asimd_store1_2_mult): Delete.
3237         (thunderx2t99_asimd_store2_mult): Delete.
3238         (thunderx2t99_asimd_store2_onelane): Delete.
3239         (thunderx2t99_asimd_store_stp): New.
3240         (thunderx2t99_asimd_store1): New.
3241         (thunderx2t99_asimd_store2): New.
3242         (thunderx2t99_asimd_store3): New.
3243         (thunderx2t99_asimd_store4): New.
3244
3245 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
3246
3247         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3248         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
3249
3250 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
3251             Segher Boessenkool  <segher@kernel.crashing.org>
3252
3253         PR target/85698
3254         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3255         operand.
3256
3257 2018-05-17  Richard Biener  <rguenther@suse.de>
3258
3259         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3260         for pruning loop and prune defs feeding only already visited PHIs.
3261
3262 2018-05-17  Richard Biener  <rguenther@suse.de>
3263
3264         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3265
3266 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
3267             Richard Biener  <rguenther@suse.de>
3268
3269         PR tree-optimization/85793
3270         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3271         for VMAT_ELEMENTWISE.
3272
3273 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
3274
3275         * internal-fn.h (lookup_internal_fn): Declare
3276         * internal-fn.c (lookup_internal_fn): New function.
3277         * gimple.c (gimple_build_call_from_tree): Handle calls to
3278         internal functions.
3279         * gimple-pretty-print.c (dump_gimple_call): Print "." before
3280         internal function names.
3281         * tree-pretty-print.c (dump_generic_node): Likewise.
3282         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3283
3284 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
3285
3286         * gimple-fold.h (gimple_build): Make the function forms take
3287         combined_fn rather than built_in_function.
3288         (gimple_simplify): Likewise.
3289         * gimple-match-head.c (gimple_simplify): Likewise.
3290         * gimple-fold.c (gimple_build): Likewise.
3291         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3292         rather than gimple_build_call_internal.
3293         (get_initial_defs_for_reduction): Likewise.
3294         (vect_create_epilog_for_reduction): Likewise.
3295         (vectorizable_live_operation): Likewise.
3296
3297 2018-05-17  Martin Liska  <mliska@suse.cz>
3298
3299         * gimple-ssa-sprintf.c (format_directive): Do not use
3300         space in between 'G_' and '('.
3301
3302 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
3303
3304         PR target/85323
3305         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3306         even if the mask is not all ones.
3307
3308         PR target/85323
3309         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3310         vector.
3311         (ix86_gimple_fold_builtin): Likewise.
3312
3313         PR target/85323
3314         * config/i386/i386.c: Include tree-vector-builder.h.
3315         (ix86_vector_shift_count): New function.
3316         (ix86_fold_builtin): Fold shift builtins by scalar count.
3317         (ix86_gimple_fold_builtin): Likewise.
3318
3319         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3320         _mm512_setzero): New intrinsics.
3321
3322 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
3323             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3324
3325         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3326         code generation for cases where splatting a value is not useful.
3327         * simplify-rtx.c (simplify_ternary_operation): Simplify
3328         vec_merge across a vec_duplicate and a paradoxical subreg forming
3329         a vector mode to a vec_concat.
3330
3331 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
3332
3333         * config.gcc: Support "goldmont-plus".
3334         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3335         "goldmont-plus".
3336         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3337         PROCESSOR_GOLDMONT_PLUS.
3338         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3339         (processor_target_table): Add "goldmont-plus".
3340         (PTA_GOLDMONT_PLUS): Define.
3341         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3342         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3343         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3344         (fold_builtin_cpu): Add "goldmont-plus".
3345         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3346         (ix86_option_override_internal): Add "goldmont-plus".
3347         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3348         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3349         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3350         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3351
3352 2018-05-17  Richard Biener  <rguenther@suse.de>
3353
3354         PR tree-optimization/85757
3355         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3356         remove defs that only feed that PHI from further processing.
3357
3358 2018-05-16  Jim Wilson  <jimw@sifive.com>
3359
3360         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3361         asterisk to name.
3362         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3363
3364 2018-05-16  Mark Wielaard  <mark@klomp.org>
3365
3366         * dwarf2out.c (count_index_strings): New function.
3367         (output_indirect_strings): Call count_index_strings and generate
3368         header for dwarf_version >= 5.
3369
3370 2018-05-16  Mark Wielaard  <mark@klomp.org>
3371
3372         * dwarf2out.c (dwarf_FORM): New function.
3373         (set_indirect_string): Use dwarf_FORM.
3374         (reset_indirect_string): Likewise.
3375         (size_of_die): Likewise.
3376         (value_format): Likewise.
3377         (output_die): Likewise.
3378         (add_skeleton_AT_string): Likewise.
3379         (output_macinfo_op): Likewise.
3380         (index_string): Likewise.
3381         (output_index_string_offset): Likewise.
3382         (output_index_string): Likewise.
3383         (count_index_strings): Likewise.
3384
3385 2018-05-16  Carl Love  <cel@us.ibm.com>
3386
3387         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3388         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3389
3390 2018-05-16  Martin Jambor  <mjambor@suse.cz>
3391
3392         * ipa-prop.c (ipa_free_all_edge_args): Remove.
3393         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3394
3395 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
3396
3397         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3398         (fnma<mode>4): Likewise.
3399         (fms<mode>4): Likewise.
3400         (fnms<mode>4): Likewise.
3401         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3402         (aarch64_fnma<mode>4): Likewise.
3403         (aarch64_fms<mode>4): Likewise.
3404         (aarch64_fnms<mode>4): Likewise.
3405         (aarch64_fnmadd<mode>4): Likewise.
3406
3407 2018-05-16  Jason Merrill  <jason@redhat.com>
3408
3409         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
3410
3411 2018-05-16  Richard Biener  <rguenther@suse.de>
3412
3413         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3414         (dump_stmt_cost): Declare.
3415         (add_stmt_cost): Dump cost we add.
3416         (add_stmt_costs): New function.
3417         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3418         No longer exported.
3419         (vect_analyze_stmt): Adjust prototype.
3420         (vectorizable_condition): Likewise.
3421         (vectorizable_live_operation): Likewise.
3422         (vectorizable_reduction): Likewise.
3423         (vectorizable_induction): Likewise.
3424         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3425         cost vector to pass to vectorizable_ and record afterwards.
3426         (vect_model_reduction_cost): Take cost vector argument and adjust.
3427         (vect_model_induction_cost): Likewise.
3428         (vectorizable_reduction): Likewise.
3429         (vectorizable_induction): Likewise.
3430         (vectorizable_live_operation): Likewise.
3431         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3432         SLP_TREE_NUMBER_OF_VEC_STMTS.
3433         (vect_analyze_slp_cost_1): Remove.
3434         (vect_analyze_slp_cost): Likewise.
3435         (vect_slp_analyze_node_operations): Take visited args and
3436         a target cost vector.  Avoid processing already visited stmt sets.
3437         (vect_slp_analyze_operations): Use a local cost vector to gather
3438         costs and register those of non-discarded instances.
3439         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3440         (vect_schedule_slp_instance): Remove copying of
3441         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
3442         zero.
3443         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3444         adding cost.  Record cost entry location.
3445         (vect_prologue_cost_for_slp_op): Function to compute cost of
3446         a constant or invariant generated for SLP vect in the prologue,
3447         split out from vect_analyze_slp_cost_1.
3448         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
3449         (vect_model_promotion_demotion_cost): Likewise.
3450         (vect_model_store_cost): Likewise, make static.
3451         (vect_model_load_cost): Likewise.
3452         (vectorizable_bswap): Add cost vector arg and adjust.
3453         (vectorizable_call): Likewise.
3454         (vectorizable_simd_clone_call): Likewise.
3455         (vectorizable_conversion): Likewise.
3456         (vectorizable_assignment): Likewise.
3457         (vectorizable_shift): Likewise.
3458         (vectorizable_operation): Likewise.
3459         (vectorizable_store): Likewise.
3460         (vectorizable_load): Likewise.
3461         (vectorizable_condition): Likewise.
3462         (vectorizable_comparison): Likewise.
3463         (can_vectorize_live_stmts): Likewise.
3464         (vect_analyze_stmt): Likewise.
3465         (vect_transform_stmt): Adjust calls to vectorizable_*.
3466         * tree-vectorizer.c: Include gimple-pretty-print.h.
3467         (dump_stmt_cost): New function.
3468
3469 2018-05-16  Richard Biener  <rguenther@suse.de>
3470
3471         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3472         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3473         * tree-ssa-dse.c: Include tree-ssa-loop.h.
3474         (check_name): New callback.
3475         (dse_classify_store): Track cycles via a visited bitmap of PHI
3476         defs and simplify handling of in-loop and across loop dead stores
3477         and properly fail for loop-variant refs.  Handle byte-tracking with
3478         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3479         limiting the walk.
3480
3481 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
3482
3483         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3484         (vect_get_mask_type_for_stmt): Likewise.
3485         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3486         split out from...
3487         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
3488         to determine the statement's vector type and the vector type that
3489         should be used for calculating nunits.  Deal with cases in which
3490         the type has to be deferred.
3491         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3492         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3493         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3494         (vect_determine_vf_for_stmt): New functions, split out from...
3495         (vect_determine_vectorization_factor): ...here.
3496         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3497         (vect_get_mask_type_for_stmt): New functions, split out from
3498         vect_determine_vectorization_factor.
3499
3500 2018-05-16  Richard Biener  <rguenther@suse.de>
3501
3502         * tree-cfg.c (verify_gimple_assign_ternary): Properly
3503         verify the [VEC_]COND_EXPR embedded comparison.
3504
3505 2018-05-15  Martin Sebor  <msebor@redhat.com>
3506
3507         PR tree-optimization/85753
3508         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3509         RECORD_TYPE in addition to ARRAY_TYPE.
3510
3511 2018-05-15  Martin Sebor  <msebor@redhat.com>
3512
3513         PR middle-end/85643
3514         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3515
3516 2018-05-15  Richard Biener  <rguenther@suse.de>
3517
3518         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3519         add by_clobber_p one.  Change algorithm to collect all defs
3520         representing uses we need to walk and try reducing them to
3521         a single one before failing.
3522         (dse_dom_walker::dse_optimize_stmt): Adjust.
3523
3524 2018-05-13  Mark Wielaard  <mark@klomp.org>
3525
3526         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3527         (size_of_loc_descr): Likewise.
3528         (output_loc_operands): Likewise.
3529         (output_loc_operands_raw): Likewise.
3530         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3531         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3532         (hash_loc_operands): Likewise.
3533         (compare_loc_operands): Likewise.
3534
3535 2018-05-14  Mark Wielaard  <mark@klomp.org>
3536
3537         * dwarf2out.c (count_index_addrs): New function.
3538         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3539
3540 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3541
3542         PR tree-optimization/83648
3543         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3544         return value as malloc candidate.
3545
3546 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3547
3548         PR ipa/85734
3549         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3550         param as true in call to suggest_attribute.
3551
3552 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
3553
3554         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3555         -mreadonly-in-sdata.
3556
3557 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3558
3559         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3560         New pattern.
3561         (aarch64_crypto_aesd_fused): Likewise.
3562
3563 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
3564
3565         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3566         (movsi_aarch64): Likewise.
3567         (load_pairsi): Likewise.
3568         (load_pairdi): Likewise.
3569         (store_pairsi): Likewise.
3570         (store_pairdi): Likewise.
3571         (load_pairsf): Likewise.
3572         (load_pairdf): Likewise.
3573         (store_pairsf): Likewise.
3574         (store_pairdf): Likewise.
3575         (zero_extend): Likewise.
3576         (trunc): Swap alternatives.
3577         (fcvt_target): Add '?' to prefer w over r.
3578
3579 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
3580
3581         PR target/85756
3582         * config/i386/i386.md: Disallow non-commutative arithmetics in
3583         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3584         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
3585         in the peephole2 before it.
3586
3587 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
3588
3589         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3590         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3591         (ix86_handle_option): Handle -mcldemote.
3592         * config.gcc: New header.
3593         * config/i386/cldemoteintrin.h: New file.
3594         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3595         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3596         -mcldemote.
3597         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3598         OPTION_MASK_ISA_CLDEMOTE.
3599         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3600         (ix86_valid_target_attribute_inner_p): Ditto.
3601         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3602         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3603         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3604         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3605         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3606         (cldemote): New.
3607         * config/i386/i386.opt: Add -mcldemote.
3608         * config/i386/x86intrin.h: New header.
3609         * doc/invoke.texi: Add -mcldemote.
3610
3611 2018-05-14  Richard Biener  <rguenther@suse.de>
3612
3613         * doc/match-and-simplify.texi: Adjust :s documentation.
3614
3615 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
3616
3617         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3618         intended memcpy size.
3619         (REORDER_45): Likewise.
3620
3621 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
3622
3623         * sort.cc: New file.
3624         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3625         * vec.c (qsort_chk): Use gcc_qsort.
3626         * Makefile.in (OBJS-libcommon): Add sort.o.
3627         (build/sort.o): New target.  Use it...
3628         (BUILD_RTL): ... here, and...
3629         (build/gencfn-macros): ... here, and...
3630         (build/genmatch): ... here.
3631
3632 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
3633             Chung-Ju Wu  <jasonwucj@gmail.com>
3634
3635         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3636         * config/nds32/nds32-graywolf.md: New file.
3637         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3638         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3639         pipeline.
3640         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3641         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3642         * config/nds32/nds32.md (pipeline_model): Add graywolf.
3643         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3644         * config/nds32/pipelines.md: Include n15 settings.
3645
3646 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
3647             Chung-Ju Wu  <jasonwucj@gmail.com>
3648
3649         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3650         * config/nds32/nds32-n13.md: New file.
3651         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3652         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3653         pipeline.
3654         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3655         * config/nds32/nds32.md (pipeline_model): Add n13.
3656         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3657         * config/nds32/pipelines.md: Include n13 settings.
3658
3659 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
3660             Chung-Ju Wu  <jasonwucj@gmail.com>
3661
3662         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3663         * config/nds32/nds32-n10.md: New file.
3664         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3665         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3666         pipeline.
3667         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3668         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3669         * config/nds32/nds32.md (pipeline_model): Add n10.
3670         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3671         * config/nds32/pipelines.md: Include n10 settings.
3672
3673 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
3674             Kito Cheng  <kito.cheng@gmail.com>
3675             Chung-Ju Wu  <jasonwucj@gmail.com>
3676
3677         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3678         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3679         Add enum values for DSP extension instructions.
3680         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3681         New constraints.
3682         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3683         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3684         New code iterators.
3685         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3686         * config/nds32/nds32-dspext.md: New file for DSP implementation.
3687         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3688         * config/nds32/nds32-intrinsic.md: Likewise.
3689         * config/nds32/nds32_intrinsic.h: Likewise.
3690         * config/nds32/nds32-md-auxiliary.c: Likewise.
3691         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3692         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3693         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3694         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3695         * config/nds32/nds32-protos.h: New declarations for DSP extension.
3696         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3697         TYPE_DMAC in switch statement.
3698         * config/nds32/nds32.c: New checking and implementation for DSP
3699         extension instructions.
3700         * config/nds32/nds32.h: Likewise.
3701         * config/nds32/nds32.md: Likewise.
3702         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3703         * config/nds32/predicates.md: Implement new predicates for DSP
3704         extension.
3705
3706 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
3707
3708         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3709         Reformat alternatives and attributes so it is easier to identify
3710         which constraints/attributes go with which instruction.
3711         (mov<mode>_hardfloat32, FMOVE64): Likewise.
3712         (mov<mode>_softfloat32, FMOVE64): Likewise.
3713         (mov<mode>_hardfloat64, FMOVE64): Likewise.
3714         (mov<mode>_softfloat64, FMOVE64): Likewise.
3715
3716 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3717
3718         * doc/extend.texi (PowerPC Built-in Functions): Rename this
3719         subsection.
3720         (Basic PowerPC Built-in Functions): The new name of the
3721         subsection previously known as "PowerPC Built-in Functions".
3722         (Basic PowerPC Built-in Functions Available on all Configurations):
3723         New subsubsection.
3724         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3725         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3726         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3727         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3728
3729 2018-05-11  Martin Jambor  <mjambor@suse.cz>
3730
3731         PR ipa/85655
3732         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3733         single const.
3734
3735 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
3736
3737         PR target/85733
3738         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3739
3740 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
3741
3742         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3743         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3744         (ix86_handle_option): Handle -mwaitpkg.
3745         * config.gcc: New header.
3746         * config/i386/cpuid.h (bit_WAITPKG): New bit.
3747         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3748         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3749         function type.
3750         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3751         OPTION_MASK_ISA_WAITPKG.
3752         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3753         (ix86_option_override_internal): Add PTA_WAITPKG.
3754         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3755         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3756         IX86_BUILTIN_TPAUSE.
3757         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3758         __builtin_ia32_umwait and __builtin_ia32_tpause.
3759         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3760         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3761         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3762         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3763         UNSPECV_TPAUSE): New.
3764         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3765         * config/i386/i386.opt: Add -mwaitpkg.
3766         * config/i386/waitpkgintrin.h: New file.
3767         * config/i386/x86intrin.h: New header.
3768         * doc/invoke.texi: Add -mwaitpkg.
3769
3770 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
3771
3772         PR target/85606
3773         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3774         equivalent.
3775         (cortex-m0): Use armv6s-m isa.
3776         (cortex-m0plus): Likewise.
3777         (cortex-m1): Likewise.
3778         (cortex-m0.small-multiply): Likewise.
3779         (cortex-m0plus.small-multiply): Likewise.
3780         (cortex-m1.small-multiply): Likewise.
3781
3782 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
3783             Jakub Jelinek  <jakub@redhat.com>
3784
3785         PR tree-optimization/85692
3786         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3787         source permute as well.
3788
3789 2018-05-11  Martin Liska  <mliska@suse.cz>
3790
3791         PR sanitizer/85556
3792         * doc/extend.texi: Document LLVM style format for no_sanitize
3793         attribute.
3794
3795 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
3796
3797         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3798         mode_supports_vsx_dform_quad to mode_supports_dq_form.
3799         (mode_supports_vsx_dform_quad): Likewise.
3800         (mode_supports_vmx_dform): Move these functions to be next to the
3801         other mode_supports functions.
3802         (mode_supports_dq_form): Likewise.
3803         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3804         mode_supports_dq_form.
3805         (reg_offset_addressing_ok_p): Likewise.
3806         (offsettable_ok_by_alignment): Likewise.
3807         (rs6000_legitimate_offset_address_p): Likewise.
3808         (legitimate_lo_sum_address_p): Likewise.
3809         (rs6000_legitimize_address): Likewise.
3810         (rs6000_legitimize_reload_address): Likewise.
3811         (rs6000_secondary_reload_inner): Likewise.
3812         (rs6000_preferred_reload_class): Likewise.
3813         (rs6000_output_move_128bit): Likewise.
3814
3815 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
3816
3817         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3818         Generate SImode target register for null target.
3819         <case IX86_BUILTIN_XGETBV>: Ditto.
3820         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3821         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3822
3823 2018-05-10  Carl Love  <cel@us.ibm.com>
3824
3825         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3826         dcbtt and dcbtstt if operands[2] is 0.
3827
3828 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
3829
3830         PR target/85693
3831         * config/i386/sse.md (usadv64qi): New expander.
3832
3833 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
3834
3835         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
3836         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
3837         -maltivec=be support.
3838         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
3839         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
3840         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
3841         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
3842         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
3843         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
3844         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
3845         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
3846         altivec_vsumsws): Adjust.
3847         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
3848         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
3849         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
3850         support.
3851         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
3852         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
3853         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
3854         (altivec_lve<VI_char>x): Delete expand.
3855         (*altivec_lve<VI_char>x_internal): Rename to...
3856         (altivec_lve<VI_char>x): ... this.
3857         (altivec_lvxl_<mode>): Delete expand.
3858         (*altivec_lvxl_<mode>_internal): Rename to ...
3859         (altivec_lvxl_<mode>): ... this.
3860         (altivec_stvxl_<mode>): Delete expand.
3861         (*altivec_stvxl_<mode>_internal): Rename to ...
3862         (altivec_stvxl_<mode>): ... this.
3863         (altivec_stve<VI_char>x): Delete expand.
3864         (*altivec_stve<VI_char>x_internal): Rename to ...
3865         (altivec_stve<VI_char>x): ... this.
3866         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
3867         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
3868         reduc_plus_scal_<mode>): Adjust.
3869         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
3870         comment.
3871         (rs6000_cpu_cpp_builtins): Adjust.
3872         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
3873         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
3874         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
3875         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
3876         -maltivec=be support.
3877         (rs6000_split_vec_extract_var): Adjust.
3878         (rs6000_split_v4si_init): Adjust.
3879         (swap_selector_for_mode): Delete.
3880         (altivec_expand_lvx_be, altivec_expand_stvx_be,
3881         altivec_expand_stvex_be): Delete.
3882         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
3883         -maltivec=be support.
3884         (rs6000_gimple_fold_builtin): Ditto.
3885         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
3886         Adjust.
3887         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
3888         (TARGET_DIRECT_MOVE_64BIT): Adjust.
3889         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
3890         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
3891         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
3892         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
3893         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
3894         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
3895         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
3896         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
3897         anonymous split): Adjust.
3898         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3899         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3900
3901 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
3902
3903         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3904         when --with-gxx-include-dir is also specified.
3905         * configure: Regenerate.
3906
3907 2018-05-09  Jim Wilson  <jimw@sifive.com>
3908
3909         PR target/84797
3910         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3911         * config/riscv/t-withmultilib: New.
3912         * config/riscv/withmultilib.h: New.
3913         * doc/install.texi: Document RISC-V --with-multilib-list support.
3914
3915 2018-05-09  Richard Biener  <rguenther@suse.de>
3916
3917         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3918         vector.
3919         (vect_bb_vectorization_profitable_p): Adjust.  Compute
3920         actual scalar cost using the cost vector and the add_stmt_cost
3921         machinery.
3922
3923 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
3924
3925         PR rtl-optimization/85645
3926         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3927         in the REG_CFA_REGISTER note for LR, don't leave it empty.
3928
3929 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
3930
3931         PR rtl-optimization/85645
3932         * shrink-wrap.c (spread_components): Return a boolean saying if
3933         anything was changed.
3934         (try_shrink_wrapping_separate): Iterate spread_components until
3935         nothing changes anymore.
3936
3937 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
3938
3939         PR rtl-optimization/85645
3940         * regrename.c (build_def_use): Also kill the chains that include the
3941         destination of a REG_CFA_REGISTER note.
3942
3943 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
3944
3945         PR rtl-optimization/85645
3946         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3947         insn that has a REG_CFA_REGISTER note.
3948
3949 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
3950
3951         * cfgexpand.c (expand_clobber): New function.
3952         (expand_gimple_stmt_1): Use it.
3953         * tree-vect-stmts.c (vect_clobber_variable): New function,
3954         split out from...
3955         (vectorizable_simd_clone_call): ...here.
3956         (vectorizable_store): Emit a clobber either side of an
3957         IFN_STORE_LANES sequence.
3958         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3959
3960 2018-05-09  Tom de Vries  <tom@codesourcery.com>
3961
3962         PR target/85626
3963         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3964         (define_insn "trap_if_false"): Add exit after trap.
3965
3966 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
3967
3968         PR rtl-optimization/85638
3969         * bb-reorder.c: Include common/common-target.h.
3970         (create_forwarder_block): New function extracted from...
3971         (fix_up_crossing_landing_pad): ...here.  Rename into...
3972         (dw2_fix_up_crossing_landing_pad): ...this.
3973         (sjlj_fix_up_crossing_landing_pad): New function.
3974         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3975         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3976         from both partitions and exit the loop after one iteration.
3977
3978 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3979
3980         Revert:
3981         * doc/extend.texi (PowerPC Built-in Functions): Rename this
3982         subsection.
3983         (Basic PowerPC Built-in Functions): The new name of the
3984         subsection previously known as "PowerPC Built-in Functions".
3985         (Basic PowerPC Built-in Functions Available on all Configurations):
3986         New subsubsection.
3987         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3988         subsubsection.
3989         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3990         subsubsection.
3991         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3992         subsubsection.
3993         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3994         subsubsection.
3995
3996 2018-05-08  Jim Wilson  <jimw@sifive.com>
3997
3998         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3999         (LD_EMUL_SUFFIX): New.
4000         (LINK_SPEC): Use it.
4001
4002 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4003
4004         * doc/extend.texi (PowerPC Built-in Functions): Rename this
4005         subsection.
4006         (Basic PowerPC Built-in Functions): The new name of the
4007         subsection previously known as "PowerPC Built-in Functions".
4008         (Basic PowerPC Built-in Functions Available on all Configurations):
4009         New subsubsection.
4010         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4011         subsubsection.
4012         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4013         subsubsection.
4014         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4015         subsubsection.
4016         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4017         subsubsection.
4018
4019 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
4020
4021         PR target/85683
4022         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4023         after cmpelim optimization.
4024
4025 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
4026
4027         * config.gcc: Support "goldmont".
4028         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4029         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4030         PROCESSOR_GOLDMONT.
4031         * config/i386/i386.c (m_GOLDMONT): Define.
4032         (processor_target_table): Add "goldmont".
4033         (PTA_GOLDMONT): Define.
4034         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4035         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4036         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4037         (fold_builtin_cpu): Add "goldmont".
4038         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4039         (ix86_option_override_internal): Add "goldmont".
4040         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4041         (processor_type): Add PROCESSOR_GOLDMONT.
4042         * config/i386/i386.md: Add CPU "glm".
4043         * config/i386/glm.md: New file.
4044         * config/i386/x86-tune.def: Add m_GOLDMONT.
4045         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4046
4047 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
4048
4049         PR target/85572
4050         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4051         E_V4DImode.
4052         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4053         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
4054         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
4055
4056         PR target/85317
4057         * config/i386/i386.c (ix86_fold_builtin): Handle
4058         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4059
4060         PR target/85480
4061         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4062         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4063
4064 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
4065
4066         PR target/85658
4067         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4068         (check_arch): Likewise.
4069         (check_fpu): Return the result rather than printing it.
4070         (end arch): Fix operator precedence.
4071         (end cpu): Likewise.
4072         (END): Print the result from check_fpu.
4073
4074 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
4075             Alan Hayward  <alan.hayward@arm.com>
4076             David Sherwood  <david.sherwood@arm.com>
4077
4078         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4079         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4080         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4081         (*fcmuo<mode>_and): New patterns.
4082
4083 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
4084
4085         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4086         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4087         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4088         (cmp_op, sve_imm_con): New code attributes.
4089         (SVE_COND_INT_CMP, imm_con): Delete.
4090         (cmp_op): Remove above unspecs from int attribute.
4091         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4092         to...
4093         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
4094         comparison-specific unspecs.
4095         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4096         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4097         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4098         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4099         (*vec_fcm<cmp_op><mode>): Rename to...
4100         (*fcm<cmp_op><mode>): ...this and adjust likewise.
4101         (*vec_fcmuo<mode>): Rename to...
4102         (*fcmuo<mode>): ...this and adjust likewise.
4103         (*pred_fcm<cmp_op><mode>): New pattern.
4104         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4105         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4106         functions.
4107         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4108         and UNORDERED.
4109         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4110         (aarch64_emit_sve_predicated_cond): New function.
4111         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4112         (aarch64_emit_unspec_cond_or): Replace with...
4113         (aarch64_emit_sve_or_conds): ...this new function.  Use
4114         aarch64_emit_sve_ptrue_op for the individual comparisons and
4115         aarch64_emit_binop to OR them together.
4116         (aarch64_emit_inverted_unspec_cond): Replace with...
4117         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
4118         aarch64_emit_sve_ptrue_op for the comparison and
4119         aarch64_emit_unop to invert the result.
4120         (aarch64_expand_sve_vec_cmp_float): Update after the above
4121         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
4122
4123 2018-05-07  Nathan Sidwell  <nathan@acm.org>
4124
4125         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4126         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4127         (Backwards Compatibility): Likewise.
4128
4129 2018-05-07  Luis Machado  <luis.machado@linaro.org>
4130
4131         PR bootstrap/85681
4132         Revert:
4133         2018-05-07  Luis Machado  <luis.machado@linaro.org>
4134
4135         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4136         <prefetch_dynamic_strides>: New const bool field.
4137         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4138         prefetch_dynamic_strides.
4139         (exynosm1_prefetch_tune): Likewise.
4140         (thunderxt88_prefetch_tune): Likewise.
4141         (thunderx_prefetch_tune): Likewise.
4142         (thunderx2t99_prefetch_tune): Likewise.
4143         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4144         to false.
4145         (aarch64_override_options_internal): Update to set
4146         PARAM_PREFETCH_DYNAMIC_STRIDES.
4147         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4148         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4149         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4150         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4151         prefetch-dynamic-strides setting.
4152
4153         2018-05-07  Luis Machado  <luis.machado@linaro.org>
4154
4155         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4156         <minimum_stride>: New const int field.
4157         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4158         minimum_stride field.
4159         (exynosm1_prefetch_tune): Likewise.
4160         (thunderxt88_prefetch_tune): Likewise.
4161         (thunderx_prefetch_tune): Likewise.
4162         (thunderx2t99_prefetch_tune): Likewise.
4163         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4164         (aarch64_override_options_internal): Update to set
4165         PARAM_PREFETCH_MINIMUM_STRIDE.
4166         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4167         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4168         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4169         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4170         stride is constant and is below the minimum stride threshold.
4171
4172 2018-05-07  Luis Machado  <luis.machado@linaro.org>
4173
4174         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4175         to 512.
4176
4177 2018-05-07  Luis Machado  <luis.machado@linaro.org>
4178
4179         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4180         <prefetch_dynamic_strides>: New const bool field.
4181         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4182         prefetch_dynamic_strides.
4183         (exynosm1_prefetch_tune): Likewise.
4184         (thunderxt88_prefetch_tune): Likewise.
4185         (thunderx_prefetch_tune): Likewise.
4186         (thunderx2t99_prefetch_tune): Likewise.
4187         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4188         to false.
4189         (aarch64_override_options_internal): Update to set
4190         PARAM_PREFETCH_DYNAMIC_STRIDES.
4191         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4192         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4193         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4194         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4195         prefetch-dynamic-strides setting.
4196
4197 2018-05-07  Luis Machado  <luis.machado@linaro.org>
4198
4199         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4200         <minimum_stride>: New const int field.
4201         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4202         minimum_stride field.
4203         (exynosm1_prefetch_tune): Likewise.
4204         (thunderxt88_prefetch_tune): Likewise.
4205         (thunderx_prefetch_tune): Likewise.
4206         (thunderx2t99_prefetch_tune): Likewise.
4207         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4208         (aarch64_override_options_internal): Update to set
4209         PARAM_PREFETCH_MINIMUM_STRIDE.
4210         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4211         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4212         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4213         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4214         stride is constant and is below the minimum stride threshold.
4215
4216 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
4217
4218         PR c++/85659
4219         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4220         the type is addressable.  Don't force op into register if it has
4221         BLKmode.
4222
4223 2018-05-05  Roland McGrath  <mcgrathr@google.com>
4224
4225         PR other/77609
4226         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4227         any section for which we don't know a specific type it should have,
4228         regardless of name.  Previously this was done only for the exact
4229         names ".init_array", ".fini_array", and ".preinit_array".
4230         (default_elf_asm_named_section): Add comment about
4231         relationship with default_section_type_flags and SECTION_NOTYPE.
4232         (get_section): Don't consider it a type conflict if one side has
4233         SECTION_NOTYPE and the other doesn't, as long as neither has the
4234         SECTION_BSS et al used in the default_section_type_flags logic.
4235
4236 2018-05-05  Tom de Vries  <tom@codesourcery.com>
4237
4238         PR target/85653
4239         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4240         (workaround_barsyncs): New function.
4241         (nvptx_reorg): Use workaround_barsyncs.
4242         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4243         (define_expand "nvptx_membar_cta"): New define_expand.
4244         (define_insn "*nvptx_membar_cta"): New insn.
4245
4246 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
4247
4248         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4249         To improve optimization opportunities.
4250         * builtin-types.def: The new needed builtin types for the above.
4251
4252 2018-05-04  Richard Biener  <rguenther@suse.de>
4253
4254         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4255         * gimple-ssa-store-merging.c
4256         (imm_store_chain_info::output_merged_store): Remove redundant create,
4257         release split_store vector contents on failure.
4258         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4259         scalar stmt vector on cache hit.
4260
4261 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
4262
4263         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4264         Xilinx FP support.
4265         * config.gcc (powerpc-xilinx-eabi*): Remove.
4266         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4267         support.
4268         (fusion_addis_mem_combo_load): Ditto.
4269         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4270         FP support.
4271         (rs6000_cpu_cpp_builtins): Ditto.
4272         * config/rs6000/rs6000-linux.c
4273         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4274         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4275         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4276         support.
4277         (rs6000_setup_reg_addr_masks): Ditto.
4278         (rs6000_init_hard_regno_mode_ok): Ditto.
4279         (rs6000_option_override_internal): Ditto.
4280         (legitimate_lo_sum_address_p): Ditto.
4281         (rs6000_legitimize_address): Ditto.
4282         (rs6000_legitimize_reload_address): Ditto.
4283         (rs6000_legitimate_address_p): Ditto.
4284         (abi_v4_pass_in_fpr): Ditto.
4285         (setup_incoming_varargs): Ditto.
4286         (rs6000_gimplify_va_arg): Ditto.
4287         (rs6000_split_multireg_move): Ditto.
4288         (rs6000_savres_strategy): Ditto.
4289         (rs6000_emit_prologue_components): Ditto.
4290         (rs6000_emit_epilogue_components): Ditto.
4291         (rs6000_emit_prologue): Ditto.
4292         (rs6000_emit_epilogue): Ditto.
4293         (rs6000_elf_file_end): Ditto.
4294         (rs6000_function_value): Ditto.
4295         (rs6000_libcall_value): Ditto.
4296         * config/rs6000/rs6000.h: Ditto.
4297         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4298         (TARGET_MINMAX): ... this.  New.
4299         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4300         * config/rs6000/rs6000.md: Remove Xilinx FP support.
4301         (*movsi_internal1_single): Delete.
4302         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4303         mfpu=, mxilinx-fpu): Delete.
4304         * config/rs6000/singlefp.h: Delete.
4305         * config/rs6000/sysv4.h: Remove Xilinx FP support.
4306         * config/rs6000/t-rs6000: Ditto.
4307         * config/rs6000/t-xilinx: Delete.
4308         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4309         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4310         (VStype_simple): Delete.
4311         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4312         * config/rs6000/xfpu.h: Delete.
4313         * config/rs6000/xfpu.md: Delete.
4314         * config/rs6000/xilinx.h: Delete.
4315         * config/rs6000/xilinx.opt: Delete.
4316         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4317         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4318
4319 2018-05-04  Tom de Vries  <tom@codesourcery.com>
4320
4321         PR libgomp/85639
4322         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4323         if ignore == 0.
4324
4325 2018-05-04  Richard Biener  <rguenther@suse.de>
4326
4327         PR middle-end/85627
4328         * tree-complex.c (update_complex_assignment): We are always in SSA form.
4329         (expand_complex_div_wide): Likewise.
4330         (expand_complex_operations_1): Likewise.
4331         (expand_complex_libcall): Preserve EH info of the original stmt.
4332         (tree_lower_complex): Handle removed blocks.
4333         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4334         on complex multiplication and division libcall builtins.
4335
4336 2018-05-04  Richard Biener  <rguenther@suse.de>
4337
4338         PR middle-end/85574
4339         * fold-const.c (negate_expr_p): Restrict negation of operand
4340         zero of a division to when we know that can happen without
4341         overflow.
4342         (fold_negate_expr_1): Likewise.
4343
4344 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
4345
4346         PR libstdc++/85466
4347         * real.h (real_nextafter): Declare.
4348         * real.c (real_nextafter): New function.
4349         * fold-const-call.c (fold_const_nextafter): New function.
4350         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4351         CASE_CFN_NEXTTOWARD.
4352         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4353         even when arg1_mode is different from arg0_mode.
4354
4355 2018-05-03  Nathan Sidwell  <nathan@acm.org>
4356
4357         * doc/extend.texi (Deprecated Features): Remove
4358         -ffriend-injection.
4359         (Backwards Compatibility): Likewise.
4360         * doc/invoke.texi (C++ Language Options): Likewise.
4361         (C++ Dialect Options): Likewise.
4362
4363 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
4364
4365         PR target/85530
4366         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4367         _mm512_mask_mullox_epi64): New intrinsics.
4368
4369 2018-05-03  Tom de Vries  <tom@codesourcery.com>
4370
4371         PR testsuite/85106
4372         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4373         dump files): Add offload-tree.
4374
4375 2018-05-03  Richard Biener  <rguenther@suse.de>
4376
4377         PR tree-optimization/85615
4378         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4379         to loops not nested in BBs loop father to avoid creating multi-entry
4380         loops.
4381
4382 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4383
4384         PR tree-optimization/70291
4385         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4386         arguments.  Change return type to tree.  Emit libcall as a new
4387         statement rather than replacing existing one when inplace_p is true.
4388         (expand_complex_multiplication_components): New function.
4389         (expand_complex_multiplication): Expand floating-point complex
4390         multiplication using the above.
4391         (expand_complex_division): Rename inner_type parameter to type.
4392         Update expand_complex_libcall call-site.
4393         (expand_complex_operations_1): Update expand_complex_multiplication
4394         and expand_complex_division call-sites.
4395
4396 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
4397
4398         PR target/85582
4399         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4400         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4401         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4402         the highest significant bit of the shift count mask is clear.  In
4403         check whether and[sq]i3 is needed verify that all significant bits
4404         of the shift count other than the highest are set.
4405
4406 2018-05-02  Tom de Vries  <tom@codesourcery.com>
4407
4408         PR libgomp/82428
4409         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4410         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4411         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4412         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4413         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4414         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4415         __builtin_goacc_parlevel_size.
4416
4417 2018-05-02  Richard Biener  <rguenther@suse.de>
4418
4419         PR tree-optimization/85597
4420         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4421         do not use split vect_get_vec_defs call but call vect_get_slp_defs
4422         directly.
4423
4424 2018-05-02  Tom de Vries  <tom@codesourcery.com>
4425
4426         PR testsuite/85106
4427         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4428         dump files): Add ltrans-tree.
4429
4430 2018-05-02  Tom de Vries  <tom@codesourcery.com>
4431
4432         PR testsuite/85106
4433         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4434         dump files): Add wpa-ipa.
4435
4436 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
4437
4438         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
4439         powerpc*-*-linux*paired* target.
4440         * config/rs6000/750cl.h: Delete.
4441         * config/rs6000/paired.h: Delete.
4442         * config/rs6000/paired.md: Delete.
4443         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4444         float support.
4445         * config/rs6000/rs6000-builtin.def: Remove paired float support.
4446         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4447         comment.  Remove paired float support.
4448         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4449         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4450         VECTOR_PAIRED.
4451         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4452         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4453         declarations.
4454         * config/rs6000/rs6000.c: Remove paired float support.
4455         (paired_expand_vector_init, paired_expand_vector_move,
4456         paired_emit_vector_compare, paired_emit_vector_cond_expr,
4457         (paired_expand_lv_builtin, paired_expand_stv_builtin,
4458         paired_expand_builtin, paired_expand_predicate_builtin,
4459         paired_init_builtins): Delete.
4460         * config/rs6000/rs6000.h: Remove paired float support.
4461         * config/rs6000/rs6000.md: Remove paired float support.
4462         (move_from_CR_ov_bit): Delete.
4463         * config/rs6000/rs6000.opt (mpaired): Delete.
4464         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4465         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4466
4467 2018-05-02  Richard Biener  <rguenther@suse.de>
4468
4469         PR middle-end/85567
4470         * gimplify.c (gimplify_save_expr): When in SSA form allow
4471         SAVE_EXPRs to compute to SSA vars.
4472
4473 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
4474
4475         PR target/85582
4476         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4477         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4478         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4479         clobber operands[2], instead use a new pseudo.  Formatting fixes.
4480
4481 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
4482
4483         PR tree-optimization/85586
4484         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4485         exit early for statements in the same group if the accesses are
4486         not strided.
4487
4488 2018-05-02  Tom de Vries  <tom@codesourcery.com>
4489
4490         PR lto/85451
4491         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4492         error message.
4493
4494 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
4495
4496         PR tree-optimization/85143
4497         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4498
4499 2018-05-01  Tom de Vries  <tom@codesourcery.com>
4500
4501         PR lto/85451
4502         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4503         not found" error message.
4504
4505 2018-05-01  Tom de Vries  <tom@codesourcery.com>
4506
4507         PR other/83786
4508         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4509         * vec.c (test_ordered_remove_if): New function.
4510         (vec_c_tests): Call test_ordered_remove_if.
4511         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4512         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4513         * tree-vect-patterns.c (vect_pattern_recog_1): Use
4514         VEC_ORDERED_REMOVE_IF.
4515
4516 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4517
4518         PR tree-optimization/82665
4519         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4520         pointer subtraction where arguments come from a memchr call.
4521
4522 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
4523
4524         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4525         --push-state --as-needed and --pop-state instead of --as-needed and
4526         --no-as-needed if ld supports it.
4527         * configure: Regenerated.
4528
4529         PR web/85578
4530         * doc/install.texi2html: Replace _002d with - and _002a with * in
4531         generated html files using sed.
4532
4533 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4534
4535         PR c++/85523
4536         * gcc-rich-location.c (blank_line_before_p): New function.
4537         (use_new_line): New function.
4538         (gcc_rich_location::add_fixit_insert_formatted): New function.
4539         * gcc-rich-location.h
4540         (gcc_rich_location::add_fixit_insert_formatted): New function.
4541
4542 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4543
4544         * selftest.c (assert_streq): Rename "expected" and "actual" to
4545         "val1" and "val2".  Extend NULL-handling to cover both inputs
4546         symmetrically, while still requiring both to be non-NULL for a pass.
4547         * selftest.h (assert_streq): Rename "expected" and "actual" to
4548         "val1" and "val2".
4549         (ASSERT_EQ): Likewise.
4550         (ASSERT_EQ_AT): Likewise.
4551         (ASSERT_KNOWN_EQ): Likewise.
4552         (ASSERT_KNOWN_EQ_AT): Likewise.
4553         (ASSERT_NE): Likewise.
4554         (ASSERT_MAYBE_NE): Likewise.
4555         (ASSERT_MAYBE_NE_AT): Likewise.
4556         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
4557         the assertion to pass.
4558         (ASSERT_STREQ_AT): Likewise.
4559
4560 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
4561
4562         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4563         interaction with -pie.
4564
4565 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4566
4567         * selftest.h: Fix alphabetization of per-source-file selftest
4568         declarations.
4569
4570 2018-04-30  Jason Merrill  <jason@redhat.com>
4571
4572         PR c++/61982 - dead stores to destroyed objects.
4573         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4574         of clobber.
4575
4576 2018-04-30  Jason Merrill  <jason@redhat.com>
4577
4578         * tree.c (build_clobber): New.
4579         * tree.h: Declare it.
4580         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4581
4582 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4583
4584         * diagnostic-show-locus.c (layout::layout): Update for
4585         location_get_source_line returning a char_span.
4586         (struct char_span): Move to input.h.
4587         (struct correction): Update for fields in char_span becoming
4588         private.
4589         (struct source_line): Update for location_get_source_line
4590         returning a char_span.
4591         (layout::print_line): Likewise.
4592         * edit-context.c (edited_file::print_content): Likewise.
4593         (edited_file::print_diff_hunk): Likewise.
4594         (edited_file::print_run_of_changed_lines): Likewise.
4595         (edited_file::get_num_lines): Likewise.
4596         (edited_line::edited_line): Likewise.
4597         * final.c (asm_show_source): Likewise.
4598         * input.c (location_get_source_line): Convert return type
4599         from const char * to char_span, losing the final "line_len"
4600         param.
4601         (dump_location_info): Update for the above.
4602         (get_substring_ranges_for_loc): Likewise.  Use a char_span
4603         when handling the literal within the line.
4604         (test_reading_source_line): Update for location_get_source_line
4605         returning a char_span.
4606         * input.h (class char_span): Move here from
4607         diagnostic-show-locus.c, converting from a struct to a class.
4608         Make data members private.
4609         (char_span::operator bool): New.
4610         (char_span::length): New.
4611         (char_span::get_buffer): New.
4612         (char_span::operator[]): New.
4613         (char_span::subspan): Make const.
4614         (char_span::xstrdup): New.
4615         (location_get_source_line): Convert return type from const char *
4616         to char_span, losing the final "line_size" param.
4617
4618 2018-04-30  Jan Hubicka  <jh@suse.cz>
4619
4620         * lto-wrapper.c (ltrans_priorities): New static var.
4621         (cmp_priority): New.
4622         (run_gcc): Read priorities and if doing parallel build order
4623         the Makefile by them.
4624
4625 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
4626
4627         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4628
4629 2018-04-30  Richard Biener  <rguenther@suse.de>
4630
4631         * tree-cfg.c (verify_address): Remove base argument, add
4632         flag whether to check TREE_ADDRESSABLE and do that.
4633         (verify_expr): Remove.
4634         (verify_types_in_gimple_reference): Add pieces from verify_expr.
4635         (verify_gimple_assign_single): Likewise.
4636         (verify_gimple_switch): Likewise.
4637         (verify_expr_location_1): Dereference tp once.  Add (disabled)
4638         piece from verify_expr.
4639         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4640
4641 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
4642
4643         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4644
4645 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
4646
4647         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4648         (small_data_pattern): Likewise.
4649         (arc_rewrite_small_data): Likewise.
4650         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4651         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4652         (get_symbol_alignment): New function.
4653         (legitimate_small_data_address_p): Likewise.
4654         (legitimate_scaled_address): Update, call
4655         legitimate_small_data_address_p.
4656         (output_sdata): New static variable.
4657         (arc_print_operand): Update how we handle small data operands.
4658         (arc_print_operand_address): Likewise.
4659         (arc_legitimate_address_p): Update, use
4660         legitimate_small_data_address_p.
4661         (arc_rewrite_small_data_p): Remove.
4662         (arc_rewrite_small_data_1): Likewise.
4663         (arc_rewrite_small_data): Likewise.
4664         (small_data_pattern): Likewise.
4665         (compact_sda_memory_operand): Update to use
4666         legitimate_small_data_address_p and get_symbol_alignment.
4667         (prepare_move_operands): Don't rewite sdata pattern.
4668         (prepare_extend_operands): Remove.
4669         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4670         pattern.
4671         (zero_extendqisi2): Likewise.
4672         (zero_extendhisi2): Likewise.
4673         (extendqihi2): Likewise.
4674         (extendqisi2): Likewise.
4675         (extendhisi2): Likewise.
4676         (addsi3): Likewise.
4677         (subsi3): Likewise.
4678         (andsi3): Likewise.
4679         * config/arc/constraints.md (Usd): Change it to memory constraint.
4680
4681 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
4682
4683         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4684         as source of std instructions.
4685         * config/arc/arc.md (movsi_insn): Update pattern predicate to
4686         allow 6-bit constants as source for store instructions.
4687         (movdi_insn): Update instruction pattern to allow 6-bit constants
4688         as source for store instructions.
4689
4690 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
4691
4692         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4693
4694 2018-04-30  Nathan Sidwell  <nathan@acm.org>
4695             Sandra Loosemore <sandra@codesourcery.com>
4696
4697         * dumpfile.c (dump_open): Allow '-' for stdout.
4698         * doc/invoke.texi (Developer Options): Document dump filename
4699         determination early.  Document stdin/stdout selection.
4700
4701 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
4702
4703         Microblaze Target: PIC data text relative
4704
4705         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4706         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4707         Add declaration.
4708         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4709         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4710         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4711         New addressing mode for data-text relative position indepenedent code.
4712         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4713         'ADDRESS_SYMBOLIC_TXT_REL'.
4714         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4715         (microblaze_legitimate_pic_operand): Exclude function calls from
4716         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4717         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4718         addresses cases.
4719         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4720         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4721         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4722         for 'address + offset'.
4723         (microblaze_expand_prologue): Add new function prologue call for
4724         'r20' assignation.
4725         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4726         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4727         table in case of TARGET_PIC_DATA_TEXT_REL.
4728         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4729         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4730         Add new macros 'UNSPEC_TEXT',
4731         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4732         + exclude function calls from 'UNSPEC_PLT' in case of data text
4733         relative mode.
4734         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4735         new target hook for generating address diff vector tables in case of
4736         flag_pic.
4737         * doc/tm.texi : Regenerate.
4738         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4739         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4740         of addr diff vector generation.
4741         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4742         target hook definition.
4743         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4744         Add default function for generate_pic_addr_diff_vec -> flag_pic.
4745         * doc/invoke.texi (Add new pic option): Add new microblaze pic
4746         option for data text relative.
4747
4748 2018-04-30  Richard Biener  <rguenther@suse.de>
4749
4750         * tree-chrec.h (evolution_function_is_constant_p): Remove
4751         redundant check.
4752         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4753
4754 2018-04-30  Richard Biener  <rguenther@suse.de>
4755
4756         PR bootstrap/85571
4757         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4758
4759 2018-04-30  Richard Biener  <rguenther@suse.de>
4760
4761         PR tree-optimization/28364
4762         PR tree-optimization/85275
4763         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4764         copying first exit test.
4765
4766 2018-04-28  Mark Wielaard  <mark@klomp.org>
4767
4768         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4769         dwarf_version >= 5.
4770         (dwarf_AT): Handle DW_AT_addr_base.
4771         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4772
4773 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
4774
4775         PR target/84431
4776         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4777         (*ashl<dwi>3_doubleword_mask_1): Ditto.
4778         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4779         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4780
4781 2018-04-28  Richard Biener  <rguenther@suse.de>
4782
4783         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4784         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4785         to reflect use.  Only add interesting stmts.
4786
4787 2018-04-27  Martin Jambor  <mjambor@suse.cz>
4788
4789         PR ipa/85549
4790         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4791         the jump function allows for passing through aggregate values.
4792
4793 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
4794
4795         * input.h (in_system_header_at): Convert from macro to inline
4796         function.
4797         (from_macro_expansion_at): Likewise.
4798         (from_macro_definition_at): Likewise.
4799
4800 2018-04-27  Jeff Law  <law@redhat.com>
4801
4802         * config.gcc: Mark tile* targets as deprecated/obsolete.
4803
4804 2018-04-27  Richard Biener  <rguenther@suse.de>
4805
4806         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4807         fix for ILP32.
4808
4809 2018-04-27  Richard Biener  <rguenther@suse.de>
4810
4811         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4812
4813 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
4814
4815         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4816         with Yd constraint. Set "preferred_for_speed" attribute from
4817         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4818         with Yd constraint.
4819         (*movdi_internal): Ditto.
4820         (movti_interunit splitters): Remove
4821         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4822         (movdi_interunit splitters): Ditto.
4823         * config/i386/constraints.md (Ye): Remove.
4824         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4825
4826 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4827
4828         PR target/85512
4829         * config/aarch64/constraints.md (Usg): Limit to 31.
4830         (Usj): Limit to 63.
4831
4832 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
4833
4834         PR tree-optimization/85529
4835         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4836         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4837         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4838         zero extension or masking of the MSB bit.
4839         (optimize_range_tests): Add FIRST_BB argument, pass it through
4840         to optimize_range_tests_var_bound.
4841         (maybe_optimize_range_tests, reassociate_bb): Adjust
4842         optimize_range_tests callers.
4843
4844 2018-04-26  Richard Biener  <rguenther@suse.de>
4845             Jakub Jelinek  <jakub@redhat.com>
4846
4847         * cgraph.h (symbol_table): Just declare debug method here.
4848         * symtab.c (symbol_table::debug): Define.
4849
4850 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
4851
4852         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
4853
4854 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
4855
4856         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4857         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
4858         (*movdi_internal): Substitute Yi and Yj constraint with x
4859         and Ym and Yn constraint with y constraint.  Update "isa"
4860         attribute and set "preferred_for_speed" attribute from
4861         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
4862         (*movsi_internal): Ditto.
4863         (*movdf_internal): Ditto.
4864         (*movsf_internal): Ditto.
4865         (*zero_extendsidi2): Ditto.
4866         * config/i386/sse.md (vec_set<mode>_0): Ditto.
4867         (sse2_loadld): Ditto.
4868         (*vec_extract<ssevecmodelower>_0): Ditto.
4869         (*vec_extractv4si_0_zext_sse4): Ditto.
4870         (vec_concatv2di): Ditto.
4871         (*vec_dup<mode>): Ditto.
4872         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
4873         * config/i386/constraints.md (Yi): Remove.
4874         (Yj): Remove.
4875         (Ym): Remove.
4876         (Yn): Remove.
4877
4878 2018-04-26  Nathan Sidwell  <nathan@acm.org>
4879
4880         * dumpfile.c (dump_open): New.
4881         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
4882         (dump_finish): Detect stdio/stderr by value not name.
4883
4884 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
4885
4886         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4887
4888 2018-04-26  Tom de Vries  <tom@codesourcery.com>
4889
4890         PR target/84952
4891         * config/nvptx/nvptx.c (verify_neutering_jumps)
4892         (verify_neutering_labels): New function
4893         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
4894
4895 2018-04-26  Tom de Vries  <tom@codesourcery.com>
4896
4897         PR target/84025
4898         * config/nvptx/nvptx.c (needs_neutering_p): New function.
4899         (nvptx_single): Use needs_neutering_p to skip over insns that do not
4900         need neutering.
4901
4902 2018-04-26  Richard Biener <rguenther@suse.de>
4903             Tom de Vries  <tom@codesourcery.com>
4904
4905         PR lto/85422
4906         * lto-streamer-out.c (output_function): Fixup loops if required to match
4907         discovery done in the reader.
4908
4909 2018-04-26  Richard Biener  <rguenther@suse.de>
4910
4911         PR tree-optimization/85116
4912         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4913         have a loop exit from the single latch predecessor.  Remove
4914         case of header with just condition.
4915         (ch_base::copy_headers): Exclude infinite loops from any
4916         processing.
4917         (pass_ch::execute): Record exits.
4918
4919 2018-04-26  Richard Biener  <rguenther@suse.de>
4920
4921         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4922         prologue cost vector and pass it to vect_get_load_cost.
4923         (vect_get_peeling_costs_all_drs): Likewise.
4924         (vect_peeling_hash_get_lowest_cost): Likewise.
4925         (vect_enhance_data_refs_alignment): Likewise.
4926
4927 2018-04-26  Richard Biener  <rguenther@suse.de>
4928
4929         PR middle-end/85450
4930         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4931         checking of integer<->pointer conversions.
4932         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4933         sign-/zero-extending pointer types.
4934         (expand_omp_for_static_chunk): Likewise.
4935
4936 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
4937             Jean Lee  <xiaoyur347@gmail.com>
4938
4939         * config/mips/mips.c (mips_asan_shadow_offset): New function.
4940         (TARGET_ASAN_SHADOW_OFFSET): Define.
4941         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4942         true for -fsanitize=address.
4943
4944 2018-04-25  Mark Wielaard  <mark@klomp.org>
4945
4946         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4947         shorter ones.
4948
4949 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
4950
4951         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4952         than "alu", remove explicit "memory" and "imm_disp" attributes.
4953         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4954
4955         PR middle-end/85414
4956         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4957         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4958         gen_lowpart_no_emit.
4959
4960 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
4961
4962         PR target/85473
4963         * config/i386/i386.c (ix86_expand_builtin): Change memory
4964         operand to XI, extend p0 to Pmode.
4965         * config/i386/i386.md: Change unspec volatile and operand
4966         1 mode to XI, change operand 0 mode to P.
4967
4968 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
4969
4970         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4971         GET_MODE_MASK before any checking.
4972         (nds32_can_use_bset_p): Likewise.
4973         (nds32_can_use_btgl_p): Likewise.
4974
4975 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
4976
4977         * config/nds32/nds32-doubleword.md: New define_split pattern for
4978         illegal register number.
4979
4980 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
4981
4982         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4983
4984 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
4985
4986         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4987
4988 2018-04-25  Richard Biener  <rguenther@suse.de>
4989
4990         * lto-streamer.h (LTO_major_version): Bump to 8.
4991
4992 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
4993
4994         * BASE-VER: Set to 9.0.0.
4995
4996 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
4997
4998         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4999         in __abskf2 and __powikf2.
5000
5001 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5002
5003         PR target/85512
5004         * config/aarch64/constraints.md (Usg, Usj): New constraints.
5005         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5006         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5007         Use the above on operand 2.  Reindent.
5008         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5009
5010 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
5011
5012         PR target/85485
5013         * common/config/i386/i386-common.c (ix86_handle_option): Don't
5014         handle OPT_mcet.
5015         * config/i386/i386.opt (mcet): Removed.
5016         * doc/install.texi: Remove -mcet documentation.
5017         * doc/invoke.texi: Likewise.
5018
5019 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
5020
5021         PR target/85485
5022         * doc/install.texi: Remove -mcet from bootstrap-cet.
5023
5024 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
5025
5026         PR target/85511
5027         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5028         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5029         if TARGET_64BIT.
5030
5031         PR target/85503
5032         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5033         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5034         containing a CONST_VECTOR.
5035
5036 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
5037
5038         * doc/install.texi: Update newlib dependency for nvptx.
5039
5040 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
5041
5042         PR target/85508
5043         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5044         instead of INTVAL when shifting x left.
5045
5046 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
5047
5048         PR tree-optimization/85478
5049         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5050         vect_grouped_store_supported for single element vectors.
5051
5052 2018-04-24  Richard Biener  <rguenther@suse.de>
5053
5054         PR target/85491
5055         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5056         load cost increase to the case of non-constant step.
5057
5058 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
5059
5060         PR target/84828
5061         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5062         destination if any_malformed_asm.
5063
5064 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
5065
5066         PR middle-end/85496
5067         * expr.c (store_field): In the bitfield case, if the value comes from
5068         a function call and is returned in registers by means of a PARALLEL,
5069         do not change the mode of the temporary unless BLKmode and VOIDmode.
5070
5071 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
5072
5073         PR rtl-optimization/85423
5074         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5075         dependencies to debug insns when the previous insn is non-debug.
5076
5077 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
5078
5079         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5080         enums into a single definition.
5081         (fls): Fix predicates and printing.
5082         (seti): Likewise.
5083
5084 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
5085
5086         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5087         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5088         and short u6 immediate.
5089         (check_if_valid_sleep_operand): Remove.
5090         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5091
5092 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5093
5094         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5095         flag_always_save_lp condition.
5096         * config/nds32/nds32.opt (malways-save-lp): New option.
5097
5098 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
5099
5100         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5101         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5102         * config/nds32/nds32.h
5103         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5104         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5105
5106 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
5107
5108         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5109         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5110
5111 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
5112             Chung-Ju Wu  <jasonwucj@gmail.com>
5113
5114         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5115         Declare.
5116         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5117         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5118
5119 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5120
5121         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5122
5123 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5124
5125         * config/nds32/nds32-protos.h (nds32_data_alignment,
5126         nds32_local_alignment): Declare.
5127         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5128         nds32_local_alignment): New functions.
5129         (TARGET_CONSTANT_ALIGNMENT): Define.
5130         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5131
5132 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5133
5134         * config/nds32/nds32.c
5135         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5136         (TARGET_MODES_TIEABLE_P): Likewise.
5137
5138 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5139
5140         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5141         level Ofast and Og.
5142
5143 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
5144             Chung-Ju Wu  <jasonwucj@gmail.com>
5145
5146         * config/nds32/constants.md (unspec_volatile_element): Add enum values
5147         for unaligned access.
5148         * config/nds32/nds32-intrinsic.c: Implementation of expanding
5149         unaligned access.
5150         * config/nds32/nds32-intrinsic.md: Likewise.
5151         * config/nds32/nds32_intrinsic.h: Likewise.
5152         * config/nds32/nds32.h (nds32_builtins): Likewise.
5153         * config/nds32/nds32.opt (munaligned-access): New option.
5154         * config/nds32/nds32.c (nds32_asm_file_start): Display
5155         flag_unaligned_access status.
5156
5157 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
5158
5159         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5160         -mno-relax is present.
5161         * config/riscv/linux.h (LINK_SPEC): Ditto.
5162
5163 2018-04-20  Martin Sebor  <msebor@redhat.com>
5164
5165         PR c/85365
5166         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5167         for null pointers.
5168         (gimple_fold_builtin_stxcpy_chk): Same.
5169         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5170
5171 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
5172
5173         PR target/85456
5174         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5175         __powikf2 when long double is IEEE 128-bit.
5176
5177 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
5178
5179         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5180         step to make sure stack always aligned.
5181
5182 2018-04-20  Carl Love  <cel@us.ibm.com>
5183
5184         PR target/83402
5185         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5186         size check for arg0.
5187
5188 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
5189             Tom de Vries  <tom@codesourcery.com>
5190
5191         PR target/85445
5192         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5193         Emit insns for calls too.
5194         (nvptx_find_par): Always look for worker-level predecessor insn.
5195         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
5196         calls.
5197         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5198         (nvptx_process_pars): Propagate frames for calls.
5199
5200 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
5201
5202         PR target/85469
5203         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5204         Removed.
5205         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5206         (ix86_handle_option): Don't handle OPT_mibt.
5207         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5208         __SHSTK__.
5209         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5210         has_ibt and ibt.
5211         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5212         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5213         (ix86_target_macros): Define __CET__ with flag_cf_protection
5214         for -fcf-protection.
5215         * config/i386/i386.c (isa2_opts): Remove -mibt.
5216         * config/i386/i386.h (TARGET_IBT): Removed.
5217         (TARGET_IBT_P): Likewise.
5218         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5219         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5220         * config/i386/i386.opt (mcet): Update help message.
5221         (mshstk): Likewise.
5222         (mibt): Removed.
5223         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
5224         -mcet as an alias for -mshstk.
5225
5226 2018-04-20  Richard Biener <rguenther@suse.de>
5227
5228         PR middle-end/85475
5229         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5230         complexity by forcing a single use of the multiply operand.
5231
5232 2018-04-20  Martin Jambor  <mjambor@suse.cz>
5233
5234         ipa/85449
5235         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5236         recursion dependency to only apply to non-clones.
5237
5238 2018-04-20  Martin Jambor  <mjambor@suse.cz>
5239
5240         ipa/85447
5241         * ipa-cp.c (create_specialized_node): Check that clones of
5242         self-recursive edges exist during IPA-CP.
5243
5244 2018-04-19  Toon Moene  <toon@moene.org>
5245
5246         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5247         by -O3.
5248
5249 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
5250
5251         PR tree-optimization/85467
5252         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5253         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
5254         VECTOR_CST element to type.
5255
5256 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
5257
5258         PR target/85397
5259         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5260         * config/i386/i386.md (builtin_setjmp_setup): Removed.
5261         (builtin_longjmp): Likewise.
5262         (save_stack_nonlocal): New pattern.
5263         (restore_stack_nonlocal): Likewise.
5264
5265 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
5266
5267         PR target/85404
5268         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5269         Replace ASM_OUTPUT_LABEL with fprintf.
5270
5271 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
5272
5273         PR target/85417
5274         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5275         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5276         * config/i386/i386-c.c (ix86_target_macros_internal): Also
5277         define __IBT__ and __SHSTK__ for -fcf-protection.
5278         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5279         TARGET_IBT.
5280         (ix86_trampoline_init): Likewise.
5281         (x86_output_mi_thunk): Likewise.
5282         (ix86_notrack_prefixed_insn_p): Likewise.
5283         (ix86_option_override_internal): Don't disallow -fcf-protection.
5284         * config/i386/i386.md (rdssp<mode>): Also enable for
5285         -fcf-protection.
5286         (incssp<mode>): Likewise.
5287         (nop_endbr): Likewise.
5288         * config/i386/i386.opt (mcet): Change help message to built-in
5289         functions only.
5290         (mibt): Likewise.
5291         (mshstk): Likewise.
5292         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5293         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
5294         enable CET built-in functions.
5295
5296 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
5297
5298         * common/config/i386/i386-common.c
5299         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5300         OPTION_MASK_ISA_MOVDIRI_UNSET,
5301         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5302         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5303         * config.gcc (movdirintrin.h): New header.
5304         * config/i386/cpuid.h (bit_MOVDIRI,
5305         bit_MOVDIR64B): New bits.
5306         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5307         and -mmvodir64b.
5308         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5309         (VOID, PVOID, PCVOID)): New function types.
5310         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5311         __builtin_ia32_directstoreu_u64,
5312         __builtin_ia32_movdir64b): New builtins.
5313         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5314         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5315         and -mmovdiri.
5316         (ix86_valid_target_attribute_inner_p): Ditto.
5317         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5318         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5319         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5320         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5321         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5322         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5323         (movdiri<mode>, movdir64b_<mode>): New.
5324         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5325         * config/i386/immintrin.h: Include movdirintrin.h.
5326         * config/i386/movdirintrin.h: New file.
5327         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5328
5329 2018-04-19  Richard Biener  <rguenther@suse.de>
5330
5331         PR middle-end/85455
5332         * cfg.c (clear_bb_flags): When loop state says we have
5333         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5334
5335 2018-04-19  Richard Biener  <rguenther@suse.de>
5336
5337         PR tree-optimization/84737
5338         * tree-vect-data-refs.c (vect_copy_ref_info): New function
5339         copying restrict info.
5340         (vect_setup_realignment): Use it.
5341         * tree-vectorizer.h (vect_copy_ref_info): Declare.
5342         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5343         the first DR to all generated stores.
5344         (vectorizable_load): Likewise for loads.
5345
5346 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
5347
5348         PR tree-optimization/85446
5349         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5350         the integral and pointer types to have the same precision.
5351
5352         * doc/install.texi: Document --disable-cet being the default and
5353         --enable-cet=auto.
5354
5355 2018-04-18  Martin Liska  <mliska@suse.cz>
5356
5357         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5358         style.
5359
5360 2018-04-18  Martin Liska  <mliska@suse.cz>
5361
5362         Revert
5363         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
5364
5365         PR ipa/83983
5366         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5367         arguments if they are comparable.
5368
5369 2018-04-18  Martin Liska  <mliska@suse.cz>
5370
5371         Revert
5372         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
5373
5374         PR lto/84805
5375         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5376         incomplete types.
5377
5378 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
5379
5380         PR target/85388
5381         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5382         ENDBR after calling __morestack.
5383
5384 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
5385
5386         PR jit/85384
5387         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5388         by using gcc_base_ver to generate a gcc_driver_version, and use
5389         it when generating GCC_DRIVER_NAME.
5390         * configure: Regenerate.
5391
5392 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
5393
5394         PR target/81084
5395         * config.gcc: Obsolete powerpc*-*-*spe*.
5396
5397 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
5398
5399         PR debug/84637
5400         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5401         (stabstr_D): Change type of unum from unsigned int to
5402         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
5403         type.
5404
5405 2018-04-17  Jim Wilson  <jimw@sifive.com>
5406
5407         PR 84856
5408         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5409         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5410         Set arg_pointer_offset after using pretend_args_size.
5411
5412 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
5413
5414         PR rtl-optimization/85431
5415         * dse.c (record_store): Ignore zero width stores.
5416
5417         PR sanitizer/85230
5418         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
5419         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5420         __builtin_stack_restore rather than after it.
5421         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5422         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5423         argument instead of virtual_dynamic_stack_rtx.
5424
5425 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5426
5427         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5428         New prototype.
5429         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5430         Add note to error message to explain internal mapping of overloaded
5431         built-in function name to non-overloaded built-in function name.
5432         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5433         function.
5434
5435 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
5436
5437         PR target/85424
5438         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5439         where the inputs overlap with the output.
5440
5441 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
5442
5443         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5444         (=v, v) alternative and explicit "memory" attribute.
5445         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
5446         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5447         attributes.
5448         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5449         "sselog1" type instead of "sselog".
5450         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5451         "sselog".  Remove explicit "memory" attribute.
5452         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5453         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5454         attributes.
5455         (vec_extract_hi_v32hi): Merge all alternatives into one, use
5456         "sselog1" type instead of "sselog".  Remove explicit "memory"
5457         attribute.
5458         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5459         use "sselog1" type instead of "sselog".  Remove explicit "memory"
5460         attribute.
5461         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5462         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5463         attributes.
5464         (vec_extract_hi_v64qi): Merge all alternatives into one, use
5465         "sselog1" type instead of "sselog".  Remove explicit "memory"
5466         attribute.
5467         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5468         use "sselog1" type instead of "sselog".  Remove explicit "memory"
5469         attribute.
5470
5471         PR target/85430
5472         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5473
5474         PR middle-end/85414
5475         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5476         on a SUBREG.
5477
5478 2018-04-17  Martin Jambor  <mjambor@suse.cz>
5479
5480         PR ipa/85421
5481         * ipa-cp.c (create_specialized_node): Call
5482         expand_all_artificial_thunks if necessary.
5483
5484 2018-04-17  Martin Liska  <mliska@suse.cz>
5485
5486         PR lto/85405
5487         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5488         in message, remote space in between '_G' and '('.
5489
5490 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
5491
5492         PR target/85281
5493         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5494         avx512f_vmcmp<mode>3<round_saeonly_name>,
5495         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5496         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5497         avx512f_rndscale<mode><round_saeonly_name>,
5498         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5499         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5500         Use %<iptr>2 instead of %2 for -masm=intel.
5501         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5502         avx512f_vcvttss2usi<round_saeonly_name>,
5503         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5504         -masm=intel.
5505         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5506         avx512f_vcvttsd2usi<round_saeonly_name>,
5507         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5508         Use %q1 instead of %1 for -masm=intel.
5509         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5510         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5511         of %3 for -masm=intel.
5512         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5513         -masm=intel.
5514         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5515         -masm=intel.
5516         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5517         -masm=intel.
5518         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5519         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5520         %g1.
5521         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5522         -masm=intel.
5523         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5524         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5525         %g1 and one with %0 and %1.
5526         (avx512er_vmrcp28<mode><round_saeonly_name>,
5527         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5528         %1 for -masm=intel.
5529         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5530         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5531         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5532         of %0 and %{%4%} for -masm=intel.
5533         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5534         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5535         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5536         order of %0 and %{%5%}%{z%} for -masm=intel.
5537
5538 2018-04-17  Jan Hubicka  <jh@suse.cz>
5539
5540         PR lto/85405
5541         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5542
5543 2018-04-17  Martin Liska  <mliska@suse.cz>
5544
5545         PR ipa/85329
5546         * multiple_target.c (create_dispatcher_calls): Set apostrophes
5547         for target_clone error message.  Make default implementation
5548         clone to be a local declaration.
5549         (separate_attrs): Add new argument and check for an empty
5550         string.
5551         (expand_target_clones): Handle it.
5552         (ipa_target_clone): Make redirection just for target_clones
5553         functions.
5554
5555 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
5556             Tom de Vries  <tom@codesourcery.com>
5557
5558         PR middle-end/84955
5559         * omp-expand.c (expand_oacc_for): Add dummy false branch for
5560         tiled basic blocks without omp continue statements.
5561
5562 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
5563
5564         PR target/83660
5565         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5566         vec_extract expression as having side effects to make sure it gets
5567         a cleanup point.
5568
5569 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
5570
5571         PR target/85403
5572         * config/i386/i386.c (get_builtin_code_for_version): Check
5573         error_mark_node.
5574
5575 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
5576
5577         PR target/84331
5578         * gcc/config.gcc: Support "skylake".
5579         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5580         PROCESSOR_SKYLAKE.
5581         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5582         (processor_target_table): Add "skylake".
5583         (ix86_option_override_internal): Add "skylake".
5584         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5585         PROCESSOR_CANNONLAKE.
5586         (get_builtin_code_for_version): Fix priority for
5587         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5588         PROCESSOR_SKYLAKE-AVX512.
5589         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5590         (processor_type): Add PROCESSOR_SKYLAKE.
5591
5592 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
5593             Jason Merrill  <jason@redhat.com>
5594
5595         PR c++/85112
5596         * convert.c (convert_to_integer_1): Use direct recursion for
5597         enumeral types and types with a precision less than the number
5598         of bits in their mode.
5599
5600 2018-04-16  Julia Koval  <julia.koval@intel.com>
5601
5602         PR target/84413
5603         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5604         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5605
5606 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
5607
5608         PR target/85293
5609         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5610         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5611         and -mno-direct-move.
5612
5613 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
5614
5615         PR target/83402
5616         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5617         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5618         Ensure negative shifts result in {0}.
5619
5620 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
5621
5622         PR rtl-optimization/79916
5623         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5624         regs (if any) to define how to gnerate SD moves when LRA is in
5625         progress.
5626
5627 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
5628
5629         PR rtl-optimization/85393
5630         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5631         * except.c (expand_dw2_landing_pad_for_region): Make static.
5632         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5633         a label and unconditional jump to old_bb, rather than
5634         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5635         basic block.
5636
5637         PR rtl-optimization/85376
5638         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5639         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5640         instead of a specific value.
5641
5642 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
5643             Bin Cheng  <bin.cheng@arm.com>
5644
5645         PR tree-optimization/82965
5646         PR tree-optimization/83991
5647         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5648         by_profile_only parameter.
5649         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5650         information if the loop was predicted to iterate too many times.
5651         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5652
5653 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
5654
5655         PR lto/71991
5656         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5657         always inline.
5658
5659 2018-04-13  Martin Liska  <mliska@suse.cz>
5660             Jakub Jelinek  <jakub@redhat.com>
5661
5662         PR middle-end/81657
5663         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5664         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5665         * builtins.c (expand_builtin_memory_copy_args): Use
5666         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5667         handle dest_addr == pc_rtx.
5668
5669 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
5670
5671         PR target/85291
5672         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5673         asked to not generate direct moves.
5674         (fix_trunc<mode>si2_stfiwx): Similar.
5675         (fix_trunc<mode>si2_internal): Similar.
5676
5677 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
5678
5679         PR debug/83157
5680         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5681         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5682         lookup if dest in some wider mode is known to be const0_rtx and
5683         if so, record permanent equivalence for it to be ZERO_EXTEND of
5684         the narrower mode destination.
5685
5686 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
5687
5688         * lto-streamer-out.c (output_function): Revert 259346.
5689         * omp-expand.c (expand_oacc_for): Likewise.
5690
5691 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
5692
5693         PR rtl-optimization/85354
5694         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5695         * sel-sched.c (sel_global_init): ... here.
5696
5697 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
5698
5699         PR target/85238
5700         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5701         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5702         mode for PE-COFF targets.
5703         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5704         (i386_pe_asm_lto_end): Likewise.
5705         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5706         (TARGET_ASM_LTO_END): Likewise.
5707         * config/i386/winnt.c (saved_debug_info_level): New static variable.
5708         (i386_pe_asm_lto_start): New function.
5709         (i386_pe_asm_lto_end): Likewise.
5710
5711 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
5712             Richard Biener  <rguenther@suse.de>
5713
5714         PR middle-end/84955
5715         * lto-streamer-out.c (output_function): Fix CFG loop state before
5716         streaming out.
5717         * omp-expand.c (expand_oacc_for): Handle calls to internal
5718         functions like regular functions.
5719
5720 2018-04-12  Richard Biener  <rguenther@suse.de>
5721
5722         PR lto/85371
5723         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5724         for the early LTO debug to properly generate references to it
5725         during DIE emission.  Do not re-use that for the skeleton for
5726         split-dwarf.
5727         (dwarf2out_early_finish): Likewise.
5728
5729 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
5730
5731         PR target/85328
5732         * config/i386/sse.md
5733         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5734         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5735         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5736         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5737         and output is a reg, avoid creating invalid lowpart subreg, but
5738         instead split into a 512-bit move.  Don't split if not AVX512VL,
5739         input is xmm16+ reg and output is a mem.
5740         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5741         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5742         xmm16+ reg and output is a mem.
5743
5744 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5745
5746         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5747         also for flag_dwarf2_cfi_asm.
5748
5749 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
5750
5751         PR rtl-optimization/85342
5752         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5753         a bool scalar var inside of the loop instead.  Don't try to update
5754         recog_data.operand after failed apply_change_group.
5755
5756 2018-04-12  Tom de Vries  <tom@codesourcery.com>
5757
5758         PR target/85296
5759         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5760         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
5761         array with flexible array member as array without given dimension.
5762         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5763         argument for undefined param to true.
5764
5765 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
5766
5767         PR target/85321
5768         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5769         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5770         from PowerPC section.
5771         * config/rs6000/sysv4.opt (mcall-): Improve help text.
5772         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5773         help text that is too long.
5774         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5775         help text that is too long.
5776         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5777         help text that is too long.
5778
5779 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
5780
5781         * config/alpha/alpha.md (stack_probe_internal): Rename
5782         from "probe_stack".  Update all callers.
5783
5784 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
5785
5786         PR rtl-optimization/84566
5787         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5788         sched_macro_fuse_insns.
5789
5790 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
5791
5792         PR target/84301
5793         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5794         (compute_block_dependences): ... from here.
5795
5796 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
5797
5798         PR tree-optimization/85331
5799         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5800         from int to HOST_WIDE_INT.
5801
5802 2018-04-11  Martin Jambor  <mjambor@suse.cz>
5803
5804         PR ipa/84149
5805         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5806         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5807         not the same as the source val.
5808         (cgraph_edge_brings_value_p): New parameter.
5809         (gather_edges_for_value): Pass destination value to
5810         cgraph_edge_brings_value_p.
5811         (perhaps_add_new_callers): Likewise.
5812         (get_info_about_necessary_edges): Likewise and exclude values brought
5813         only by self-recursive edges.
5814         (create_specialized_node): Redirect only clones of self-calling edges.
5815         (+self_recursive_pass_through_p): New function.
5816         (find_more_scalar_values_for_callers_subset): Use it.
5817         (find_aggregate_values_for_callers_subset): Likewise.
5818         (known_aggs_to_agg_replacement_list): Removed.
5819         (decide_whether_version_node): Re-calculate known constants for all
5820         remaining context clones.
5821
5822 2018-04-11  Richard Biener  <rguenther@suse.de>
5823
5824         PR lto/85339
5825         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5826         from early DWARF output.
5827         (dwarf2out_early_finish): Output line info unconditionally into
5828         early DWARF and add reference to it.
5829
5830 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
5831
5832         PR target/85281
5833         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5834         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5835         other than V2DFmode using iptr mode attribute.
5836         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5837
5838 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
5839
5840         PR rtl-optimization/84659
5841         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5842
5843 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
5844
5845         PR debug/85302
5846         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5847         SIZEP is NULL.
5848         (output_loc_list): Pass address of a dummy size variable even in the
5849         locview handling loop.
5850         (index_location_lists): Add comment on why skip_loc_list_entry can't
5851         call size_of_locs.
5852
5853 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
5854
5855         PR target/85261
5856         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5857         into register.
5858
5859 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
5860
5861         PR target/85321
5862         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5863         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5864         and -mstring-compare-inline-limit.
5865
5866 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
5867
5868         PR target/85287
5869         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5870         for stack clash protection in a register whenever we need it to be in
5871         a register.
5872
5873 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
5874
5875         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5876         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5877
5878 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
5879
5880         PR target/85321
5881         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5882         the help text.
5883         (mlong-double-): Ditto.
5884         * config/rs6000/sysv4.opt (msdata=): Ditto.
5885         (mtls-size=): Ditto.
5886
5887 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5888
5889         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5890         erroneous entries for
5891         "vector int vec_ldl (int, long int *)", and
5892         "vector unsigned int vec_ldl (int, unsigned long int *)".
5893         Add comments and entries for
5894         "vector bool char vec_ldl (int, bool char *)",
5895         "vector bool short vec_ldl (int, bool short *)",
5896         "vector bool int vec_ldl (int, bool int *)",
5897         "vector bool long long vec_ldl (int, bool long long *)",
5898         "vector pixel vec_ldl (int, pixel *)",
5899         "vector long long vec_ldl (int, long long *)",
5900         "vector unsigned long long vec_ldl (int, unsigned long long *)".
5901         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5902         type tree bool_long_long_type_node and correct definition of
5903         bool_V2DI_type_node to make reference to this new type tree.
5904         (rs6000_mangle_type): Replace erroneous reference to
5905         bool_long_type_node with bool_long_long_type_node.
5906         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5907         comments to emphasize sign distinctions for char and int types and
5908         replace RS6000_BTI_bool_long constant with
5909         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
5910         use of RS6000_BTI_pixel.
5911         (bool_long_type_node): Remove this macro definition.
5912         (bool_long_long_type_node): New macro definition
5913
5914 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
5915
5916         PR rtl-optimization/85300
5917         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5918         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5919         simplify_unary_operation fails.
5920
5921 2018-04-10  Martin Liska  <mliska@suse.cz>
5922
5923         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5924         cgraph_edge and ipa_ref.
5925
5926 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
5927
5928         PR target/85177
5929         PR target/85255
5930         * config/i386/sse.md
5931         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5932         computation of the VEC_MERGE selector from mask.
5933         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5934         Fix decoding of the VEC_MERGE selector into mask.
5935
5936 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
5937
5938         PR tree-optimization/85286
5939         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5940
5941 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
5942
5943         * final.c (final_1): Set insn_last_address as well as
5944         insn_current_address.
5945
5946 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5947
5948         PR target/85173
5949         * explow.c (emit_stack_probe): Call validize_mem on memory location
5950         before passing it to gen_probe_stack.  Create address operand and
5951         legitimize it for the probe_stack_address case.
5952
5953 2018-04-09  Jan Hubicka  <jh@suse.cz>
5954
5955         PR lto/85078
5956         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5957         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5958         * tree.c (free_lang_data_in_type): Fix handling of binfos;
5959         walk basetypes.
5960         (free_lang_data): Rebuild type inheritance graph.
5961
5962 2018-04-09  Martin Sebor  <msebor@redhat.com>
5963
5964         * invoke.texi (-finline-small-functions): Mention other optimization
5965         options.
5966         (-findirect-inlining, -fpartial-inlining): Same.
5967         (-finline-functions-called-once): Same.
5968         (-freorder-blocks-and-partition): Same.
5969
5970 2018-04-09  Jan Hubicka  <jh@suse.cz>
5971
5972         PR rtl/84058
5973         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5974         jumps; choose last target that matches the criteria (i.e.
5975         no partition changes for non-crossing jumps).
5976         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5977         support for redirecting crossing jumps to non-crossing.
5978
5979 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
5980
5981         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5982         also for naked functions.
5983
5984 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
5985
5986         * config/arc/arc.md (add_shift): New pattern.
5987         (add_shift2): Likewise.
5988         (sub_shift): Likewise.
5989         (sub_shift_cmp0_noout): Likewise.
5990         (compare_si_ashiftsi): Likewise.
5991         (xbfu_cmp0_noout): New combine pattern.
5992         (xbfu_cmp0"): Likewise.
5993         (movsi_set_cc_insn): Place the predicable variant first.
5994         (commutative_binary_cmp0_noout): Remove clobber.
5995         (commutative_binary_cmp0): New pattern.
5996         (noncommutative_binary_cmp0): Likewise.
5997         (noncommutative_binary_cmp0_noout): Likewise.
5998         (noncommutative_binary_comparison_result_used): Removed.
5999         (rsub_cmp0): New pattern.
6000         (rsub_cmp0_noout): Likewise.
6001         (extzvsi): Changed, keep only meaningful variants.
6002         (SQH, SEZ): New iterators.
6003         (SQH_postfix): New mode attribute.
6004         (SEZ_prefix): New code attribute.
6005         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6006         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6007         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6008         of numerical value.
6009         (noncommutative_operator): Check the availability of barrel
6010         shifter option.
6011
6012 2018-04-09  Richard Biener  <rguenther@suse.de>
6013
6014         PR tree-optimization/85284
6015         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6016         Only use the niter constraining form of simple_iv when the exit
6017         is always executed.
6018
6019 2018-04-09  Tom de Vries  <tom@codesourcery.com>
6020
6021         PR target/84041
6022         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6023         (define_expand "*memory_barrier"): New define_expand.
6024         (define_insn "memory_barrier"): New insn.
6025
6026 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
6027
6028         PR rtl-optimization/80463
6029         PR rtl-optimization/83972
6030         PR rtl-optimization/83480
6031
6032         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6033         correct producer for the insn.
6034         (tidy_control_flow): Fixup seqnos in case of debug insns.
6035
6036 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
6037
6038         PR rtl-optimization/83913
6039
6040         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6041         different sched-times when merging exprs.
6042
6043 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
6044
6045         PR rtl-optimization/83962
6046
6047         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6048         tidy_fallthru_edge and tidy_control_flow.
6049
6050 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
6051
6052         PR rtl-optimization/83530
6053
6054         * sel-sched.c (force_next_insn): New global variable.
6055         (remove_insn_for_debug): When force_next_insn is true, also leave only
6056         next insn in the ready list.
6057         (sel_sched_region): When the region wasn't scheduled, make another pass
6058         over it with force_next_insn set to 1.
6059
6060 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
6061
6062         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6063         into tm_file.
6064         * config/nds32/constants.md (unspec_volatile_element): Add enum values
6065         for interrupt control.
6066         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6067         functions for interrupt control.
6068         * config/nds32/nds32-intrinsic.md: Likewise.
6069         * config/nds32/nds32_intrinsic.h: Likewise.
6070         * config/nds32/nds32.h (nds32_builtins): Likewise.
6071
6072 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
6073
6074         * config/nds32/nds32.c (nds32_init_machine_status,
6075         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6076         strict_aligned_p field.
6077         (nds32_expand_to_rtl_hook): New function.
6078         (TARGET_EXPAND_TO_RTL_HOOK): Define.
6079         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6080
6081 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
6082             Chung-Ju Wu  <jasonwucj@gmail.com>
6083
6084         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6085         * config/nds32/nds32-n7.md: New file.
6086         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6087         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6088         pipeline.
6089         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6090         * config/nds32/nds32.md (pipeline_model): Add n7.
6091         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6092         * config/nds32/pipelines.md: Include n7 settings.
6093
6094 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
6095             Chung-Ju Wu  <jasonwucj@gmail.com>
6096
6097         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6098         * config/nds32/nds32-e8.md: New file.
6099         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6100         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6101         pipeline.
6102         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6103         * config/nds32/nds32.md (pipeline_model): Add e8.
6104         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6105         * config/nds32/pipelines.md: Include e8 settings.
6106
6107 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
6108             Chung-Ju Wu  <jasonwucj@gmail.com>
6109
6110         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6111         * config/nds32/nds32-n8.md: New file.
6112         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6113         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6114         pipeline.
6115         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6116         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6117         * config/nds32/nds32.md (pipeline_model): Add n8.
6118         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6119         * config/nds32/pipelines.md: Include n8 settings.
6120
6121 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
6122             Chung-Ju Wu  <jasonwucj@gmail.com>
6123
6124         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6125         * config/nds32/nds32-n9-2r1w.md: New file.
6126         * config/nds32/nds32-n9-3r2w.md: New file.
6127         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6128         nds32_register_ports): New or modify for cpu n9.
6129         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6130         pipeline.
6131         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6132         * config/nds32/nds32-utils.c: New file.
6133         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6134         TARGET_MUL_SLOW): Define.
6135         * config/nds32/nds32.md (pipeline_model): New attribute.
6136         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6137         New options that support cpu n9.
6138         * config/nds32/pipelines.md: Include n9 settings.
6139         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6140
6141 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
6142
6143         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6144         information if necessary.
6145         (output_cond_branch_compare_zero): Likewise.
6146         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6147         (nds32_target_alignment): Refine for alignment.
6148         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6149         (FUNCTION_BOUNDARY): Modify.
6150         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6151         align case.
6152         * config/nds32/nds32.opt (malways-align, malign-functions): New.
6153
6154 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
6155
6156         * config/nds32/constants.md (unspec_volatile_element): Add values for
6157         TLB operation and data prefetch.
6158         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6159         functions for TLB operation and data prefetch.
6160         * config/nds32/nds32-intrinsic.md: Likewise.
6161         * config/nds32/nds32_intrinsic.h: Likewise.
6162         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6163         (nds32_print_operand): Likewise.
6164         * config/nds32/nds32.h (nds32_builtins): Likewise.
6165
6166 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
6167         Andrew Pinski <pinsika@gcc.gnu.org>
6168
6169         PR middle-end/82976
6170         * match.pd: Use constant_boolean_node of correct type instead of
6171         boolean_true_node or boolean_false_node for simplifying
6172         pointer comparisons to zero.
6173
6174 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
6175
6176         PR tree-optimization/80021
6177         * tree.c (verify_type_variant): Make error call in verify_variant_match
6178         translatable and remove final full stop.
6179
6180 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
6181
6182         * config/nds32/constants.md (unspec_volatile_element): Add
6183         UNSPEC_VOLATILE_EH_RETURN.
6184         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6185         nds32_output_stack_pop): Support dwarf exception handling process.
6186         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6187         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6188         exception handling process.
6189         (nds32_compute_stack_frame): Likewise.
6190         (nds32_return_addr_rtx): Likewise.
6191         (nds32_initial_elimination_offset): Likewise.
6192         (nds32_expand_prologue): Likewise.
6193         (nds32_expand_epilogue): Likewise.
6194         (nds32_dynamic_chain_address): New function.
6195         * config/nds32/nds32.h (machine_function): Add fields for dwarf
6196         exception handling.
6197         (DYNAMIC_CHAIN_ADDRESS): Define.
6198         (EH_RETURN_DATA_REGNO): Define.
6199         (EH_RETURN_STACKADJ_RTX): Define.
6200         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6201         patterns for dwarf exception handling.
6202
6203 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
6204
6205         * config/nds32/nds32.h: Clean up obsolete macros.
6206
6207 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6208
6209         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6210         Add enum values for particular instructions.
6211         * config/nds32/nds32-intrinsic.c: Implementation of expanding
6212         particular intrinsic functions.
6213         * config/nds32/nds32-intrinsic.md: Likewise.
6214         * config/nds32/nds32_intrinsic.h: Likewise.
6215         * config/nds32/nds32.h (nds32_builtins): Likewise.
6216         * config/nds32/nds32.md (type): Add pbsad and pbsada.
6217         (btst, ave): New patterns for particular instructions.
6218
6219 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6220
6221         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6222         Add enum values for atomic load/store and memory sync.
6223         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6224         and memory sync.
6225         * config/nds32/nds32-intrinsic.md: Likewise.
6226         * config/nds32/nds32_intrinsic.h: Likewise.
6227         * config/nds32/nds32.h (nds32_builtins): Likewise.
6228
6229 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
6230
6231         PR tree-optimization/85257
6232         * fold-const.c (native_encode_vector): If not all elts could fit
6233         and off is -1, return 0 rather than offset.
6234         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6235         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6236         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
6237         adjust buffer in native_interpret_expr call.
6238
6239 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6240
6241         * config/nds32/constants.md (unspec_volatile_element): Add cache
6242         control enum values.
6243         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6244         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6245         * config/nds32/nds32.c (nds32_cctl_names): New.
6246         (nds32_print_operand): Handle cache control register names.
6247         * config/nds32/nds32.h (nds32_builtins): New enum values.
6248         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6249         macros.
6250         * config/nds32/nds32.md (type): Add mmu.
6251         * config/nds32/pipelines.md (simple_insn): Add mmu.
6252
6253 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
6254
6255         * config/nds32/nds32.md (type): Remove call.
6256         * config/nds32/pipelines.md (simple_insn): Likewise.
6257
6258 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6259
6260         * config/nds32/constants.md (unspec_volatile_element): Add
6261         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6262         UNSPEC_VOLATILE_FMFCFG.
6263         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6264         description for fmfcfg and fmfcsr.
6265         (bdesc_1arg): Add fmtcsr.
6266         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6267         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6268         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6269         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6270         unspec_fmfcfg): New patterns.
6271         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6272         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6273         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6274         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6275         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6276         __nds32__fmfcfg): Define.
6277
6278 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
6279
6280         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6281         intrinsic register names.
6282         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6283         intrinsic register enum values and macros.
6284
6285 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
6286
6287         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6288         for load/store addressing form.
6289         (nds32_print_operand_address): Likewise.
6290
6291 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
6292
6293         PR target/85196
6294         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6295         based on LABEL_REF.  Remove useless assertion.
6296         (pic_address_needs_scratch): Fix formatting.
6297         (sparc_legitimize_pic_address): Minor tweaks.
6298         (sparc_delegitimize_address): Adjust assertion accordingly.
6299         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6300         into symbolic_operand.
6301         (movsi_high_pic_label_ref): Likewise.
6302         (movsi_lo_sum_pic_label_ref): Likewise.
6303         (movdi_pic_label_ref): Likewise.
6304         (movdi_high_pic_label_ref): Likewise.
6305         (movdi_lo_sum_pic_label_ref): Likewise.
6306
6307 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
6308
6309         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6310         custom LIB_SPEC setup.
6311
6312 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
6313             Kito Cheng  <kito.cheng@gmail.com>
6314
6315         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6316         * config/riscv/freebsd.h: New.
6317
6318 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
6319
6320         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6321         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6322         file.
6323
6324 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
6325             Kito Cheng  <kito.cheng@gmail.com>
6326
6327         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6328         nds32_output_call, nds32_symbol_binds_local_p): New functions.
6329         * config/nds32/nds32-protos.h (nds32_output_call,
6330         nds32_output_return): Declare.
6331         * config/nds32/nds32.md: Refine all the call and return patterns.
6332
6333 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
6334
6335         PR debug/85252
6336         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6337         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6338
6339         PR rtl-optimization/84872
6340         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6341         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6342         EDGE_CROSSING edge.
6343
6344 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
6345
6346         * expr.c (copy_blkmode_to_reg): Revert 254862.
6347         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6348
6349 2018-04-06  Richard Biener  <rguenther@suse.de>
6350
6351         PR middle-end/85244
6352         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6353         after seeing a component reference with an adjacent field.  Treat
6354         refs to arrays at struct end of external decls similar to
6355         refs to unconstrained commons.
6356
6357 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
6358
6359         PR sanitizer/85213
6360         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6361         look through SAVE_EXPRs with non-side-effects argument.  Adjust
6362         recursive calls.
6363         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6364         save_p here.
6365
6366 2018-04-06  Richard Biener  <rguenther@suse.de>
6367
6368         PR middle-end/85180
6369         * alias.c (find_base_term): New wrapper around find_base_term
6370         unwinding CSELIB_VAL_PTR changes.
6371         (find_base_term): Do not restore CSELIB_VAL_PTR during the
6372         recursion.
6373
6374 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
6375
6376         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6377         instructions.
6378         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6379         constant definitions.
6380         ("nop"): lr 0,0 -> nopr r0
6381         ("nop_lr0", "nop_lr1"): New insn definitions.
6382
6383 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
6384
6385         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6386         NDS32_V3PUSH_AVAILABLE_P macro.
6387
6388 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
6389             Chung-Ju Wu  <jasonwucj@gmail.com>
6390
6391         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6392         (nds32*-*-*): Add float and fpu_config into supported_defaults.
6393         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6394         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6395         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6396         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6397         * config/nds32/constraints.md: New constraints and checking for hard
6398         float configuration.
6399         * config/nds32/iterators.md: New mode iterator and attribute for hard
6400         float configuration.
6401         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6402         patterns.
6403         * config/nds32/nds32-fpu.md: New file.
6404         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6405         deal with hard float code generation.
6406         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6407         ARCH_V3S.
6408         (abi_type, float_reg_number): New enum type.
6409         * config/nds32/nds32-predicates.c: New predicates for hard float.
6410         * config/nds32/nds32-protos.h: Declare functions for hard float.
6411         * config/nds32/nds32.c: Implementation for hard float configuration.
6412         * config/nds32/nds32.h: Definitions for hard float configuration.
6413         * config/nds32/nds32.md: Include hard float machine description and
6414         modify patterns for hard float configuration.
6415         * config/nds32/nds32.opt: New options for hard float configuration.
6416         * config/nds32/predicates.md: New predicates for hard float
6417         configuration.
6418
6419 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
6420
6421         * common/config/nds32/nds32-common.c
6422         (nds32_option_optimization_table): Enable -mreleax-hint by default.
6423
6424 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
6425
6426         PR middle-end/85195
6427         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6428         CONSTRUCTOR_ELT (ctor, ...)->value.
6429
6430 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
6431
6432         PR target/85193
6433         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6434
6435 2018-04-05  Tom de Vries  <tom@codesourcery.com>
6436
6437         PR target/85204
6438         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6439         cond jump.
6440
6441 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
6442             Kito Cheng  <kito.cheng@gmail.com>
6443
6444         * config/nds32/constraints.md (U33): Fine-tune checking condition.
6445         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6446         * config/nds32/nds32.h (nds32_16bit_address_type): Add
6447         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6448
6449 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
6450             Kito Cheng  <kito.cheng@gmail.com>
6451
6452         * config/nds32/constraints.md (Ufe): New memory constraint.
6453         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6454         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6455         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6456         operands.
6457         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6458         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6459
6460 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6461
6462         * config/nds32/nds32.md: Use optimize_size in the condition for
6463         alu-shift instructions.
6464
6465 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6466
6467         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6468
6469 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6470
6471         * config/nds32/nds32.md (negsi2): Refine pattern.
6472
6473 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
6474             Chung-Ju Wu  <jasonwucj@gmail.com>
6475
6476         * config/nds32/iterators.md (shift_rotate): New code iterator.
6477         (shift): New code attribute.
6478         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6479         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6480         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6481         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6482         bit-wise operations.
6483         (andsi3, *andsi3): Ditto.
6484         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6485         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6486         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6487         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6488         nds32_ior_operand, nds32_xor_operand): New predicates.
6489
6490 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6491
6492         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6493         (addsi3, subsi3): ... this.
6494
6495 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6496
6497         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6498
6499 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6500
6501         * config/nds32/nds32.md: Adjust indention.
6502
6503 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
6504
6505         * config/nds32/nds32.md (feature): New attribute.
6506
6507 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
6508
6509         * config/nds32/nds32.md (subtype): New attribute.
6510
6511 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6512
6513         PR target/85203
6514         * config/arm/arm-builtins.c (arm_expand_builtin): Change
6515         expansion to perform a bitwise AND of the argument followed by a
6516         boolean negation of the result.
6517
6518 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
6519
6520         PR rtl-optimization/84878
6521         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6522         the basic block.  Assert the use reference is not artificial and that
6523         it has an associated insn.
6524
6525 2018-04-04  Michael Matz  <matz@suse.de>
6526
6527         * builtins.c (compute_objsize): Pass correct operand
6528         to array_at_struct_end_p.
6529
6530 2018-04-04  Richard Biener  <rguenther@suse.de>
6531
6532         PR lto/85176
6533         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6534         from contexts for DINFO_LEVEL_TERSE and below.
6535
6536 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
6537
6538         * config/nds32/nds32-doubleword.md (move_<mode>): Require
6539         resiter_operand condition.
6540         * config/nds32/nds32.md (*move<mode>): Ditto.
6541
6542 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
6543             Monk Chiang  <sh.chiang04@gmail.com>
6544
6545         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6546
6547 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
6548
6549         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6550
6551 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
6552             Kito Cheng  <kito.cheng@gmail.com>
6553
6554         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6555         nds32_cond_code_str, output_cond_branch,
6556         output_cond_branch_compare_zero, nds32_expand_cbranch,
6557         nds32_expand_cstore, nds32_expand_movcc,
6558         nds32_output_cbranchsi4_equality_zero,
6559         nds32_output_cbranchsi4_equality_reg,
6560         nds32_output_cbranchsi4_equality_reg_or_const_int,
6561         nds32_output_cbranchsi4_greater_less_zero: New functions.
6562         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6563         nds32_expand_cstore, nds32_expand_movcc,
6564         nds32_output_cbranchsi4_equality_zero,
6565         nds32_output_cbranchsi4_equality_reg,
6566         nds32_output_cbranchsi4_equality_reg_or_const_int,
6567         nds32_output_cbranchsi4_greater_less_zero): Declare.
6568         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6569         nds32_rimm11s_operand): New predicates.
6570         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6571         * config/nds32/nds32.md: Rewrite all the branch and conditional move
6572         patterns.
6573
6574 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
6575
6576         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6577         * config/nds32/nds32.md: Ditto.
6578         * config/nds32/pipelines.md: Ditto.
6579
6580 2018-04-04  Richard Biener  <rguenther@suse.de>
6581
6582         PR tree-optimization/85168
6583         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6584         propagating abnormals.
6585
6586 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
6587
6588         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6589
6590 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
6591             Kito Cheng  <kito.cheng@gmail.com>
6592
6593         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6594         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6595         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6596         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6597         * config/nds32/nds32.md (sibcall_internal): New.
6598         (sibcall_register): Remove.
6599         (sibcall_immediate): Remove.
6600         (sibcall_value_internal): New.
6601         (sibcall_value_register): Remove.
6602         (sibcall_value_immediate): Remove.
6603         * config/nds32/predicates.md (nds32_general_register_operand): New.
6604         (nds32_call_address_operand): New.
6605
6606 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
6607
6608         PR rtl-optimization/85167
6609         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6610         bb_defs if *split_p, instead preinitialize it to NULL.
6611
6612         PR tree-optimization/85156
6613         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6614         evaluating the argument multiple times.
6615
6616 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
6617
6618         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6619         than vector.
6620         (_mm_cvtpd_ps): Likewise.
6621         (_mm_cvttpd_epi32): Likewise.
6622         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6623         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6624         vector, pixel, and bool following altivec.h include.
6625
6626 2018-04-03  Martin Sebor  <msebor@redhat.com>
6627
6628         * doc/extend.texi (Common Function Attributes): Clarify.
6629         (const attribute): Likewise.
6630         (pure attribute): Likewise.
6631
6632 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
6633
6634         PR target/85169
6635         * config/i386/i386.c (ix86_expand_vector_set): Use
6636         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
6637
6638 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
6639
6640         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6641         instructions when changing rounding bits to preserve precision bits
6642         in the x87 control word.
6643
6644 2018-04-03  Martin Liska  <mliska@suse.cz>
6645
6646         PR tree-optimization/82491
6647         * rtl.h (strip_offset_and_add): Replace += suboffset with
6648         poly_uint64 () + suboffset.
6649
6650 2018-03-29  Martin Liska  <mliska@suse.cz>
6651             Martin Jambor  <mjambor@suse.cz>
6652
6653         PR ipa/84947
6654         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6655         param_type is not an integral or pointer type.
6656
6657 2018-04-03  Richard Biener  <rguenther@suse.de>
6658
6659         * sese.h (recompute_all_dominators): Remove.
6660
6661 2018-04-02  Martin Sebor  <msebor@redhat.com>
6662
6663         * doc/invoke.texi (-Wrestrict): Fix typos.
6664
6665 2018-04-02  Jim Wilson  <jimw@sifive.com>
6666
6667         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6668         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6669         (<optab>di3, <optab>si3_extend): Likewise.
6670         (<optab>si3_mask, <optab>si3_mask_1): New.
6671         (<optab>di3_mask, <optab>di3_mask_1): New.
6672         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6673         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6674         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6675
6676 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
6677
6678         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6679         example.
6680
6681 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
6682
6683         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6684         (nds32_canonicalize_comparison): New function.
6685
6686 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
6687             Kito Cheng  <kito.cheng@gmail.com>
6688             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
6689
6690         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6691         * config/nds32/constants.md (unspec_volatile_element): Add
6692         UNSPEC_VOLATILE_RELAX_GROUP.
6693         * config/nds32/nds32-relax-opt.c: New file.
6694         * config/nds32/nds32-predicates.c
6695         (nds32_symbol_load_store_p): New function.
6696         * config/nds32/nds32-protos.h
6697         (nds32_symbol_load_store_p): Declare function.
6698         (make_pass_nds32_relax_opt): Declare new rtl pass function.
6699         * config/nds32/nds32.c
6700         (nds32_register_pass): New function to register pass.
6701         (nds32_register_passes): New function to register passes.
6702         * config/nds32/nds32.md (relax_group): New pattern.
6703         * config/nds32/nds32.opt (mrelax-hint): New option.
6704         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6705
6706 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
6707
6708         * config/nds32/t-nds32: Modify files dependency.
6709
6710 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
6711
6712         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6713         (PROFILE_HOOK): Define its implementation.
6714
6715 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
6716
6717         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6718         type and 32-bit size.
6719
6720 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
6721
6722         PR middle-end/85090
6723         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6724         (V_128_256): New mode iterator.
6725         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6726         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6727         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6728         of V.
6729         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6730         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6731
6732 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
6733
6734         PR target/83315
6735         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6736         NaN inputs correctly.
6737
6738 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
6739
6740         PR target/80546
6741         * config/rs6000/vsx.md (??r): New mode attribute.
6742         (*vsx_mov<mode>_64bit): Use it.
6743         (*vsx_mov<mode>_32bit): Likewise.
6744
6745 2018-03-30  Martin Sebor  <msebor@redhat.com>
6746
6747         PR tree-optimization/84818
6748         * builtins.c (check_access): Use warning_n.
6749
6750 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
6751
6752         PR target/83822
6753         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6754         condition.
6755         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6756         condition.
6757
6758 2018-03-30  Julia Koval  <julia.koval@intel.com>
6759
6760         PR target/84413
6761         * x86-tune.def (movx, partial_reg_dependency): Enable for
6762         m_SKYLAKE_AVX512.
6763
6764 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
6765
6766         PR inline-asm/84985
6767         * lra-constraints.c (process_alt_operands): Move setting
6768         this_alternative_matches below.
6769
6770 2018-03-29  Martin Liska  <mliska@suse.cz>
6771
6772         PR lto/84995.
6773         * doc/invoke.texi: Document how LTO works with debug info.
6774         Describe auto-load support of binutils.  Mention 'x86-64'
6775         as valid option value of -march option.
6776
6777 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
6778
6779         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6780
6781         PR c/85094
6782         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6783         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6784         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6785         checking.
6786
6787 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
6788
6789         PR target/84912
6790         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6791         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6792         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6793         for RS6000_BTM_POWERPC64.
6794         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6795         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6796         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6797         definition.
6798         (DIVDE): Use it.
6799         (DIVDEU): Likewise.
6800
6801 2018-03-28 Carl Love  <cel@us.ibm.com>
6802
6803         Revert
6804         2017-09-27  Carl Love  <cel@us.ibm.com>
6805
6806         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6807         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6808         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6809         fctiw instruction.
6810
6811 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6812
6813         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6814         instead of __vector bool.
6815         (_mm_max_pu8): Likewise.
6816         (_mm_min_pi16): Likewise.
6817
6818 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
6819
6820         PR target/84912
6821         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6822         (DIVWEUO): Likewise.
6823         (DIVDEO): Likewise.
6824         (DIVDEUO): Likewise.
6825         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6826         DIVWEUO and DIVDEUO.
6827         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6828         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6829         (div_extend): Likewise.
6830         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6831         builtin function.
6832         (__builtin_divweuo): Likewise.
6833         (__builtin_divdeo): Likewise.
6834         (__builtin_divdeuo): Likewise.
6835
6836 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
6837
6838         PR target/85095
6839         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6840         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6841
6842         PR tree-optimization/82004
6843         * gimple-match-head.c (optimize_pow_to_exp): New function.
6844         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6845         Don't fold to exp if optimize_pow_to_exp is false.
6846
6847 2018-03-28  Martin Liska  <mliska@suse.cz>
6848
6849         PR other/84819
6850         * calls.c (initialize_argument_information): Fix trailing space.
6851         * common.opt: Fix typo and provide better explanation for
6852         -fsanitize-coverage option.
6853         * config/i386/i386.opt: Fix typo.
6854
6855 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
6856             Martin Liska  <mliska@suse.cz>
6857
6858         PR sanitizer/85081
6859         * gimplify.c (asan_poison_variable): Don't do the check for
6860         gimplify_omp_ctxp here.
6861         (gimplify_decl_expr): Do it here.
6862         (gimplify_target_expr): Likewise.
6863
6864 2018-03-28  Martin Liska  <mliska@suse.cz>
6865
6866         PR target/84988
6867         * config/i386/i386.c (ix86_function_arg_advance): Do not call
6868         chkp_type_bounds_count if MPX is not enabled.
6869
6870 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
6871
6872         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6873
6874 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
6875
6876         PR target/84914
6877         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6878         function to create the function decl for complex long double
6879         multiply and divide for -mabi=ieeelongdouble.
6880         (init_float128_ieee): Call it.
6881
6882 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
6883
6884         PR target/85044
6885         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6886         -fcf-protection=branch -mibt.
6887         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6888
6889 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6890
6891         PR target/81863
6892         * config/arm/arm.c (arm_valid_symbolic_address): Handle
6893         arm_word_relocations.
6894
6895 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
6896
6897         PR target/85056
6898         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6899         extern array declarations.
6900
6901 2018-03-27  Richard Biener  <rguenther@suse.de>
6902
6903         PR middle-end/84067
6904         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6905         explicit single_use checks.
6906
6907 2018-03-27  Richard Biener  <rguenther@suse.de>
6908
6909         PR tree-optimization/85082
6910         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6911         Valueize the VUSE.
6912
6913 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6914
6915         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6916         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6917         Turn on fasynchronous-unwind-tables and funwind-tables.
6918
6919 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
6920
6921         PR target/85073
6922         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6923         (*bmi_blsr_<mode>_ccz): Ditto.
6924
6925 2018-03-26  Tom de Vries  <tom@codesourcery.com>
6926
6927         PR tree-optimization/85063
6928         * omp-general.c (offloading_function_p): New function.  Factor out
6929         of ...
6930         * omp-offload.c (pass_omp_target_link::gate): ... here.
6931         * omp-general.h (offloading_function_p): Declare.
6932         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6933         with attribute omp declare target for offloading functions.
6934
6935 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
6936
6937         PR tree-optimization/84005
6938         * tree-data-ref.h (get_base_for_alignment): Declare.
6939         * tree-data-ref.c (get_base_for_alignment_1): New function.
6940         (get_base_for_alignment): Likewise.
6941         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6942         get_base_for_alignment to find a suitable base object, instead
6943         of always using drb->base_address.
6944
6945 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
6946
6947         PR inline-asm/85022
6948         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6949         known size by default.
6950
6951 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
6952
6953         PR inline-asm/85030
6954         * lra-constraints.c (process_alt_operands): Don't match BLKmode
6955         and non BLKmode operands.
6956
6957 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6958
6959         PR target/85026
6960         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6961         Clean up attributes.
6962
6963 2018-03-23  Richard Biener  <rguenther@suse.de>
6964
6965         PR debug/85020
6966         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6967         we are going to emit early debug for LTO.
6968
6969 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
6970
6971         PR inline-asm/85034
6972         * function.c (match_asm_constraints_1): Don't optimize if input
6973         doesn't satisfy general_operand predicate for output's mode.
6974
6975         PR inline-asm/85022
6976         * alias.c (write_dependence_p): Don't require for x_canonicalized
6977         non-VOIDmode if x has VOIDmode.
6978
6979         PR sanitizer/85029
6980         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6981         just don't try to optimize it rather than assert it never happens.
6982
6983 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6984
6985         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6986         macro expansions for definition of ST_INTERNAL_<mode> and
6987         LD_INTERNAL_<mode> builtins.
6988         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6989         Remove prototype.
6990         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6991         function.
6992         (altivec_expand_st_builtin): Likewise.
6993         (altivec_expand_builtin): Remove calls to deleted functions.
6994         (rs6000_address_for_altivec): Delete this function.
6995         * config/rs6000/vector.md: Remove expands for
6996         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6997
6998 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
6999
7000         PR target/84826
7001         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7002         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7003         re-computing once computed.
7004         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7005         (arm_init_machine_status): Initialize
7006         machine->static_chain_stack_bytes.
7007
7008 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7009
7010         PR target/84760
7011         * doc/extend.texi: Add four new prototypes for vec_ld.
7012         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7013         definitions for more logical presentation.
7014         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7015         entries for V1TI variants of __builtin_altivec_ld builtin.
7016         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7017         handling of V1TI variant of LVX icode pattern.
7018         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7019         (rs6000_gimple_fold_builtin): Likewise.
7020         (altivec_init_builtins): Add code to define
7021         __builtin_altivec_lvx_v1ti function.
7022
7023 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
7024
7025         PR inline-asm/84941
7026         * function.c (match_asm_constraints_1): Don't do the optimization
7027         if input isn't a REG, SUBREG, MEM or constant.
7028
7029 2018-03-22  Tom de Vries  <tom@codesourcery.com>
7030
7031         PR tree-optimization/84956
7032         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7033         bb_has_abnormal_pred.
7034
7035 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
7036
7037         PR sanitizer/85018
7038         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7039         DECL_INITIAL (decl) to decl at the end.
7040         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7041         adjust the comment.
7042
7043 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
7044
7045         * doc/extend.texi (__builtin_tgmath): Document when complex
7046         integer types are treated as _Complex _Float64.
7047
7048 2018-03-21  Tom de Vries  <tom@codesourcery.com>
7049
7050         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7051
7052 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
7053
7054         PR tree-optimization/84960
7055         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7056         if it is ENTRY block, move them into single succ of ENTRY in that case.
7057
7058 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
7059
7060         PR tree-optimization/84811
7061         * poly-int.h (poly_span_traits): Remove the T3 parameter and
7062         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7063         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7064         (known_subrange_p): Update accordingly.  Cast each value involved
7065         in the size comparison, rather than casting the result of the
7066         subtraction.
7067
7068 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
7069
7070         PR tree-optimization/84982
7071         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7072         by flipping the least significant bit rather than all bits from
7073         bitpos to bitpos + bitsize - 1.
7074
7075 2018-03-21  Nathan Sidwell  <nathan@acm.org>
7076
7077         * doc/extend.texi (Deprecated Features): Remove mention of
7078         long-deleted deprecations.
7079
7080 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7081
7082         PR jit/84288
7083         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7084         * configure: Regenerate.
7085
7086 2018-03-21  Tom de Vries  <tom@codesourcery.com>
7087
7088         PR tree-optimization/83126
7089         * tree-parloops.c (num_phis): New function.
7090         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7091
7092 2018-03-21  Nathan Sidwell  <nathan@acm.org>
7093
7094         * doc/extend.texi (Deprecated Features): Update deprecated flags,
7095         mention anon-struct/union members and trailing attributes.
7096
7097 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
7098
7099         PR tree-optimization/84969
7100         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7101         builtin memset partitions if they set different rhs values.
7102
7103 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
7104
7105         PR rtl-optimization/84989
7106         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7107         VEC_DUPLICATE with scalar result mode.
7108
7109 2018-03-21  Martin Liska  <mliska@suse.cz>
7110
7111         PR ipa/84963
7112         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7113         not intended return statement.
7114
7115 2018-03-21  Martin Liska  <mliska@suse.cz>
7116
7117         PR target/84988
7118         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7119         (chkp_find_bound_slots_1): Limit number of iterations.
7120
7121 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
7122
7123         PR target/84838
7124         * Minor grammar fixes for x86 options.
7125
7126 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
7127
7128         PR debug/84875
7129         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7130         holding REG_CFA_RESTORE notes, instead turn them into a USE.
7131
7132 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
7133
7134         PR target/83789
7135         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7136         (altivec_lvx_<mode>_1op): Likewise.
7137         (altivec_stvx_<mode>_2op): Likewise.
7138         (altivec_stvx_<mode>_1op): Likewise.
7139         (altivec_lvx_<VM2:mode>): New define_expand.
7140         (altivec_stvx_<VM2:mode>): Likewise.
7141         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7142         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7143         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7144         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7145         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7146         (rs6000_gen_lvx): Likewise.
7147         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7148         (altivec_expand_stv_builtin): Likewise.
7149         (altivec_expand_builtin): Likewise.
7150         * config/rs6000/vector.md: Likewise.
7151
7152 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7153
7154         PR target/82518
7155         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7156         BYTES_BIG_ENDIAN.
7157
7158 2018-03-20  Richard Biener  <rguenther@suse.de>
7159
7160         PR target/84986
7161         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7162         sign-conversions as zero, fall back to standard scalar_stmt
7163         cost for the rest.
7164
7165 2018-03-20  Martin Liska  <mliska@suse.cz>
7166
7167         PR ipa/84825
7168         * predict.c (rebuild_frequencies): Handle case when we have
7169         PROFILE_ABSENT, but flag_guess_branch_prob is false.
7170
7171 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
7172
7173         PR target/84990
7174         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7175         flag_section_anchors.
7176         * varasm.c (use_blocks_for_decl_p): Remove hack for
7177         dw2_force_const_mem.
7178
7179         PR target/84845
7180         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7181         to ...
7182         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
7183         be created, use lowpart_subreg of operands[0] rather than operands[0]
7184         itself.
7185         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7186         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7187         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7188         and n constraint instead of aarch64_shift_imm_di and Usd.
7189         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7190         (*aarch64_<optab>_reg_minus<mode>3): ... this.
7191
7192 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
7193
7194         PR target/82989
7195         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7196         to favor GPR over NEON registers.
7197         (<shift>di3_neon): Likewise.
7198
7199 2018-03-20  Tom de Vries  <tom@codesourcery.com>
7200
7201         PR target/84952
7202         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7203         (nvptx_process_pars): Emit bar.sync asap and alap.
7204
7205 2018-03-20  Tom de Vries  <tom@codesourcery.com>
7206
7207         PR target/84954
7208         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7209         seen_label if seen_label is already set.
7210
7211 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
7212
7213         PR target/84945
7214         * config/i386/i386.c (fold_builtin_cpu): For features above 31
7215         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7216         Use 1U instead of 1.  Formatting fixes.
7217
7218         PR c/84953
7219         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7220         instead of TREE_TYPE (s1) for the return value.
7221
7222 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
7223
7224         PR tree-optimization/84946
7225         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7226         bitsize + bitsize in poly_uint64 rather than poly_int64.
7227
7228         PR sanitizer/78651
7229         * dwarf2asm.c: Include fold-const.c.
7230         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7231         of decl rather than decl itself.
7232
7233         PR rtl-optimization/84643
7234         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7235
7236 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
7237
7238         PR sanitizer/78651
7239         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7240         calling assemble_variable.
7241
7242 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
7243
7244         PR target/81647
7245         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7246         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7247
7248 2018-03-19  Jim Wilson  <jimw@sifive.com>
7249
7250         PR bootstrap/84856
7251         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7252         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7253         (riscv_first_stack_step): Likewise.
7254         (riscv_option_override): Use STACK_BOUNDARY instead of
7255         MIN_STACK_BOUNDARY.
7256         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7257         MIN_STACK_BOUNDARY.
7258         (BIGGEST_ALIGNMENT): Set to 128.
7259         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7260         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7261         STACK_BOUNDARY.
7262
7263 2018-03-19  Richard Biener  <rguenther@suse.de>
7264
7265         PR tree-optimization/84933
7266         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7267         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7268
7269 2018-03-19  Richard Biener  <rguenther@suse.de>
7270
7271         PR tree-optimization/84859
7272         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7273         (cond_if_else_store_replacement): Perform sinking operation on
7274         single-store BBs regardless of MAX_STORES_TO_SINK setting.
7275         Generalize what a BB with a single eligible store is.
7276
7277 2018-03-19  Richard Biener  <rguenther@suse.de>
7278
7279         PR tree-optimization/84929
7280         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7281         chrec_is_positive against non-chrec arg.
7282
7283 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
7284
7285         PR target/84711
7286         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7287
7288 2018-03-18  Martin Liska  <mliska@suse.cz>
7289
7290         PR rtl-optimization/84635
7291         * regrename.c (build_def_use): Use matches_mode only when
7292         matches >= 0.
7293
7294 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
7295
7296         PR tree-optimization/84913
7297         * tree-vect-loop.c (vectorizable_reduction): Don't try to
7298         vectorize chains of COND_EXPRs.
7299
7300 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
7301
7302         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7303
7304 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
7305
7306         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7307
7308 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
7309
7310         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7311
7312 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7313             Kito Cheng  <kito.cheng@gmail.com>
7314
7315         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7316         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7317         (nds32_adjust_reg_alloc_order): New function.
7318         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7319
7320 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
7321
7322         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7323         nds32_print_operand, nds32_print_operand_address): Use
7324         HOST_WIDE_INT_PRINT_DEC instead.
7325
7326 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7327
7328         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7329
7330 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
7331
7332         PR target/84902
7333         * config/i386/i386.c (initial_ix86_tune_features,
7334         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7335         unsigned long long.
7336         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7337         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7338         rather than 1u << ix86_tune.  Formatting fix.
7339         (ix86_option_override_internal): Change ix86_arch_mask from
7340         unsigned int to unsigned HOST_WIDE_INT, initialize to
7341         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7342         (ix86_function_specific_restore): Likewise.
7343
7344 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
7345
7346         PR target/84899
7347         * postreload.c (reload_combine_recognize_pattern): Perform
7348         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7349         truncate_int_for_mode the result for the destination's mode.
7350
7351         PR c/84909
7352         * hsa-gen.c (mem_type_for_type): Fix comment typo.
7353         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7354         Likewise.
7355         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7356         Likewise.
7357
7358 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
7359
7360         PR target/84876
7361         * lra-assigns.c (lra_split_hard_reg_for): Don't use
7362         regno_allocno_class_array and sorted_pseudos.
7363         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7364         insns where regno is used.
7365
7366 2018-03-16  Martin Liska  <mliska@suse.cz>
7367
7368         PR ipa/84833
7369         * multiple_target.c (create_dispatcher_calls): Redirect
7370         reference in the symbol table.
7371
7372 2018-03-16  Martin Liska  <mliska@suse.cz>
7373
7374         PR ipa/84722
7375         * multiple_target.c (create_dispatcher_calls): Redirect also
7376         an alias.
7377
7378 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
7379
7380         PR c++/79937
7381         PR c++/82410
7382         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7383         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7384         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7385
7386 2018-03-16  Julia Koval  <julia.koval@intel.com>
7387
7388         * doc/invoke.texi (Skylake Server): Add CLWB.
7389         Cannonlake): Remove CLWB.
7390
7391 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
7392
7393         PR tree-optimization/84841
7394         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7395         1 << 3.
7396         (FLOAT_ONE_CONST_TYPE): Define.
7397         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7398         (sort_by_operand_rank): Put entries with higher constant_type last
7399         rather than first to match comments.
7400
7401 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
7402
7403         * config/nios2/nios2.md (movsi_internal): Fix thinko in
7404         split predicate.
7405
7406 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
7407
7408         PR c++/79085
7409         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7410         check and use address of target always.
7411
7412 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
7413
7414         PR target/84574
7415         * config/i386/i386.c (indirect_thunk_needed): Update comments.
7416         (indirect_thunk_bnd_needed): Likewise.
7417         (indirect_thunks_used): Likewise.
7418         (indirect_thunks_bnd_used): Likewise.
7419         (indirect_return_needed): New.
7420         (indirect_return_bnd_needed): Likewise.
7421         (output_indirect_thunk_function): Add a bool argument for
7422         function return.
7423         (output_indirect_thunk_function): Don't generate alias for
7424         function return thunk.
7425         (ix86_code_end): Call output_indirect_thunk_function to generate
7426         function return thunks.
7427         (ix86_output_function_return): Set indirect_return_bnd_needed
7428         and indirect_return_needed instead of indirect_thunk_bnd_needed
7429         and indirect_thunk_needed.
7430
7431 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
7432
7433         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7434         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7435         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7436
7437 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
7438             Paul Hua <paul.hua.gm@gmail.com>
7439
7440         PR c/84852
7441         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7442
7443 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
7444
7445         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7446         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7447         resp. SFmode cases.
7448
7449 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
7450
7451         PR target/84711
7452         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7453         instead of GET_MODE_SIZE when comparing Units.
7454
7455 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
7456
7457         PR target/68256
7458         * varasm.c (hash_section): Return an unchangeble hash value
7459         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7460         Return !aarch64_can_use_per_function_literal_pools_p ().
7461
7462 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
7463
7464         PR target/84860
7465         * optabs.c (emit_conditional_move): Pass address of cmode's copy
7466         rather than address of cmode as last argument to prepare_cmp_insn.
7467
7468 2018-03-15  Julia Koval  <julia.koval@intel.com>
7469
7470         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7471         F_AVX512VNNI, F_AVX512BITALG): New.
7472
7473 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
7474
7475         PR target/83451
7476         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
7477         insn for floating-point loads and stores.
7478
7479 2018-03-14  Carl Love  <cel@us.ibm.com>
7480
7481         * config/rs6000/rs6000-c.c: Add macro definitions for
7482         ALTIVEC_BUILTIN_VEC_PERMXOR.
7483         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7484         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7485         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7486         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7487         UNSPEC_VPERMXOR.
7488         * config/doc/extend.texi: Add prototypes for vec_permxor.
7489
7490 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
7491
7492         PR c/84852
7493         * diagnostic-show-locus.c (class layout_point): Convert m_line
7494         from int to linenum_type.
7495         (line_span::comparator): Use linenum "compare" function when
7496         comparing line numbers.
7497         (test_line_span): New function.
7498         (layout_range::contains_point): Convert param "row" from int to
7499         linenum_type.
7500         (layout_range::intersects_line_p): Likewise.
7501         (layout::will_show_line_p): Likewise.
7502         (layout::print_source_line): Likewise.
7503         (layout::should_print_annotation_line_p): Likewise.
7504         (layout::print_annotation_line): Likewise.
7505         (layout::print_leading_fixits): Likewise.
7506         (layout::annotation_line_showed_range_p): Likewise.
7507         (struct line_corrections): Likewise for field m_row.
7508         (line_corrections::line_corrections): Likewise for param "row".
7509         (layout::print_trailing_fixits): Likewise.
7510         (layout::get_state_at_point): Likewise.
7511         (layout::get_x_bound_for_row): Likewise.
7512         (layout::print_line): Likewise.
7513         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7514         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7515         * input.c (selftest::test_linenum_comparisons): New function.
7516         (selftest::input_c_tests): Call it.
7517         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7518         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7519         * selftest.h (ASSERT_GT): New macro.
7520         (ASSERT_GT_AT): New macro.
7521         (ASSERT_LT): New macro.
7522         (ASSERT_LT_AT): New macro.
7523
7524 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
7525
7526         PR rtl-optimization/84780
7527         * combine.c (distribute_links): Don't make a link based on pc_rtx.
7528
7529 2018-03-14  Martin Liska  <mliska@suse.cz>
7530
7531         * tree.c (record_node_allocation_statistics): Use
7532         get_stats_node_kind.
7533         (get_stats_node_kind): New function extracted from
7534         record_node_allocation_statistics.
7535         (free_node): Use get_stats_node_kind.
7536
7537 2018-03-14  Richard Biener  <rguenther@suse.de>
7538
7539         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7540         that the value-set of ANTIC_IN doesn't grow.
7541
7542         Revert
7543         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7544         member.
7545         (BB_VISITED_WITH_VISITED_SUCCS): New define.
7546         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7547
7548 2018-03-14  Julia Koval  <julia.koval@intel.com>
7549
7550         * config.gcc (icelake-client, icelake-server): New.
7551         (icelake): Remove.
7552         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7553         (initial_ix86_arch_features): Ditto.
7554         (PTA_SKYLAKE): Add SGX.
7555         (PTA_ICELAKE): Remove.
7556         (PTA_ICELAKE_CLIENT): New.
7557         (PTA_ICELAKE_SERVER): New.
7558         (ix86_option_override_internal): Split up icelake on icelake client and
7559         icelake server.
7560         (get_builtin_code_for_version): Ditto.
7561         (fold_builtin_cpu): Ditto.
7562         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7563         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7564         * config/i386/i386.h (processor_type): Ditto.
7565         * doc/invoke.texi: Ditto.
7566
7567 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
7568
7569         PR sanitizer/83392
7570         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7571         INTEGER_CST offset, add it together with bitpos / 8 and
7572         sign extend based on POINTER_SIZE.
7573
7574         PR target/84844
7575         Revert
7576         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
7577
7578         PR target/78090
7579         * config/i386/constraints.md (Yc): New register constraint.
7580         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7581         Use Yc constraint for alternative 2 of operand 0.  Remove
7582         preferred_for_speed attribute.
7583
7584 2018-03-14  Richard Biener  <rguenther@suse.de>
7585
7586         PR tree-optimization/84830
7587         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7588         with the old one to avoid oscillations.
7589
7590 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
7591
7592         PR target/83712
7593         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7594         pseudos.
7595         (assign_by_spills): Return a flag of reload assignment failure.
7596         Do not process the reload assignment failures.  Do not spill other
7597         reload pseudos if they has the same reg class.  Update n if
7598         necessary.
7599         (lra_assign): Add a return arg.  Set up from the result of
7600         assign_by_spills call.
7601         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7602         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
7603         usage_insns if it is not NULL.
7604         (spill_hard_reg_in_range): New function.
7605         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7606         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7607         function prototypes.
7608         (lra_assign): Change prototype.
7609         * lra.c (lra): Add code to deal with fails by splitting hard reg
7610         live ranges.
7611
7612 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
7613
7614         * config/riscv/riscv.opt (mrelax): New option.
7615         * config/riscv/riscv.c (riscv_file_start): Emit ".option
7616         "norelax" when riscv_mrelax is disabled.
7617         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7618
7619 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
7620
7621         PR target/84743
7622         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7623         reassociation for int modes.
7624
7625 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
7626
7627         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7628         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7629         for big-endian.
7630         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7631         * config/aarch64/aarch64-sve.md
7632         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7633         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7634         (*extend<mode><Vwide>2): Rename to...
7635         (aarch64_sve_extend<mode><Vwide>2): ...this.
7636         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7637         renaming the old pattern to...
7638         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
7639         unsigned packs.
7640         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7641         define_expand, renaming the old pattern to...
7642         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7643         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7644         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7645         account when deciding which SVE instruction the optab should use.
7646         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7647
7648 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
7649
7650         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7651         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7652         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7653         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7654         (tlsdesc_small_<mode>): Turn a define_expand and use
7655         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
7656         (tlsdesc_small_advsimd_<mode>): ...this.
7657         (tlsdesc_small_sve_<mode>): New pattern.
7658
7659 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
7660
7661         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7662         (UNSPEC_UMUL_HIGHPART): New constants.
7663         (MUL_HIGHPART): New int iteraor.
7664         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7665         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7666         define_expand.
7667         (*<su>mul<mode>3_highpart): New define_insn.
7668
7669 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
7670
7671         PR lto/84805
7672         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7673         incomplete types.
7674
7675 2018-03-13  Martin Liska  <mliska@suse.cz>
7676
7677         PR ipa/84658.
7678         * (sem_item_optimizer::sem_item_optimizer): Initialize new
7679         vector.
7680         (sem_item_optimizer::~sem_item_optimizer): Release it.
7681         (sem_item_optimizer::merge_classes): Register variable aliases.
7682         (sem_item_optimizer::fixup_pt_set): New function.
7683         (sem_item_optimizer::fixup_points_to_sets): Likewise.
7684         * ipa-icf.h: Declare new variables and functions.
7685
7686 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
7687
7688         PR middle-end/84834
7689         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7690         integer_pow2p@2 and test integer_pow2p in condition.
7691         (A < 0 ? C : 0): Similarly for @1.
7692
7693         PR middle-end/84831
7694         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7695         characters starting at p contain '\0' character, don't look beyond
7696         that.
7697
7698         PR target/84827
7699         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7700         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7701
7702         PR target/84828
7703         * reg-stack.c (change_stack): Change update_end var from int to
7704         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7705         also call set_block_for_insn on the newly added insns and rescan.
7706
7707         PR target/84786
7708         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7709         on the last operand.
7710
7711         PR c++/84704
7712         * tree.c (stabilize_reference_1): Return save_expr (e) for
7713         STATEMENT_LIST even if it doesn't have side-effects.
7714
7715 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
7716
7717         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7718
7719 2018-03-12  Renlin Li  <renlin.li@arm.com>
7720
7721         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7722         aarch64_output_scalar_simd_mov_immediate.
7723
7724 2018-03-12  Martin Sebor  <msebor@redhat.com>
7725
7726         PR tree-optimization/83456
7727         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7728         for perfectly overlapping calls to memcpy.
7729         (gimple_fold_builtin_memory_chk): Same.
7730         (gimple_fold_builtin_strcpy): Handle no-warning.
7731         (gimple_fold_builtin_stxcpy_chk): Same.
7732         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7733
7734 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
7735
7736         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7737         parameter.  Use it for SFmode.
7738         (rs6000_function_arg_advance_1): Adjust.
7739         (rs6000_function_arg): Adjust.
7740         (rs6000_gimplify_va_arg): Pass false for that new parameter.
7741
7742 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
7743
7744         PR rtl-optimization/84169
7745         PR rtl-optimization/84780
7746         * combine.c (can_combine_p): Check for a 2-insn combination whether
7747         the destination register is used between the two insns, too.
7748
7749 2018-03-12  Richard Biener  <rguenther@suse.de>
7750
7751         PR tree-optimization/84803
7752         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7753         for refs DR analysis didn't process.
7754
7755 2018-03-12  Richard Biener  <rguenther@suse.de>
7756
7757         PR tree-optimization/84777
7758         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7759         force-vectorize loops ignore whether we are optimizing for size.
7760
7761 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7762
7763         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7764         (TARGET_MD_ASM_ADJUST): Define.
7765
7766 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
7767             Kito Cheng  <kito.cheng@gmail.com>
7768             Chung-Ju Wu  <jasonwucj@gmail.com>
7769
7770         * config/nds32/nds32.c (nds32_compute_stack_frame,
7771         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7772         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7773         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7774         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7775         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7776         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7777         * config/nds32/nds32.md (prologue, epilogue): Use macro
7778         NDS32_V3PUSH_AVAILABLE_P to do checking.
7779
7780 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
7781
7782         PR debug/58150
7783         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7784         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7785         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
7786         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
7787         addition of most attributes on !orig_type_die or the attribute not
7788         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7789
7790 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
7791             Chung-Ju Wu  <jasonwucj@gmail.com>
7792
7793         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7794         __NDS32_VH__ macro.
7795         * config/nds32/nds32.opt (mvh): New option.
7796
7797 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
7798             Chung-Ju Wu  <jasonwucj@gmail.com>
7799
7800         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7801         function.
7802         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7803         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7804         definition.
7805
7806 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
7807             Chung-Ju Wu  <jasonwucj@gmail.com>
7808
7809         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7810         function.
7811         * config/nds32/nds32-multiple.md (strlensi): New pattern.
7812         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7813
7814 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
7815             Kito Cheng  <kito.cheng@gmail.com>
7816             Chung-Ju Wu  <jasonwucj@gmail.com>
7817
7818         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7819         UNSPEC_FFMISM and UNSPEC_FLMISM.
7820         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7821         for ffb, ffmism and flmism.
7822         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7823         (unspec_ffmism): Ditto.
7824         (unspec_flmism): Ditto.
7825         (nds32_expand_builtin_impl): Check if string extension is available.
7826         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7827         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7828
7829 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
7830
7831         Reverting patch:
7832         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
7833
7834         PR target/83712
7835         * lra-assigns.c (assign_by_spills): Return a flag of reload
7836         assignment failure.  Do not process the reload assignment
7837         failures.  Do not spill other reload pseudos if they has the same
7838         reg class.
7839         (lra_assign): Add a return arg.  Set up from the result of
7840         assign_by_spills call.
7841         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7842         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
7843         usage_insns if it is not NULL.
7844         (spill_hard_reg_in_range): New function.
7845         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7846         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7847         function prototypes.
7848         (lra_assign): Change prototype.
7849         * lra.c (lra): Add code to deal with fails by splitting hard reg
7850         live ranges.
7851
7852 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
7853
7854         PR target/84807
7855         * config/i386/i386.opt: Replace Enforcment with Enforcement.
7856
7857 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
7858
7859         PR debug/84620
7860         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7861         (dw_val_node): Add val_symbolic_view.
7862         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7863         (symview_upper_bound): New.
7864         (new_line_info_table): Initialize symviews_since_reset.
7865         (dwarf2out_source_line): Count symviews_since_reset and set
7866         symview_upper_bound.
7867         (dw_val_equal_p): Handle symview.
7868         (add_AT_symview): New.
7869         (print_dw_val): Handle symview.
7870         (attr_checksum, attr_checksum_ordered): Likewise.
7871         (same_dw_val_p, size_of_die): Likewise.
7872         (value_format, output_die): Likewise.
7873         (add_high_low_attributes): Use add_AT_symview for entry_view.
7874         (dwarf2out_finish): Reset symview_upper_bound, clear
7875         zero_view_p.
7876
7877 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
7878
7879         PR target/83969
7880         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7881         Add strict argument and use it.
7882         (rs6000_split_multireg_move): Update for new strict argument.
7883         (mem_operand_gpr): Disallow all non-offsettable addresses.
7884         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7885
7886 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
7887
7888         PR target/84772
7889         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7890         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7891         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7892
7893         PR c++/84767
7894         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7895         decl, use remap_type if we want to use the type.
7896
7897 2018-03-09  Martin Sebor  <msebor@redhat.com>
7898
7899         PR tree-optimization/84526
7900         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7901         Remove dead code.
7902         (builtin_access::generic_overlap): Be prepared to handle non-array
7903         base objects.
7904
7905 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
7906
7907         PR rtl-optimization/84682
7908         * lra-constraints.c (process_address_1): Check is_address flag
7909         for address constraints.
7910         (process_alt_operands): Likewise.
7911         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7912         preprocess_constraints.
7913         * recog.h (preprocess_constraints): Add oploc parameter.
7914         Adjust callers.
7915         * recog.c (preprocess_constraints): Test address_operand for
7916         CT_ADDRESS constraints.
7917
7918 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
7919
7920         PR target/83712
7921         * lra-assigns.c (assign_by_spills): Return a flag of reload
7922         assignment failure.  Do not process the reload assignment
7923         failures.  Do not spill other reload pseudos if they has the same
7924         reg class.
7925         (lra_assign): Add a return arg.  Set up from the result of
7926         assign_by_spills call.
7927         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7928         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
7929         usage_insns if it is not NULL.
7930         (spill_hard_reg_in_range): New function.
7931         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7932         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7933         function prototypes.
7934         (lra_assign): Change prototype.
7935         * lra.c (lra): Add code to deal with fails by splitting hard reg
7936         live ranges.
7937
7938 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7939
7940         PR target/83193
7941         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7942         Accept complain bool parameter.  Only emit errors if it is true.
7943         (arm_parse_cpu_option_name): Likewise.
7944         (arm_target_thumb_only): Adjust callers of the above.
7945         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7946         prototype to take a default true bool parameter.
7947         (arm_parse_arch_option_name): Likewise.
7948
7949 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
7950             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
7951
7952         PR jit/64089
7953         PR jit/84288
7954         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7955         * configure: Regenerate.
7956         * configure.ac ("linker --version-script option"): New.
7957         ("linker soname option"): New.
7958
7959 2018-03-09  Richard Biener  <rguenther@suse.de>
7960
7961         PR tree-optimization/84775
7962         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7963         immediate uses of predicate stmts and mark them modified.
7964
7965         Revert
7966         PR tree-optimization/84178
7967         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7968         to caller.
7969         (version_loop_for_if_conversion): Delay update_ssa call.
7970         (tree_if_conversion): Delay update_ssa until after predicate
7971         insertion.
7972
7973 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
7974
7975         PR target/84763
7976         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7977         when the function accesses prior frames.
7978
7979 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
7980
7981         PR debug/84456
7982         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7983         gen_llsym, otherwise call maybe_gen_llsym.
7984
7985         PR inline-asm/84742
7986         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7987         has ',' character inside of it.
7988
7989 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7990
7991         PR target/84748
7992         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7993         as clobbering CC_REGNUM.
7994
7995 2018-03-08  Richard Biener  <rguenther@suse.de>
7996
7997         PR middle-end/84552
7998         * tree-scalar-evolution.c: Include tree-into-ssa.h.
7999         (follow_copies_to_constant): Do not follow SSA names registered
8000         for update.
8001
8002 2018-03-08  Richard Biener  <rguenther@suse.de>
8003
8004         PR tree-optimization/84178
8005         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8006         to caller.
8007         (version_loop_for_if_conversion): Delay update_ssa call.
8008         (tree_if_conversion): Delay update_ssa until after predicate
8009         insertion.
8010
8011 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
8012
8013         PR tree-optimization/84178
8014         * tree-if-conv.c (release_bb_predicate): Remove the
8015         the assertion that the stmts have NULL use_ops.
8016         Discard the statements, asserting that they haven't
8017         yet been added to a BB.
8018
8019 2018-03-08  Richard Biener  <rguenther@suse.de>
8020
8021         PR tree-optimization/84746
8022         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8023         (phi_translate): Pass in destination ANTIC_OUT set.
8024         (phi_translate_1): Likewise.  For a simplified result lookup
8025         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8026         (phi_translate_set): Adjust.
8027         (do_pre_regular_insertion): Likewise.
8028         (do_pre_partial_partial_insertion): Likewise.
8029
8030 2018-03-08  Martin Liska  <mliska@suse.cz>
8031
8032         PR gcov-profile/84735
8033         * doc/gcov.texi: Document usage of profile files.
8034         * gcov-io.h: Document changes in the format.
8035
8036 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
8037
8038         PR debug/84404
8039         PR debug/84408
8040         * dwarf2out.c (struct dw_line_info_table): Update comments for
8041         view == -1.
8042         (FORCE_RESET_NEXT_VIEW): New.
8043         (FORCE_RESETTING_VIEW_P): New.
8044         (RESETTING_VIEW_P): Check for -1 too.
8045         (ZERO_VIEW_P): Likewise.
8046         (new_line_info_table): Force-reset next view.
8047         (dwarf2out_begin_function): Likewise.
8048         (dwarf2out_source_line): Simplify zero_view_p initialization.
8049         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8050         view directly.  Omit view when omitting .loc at line 0.
8051
8052 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
8053
8054         PR tree-optimization/84740
8055         * tree-switch-conversion.c (process_switch): Call build_constructors
8056         only if info.phi_count is non-zero.
8057
8058         PR tree-optimization/84739
8059         * tree-tailcall.c (find_tail_calls): Check call arguments against
8060         DECL_ARGUMENTS (current_function_decl) rather than
8061         DECL_ARGUMENTS (func) when checking for tail recursion.
8062
8063 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
8064
8065         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8066         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8067         Volker Reichelt's entry and add entries for people that perform
8068         GCC fuzzy testing and report numerous bugs.
8069
8070 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
8071
8072         PR target/82411
8073         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8074         readonly data in sdata, if that is disabled.
8075         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8076         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8077         -mreadonly-in-sdata option.
8078
8079 2018-03-07  Martin Sebor  <msebor@redhat.com>
8080
8081         PR tree-optimization/84468
8082         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8083         basic block when looking for nul assignment.
8084
8085 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
8086
8087         PR target/84277
8088         * except.h (output_function_exception_table): Adjust prototype.
8089         * except.c (output_function_exception_table): Remove FNNAME parameter
8090         and add SECTION parameter.  Ouput one part of the table at a time.
8091         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8092         the first part of the exception table and emit unwind directives.
8093         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8094         (i386_pe_seh_cold_init): Likewise.
8095         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8096         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8097         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8098         (ix86_output_call_insn): Emit a nop in one more case for SEH.
8099         * config/i386/winnt.c: Include except.h.
8100         (struct seh_frame_state): Add reg_offset, after_prologue and
8101         in_cold_section fields.
8102         (i386_pe_seh_end_prologue): Set seh->after_prologue.
8103         (i386_pe_seh_cold_init): New function.
8104         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8105         to seh->in_cold_section.
8106         (seh_emit_push): Record the offset of the push.
8107         (seh_emit_save): Record the offet of the save.
8108         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8109         Test seh->after_prologue to disregard the epilogue.
8110         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8111         (i386_pe_end_cold_function): New function.
8112
8113 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
8114
8115         PR fortran/84565
8116         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8117         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8118
8119         PR c++/84704
8120         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8121         on tmp_var.
8122         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8123         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8124
8125         PR middle-end/84723
8126         * multiple_target.c: Include tree-inline.h and intl.h.
8127         (expand_target_clones): Diagnose and fail if node->definition and
8128         !tree_versionable_function_p (node->decl).
8129
8130 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
8131
8132         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8133         sprint_ul.
8134         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8135         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8136         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8137
8138 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
8139
8140         PR target/84710
8141         * combine.c (try_combine): Use reg_or_subregno instead of handling
8142         just paradoxical SUBREGs and REGs.
8143
8144 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
8145
8146         * config/arc/arc.c (arc_finalize_pic): Remove function.
8147         (arc_must_save_register): We use single base PIC register, remove
8148         checks to save/restore the PIC register.
8149         (arc_expand_prologue): Likewise.
8150         * config/arc/arc-protos.h (arc_set_default_type_attributes):
8151         Remove.
8152         (arc_verify_short): Likewise.
8153         (arc_attr_type): Likewise.
8154         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8155         (walk_stores): Likewise.
8156         (arc_address_cost): Make it static.
8157         (arc_verify_short): Likewise.
8158         (branch_dest): Likewise.
8159         (arc_attr_type): Likewise.
8160         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8161         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8162         (arc_final_prescan_insn): Remove inserting the nops due to
8163         hardware hazards.  It is done in reorg step.
8164         (insn_length_variant_t): Remove.
8165         (insn_length_parameters_t): Likewise.
8166         (arc_insn_length_parameters): Likewise.
8167         (arc_get_insn_variants): Likewise.
8168         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8169
8170 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
8171
8172         PR inline-asm/84683
8173         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8174         assertion failure.
8175
8176         PR tree-optimization/84687
8177         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8178         on new_node->decl.
8179         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8180
8181 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8182
8183         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8184         Rename to ppc_speculation_barrier.
8185         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8186         __builtin_ppc_speculation_barrier.
8187
8188 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
8189
8190         PR target/84700
8191         * combine.c (combine_simplify_rtx): Don't try to simplify if
8192         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8193         are equal to x.
8194
8195 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
8196
8197         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8198         to 32 bytes when compiling for POWER9.
8199
8200 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
8201
8202         PR target/84564
8203         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8204         regparm >= 3 with no arg reg available also for calls with
8205         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
8206
8207         PR target/84524
8208         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8209         orig,vex.
8210         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
8211
8212 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
8213
8214         PR target/84264
8215         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8216
8217 2018-03-05  Richard Biener  <rguenther@suse.de>
8218
8219         PR tree-optimization/84486
8220         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8221         When inserting a __builtin_assume_aligned call set the LHS
8222         SSA name alignment info accordingly.
8223
8224 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
8225
8226         PR tree-optimization/84114
8227         * config/aarch64/aarch64.c (aarch64_reassociation_width)
8228         Avoid reassociation of FLOAT_MODE addition.
8229
8230 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
8231
8232         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8233         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8234         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8235         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8236         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8237         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8238         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8239         and -mwbnoinvd.
8240         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8241         __builtin_ia32_wbinvd): New builtins.
8242         (SPECIAL_ARGS2): New.
8243         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8244         (SPECIAL_ARGS2): New.
8245         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8246         (ix86_valid_target_attribute_inner_p): Ditto.
8247         (ix86_init_mmx_sse_builtins): Add special_args2.
8248         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8249         TARGET_WBNOINVD_P): New.
8250         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8251         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8252         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8253         * config/i386/immintrin.h (_wbinvd): New intrinsic.
8254         * config/i386/pconfigintrin.h: New file.
8255         * config/i386/wbnoinvdintrin.h: Ditto.
8256         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8257         wbnoinvdintrin.h.
8258         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8259
8260 2018-03-05  Richard Biener  <rguenther@suse.de>
8261
8262         PR tree-optimization/84670
8263         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8264         member.
8265         (BB_VISITED_WITH_VISITED_SUCCS): New define.
8266         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8267         (compute_antic_aux): Only assert the number of values in ANTIC_IN
8268         doesn't grow if all successors (recursively) were visited at least
8269         once.
8270
8271 2018-03-05  Richard Biener  <rguenther@suse.de>
8272
8273         PR tree-optimization/84650
8274         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8275         if executed in the loop pipeline.
8276
8277 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
8278
8279         * doc/configfiles.texi (Configuration Files): Move info about
8280         conditionalizing $target-protos.h to...
8281         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
8282         differs from $target-protos.h.
8283
8284 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
8285             Chung-Ju Wu  <jasonwucj@gmail.com>
8286
8287         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8288         * config/nds32/nds32-multiple.md (setmemsi): Define.
8289         * config/nds32/nds32-memory-manipulation.c
8290         (nds32_gen_dup_4_byte_to_word_value): New.
8291         (emit_setmem_word_loop): New.
8292         (emit_setmem_byte_loop): New.
8293         (nds32_expand_setmem_loop): New.
8294         (nds32_expand_setmem_loop_v3m): New.
8295         (nds32_expand_setmem_unroll): New.
8296         (nds32_expand_setmem): New.
8297
8298 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
8299             Chung-Ju Wu  <jasonwucj@gmail.com>
8300
8301         * config/nds32/nds32-memory-manipulation.c
8302         (nds32_emit_load_store): New.
8303         (nds32_emit_post_inc_load_store): New.
8304         (nds32_emit_mem_move): New.
8305         (nds32_emit_mem_move_block): New.
8306         (nds32_expand_movmemsi_loop_unknown_size): New.
8307         (nds32_expand_movmemsi_loop_known_size): New.
8308         (nds32_expand_movmemsi_loop): New.
8309         (nds32_expand_movmemsi_unroll): New.
8310         (nds32_expand_movmemqi): Rename ...
8311         (nds32_expand_movmemsi): ... to this.
8312         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8313         (movmemsi): ... to this.
8314         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8315         (nds32_expand_movmemsi): ... to this.
8316
8317 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
8318             Monk Chiang  <sh.chiang04@gmail.com>
8319             Chung-Ju Wu  <jasonwucj@gmail.com>
8320
8321         * config/nds32/nds32-protos.h
8322         (nds32_expand_load_multiple): New arguments.
8323         (nds32_expand_store_multiple): Ditto.
8324         (nds32_valid_multiple_load_store): Rename ...
8325         (nds32_valid_multiple_load_store_p): ... to this.
8326         * config/nds32/nds32-memory-manipulation.c
8327         (nds32_expand_load_multiple): Refine implementation.
8328         (nds32_expand_store_multiple): Ditto.
8329         * config/nds32/nds32-multiple.md
8330         (load_multiple): Update nds32_expand_load_multiple interface.
8331         (store_multiple): Update nds32_expand_store_multiple interface.
8332         * config/nds32/nds32-predicates.c
8333         (nds32_valid_multiple_load_store): Rename ...
8334         (nds32_valid_multiple_load_store_p): ... to this and refine
8335         implementation.
8336         * config/nds32/predicates.md
8337         (nds32_load_multiple_and_update_address_operation): New predicate.
8338         (nds32_store_multiple_and_update_address_operation): New predicate.
8339
8340 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
8341             Chung-Ju Wu  <jasonwucj@gmail.com>
8342
8343         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8344         (combo): New attribute.
8345         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8346
8347 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
8348
8349         * config/nds32/nds32.opt: Change -mcmodel= default value.
8350
8351 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
8352             Monk Chiang  <sh.chiang04@gmail.com>
8353             Chung-Ju Wu  <jasonwucj@gmail.com>
8354
8355         * config/nds32/constants.md (unspec_element): New enum.
8356         * config/nds32/constraints.md (Umw): New constraint.
8357         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8358         * config/nds32/nds32-intrinsic.md: Likewise.
8359         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8360         (nds32_valid_smw_lwm_base_p): New.
8361         (nds32_output_smw_single_word): New.
8362         (nds32_output_lmw_single_word): New.
8363         (nds32_expand_unaligned_load): New.
8364         (nds32_expand_unaligned_store): New.
8365         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8366         (nds32_output_smw_single_word): Declare.
8367         (nds32_output_lmw_single_word): Declare.
8368         (nds32_expand_unaligned_load): Declare.
8369         (nds32_expand_unaligned_store): Declare.
8370         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8371         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8372         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8373         NDS32_BUILTIN_UASTORE_DW.
8374         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8375         predicate.
8376
8377 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
8378             Kito Cheng  <kito.cheng@gmail.com>
8379             Chung-Ju Wu  <jasonwucj@gmail.com>
8380
8381         * config/nds32/nds32-intrinsic.c
8382         (nds32_expand_builtin_null_ftype_reg): Delete.
8383         (nds32_expand_builtin_reg_ftype_imm): Ditto.
8384         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8385         (nds32_read_argument): New.
8386         (nds32_legitimize_target): Ditto.
8387         (nds32_legitimize_argument): Ditto.
8388         (nds32_check_constant_argument): Ditto.
8389         (nds32_expand_unop_builtin): Ditto.
8390         (nds32_expand_unopimm_builtin): Ditto.
8391         (nds32_expand_binop_builtin): Ditto.
8392         (nds32_builtin_decl_impl): Ditto.
8393         (builtin_description): Ditto.
8394         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8395         (nds32_init_builtins_impl): Ditto.
8396         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8397         (nds32_builtin_decl): New.
8398         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8399         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8400
8401 2018-03-02  Jeff Law  <law@redhat.com>
8402
8403         * reorg.c (stop_search_p): Handle DEBUG_INSN.
8404         (redundant_insn, fill_simple_delay_slots): Likewise.
8405         (fill_slots_from_thread): Likewise.
8406         * resource.c (mark_referenced_resources): Likewise.
8407         (mark_set_resources, find_dead_or_set_registers): Likewise.
8408
8409 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
8410
8411         * substring-locations.h (format_warning_va): Formatting fix for
8412         ATTRIBUTE_GCC_DIAG.
8413         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8414         argument.
8415         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8416         * substring-locations.c: Include intl.h.
8417         (format_warning_va): Turned into small wrapper around
8418         format_warning_n_va, renamed to ...
8419         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8420         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8421         use ngettext.
8422         (format_warning_at_substring_n): New function.
8423         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8424         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
8425         format_warning_at_substring with just a shorter name instead of
8426         const function pointer.
8427         (fmtwarn_n): New function.
8428         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8429         appropriate, get rid of all the fmtstr temporaries, move conditionals
8430         with G_() wrapped string literals directly into fmtwarn arguments,
8431         cast dir.len to (int), formatting fixes.
8432
8433 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
8434
8435         * doc/invoke.texi: Remove "Cilk Plus" references.
8436
8437 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
8438             Richard Biener  <rguenther@suse.de>
8439
8440         PR ipa/84628
8441         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8442         for error or warning attributes if CALL_FROM_THUNK_P is set.
8443         Formatting fixes.
8444
8445 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
8446
8447         PR target/56540
8448         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8449         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8450
8451         PR target/56540
8452         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8453         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8454
8455         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8456         instead of -1U in last predictors element's probability member.
8457
8458 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
8459
8460         PR ipa/83983
8461         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8462         arguments if they are comparable.
8463
8464 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
8465
8466         PR tree-optimization/84634
8467         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8468         masks and masked_loop_p with a single loop_masks, making sure it's
8469         null for bb vectorization.
8470
8471 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
8472
8473         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8474         (vect_analyze_data_ref_access): Use loop->safe_len rather than
8475         loop->force_vectorize to check whether there is no alias.
8476
8477 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
8478
8479         PR target/84614
8480         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8481         prototypes.
8482         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8483         comments.
8484         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8485         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8486         instead of a loop around prev_real_insn.
8487         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8488         prev_real_insn.
8489
8490         PR inline-asm/84625
8491         * config/i386/i386.c (ix86_print_operand): Use conditional
8492         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8493         zero vector.
8494
8495 2018-03-02  Richard Biener  <rguenther@suse.de>
8496
8497         PR tree-optimization/84427
8498         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8499         (bitmap_set_subtract_values): Rewrite to handle multiple
8500         exprs per value.
8501         (clean): Likewise.
8502         (prune_clobbered_mems): Likewise.
8503         (phi_translate): Take edge instead of pred/phiblock.
8504         (phi_translate_1): Likewise.
8505         (phi_translate_set): Likewise.  Insert all translated
8506         exprs for a value into the set, keeping possibly multiple
8507         expressions per value.
8508         (compute_antic_aux): Adjust for phi_translate changes.
8509         When intersecting union the expressions and prune those
8510         not in the final value set, keeping possibly multiple
8511         expressions per value.  Do not use value-insertion
8512         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8513         all expressions.  Add verification that the value-sets
8514         only shrink during iteration.
8515         (compute_partial_antic_aux): Adjust for the phi_translate changes.
8516         (do_pre_regular_insertion): Likewise.
8517         (do_pre_partial_partial_insertion): Likewise.
8518
8519 2018-03-02  Richard Biener  <rguenther@suse.de>
8520
8521         PR target/82005
8522         * config/darwin.c (saved_debug_info_level): New static global.
8523         (darwin_asm_lto_start): Disable debug info generation for LTO out.
8524         (darwin_asm_lto_end): Restore debug info generation settings.
8525
8526 2018-03-01  Martin Liska  <mliska@suse.cz>
8527
8528         PR sanitizer/82484
8529         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8530         volatile arguments.
8531
8532 2018-03-01  Richard Biener  <rguenther@suse.de>
8533
8534         PR debug/84645
8535         * dwarf2out.c (gen_variable_die): Properly handle late VLA
8536         type annotation with LTO when debug was disabled at compile-time.
8537
8538 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
8539
8540         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8541         XINT with INTVAL.
8542         (mips_final_postscan_insn): Likewise.
8543
8544 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
8545
8546         PR rtl-optimization/84528
8547         * alias.c (init_alias_target): Add commentary.
8548         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8549         a unique base value if the frame pointer is not eliminated
8550         to the stack pointer.
8551
8552 2018-03-01  Tom de Vries  <tom@codesourcery.com>
8553
8554         PR rtl-optimization/83327
8555         * lra-int.h (hard_regs_spilled_into): Declare.
8556         * lra.c (hard_regs_spilled_into): Define.
8557         (init_reg_info): Init hard_regs_spilled_into.
8558         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8559         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8560         (process_bb_lives): Handle hard_regs_spilled_into.
8561         (lra_create_live_ranges_1): Before doing liveness propagation, clear
8562         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8563
8564 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
8565
8566         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8567         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8568         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8569         * config/rs6000/aix72.h: New file.
8570
8571 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
8572
8573         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8574         instead of warning_at with conditional singular and plural messages
8575         where possible.
8576
8577         PR target/52991
8578         * stor-layout.c (update_alignment_for_field): For
8579         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8580         && !DECL_PACKED (field), do the alignment update, just use
8581         only desired_align instead of MAX (type_align, desired_align)
8582         as the alignment.
8583         (place_field): Don't do known_align < desired_align handling
8584         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8585         is non-NULL, instead do it after rli->prev_field handling and
8586         only if not within a bitfield word.  For DECL_PACKED (field)
8587         use type_align of BITS_PER_UNIT.
8588
8589 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
8590
8591         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8592         superfluous parentheses and trailing spaces.
8593
8594 2018-02-28  Richard Biener  <rguenther@suse.de>
8595
8596         PR tree-optimization/84584
8597         * graphite-scop-detection.c (scop_detection::add_scop): Discard
8598         SCoPs with fake exit edge.
8599
8600 2018-02-28  Martin Liska  <mliska@suse.cz>
8601
8602         PR testsuite/84597
8603         * timevar.c (timer::print): Fix format to properly print 100%
8604         values.
8605
8606 2018-02-28  Richard Biener  <rguenther@suse.de>
8607
8608         PR middle-end/84607
8609         * genmatch.c (capture_info::walk_match): Do not mark
8610         captured expressions without operands as expr_p given
8611         they act more like predicates and should be subject to
8612         "lost tail" side-effect preserving.
8613
8614 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
8615
8616         PR rtl-optimization/81611
8617         * auto-inc-dec.c (attempt_change): Move dead note from
8618         mem_insn if it's the next use of regno
8619         (find_address): Take address use of reg holding
8620         non-incremented value.  Add parm to limit search to the named
8621         reg only.
8622         (merge_in_block): Attempt to use a mem insn that is the next
8623         use of the original regno.
8624
8625 2018-02-27  Martin Sebor  <msebor@redhat.com>
8626
8627         PR c++/83871
8628         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8629         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8630
8631 2018-02-27  Martin Sebor  <msebor@redhat.com>
8632
8633         PR translation/84207
8634         * diagnostic-core.h (warning_n, error_n, inform_n): Change
8635         n argument to unsigned HOST_WIDE_INT.
8636         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8637         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
8638         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8639         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8640
8641 2018-02-27  Richard Biener  <rguenther@suse.de>
8642
8643         PR tree-optimization/84512
8644         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8645         Do not use the estimate returned from record_stmt_cost for
8646         the scalar iteration cost but sum properly using add_stmt_cost.
8647
8648 2018-02-27  Richard Biener  <rguenther@suse.de>
8649
8650         PR tree-optimization/84466
8651         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8652         Adjust last change to less strictly validate use operands.
8653
8654 2018-02-27  Martin Liska  <mliska@suse.cz>
8655
8656         PR gcov-profile/84548
8657         * gcov.c (process_file): Allow partial overlap and consider it
8658         also as group functions.
8659         (output_lines): Properly calculate range of lines for a group.
8660
8661 2018-02-27  Martin Liska  <mliska@suse.cz>
8662
8663         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8664         'ggc' suffixes.  Change first column width.
8665         (timer::print): Fix formatting of the column.
8666
8667 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
8668
8669         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8670         preserve inline entry blocks for the sake of debug inline
8671         entry point markers alone.
8672         (remove_unused_locals): Suggest in comments a better place to
8673         force the preservation of inline entry blocks that are
8674         otherwise unused, but do not preserve them.
8675
8676 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
8677
8678         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8679
8680 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
8681
8682         PR target/84039
8683         * config/i386/constraints.md (Bs): Replace
8684         ix86_indirect_branch_register with
8685         TARGET_INDIRECT_BRANCH_REGISTER.
8686         (Bw): Likewise.
8687         * config/i386/i386.md (indirect_jump): Likewise.
8688         (tablejump): Likewise.
8689         (*sibcall_memory): Likewise.
8690         (*sibcall_value_memory): Likewise.
8691         Peepholes of indirect call and jump via memory: Likewise.
8692         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8693         (*sibcall_value_GOT_32): Likewise.
8694         * config/i386/predicates.md (indirect_branch_operand): Likewise.
8695         (GOT_memory_operand): Likewise.
8696         (call_insn_operand): Likewise.
8697         (sibcall_insn_operand): Likewise.
8698         (GOT32_symbol_operand): Likewise.
8699         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8700
8701 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
8702
8703         PR rtl-optimization/83496
8704         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8705         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
8706         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8707         redundant insn, if any.
8708         (relax_delay_slots): Likewise.
8709         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8710
8711 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
8712
8713         PR tree-optimization/83965
8714         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8715         that grouped statements are part of a reduction chain.  Return
8716         true if the statement is not marked as a reduction itself but
8717         is part of a group.
8718         (vect_recog_dot_prod_pattern): Don't check whether the statement
8719         is part of a group here.
8720         (vect_recog_sad_pattern): Likewise.
8721         (vect_recog_widen_sum_pattern): Likewise.
8722
8723 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
8724
8725         PR debug/84545
8726         * final.c (rest_of_clean_state): Also look for calls inside sequences.
8727
8728 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
8729
8730         PR target/84530
8731         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8732         the bool argument.
8733         (ix86_output_indirect_function_return): New prototype.
8734         (ix86_split_simple_return_pop_internal): Likewise.
8735         * config/i386/i386.c (indirect_return_via_cx): New.
8736         (indirect_return_via_cx_bnd): Likewise.
8737         (indirect_thunk_name): Handle return va CX_REG.
8738         (output_indirect_thunk_function): Create alias for
8739         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8740         (ix86_output_indirect_jmp): Remove the bool argument.
8741         (ix86_output_indirect_function_return): New function.
8742         (ix86_split_simple_return_pop_internal): Likewise.
8743         * config/i386/i386.md (*indirect_jump): Don't pass false
8744         to ix86_output_indirect_jmp.
8745         (*tablejump_1): Likewise.
8746         (simple_return_pop_internal): Change it to define_insn_and_split.
8747         Call ix86_split_simple_return_pop_internal to split it for
8748         -mfunction-return=.
8749         (simple_return_indirect_internal): Call
8750         ix86_output_indirect_function_return instead of
8751         ix86_output_indirect_jmp.
8752
8753 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
8754
8755         PR bootstrap/84405
8756         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8757         memset and value initialization afterwards.
8758
8759 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
8760
8761         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8762
8763 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
8764
8765         PR target/84521
8766         * common/config/aarch64/aarch64-common.c
8767         (aarch_option_optimization_table[]): Switch
8768         off fomit-frame-pointer
8769
8770 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
8771             Chung-Ju Wu  <jasonwucj@gmail.com>
8772
8773         * config/nds32/nds32-multiple.md (load_multiple): Disallow
8774         volatile memory.
8775         (store_multiple): Ditto.
8776
8777 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
8778
8779         * config.gcc: Add --with-cpu support for nds32 target.
8780         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8781         * config/nds32/nds32.opt: Add -mcpu= option.
8782
8783 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
8784
8785         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8786         isel=yes): Warn for these deprecated options.
8787
8788 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
8789
8790         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8791         ISA_2_5_MASKS_EMBEDDED.
8792
8793 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
8794
8795         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8796         p->max as pointers rather than using iterative_hash_expr.
8797
8798 2018-02-23  Carl Love  <cel@us.ibm.com>
8799
8800         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8801         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8802         BU_P8V_OVERLOAD_2.
8803         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8804         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8805         P8V_BUILTIN_VEC_VUNSIGNED2.
8806
8807 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
8808
8809         PR target/81572
8810         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8811         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8812         LRA_UNKNOWN_ALT.
8813         * lra-constraints.c (curr_insn_transform): Set up
8814         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
8815         LRA_UNKNOWN_ALT.
8816         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8817         * lra-eliminations.c (spill_pseudos): Ditto.
8818         (process_insn_for_elimination): Ditto.
8819         * lra-lives.c (reg_early_clobber_p): Use the new macros.
8820         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8821         LRA_NON_CLOBBERED_ALT.
8822
8823 2018-02-22  Martin Sebor  <msebor@redhat.com>
8824
8825         PR tree-optimization/84480
8826         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8827         to maybe_diag_stxncpy_trunc.  Call it.
8828         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8829         from gimple_fold_builtin_strcpy.  Print inlining stack.
8830         (handle_builtin_stxncpy): Print inlining stack.
8831         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8832
8833 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
8834
8835         PR target/84176
8836         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8837         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8838         and -fcheck-pointer-bounds are used together.
8839         (indirect_thunk_prefix): New enum.
8840         (indirect_thunk_need_prefix): New function.
8841         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
8842         "_nt" instead of "_bnd" for NOTRACK prefix.
8843         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8844         (output_indirect_thunk_function): Likewise.
8845         (): Likewise.
8846         (ix86_code_end): Update output_indirect_thunk_function calls.
8847         (ix86_output_indirect_branch_via_reg): Replace
8848         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8849         (ix86_output_indirect_branch_via_push): Likewise.
8850         (ix86_output_function_return): Likewise.
8851         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8852         incompatible with -fcf-protection=branch and
8853         -fcheck-pointer-bounds.
8854
8855 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
8856
8857         PR target/83335
8858         * config/aarch64/aarch64.c (aarch64_print_address_internal):
8859         Change gcc_assert call to output_operand_lossage.
8860
8861 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
8862
8863         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8864
8865 2018-02-22  DJ Delorie  <dj@redhat.com>
8866             Sebastian Perta  <sebastian.perta@renesas.com>
8867             Oleg Endo  <olegendo@gcc.gnu.org>
8868
8869         * config/rx/rx.c (rx_rtx_costs): New function.
8870         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8871
8872 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8873
8874         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8875
8876 2018-02-22  Martin Liska  <mliska@suse.cz>
8877
8878         PR driver/83193
8879         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8880         Add "native" as a possible value.
8881
8882 2018-02-22  Martin Liska  <mliska@suse.cz>
8883
8884         PR driver/83193
8885         * config/i386/i386.c (ix86_option_override_internal):
8886         Add "native" as a possible value for -march and -mtune.
8887
8888 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
8889
8890         PR target/84502
8891         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8892         to all type variants.
8893
8894         PR tree-optimization/84503
8895         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8896         width as info->bitpos + info->bitsize - start.
8897         (merged_store_group::merge_overlapping): Simplify width computation.
8898         (check_no_overlap): New function.
8899         (imm_store_chain_info::try_coalesce_bswap): Compute expected
8900         start + width and last_order of the group, fail if check_no_overlap
8901         fails.
8902         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8903         to group if check_no_overlap fails.
8904
8905 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
8906
8907         * config/rs6000/altivec.md: Delete contraint arguments to
8908         define_expand, define_split, and define_peephole2, and in
8909         define_insn_and_split if always unused.
8910         * config/rs6000/darwin.md: Ditto.
8911         * config/rs6000/dfp.md: Ditto.
8912         * config/rs6000/rs6000.md: Ditto.
8913         * config/rs6000/sync.md: Ditto.
8914         * config/rs6000/vector.md: Ditto.
8915         * config/rs6000/vsx.md: Ditto.
8916
8917 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
8918
8919         * config/rs6000/altivec.md: Write output control strings as braced
8920         blocks instead of double-quoted strings.
8921         * config/rs6000/darwin.md: Ditto.
8922         * config/rs6000/rs6000.md: Ditto.
8923         * config/rs6000/vector.md: Ditto.
8924         * config/rs6000/vsx.md: Ditto.
8925
8926 2018-02-21  Jason Merrill  <jason@redhat.com>
8927
8928         PR c++/84314 - ICE with templates and fastcall attribute.
8929         * attribs.c (build_type_attribute_qual_variant): Remove assert.
8930
8931 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
8932
8933         * ipa-cp.c (determine_versionability): Fix comment typos.
8934
8935 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
8936
8937         PR c/84229
8938         * ipa-cp.c (determine_versionability): Do not version functions caling
8939         va_arg_pack.
8940
8941 2018-02-21  Martin Liska  <mliska@suse.cz>
8942
8943         PR driver/83193
8944         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8945         Add "native" as a possible value.
8946         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
8947         the macro when native cpu detection is available.
8948
8949 2018-02-21  Martin Liska  <mliska@suse.cz>
8950
8951         PR driver/83193
8952         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8953         Add "native" as a possible value.
8954         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8955         when native cpu detection is available.
8956
8957 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
8958             Martin Sebor  <msebor@redhat.com>
8959
8960         PR tree-optimization/84478
8961         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8962         false.
8963         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8964         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8965         support which is conservatively correct, for 2 only stay conservative
8966         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
8967         argument to the 2 argument get_range_strlen, adjust 6 arg
8968         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8969         false.
8970         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8971         (gimple_fold_builtin_strlen): Pass true as last argument to
8972         get_range_strlen.
8973
8974 2018-02-20  Martin Sebor  <msebor@redhat.com>
8975
8976         PR middle-end/84095
8977         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8978         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
8979         (builtin_memref::builtin_memref): Factor out parts into
8980         set_base_and_offset and call it.
8981
8982 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
8983
8984         PR middle-end/84406
8985         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8986         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8987         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
8988         search at the associated MODE_INT.
8989
8990 2018-02-20  Jeff Law  <law@redhat.com>
8991
8992         PR middle-end/82123
8993         PR tree-optimization/81592
8994         PR middle-end/79257
8995         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8996         for range data rather than using global data.
8997         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8998         range data rather than using global data.
8999         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9000         pass it to children as needed.
9001         (struct directive::fmtresult): Similarly.
9002         (struct directive::set_width): Similarly.
9003         (struct directive::set_precision): Similarly.
9004         (format_integer, format_directive, parse_directive): Similarly.
9005         (format_none): Accept unnamed vr_values parameter.
9006         (format_percent, format_floating, format_character): Similarly.
9007         (format_string, format_plain): Similarly.
9008         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9009         the EVRP range analyzer for range data rather than using global data.
9010         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9011         gimple-ssa-evrp-analyze.h
9012         (class sprintf_dom_walker): Add after_dom_children member function.
9013         Add evrp_range_analyzer member.
9014         (sprintf_dom_walker::before_dom_children): Call into the EVRP
9015         range analyzer as needed.
9016         (sprintf_dom_walker::after_dom_children): New member function.
9017         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9018         if not optimizing.
9019         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9020         (evrp_range_analyzer::pop_to_marker): Likewise.
9021
9022 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
9023
9024         PR tree-optimization/84419
9025         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9026         with the required type if its current type is compatible but
9027         different.
9028
9029 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
9030
9031         PR middle-end/82004
9032         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9033         after vectorization.
9034
9035 2018-02-20  Martin Liska  <mliska@suse.cz>
9036
9037         PR driver/83193
9038         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9039         possible values if we don't have a hint.
9040
9041 2018-02-20  Martin Liska  <mliska@suse.cz>
9042
9043         PR c/84310
9044         PR target/79747
9045         * final.c (shorten_branches): Build align_tab array with one
9046         more element.
9047         * opts.c (finish_options): Add alignment option limit check.
9048         (MAX_CODE_ALIGN): Likewise.
9049         (MAX_CODE_ALIGN_VALUE): Likewise.
9050         * doc/invoke.texi: Document maximum allowed option value for
9051         all -falign-* options.
9052
9053 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
9054
9055         PR target/84146
9056         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9057         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9058         * var-tracking.c (emit_note_insn_var_location): Remove all references
9059         to NOTE_INSN_CALL_ARG_LOCATION.
9060         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9061         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9062         Use copy_rtx_if_shared.
9063         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9064         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9065         (dwarf2out_var_location): Remove handling of
9066         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9067         on call_insn.
9068         * final.c (final_scan_insn): Remove all references to
9069         NOTE_INSN_CALL_ARG_LOCATION.
9070         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
9071         before dumping final insns.
9072         * except.c (emit_note_eh_region_end): Remove all references to
9073         NOTE_INSN_CALL_ARG_LOCATION.
9074         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9075         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9076         * config/arc/arc.c (hwloop_optimize): Likewise.
9077         * config/arm/arm.c (create_fix_barrier): Likewise.
9078         * config/s390/s390.c (s390_chunkify_start): Likewise.
9079         * config/sh/sh.c (find_barrier): Likewise.
9080         * config/i386/i386.c (rest_of_insert_endbranch,
9081         ix86_seh_fixup_eh_fallthru): Likewise.
9082         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9083         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9084         * config/frv/frv.c (frv_function_prologue): Likewise.
9085         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
9086         reg note.
9087         (note_outside_basic_block_p): Remove all references to
9088         NOTE_INSN_CALL_ARG_LOCATION.
9089         * gengtype.c (adjust_field_rtx_def): Likewise.
9090         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9091         Likewise.
9092         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9093         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9094
9095         PR c++/84444
9096         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9097         is ADDR_EXPR.
9098
9099         PR tree-optimization/84452
9100         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9101         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9102         is NULL.
9103
9104 2018-02-19  Martin Liska  <mliska@suse.cz>
9105
9106         PR sanitizer/82183
9107         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9108
9109 2018-02-19  Martin Liska  <mliska@suse.cz>
9110             Richard Sandiford  <richard.sandiford@linaro.org>
9111
9112         PR tree-optimization/82491
9113         * gimple-fold.c (get_base_constructor): Make earlier bail out
9114         to prevent ubsan.
9115
9116 2018-02-19  Carl Love  <cel@us.ibm.com>
9117
9118         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9119         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9120         BU_P8V_OVERLOAD_1.
9121         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9122         P8V_BUILTIN_VEC_NEG.
9123
9124 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
9125
9126         * config/rl78/rl78.md (movdf): New define expand.
9127
9128 2018-02-19  Martin Liska  <mliska@suse.cz>
9129
9130         PR other/80589
9131         * doc/invoke.texi: Fix typo.
9132         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9133
9134 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
9135
9136         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9137         handle rs6000_single_float and rs6000_double_float specially for
9138         e500 family CPUs.
9139
9140 2018-02-16  Jeff Law  <law@redhat.com>
9141
9142         * config/rx/rx.c (add_pop_cfi_notes): New function.;
9143         (pop_regs): Use it.
9144
9145 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
9146
9147         PR ipa/84425
9148         * ipa-inline.c (inline_small_functions): Fix a typo.
9149
9150 2018-02-16  Nathan Sidwell  <nathan@acm.org>
9151
9152         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9153
9154 2018-02-16  Carl Love  <cel@us.ibm.com>
9155
9156         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9157         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9158         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9159         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9160         expansion to P8V_BUILTIN_VEC_FLOAT2.
9161
9162 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
9163
9164         PR rtl-optimization/70023
9165         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9166         src_regno into account.
9167
9168 2018-02-16  Carl Love  <cel@us.ibm.com>
9169
9170         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9171         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9172         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9173         * config/rs6000/rs6000.c: Remove case statements for
9174         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9175         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9176         and P9V_BUILTIN_VEC_VINSERT4B.
9177         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9178         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9179         * config/rs6000/vsx.md:
9180         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9181         vec_insert4b.
9182
9183 2018-02-16  Carl Love  <cel@us.ibm.com>
9184
9185         * config/rs6000/altivec.h: Add builtin names vec_extract4b
9186         vec_insert4b.
9187         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9188         definitions.
9189         * config/rs6000/rs6000-c.c: Add the definitions for
9190         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9191         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9192         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9193         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
9194         definition for insert4b and define insn *insert3b_internal.
9195         * doc/extend.texi: Add documentation for vec_extract4b.
9196
9197 2018-02-16  Nathan Sidwell  <nathan@acm.org>
9198
9199         * doc/extend.texi (Backwards Compatibility): Mention friend
9200         injection.  Note for-scope is deprecated.
9201         * doc/invoke.texi (-ffriend-injection): Deprecate.
9202
9203 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
9204
9205         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9206         that moved to I2, also allow destinations that are a paradoxical
9207         subreg (instead of a normal reg).
9208
9209 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
9210
9211         PR target/83831
9212         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9213         to QImode.
9214
9215 2018-02-16  Richard Biener  <rguenther@suse.de>
9216
9217         PR tree-optimization/84037
9218         PR tree-optimization/84016
9219         PR target/82862
9220         * config/i386/i386.c (ix86_builtin_vectorization_cost):
9221         Adjust vec_construct for the fact we need additional higher latency
9222         128bit inserts for AVX256 and AVX512 vector builds.
9223         (ix86_add_stmt_cost): Scale vector construction cost for
9224         elementwise loads.
9225
9226 2018-02-16  Richard Biener  <rguenther@suse.de>
9227
9228         PR tree-optimization/84417
9229         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9230         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9231         (non_rewritable_lvalue_p): Likewise, use poly-ints.
9232
9233 2018-02-16  Martin Liska  <mliska@suse.cz>
9234
9235         PR sanitizer/84307
9236         * internal-fn.def (ASAN_CHECK): Set proper flags.
9237         (ASAN_MARK): Likewise.
9238
9239 2018-02-16  Julia Koval  <julia.koval@intel.com>
9240
9241         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9242         from PTA_CANNONLAKE.
9243
9244 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
9245
9246         PR target/84272
9247         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9248         Use ++iter rather than iter++ for std::list iterators.
9249         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
9250         defer deleting them until all nodes in the forest are processed.  Do
9251         free even leaf nodes.  Change to_process into auto_vec.
9252
9253         PR bootstrap/84405
9254         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9255         * vec.h (vec_default_construct): Use memset instead of placement new
9256         if BROKEN_VALUE_INITIALIZATION is defined.
9257         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9258         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9259         is defined.
9260
9261         PR rtl-optimization/83723
9262         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9263         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
9264         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
9265         recursive calls.
9266         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9267         callers.
9268         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9269
9270 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
9271
9272         PR rtl-optimization/81443
9273         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9274         from inner REGs to paradoxical SUBREGs.
9275
9276 2018-02-16  Richard Biener  <rguenther@suse.de>
9277
9278         PR tree-optimization/84399
9279         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9280         For operands we can analyze at their definition make sure we can
9281         analyze them at each use as well.
9282
9283 2018-02-16  Richard Biener  <rguenther@suse.de>
9284
9285         PR tree-optimization/84190
9286         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9287         volatile accesses if the decl isn't volatile.
9288
9289 2018-02-15  Jason Merrill  <jason@redhat.com>
9290
9291         PR c++/84314 - ICE with templates and fastcall attribute.
9292         * attribs.c (build_type_attribute_qual_variant): Don't clobber
9293         TYPE_CANONICAL on an existing type.
9294
9295 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
9296
9297         PR tree-optimization/84383
9298         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9299         dstoff nor call operand_equal_p if dstbase is NULL.
9300
9301         PR tree-optimization/84334
9302         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9303         also a CONSTANT_CLASS_P, punt.
9304
9305 2018-02-14  Jim Wilson  <jimw@sifive.com>
9306
9307         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9308         first SMALL_OPERAND check.  New local min_second_step.  Move assert
9309         to where locals are set.  Add TARGET_RVC support.
9310         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9311
9312 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
9313
9314         * doc/invoke.texi: Correct -Wformat-overflow code sample.
9315
9316 2018-02-14  Martin Sebor  <msebor@redhat.com>
9317
9318         PR tree-optimization/83698
9319         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9320         arrays constrain the offset range to their bounds.
9321         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9322         (builtin_access::overlap): Avoid setting the size of overlap if it's
9323         already been set.
9324         (maybe_diag_overlap): Also consider arrays when deciding what values
9325         of offsets to include in diagnostics.
9326
9327 2018-02-14  Martin Sebor  <msebor@redhat.com>
9328
9329         PR c/84108
9330         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9331         that correspond to the kind of a declaration.
9332
9333 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
9334
9335         PR target/83984
9336         * config/pa/pa.md: Load address of PIC label using the linkage table
9337         if the label is nonlocal.
9338
9339 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9340
9341         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9342         warning message if user requests -maltivec=be.
9343         * doc/invoke.texi: Document deprecation of -maltivec=be.
9344
9345 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
9346
9347         PR target/84220
9348         * config/rs6000/rs6000-c.c: Update definitions for
9349         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9350         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9351
9352 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
9353
9354         PR target/84239
9355         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9356         add _get_ssp intrinsics. Remove argument from
9357         __builtin_ia32_rdssp[d|q].
9358         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9359         * config/i386/i386-builtin.def: Remove argument from
9360         __builtin_ia32_rdssp[d|q].
9361         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9362         ix86_expand_special_args_builtin for _rdssp[d|q].
9363         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9364         Clear register before usage.
9365         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9366         Add documentation for new _get_ssp and _inc_ssp intrinsics.
9367
9368 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
9369
9370         PR tree-optimization/84357
9371         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9372         operand 1 of an ARRAY_REF too.
9373
9374 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
9375
9376         PR target/83831
9377         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9378         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9379         declarations.
9380         (set_of_reg): New struct.
9381         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9382         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9383         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9384         functions.
9385         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9386         Split into bitclr, bitset, bitinvert patterns if appropriate.
9387         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9388         use rx_fuse_in_memory_bitop.
9389         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9390         to named insn, correct maximum insn length.
9391
9392 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
9393
9394         PR target/79242
9395         * machmode.def: Define a complex mode for PARTIAL_INT.
9396         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9397         MODE_PARTIAL_INT.
9398         * doc/rtl.texi: Document CSPImode.
9399         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9400         handling.
9401         (msp430_hard_regno_nregs_with_padding): Likewise.
9402
9403 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
9404
9405         PR target/84279
9406         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9407
9408 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
9409
9410         PR rtl-optimization/84169
9411         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
9412         we generated a parallel as new i3 and we split that to new i2 and i3
9413         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
9414         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
9415         those to i2, not i1.  Partially rewrite this scan code.
9416
9417 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
9418
9419         PR c/82210
9420         * stor-layout.c (place_field): For variable length fields, adjust
9421         offset_align afterwards not just based on the field's alignment,
9422         but also on the size.
9423
9424         PR middle-end/84309
9425         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9426         of exps and logs in the use_exp2 case.
9427
9428 2018-02-13  Jeff Law  <law@redhat.com>
9429
9430         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9431         entry for "vector".
9432
9433         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9434         ARGS as unused.
9435
9436 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
9437
9438         PR debug/84342
9439         PR debug/84319
9440         * common.opt (gas-loc-support, gas-locview-support): New.
9441         (ginline-points, ginternal-reset-location-views): New.
9442         * doc/invoke.texi: Document them.  Use @itemx where intended.
9443         (gvariable-location-views): Adjust.
9444         * target.def (reset_location_view): New.
9445         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9446         (TARGET_RESET_LOCATION_VIEW): New.
9447         * doc/tm.texi: Rebuilt.
9448         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9449         (dwarf2out_default_as_locview_support): New.
9450         (output_asm_line_debug_info): Use option variables.
9451         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9452         (output_loc_list): Likewise.
9453         (add_high_low_attributes): Check option variables.
9454         Don't output entry view attribute in strict mode.
9455         (gen_inlined_subroutine_die): Check option variables.
9456         (dwarf2out_inline_entry): Likewise.
9457         (init_sections_and_labels): Likewise.
9458         (dwarf2out_early_finish): Likewise.
9459         (maybe_reset_location_view): New, from...
9460         (dwarf2out_var_location): ... here.  Call it.
9461         * debug.h (dwarf2out_default_as_loc_support): Declare.
9462         (dwarf2out_default_as_locview_support): Declare.
9463         * hooks.c (hook_int_rtx_insn_0): New.
9464         * hooks.h (hook_int_rtx_insn_0): Declare.
9465         * toplev.c (process_options): Take -gas-loc-support and
9466         -gas-locview-support from dwarf2out.  Enable
9467         -gvariable-location-views by default only with locview
9468         assembler support.  Enable -ginternal-reset-location-views by
9469         default only if the target defines the corresponding hook.
9470         Enable -ginline-points by default if location views are
9471         enabled; force it disabled if statement frontiers are
9472         disabled.
9473         * tree-inline.c (expand_call_inline): Check option variables.
9474         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9475
9476 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
9477
9478         PR tree-optimization/84321
9479         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9480         handling.  Also check whether the anti-range contains any values
9481         that satisfy the mask; switch to a VR_RANGE if not.
9482
9483 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
9484
9485         PR sanitizer/84340
9486         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9487
9488 2018-02-13  Martin Jambor  <mjambor@suse.cz>
9489
9490         PR c++/83990
9491         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9492         of call statements, also set location of a load to a temporary.
9493
9494 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
9495
9496         * config/rl78/rl78.c (add_vector_labels): New function.
9497         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9498         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9499         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9500         which checks that no arguments are passed.
9501         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9502         * doc/extend.texi: Documentation for the new attribute.
9503
9504 2018-02-13  Andreas Schwab  <schwab@suse.de>
9505
9506         * config/riscv/linux.h (CPP_SPEC): Define.
9507
9508 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
9509
9510         PR target/84335
9511         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9512         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9513         OPTION_MASK_ISA_AES as first argument to def_builtin_const
9514         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9515         instead of OPTION_MASK_ISA_PCLMUL as first argument to
9516         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9517         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9518         temporarily for AES and PCLMUL builtins.
9519
9520         PR tree-optimization/84339
9521         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9522         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9523         Formatting fixes.
9524
9525         PR middle-end/84309
9526         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9527         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9528         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9529         inline function.
9530         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9531         inline function.
9532         * omp-simd-clone.h: New file.
9533         * omp-simd-clone.c: Include omp-simd-clone.h.
9534         (expand_simd_clones): No longer static.
9535         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9536         cgraph.h and omp-simd-clone.h.
9537         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9538         (vect_recog_widen_shift_pattern): Formatting fix.
9539         (vect_pattern_recog_1): Don't check optab for calls.
9540
9541         PR target/84336
9542         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9543         operands[2] into a REG before using gen_lowpart on it.
9544
9545 2018-02-12  Jeff Law  <law@redhat.com>
9546
9547         PR target/83760
9548         * config/sh/sh.c (find_barrier): Consider a sibling call
9549         a barrier as well.
9550
9551         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9552         successfully back substituting a reg.
9553
9554 2018-02-12  Richard Biener  <rguenther@suse.de>
9555
9556         PR tree-optimization/84037
9557         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9558         parameter, move visited init to caller.
9559         (vect_slp_analyze_operations): Separate cost from validity
9560         check, initialize visited once for all instances.
9561         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9562         for all instances.
9563         * tree-vect-stmts.c (vect_model_simple_cost): Make early
9564         out an assert.
9565         (vect_model_promotion_demotion_cost): Likewise.
9566         (vectorizable_bswap): Guard cost modeling with !slp_node
9567         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9568         SLP stmts.
9569         (vectorizable_call): Likewise.
9570         (vectorizable_conversion): Likewise.
9571         (vectorizable_assignment): Likewise.
9572         (vectorizable_shift): Likewise.
9573         (vectorizable_operation): Likewise.
9574         (vectorizable_store): Likewise.
9575         (vectorizable_load): Likewise.
9576         (vectorizable_condition): Likewise.
9577         (vectorizable_comparison): Likewise.
9578
9579 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
9580
9581         PR sanitizer/84307
9582         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9583         (ASAN_MARK): Fix fnspec to account for return value, change pointer
9584         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9585
9586 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
9587
9588         PR middle-end/83665
9589         * params.def (inline-min-speedup): Increase from 8 to 15.
9590         (max-inline-insns-auto): Decrease from 40 to 30.
9591         * ipa-split.c (consider_split): Add some buffer for function to
9592         be considered inlining candidate.
9593         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9594         default values.
9595
9596 2018-02-12  Richard Biener  <rguenther@suse.de>
9597
9598         PR tree-optimization/84037
9599         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9600         matched stmts if we cannot swap the non-matched ones.
9601
9602 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
9603
9604         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9605         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9606         _mm_maskz_scalef_round_ss): New intrinsics.
9607         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9608         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9609         __builtin_ia32_scalefss_round): Remove.
9610         (__builtin_ia32_scalefsd_mask_round,
9611         __builtin_ia32_scalefss_mask_round): New intrinsics.
9612         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9613         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9614         ((match_operand:VF_128 2 "<round_nimm_predicate>"
9615         "<round_constraint>")): Changed to ...
9616         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9617         "<round_scalar_constraint>")): ... this.
9618         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9619         %0, %1, %2<round_op3>}"): Changed to ...
9620         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9621         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9622         %2<round_scalar_mask_op3>}"): ... this.
9623         * config/i386/subst.md (round_scalar_nimm_predicate): New.
9624
9625 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
9626
9627         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9628         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9629         (_mm_maskz_sqrt_round_ss): New intrinsics.
9630         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9631         (__builtin_ia32_sqrtsd_mask_round)
9632         (__builtin_ia32_sqrtss_mask_round): New builtins.
9633         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9634         (__builtin_ia32_sqrtss_round): Remove.
9635         (__builtin_ia32_sqrtsd_mask_round)
9636         (__builtin_ia32_sqrtss_mask_round): New builtins.
9637         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9638         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9639         ((match_operand:VF_128 1 "vector_operand"
9640         "xBm,<round_constraint>")): Changed to ...
9641         ((match_operand:VF_128 1 "vector_operand"
9642         "xBm,<round_scalar_constraint>")): ... this.
9643         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9644         %0, %2, %<iptr>1<round_op3>}): Changed to ...
9645         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9646         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9647         %<iptr>1<round_scalar_mask_op3>}): ... this.
9648         ((set_attr "prefix" "<round_prefix>")): Changed to ...
9649         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9650
9651 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
9652
9653         PR target/84266
9654         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9655         Cast vec_cmpeq result to correct type.
9656         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9657         Cast vec_cmpgt result to correct type.
9658
9659 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
9660
9661         * final.c (final_scan_insn_1): Renamed from...
9662         (final_scan_insn): ... this.  New wrapper, to recover
9663         seen from the outermost call in recursive ones.
9664         * config/sparc/sparc.c (output_return): Drop seen from call.
9665         (output_sibcall): Likewise.
9666         * config/visium/visium.c (output_branch): Likewise.
9667
9668 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
9669
9670         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9671         function label.
9672
9673 2018-02-10  Alan Modra  <amodra@gmail.com>
9674
9675         PR target/84300
9676         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9677         Specify LR as an input.
9678
9679 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
9680
9681         PR sanitizer/83987
9682         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9683         remove_member_access_dummy_vars): New functions.
9684         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9685         lower_omp_1, execute_lower_omp): Use them.
9686
9687         PR rtl-optimization/84308
9688         * shrink-wrap.c (spread_components): Release todo vector.
9689
9690 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
9691
9692         PR rtl-optimization/57193
9693         * ira-color.c (struct allocno_color_data): Add member
9694         conflict_allocno_hard_prefs.
9695         (update_conflict_allocno_hard_prefs): New.
9696         (bucket_allocno_compare_func): Add a preference based on
9697         conflict_allocno_hard_prefs.
9698         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9699         (color_allocnos): Remove a dead code.  Initiate
9700         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
9701
9702 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
9703
9704         PR target/84226
9705         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9706         constraint from =wa to wa.  Avoid a subreg on the output operand,
9707         instead use a pseudo and subreg it in a move.
9708         (p9_xxbrd_<mode>): Changed to ...
9709         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9710         (p9_xxbrd_v2df): New expander.
9711         (p9_xxbrw_<mode>): Changed to ...
9712         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9713         (p9_xxbrw_v4sf): New expander.
9714
9715 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
9716
9717         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9718
9719 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
9720
9721         PR target/83926
9722         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9723         multiply in 32-bit mode.
9724         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9725         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9726         mode.
9727
9728 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
9729
9730         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9731         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9732         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9733         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9734
9735 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
9736
9737         PR lto/84213
9738         * dwarf2out.c (is_trivial_indirect_ref): New function.
9739         (dwarf2out_late_global_decl): Do not generate a location
9740         attribute for variables that have a non-trivial DECL_VALUE_EXPR
9741         and that are not defined in the current unit.
9742
9743 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
9744
9745         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9746         instead of a libcall for UNORDERED.
9747
9748 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
9749
9750         PR target/82641
9751         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9752         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9753
9754 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
9755
9756         PR target/PR84295
9757         * config/s390/s390.c (s390_set_current_function): Invoke
9758         s390_indirect_branch_settings also if fndecl didn't change.
9759
9760 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
9761
9762         * config/rs6000/rs6000.md (blockage): Set length to zero.
9763
9764 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
9765
9766         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9767
9768 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
9769
9770         PR sanitizer/84285
9771         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9772         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9773         -static-lib*san.
9774
9775         PR debug/84252
9776         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9777         PARALLEL incoming that failed vt_get_decl_and_offset check.
9778
9779         PR middle-end/84237
9780         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9781         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9782         TREE_READONLY bit.
9783         (get_variable_section): For decls in named .bss* sections pass true as
9784         second argument to bss_initializer_p.
9785
9786 2018-02-09  Marek Polacek  <polacek@redhat.com>
9787             Jakub Jelinek  <jakub@redhat.com>
9788
9789         PR c++/83659
9790         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9791         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
9792         Sync some changes from cxx_fold_indirect_ref.
9793
9794 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
9795
9796         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9797         markers.
9798         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9799         (BLOCK_INLINE_ENTRY_LABEL): New.
9800         (dwarf2out_var_location): Disregard inline entry markers.
9801         (inline_entry_data): New struct.
9802         (inline_entry_data_hasher): New hashtable type.
9803         (inline_entry_data_hasher::hash): New.
9804         (inline_entry_data_hasher::equal): New.
9805         (inline_entry_data_table): New variable.
9806         (add_high_low_attributes): Add DW_AT_entry_pc and
9807         DW_AT_GNU_entry_view attributes if a pending entry is found
9808         in inline_entry_data_table.  Add old entry_pc attribute only
9809         if debug nonbinding markers are disabled.
9810         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9811         markers are enabled.
9812         (block_within_block_p, dwarf2out_inline_entry): New.
9813         (dwarf2out_finish): Check that no entries remained in
9814         inline_entry_data_table.
9815         * final.c (reemit_insn_block_notes): Handle inline entry notes.
9816         (final_scan_insn, notice_source_line): Likewise.
9817         (rest_of_clean_state): Skip inline entry markers.
9818         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9819         markers.
9820         * gimple.c (gimple_build_debug_inline_entry): New.
9821         * gimple.h (enum gimple_debug_subcode): Add
9822         GIMPLE_DEBUG_INLINE_ENTRY.
9823         (gimple_build_debug_inline_entry): Declare.
9824         (gimple_debug_inline_entry_p): New.
9825         (gimple_debug_nonbind_marker_p): Adjust.
9826         * insn-notes.def (INLINE_ENTRY): New.
9827         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9828         inline entry marker notes.
9829         (print_insn): Likewise.
9830         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9831         (INSN_DEBUG_MARKER_KIND): Likewise.
9832         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9833         * tree-inline.c (expand_call_inline): Build and insert
9834         debug_inline_entry stmt.
9835         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9836         inline entry blocks early, if nonbind markers are enabled.
9837         (dump_scope_block): Dump fragment info.
9838         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9839         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9840         (gimple_build_debug_inline_entry): New.
9841         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9842         Enable/disable inline entry points too.
9843         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9844         (DEBUG_INSN): Describe inline entry markers.
9845
9846         * common.opt (gvariable-location-views): New.
9847         (gvariable-location-views=incompat5): New.
9848         * config.in: Rebuilt.
9849         * configure: Rebuilt.
9850         * configure.ac: Test assembler for view support.
9851         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9852         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9853         * dwarf2out.c (var_loc_view): New typedef.
9854         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9855         (dwarf2out_locviews_in_attribute): New.
9856         (dwarf2out_locviews_in_loclist): New.
9857         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9858         (enum dw_line_info_opcode): Add LI_adv_address.
9859         (struct dw_line_info_table): Add view.
9860         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9861         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9862         (zero_view_p): New variable.
9863         (ZERO_VIEW_P): New macro.
9864         (output_asm_line_debug_info): New.
9865         (struct var_loc_node): Add view.
9866         (add_AT_view_list, AT_loc_list): New.
9867         (add_var_loc_to_decl): Add view param.  Test it against last.
9868         (new_loc_list): Add view params.  Record them.
9869         (AT_loc_list_ptr): Handle loc and view lists.
9870         (view_list_to_loc_list_val_node): New.
9871         (print_dw_val): Handle dw_val_class_view_list.
9872         (size_of_die): Likewise.
9873         (value_format): Likewise.
9874         (loc_list_has_views): New.
9875         (gen_llsym): Set vl_symbol too.
9876         (maybe_gen_llsym, skip_loc_list_entry): New.
9877         (dwarf2out_maybe_output_loclist_view_pair): New.
9878         (output_loc_list): Output view list or entries too.
9879         (output_view_list_offset): New.
9880         (output_die): Handle dw_val_class_view_list.
9881         (output_dwarf_version): New.
9882         (output_compilation_unit_header): Use it.
9883         (output_skeleton_debug_sections): Likewise.
9884         (output_rnglists, output_line_info): Likewise.
9885         (output_pubnames, output_aranges): Update version comments.
9886         (output_one_line_info_table): Output view numbers in asm comments.
9887         (dw_loc_list): Determine current endview, pass it to new_loc_list.
9888         Call maybe_gen_llsym.
9889         (loc_list_from_tree_1): Adjust.
9890         (add_AT_location_description): Create view list attribute if
9891         needed, check it's absent otherwise.
9892         (convert_cfa_to_fb_loc_list): Adjust.
9893         (maybe_emit_file): Call output_asm_line_debug_info for test.
9894         (dwarf2out_var_location): Reset views as needed.  Precompute
9895         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
9896         attribute.  Set view.
9897         (new_line_info_table): Reset next view.
9898         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9899         (dwarf2out_source_line): Likewise.  Output view resets and labels to
9900         the assembler, or select appropriate line info opcodes.
9901         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9902         (optimize_string_length): Catch it.  Adjust.
9903         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
9904         dw_val_class_view_list, and remove it if no longer needed.
9905         (hash_loc_list): Hash view numbers.
9906         (loc_list_hasher::equal): Compare them.
9907         (optimize_location_lists): Check whether a view list symbol is
9908         needed, and whether the locview attribute is present, and
9909         whether they match.  Remove the locview attribute if no longer
9910         needed.
9911         (index_location_lists): Call skip_loc_list_entry for test.
9912         (dwarf2out_finish): Call output_asm_line_debug_info for test.
9913         Use output_dwarf_version.
9914         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9915         (struct dw_val_node): Add val_view_list.
9916         * final.c (SEEN_NEXT_VIEW): New.
9917         (set_next_view_needed): New.
9918         (clear_next_view_needed): New.
9919         (maybe_output_next_view): New.
9920         (final_start_function): Rename to...
9921         (final_start_function_1): ... this.  Take pointer to FIRST,
9922         add SEEN parameter.  Emit param bindings in the initial view.
9923         (final_start_function): Reintroduce SEEN-less interface.
9924         (final): Rename to...
9925         (final_1): ... this.  Take SEEN parameter.  Output final pending
9926         next view at the end.
9927         (final): Reintroduce seen-less interface.
9928         (final_scan_insn): Output pending next view before switching
9929         sections or ending a block.  Mark the next view as needed when
9930         outputting variable locations.  Notify debug backend of section
9931         changes, and of location view changes.
9932         (rest_of_handle_final): Adjust.
9933         * toplev.c (process_options): Autodetect value for debug variable
9934         location views option.  Warn on incompat5 without -gdwarf-5.
9935         * doc/invoke.texi (gvariable-location-views): New.
9936         (gvariable-location-views=incompat5): New.
9937         (gno-variable-location-views): New.
9938
9939 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
9940
9941         PR tree-optimization/84136
9942         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9943         that the result of find_edge is non-NULL.
9944
9945 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
9946
9947         PR target/83008
9948         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9949         storing integer register in SImode.  Fix cost of 256 and 512
9950         byte aligned SSE register store.
9951
9952 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
9953
9954         * config/i386/i386.c (ix86_multiplication_cost): Fix
9955         multiplication cost for TARGET_AVX512DQ.
9956
9957 2018-02-08  Marek Polacek  <polacek@redhat.com>
9958
9959         PR tree-optimization/84238
9960         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9961         get_range_strlen.
9962
9963 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
9964
9965         PR tree-optimization/84265
9966         * tree-vect-stmts.c (vectorizable_store): Don't treat
9967         VMAT_CONTIGUOUS accesses as grouped.
9968         (vectorizable_load): Likewise.
9969
9970 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
9971
9972         PR tree-optimization/81635
9973         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9974         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9975         (test_round_for_mask): New functions.
9976         (wide_int_cc_tests): Call test_round_for_mask.
9977         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9978         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9979         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9980         range returned by get_range_info.
9981
9982 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
9983
9984         PR ipa/81360
9985         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9986         * symtab.c: Include builtins.h
9987         (symtab_node::output_to_lto_symbol_table_p): Move here
9988         from lto-streamer-out.c:output_symbol_p.
9989         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9990         (output_symbol_p): Move all logic to symtab.c
9991         (produce_symtab): Update.
9992
9993 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
9994
9995         * config/s390/s390-opts.h (enum indirect_branch): Define.
9996         * config/s390/s390-protos.h (s390_return_addr_from_memory)
9997         (s390_indirect_branch_via_thunk)
9998         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9999         (enum s390_indirect_branch_type): Define.
10000         * config/s390/s390.c (struct s390_frame_layout, struct
10001         machine_function): Remove.
10002         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10003         (indirect_branch_table_label_no, indirect_branch_table_name):
10004         Define variables.
10005         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10006         (enum s390_indirect_branch_option): Define.
10007         (s390_return_addr_from_memory): New function.
10008         (s390_handle_string_attribute): New function.
10009         (s390_attribute_table): Add new attribute handler.
10010         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10011         (s390_indirect_branch_via_thunk): New function.
10012         (s390_indirect_branch_via_inline_thunk): New function.
10013         (s390_function_ok_for_sibcall): When jumping via thunk disallow
10014         sibling call optimization for non z10 compiles.
10015         (s390_emit_call): Force indirect branch target to be a single
10016         register.  Add r1 clobber for non-z10 compiles.
10017         (s390_emit_epilogue): Emit return jump via return_use expander.
10018         (s390_reorg): Handle JUMP_INSNs as execute targets.
10019         (s390_option_override_internal): Perform validity checks for the
10020         new command line options.
10021         (s390_indirect_branch_attrvalue): New function.
10022         (s390_indirect_branch_settings): New function.
10023         (s390_set_current_function): Invoke s390_indirect_branch_settings.
10024         (s390_output_indirect_thunk_function):  New function.
10025         (s390_code_end): Implement target hook.
10026         (s390_case_values_threshold): Implement target hook.
10027         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10028         macros.
10029         * config/s390/s390.h (struct s390_frame_layout)
10030         (struct machine_function): Move here from s390.c.
10031         (TARGET_INDIRECT_BRANCH_NOBP_RET)
10032         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10033         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10034         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10035         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10036         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10037         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10038         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10039         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10040         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10041         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10042         (mnemonic attribute): Add values which aren't recognized
10043         automatically.
10044         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10045         pattern for branch conversion.  Fix mnemonic attribute.
10046         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10047         indirect branch via thunk if requested.
10048         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10049         ("*indirect_jump"): Disable for branch conversion using out of
10050         line thunks.
10051         ("indirect_jump_via_thunk<mode>_z10")
10052         ("indirect_jump_via_thunk<mode>")
10053         ("indirect_jump_via_inlinethunk<mode>_z10")
10054         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10055         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10056         ("casesi_jump_via_inlinethunk<mode>_z10")
10057         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10058         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10059         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10060         ("*indirect2_jump"): Disable for branch conversion.
10061         ("casesi_jump"): Turn into expander and expand patterns for branch
10062         conversion.
10063         ("return_use"): New expander.
10064         ("*return"): Emit return via thunk and rename it to ...
10065         ("*return<mode>"): ... this one.
10066         * config/s390/s390.opt: Add new options and and enum for the
10067         option values.
10068
10069 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
10070
10071         * lra-constraints.c (match_reload): Unconditionally use
10072         gen_lowpart_SUBREG, rather than selecting between that
10073         and equivalent gen_rtx_SUBREG code.
10074
10075 2018-02-08  Richard Biener  <rguenther@suse.de>
10076
10077         PR tree-optimization/84233
10078         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10079         changed flag instead of boguously re-using phi_inserted.
10080
10081 2018-02-08  Martin Jambor  <mjambor@suse.cz>
10082
10083         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10084         static local variables.
10085
10086 2018-02-08  Richard Biener  <rguenther@suse.de>
10087
10088         PR tree-optimization/84278
10089         * tree-vect-stmts.c (vectorizable_store): When looking for
10090         smaller vector types to perform grouped strided loads/stores
10091         make sure the mode is supported by the target.
10092         (vectorizable_load): Likewise.
10093
10094 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
10095
10096         * config/aarch64/aarch64.c (aarch64_components_for_bb):
10097         Increase LDP/STP opportunities by adding adjacent callee-saves.
10098
10099 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
10100
10101         PR rtl-optimization/84068
10102         PR rtl-optimization/83459
10103         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10104
10105 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
10106
10107         PR tree-optimization/84224
10108         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10109         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10110         non-zero arguments.
10111
10112 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
10113
10114         PR target/84113
10115         * config/rs6000/altivec.md (*restore_world): Remove LR use.
10116         * config/rs6000/predicates.md (restore_world_operation): Adjust op
10117         count, remove one USE.
10118
10119 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
10120
10121         * doc/install.texi (Configuration): Document the
10122         --with-long-double-format={ibm,ieee} PowerPC configuration
10123         options.
10124
10125         PR target/84154
10126         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10127         Convert from define_expand to be define_insn_and_split.  Rework
10128         float/double/_Float128 conversions to QI/HI/SImode to work with
10129         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
10130         conversions to QI/HImode types did a store and then a load to
10131         truncate the value.  For conversions to VSX registers, don't split
10132         the insn, instead emit the code directly.  Use the code iterator
10133         any_fix to combine signed and unsigned conversions.
10134         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10135         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10136         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10137         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10138         (fix_<mode>di2_hw): Likewise.
10139         (fixuns_<mode>di2_hw): Likewise.
10140         (fix_<mode>si2_hw): Likewise.
10141         (fixuns_<mode>si2_hw): Likewise.
10142         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10143         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10144         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10145         fix<uns>_trunc<SFDF:mode>si2_p8.
10146         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10147         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10148         (fix<uns>_<mode>_mem): Likewise.
10149         (fctiw<u>z_<mode>_mem): Likewise.
10150         (fix<uns>_<mode>_mem): Likewise.
10151         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10152         the register allocator from doing a direct move to the GPRs to do
10153         a store, and instead use the ISA 3.0 store byte/half-word from
10154         vector register instruction.  For IEEE 128-bit floating point,
10155         also optimize stores of 32-bit ints.
10156         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10157
10158 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
10159
10160         * genextract.c (push_pathstr_operand): New function to support
10161         [a-zA-Z].
10162         (walk_rtx): Call push_pathstr_operand.
10163         (print_path): Support [a-zA-Z].
10164
10165 2018-02-07  Richard Biener  <rguenther@suse.de>
10166
10167         PR tree-optimization/84037
10168         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10169         (cse_and_gimplify_to_preheader): Declare.
10170         (vect_get_place_in_interleaving_chain): Likewise.
10171         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10172         ivexpr_map.
10173         (_loop_vec_info::~_loop_vec_info): Delete it.
10174         (cse_and_gimplify_to_preheader): New function.
10175         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10176         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10177         (vectorizable_load): Likewise.  For grouped stores always base
10178         the IV on the first element.
10179         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10180         condition before gimplifying.
10181
10182 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
10183
10184         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10185         *DIV_EXPR and *MOD_EXPR.
10186
10187 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
10188
10189         PR target/84248
10190         * config/i386/i386.c (ix86_option_override_internal): Mask out
10191         the CF_SET bit when checking -fcf-protection.
10192
10193 2018-02-07  Tom de Vries  <tom@codesourcery.com>
10194
10195         PR libgomp/84217
10196         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10197         enough.
10198
10199 2018-02-07  Richard Biener  <rguenther@suse.de>
10200
10201         PR tree-optimization/84204
10202         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10203         this place.
10204
10205         PR tree-optimization/84205
10206         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10207         special-case isl_ast_op_zdiv_r.
10208
10209         PR tree-optimization/84223
10210         * graphite-scop-detection.c (gather_bbs::before_dom_children):
10211         Only add conditions from within the region.
10212         (gather_bbs::after_dom_children): Adjust.
10213
10214 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
10215
10216         PR target/84209
10217         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10218         * config/avr/avr.md: Only post-reload split REG-REG moves if
10219         either register is GENERAL_REG_P.
10220
10221 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
10222
10223         PR tree-optimization/84235
10224         * tree-ssa-scopedtables.c
10225         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10226         if the subtraction is performed in floating point type where NaNs are
10227         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10228         build 1.  Formatting fix.
10229
10230 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
10231
10232         PR target/84146
10233         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10234         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10235         and skip it regardless of bb boundaries.  Use CALL_P macro,
10236         don't test INSN_P (insn) together with CALL_P or JUMP_P check
10237         unnecessarily, formatting fix.
10238
10239 2018-02-06  Michael Collison  <michael.collison@arm.com>
10240
10241         * config/arm/thumb2.md:
10242         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10243         (*thumb_mov_notscc): Ditto.
10244
10245 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
10246
10247         PR target/84154
10248         * config/rs6000/rs6000.md (su code attribute): Use "u" for
10249         unsigned_fix, not "s".
10250
10251 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10252
10253         * configure.ac (gcc_fn_eh_frame_ro): New function.
10254         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10255         correct .eh_frame permissions.
10256         * configure: Regenerate.
10257
10258 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
10259
10260         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10261         irrelevant options.
10262
10263 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10264
10265         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10266         Display warning message for -mno-speculate-indirect-jumps.
10267
10268 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
10269
10270         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10271         Undocumented.
10272         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10273
10274 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
10275
10276         PR tree-optimization/84225
10277         * tree-eh.c (find_trapping_overflow): Only call
10278         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10279
10280 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
10281
10282         PR target/84145
10283         * config/i386/i386.c: Reimplement the check of possible options
10284         -mibt/-mshstk conbination. Change error messages.
10285         * doc/invoke.texi: Fix a typo: remove extra '='.
10286
10287 2018-02-06  Marek Polacek  <polacek@redhat.com>
10288
10289         PR tree-optimization/84228
10290         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10291
10292 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
10293
10294         PR target/82641
10295         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10296         emitted arch directives.
10297         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10298         __ARM_FEATURE_COPROC before changing architectures.
10299
10300 2018-02-06  Richard Biener  <rguenther@suse.de>
10301
10302         * config/i386/i386.c (print_reg): Fix typo.
10303         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10304
10305 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
10306
10307         * configure: Regenerate.
10308
10309 2018-02-05  Martin Sebor  <msebor@redhat.com>
10310
10311         PR tree-optimization/83369
10312         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10313         inlining context.
10314
10315 2018-02-05  Martin Liska  <mliska@suse.cz>
10316
10317         * doc/invoke.texi: Cherry-pick upstream r323995.
10318
10319 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
10320
10321         * ira.c (ira_init_register_move_cost): Adjust comment.
10322
10323 2018-02-05  Martin Liska  <mliska@suse.cz>
10324
10325         PR gcov-profile/84137
10326         * doc/gcov.texi: Fix typo in documentation.
10327
10328 2018-02-05  Martin Liska  <mliska@suse.cz>
10329
10330         PR gcov-profile/83879
10331         * doc/gcov.texi: Document necessity of --dynamic-list-data when
10332         using dlopen functionality.
10333
10334 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
10335
10336         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10337         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10338         _mm_maskz_range_ss, _mm_mask_range_round_ss,
10339         _mm_maskz_range_round_ss): New intrinsics.
10340         (__builtin_ia32_rangesd128_round)
10341         (__builtin_ia32_rangess128_round): Remove.
10342         (__builtin_ia32_rangesd128_mask_round,
10343         __builtin_ia32_rangess128_mask_round): New builtins.
10344         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10345         __builtin_ia32_rangess128_round): Remove.
10346         (__builtin_ia32_rangesd128_mask_round,
10347         __builtin_ia32_rangess128_mask_round): New builtins.
10348         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10349         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10350         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10351         "<round_saeonly_constraint>")): Changed to ...
10352         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10353         "<round_saeonly_scalar_constraint>")): ... this.
10354         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10355         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10356         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10357         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10358         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10359
10360 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
10361
10362         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10363         options.
10364         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10365         Remove all values except native, 8540 and 8548.
10366
10367 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
10368
10369         * config/i386/i386.c (ix86_output_function_return): Pass
10370         INVALID_REGNUM, instead of -1, as invalid register number to
10371         indirect_thunk_name and output_indirect_thunk.
10372
10373 2018-02-02  Julia Koval  <julia.koval@intel.com>
10374
10375         * config.gcc: Add -march=icelake.
10376         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10377         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10378         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10379         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10380         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10381         (processor_target_table): Add icelake.
10382         (ix86_option_override_internal): Handle new PTAs.
10383         (get_builtin_code_for_version): Handle icelake.
10384         (M_INTEL_COREI7_ICELAKE): New.
10385         (fold_builtin_cpu): Handle icelake.
10386         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10387         * doc/invoke.texi: Add -march=icelake.
10388
10389 2018-02-02  Julia Koval  <julia.koval@intel.com>
10390
10391         * config/i386/i386.c (ix86_option_override_internal): Change flags type
10392         to wide_int_bitmask.
10393         * wide-int-bitmask.h: New.
10394
10395 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
10396
10397         PR target/84066
10398         * config/i386/i386.md: Replace Pmode with word_mode in
10399         builtin_setjmp_setup and builtin_longjmp to support x32.
10400
10401 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
10402
10403         PR target/56010
10404         PR target/83743
10405         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10406         #include "opts.h".
10407         (rs6000_supported_cpu_names): New static variable.
10408         (linux_cpu_translation_table): Likewise.
10409         (elf_platform) <cpu>: Define new static variable and use it.
10410         Translate kernel AT_PLATFORM name to canonical name if needed.
10411         Error if platform name is unknown.
10412
10413 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
10414
10415         PR target/84089
10416         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10417
10418 2018-02-01  Jeff Law  <law@redhat.com>
10419
10420         PR target/84128
10421         * config/i386/i386.c (release_scratch_register_on_entry): Add new
10422         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
10423         the scratch if RELEASE_VIA_POP is false.
10424         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10425         If we have to save a temporary register, decrement SIZE appropriately.
10426         Pass new arguments to release_scratch_register_on_entry.
10427         (ix86_adjust_stack_and_probe): Likewise.
10428         (ix86_emit_probe_stack_range): Pass new arguments to
10429         release_scratch_register_on_entry.
10430
10431 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
10432
10433         PR rtl-optimization/84157
10434         * combine.c (change_zero_ext): Use REG_P predicate in
10435         front of HARD_REGISTER_P predicate.
10436
10437 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
10438
10439         * config/avr/avr.c (avr_option_override): Move disabling of
10440         -fdelete-null-pointer-checks to...
10441         * common/config/avr/avr-common.c (avr_option_optimization_table):
10442         ...here.
10443
10444 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10445
10446         PR tree-optimization/81635
10447         * tree-data-ref.c (split_constant_offset_1): For types that
10448         wrap on overflow, try to use range info to prove that wrapping
10449         cannot occur.
10450
10451 2018-02-01  Renlin Li  <renlin.li@arm.com>
10452
10453         PR target/83370
10454         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10455         TAILCALL_ADDR_REGS.
10456         (aarch64_register_move_cost): Likewise.
10457         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10458         TAILCALL_ADDR_REGS.
10459         (REG_CLASS_NAMES): Likewise.
10460         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10461         TAILCALL_ADDR_REGS. Remove IP registers.
10462         * config/aarch64/aarch64.md (Ucs): Update register constraint.
10463
10464 2018-02-01  Richard Biener  <rguenther@suse.de>
10465
10466         * domwalk.h (dom_walker::dom_walker): Add additional constructor
10467         for specifying RPO order and allow NULL for that.
10468         * domwalk.c (dom_walker::dom_walker): Likewise.
10469         (dom_walker::walk): Handle NULL RPO order.
10470         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10471         in RPO order.
10472         (rewrite_update_dom_walker): Likewise.
10473         (mark_def_dom_walker): Likewise.
10474
10475 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10476
10477         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10478         (aarch64_maybe_expand_sve_subreg_move): Declare.
10479         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10480         * config/aarch64/predicates.md (aarch64_any_register_operand): New
10481         predicate.
10482         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10483         that are semantically a reverse operation.
10484         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10485         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10486         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10487         functions.
10488         (aarch64_can_change_mode_class): For big-endian, forbid changes
10489         between two SVE modes if they have different element sizes.
10490
10491 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10492
10493         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10494         the TImode handling for big-endian targets.
10495
10496 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10497
10498         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10499         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
10500         not just bytes.
10501         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10502         Remove BSWAP handing for big-endian targets and use the form of
10503         LD1RQ appropariate for the mode.
10504
10505 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10506
10507         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10508         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10509         duplicated element.
10510
10511 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
10512
10513         PR tearget/83845
10514         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10515         check for operands that need to go through aarch64_sve_reload_be.
10516
10517 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
10518
10519         PR tree-optimization/81661
10520         PR tree-optimization/84117
10521         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10522         * tree-eh.c: Include gimplify.h.
10523         (find_trapping_overflow, replace_trapping_overflow,
10524         rewrite_to_non_trapping_overflow): New functions.
10525         * tree-vect-loop.c: Include tree-eh.h.
10526         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10527         * tree-data-ref.c: Include tree-eh.h.
10528         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10529
10530 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
10531
10532         PR rtl-optimization/84123
10533         * combine.c (change_zero_ext): Check if hard register satisfies
10534         can_change_dest_mode before calling gen_lowpart_SUBREG.
10535
10536 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
10537
10538         PR target/82444
10539         * ira.c (ira_init_register_move_cost): Remove assert.
10540
10541 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
10542
10543         PR rtl-optimization/84071
10544         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10545         * doc/tm.texi: Regenerate.
10546
10547 2018-01-31  Richard Biener  <rguenther@suse.de>
10548
10549         PR tree-optimization/84132
10550         * tree-data-ref.c (analyze_miv_subscript): Properly
10551         check whether evolution_function_is_affine_multivariate_p
10552         before calling gcd_of_steps_may_divide_p.
10553
10554 2018-01-31  Julia Koval  <julia.koval@intel.com>
10555
10556         PR target/83618
10557         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10558         * config/i386/i386.md (rdpid_rex64) New.
10559         (rdpid): Make 32bit only.
10560
10561 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
10562
10563         PR lto/84105
10564         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10565         an IDENTIFIER_NODE for FUNCTION_TYPE's.
10566
10567 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
10568
10569         Revert
10570         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
10571
10572         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10573
10574 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
10575
10576         PR rtl-optimization/84071
10577         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10578         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10579
10580 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
10581
10582         * config/arc/arc.c (arc_handle_aux_attribute): New function.
10583         (arc_attribute_table): Add 'aux' attribute.
10584         (arc_in_small_data_p): Consider aux like variables.
10585         (arc_is_aux_reg_p): New function.
10586         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10587         (arc_get_aux_arg): New function.
10588         (prepare_move_operands): Handle aux-register access.
10589         (arc_handle_aux_attribute): New function.
10590         * doc/extend.texi (ARC Variable attributes): Add subsection.
10591
10592 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
10593
10594         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10595         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10596         (arc_attribute_table): Add 'uncached' attribute.
10597         (arc_print_operand): Print '.di' flag for uncached memory
10598         accesses.
10599         (arc_in_small_data_p): Do not consider for small data the uncached
10600         types.
10601         (arc_is_uncached_mem_p): New function.
10602         * config/arc/predicates.md (compact_store_memory_operand): Check
10603         for uncached memory accesses.
10604         (nonvol_nonimm_operand): Likewise.
10605         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10606
10607 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
10608
10609         PR c/84100
10610         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10611         falign-loops=): Add Optimization flag.
10612
10613 2018-01-30  Jeff Law  <law@redhat.com>
10614
10615         PR target/84064
10616         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10617         INT_REGISTERS_SAVED.  Check it prior to calling
10618         get_scratch_register_on_entry.
10619         (ix86_adjust_stack_and_probe): Similarly.
10620         (ix86_emit_probe_stack_range): Similarly.
10621         (ix86_expand_prologue): Corresponding changes.
10622
10623 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10624
10625         PR target/40411
10626         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10627         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10628
10629 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
10630
10631         PR target/84112
10632         * lra-constraints.c (curr_insn_transform): Process AND in the
10633         address.
10634
10635 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
10636
10637         PR rtl-optimization/83986
10638         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10639         dependence against last_pending_memory_flush in addition to
10640         pending_jump_insns.
10641
10642 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
10643
10644         PR tree-optimization/81611
10645         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10646         copies.
10647
10648 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
10649
10650         PR target/83758
10651         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10652         a reg rtx.
10653
10654 2018-01-30  Richard Biener  <rguenther@suse.de>
10655             Jakub Jelinek  <jakub@redhat.com>
10656
10657         PR tree-optimization/84111
10658         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10659         inner loops added during recursion, as they don't have up-to-date
10660         SSA form.
10661
10662 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
10663
10664         PR ipa/81360
10665         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10666         (can_inline_edge_by_limits_p): ... here.
10667         (can_early_inline_edge_p, check_callers,
10668         update_caller_keys, update_callee_keys, recursive_inlining,
10669         add_new_edges_to_heap, speculation_useful_p,
10670         inline_small_functions,
10671         inline_small_functions, flatten_function,
10672         inline_to_all_callers_1): Update.
10673
10674 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
10675
10676         * profile-count.c (profile_count::combine_with_ipa_count): Handle
10677         zeros correctly.
10678
10679 2018-01-30  Richard Biener  <rguenther@suse.de>
10680
10681         PR tree-optimization/83008
10682         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10683         invariant and constant vector uses in stmts when they need
10684         more than one stmt.
10685
10686 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10687
10688         PR bootstrap/84017
10689         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10690         * configure: Regenerate.
10691
10692 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
10693
10694         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10695         pattern.
10696         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10697         Use gen_rtx_REG rather than gen_lowpart.
10698
10699 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
10700
10701         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10702         rather than 0 when creating partial subregs.
10703
10704 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
10705
10706         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10707         of usage.
10708
10709 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
10710
10711         PR target/81550
10712         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10713         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10714         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10715         flags.  This restores the settings used before the 2017-07-24.
10716         Turning off pre increment/decrement/modify allows IVOPTS to
10717         optimize DF/SF loops where the index is an int.
10718
10719 2018-01-29  Richard Biener  <rguenther@suse.de>
10720             Kelvin Nilsen  <kelvin@gcc.gnu.org>
10721
10722         PR bootstrap/80867
10723         * tree-vect-stmts.c (vectorizable_call): Don't call
10724         targetm.vectorize_builtin_md_vectorized_function if callee is
10725         NULL.
10726
10727 2018-01-22  Carl Love  <cel@us.ibm.com>
10728
10729         * doc/extend.tex: Fix typo in second arg in
10730         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10731
10732 2018-01-29  Richard Biener  <rguenther@suse.de>
10733
10734         PR tree-optimization/84086
10735         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10736         (flush_ssaname_freelist): When SSA names were released reset
10737         the SCEV hash table.
10738
10739 2018-01-29  Richard Biener  <rguenther@suse.de>
10740
10741         PR tree-optimization/84057
10742         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10743         removed paths when removing edges.
10744
10745 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
10746
10747         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10748         -mfunction-return=@var{choice}.
10749
10750 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10751
10752         PR diagnostic/84034
10753         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10754         Handle CR like TAB.
10755         (layout::print_source_line): Likewise.
10756         (test_get_line_width_without_trailing_whitespace): Add test cases.
10757
10758 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
10759
10760         PR middle-end/84040
10761         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10762         debug insns.
10763
10764 2018-01-26  Jim Wilson  <jimw@sifive.com>
10765
10766         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10767
10768         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10769         specified.
10770
10771 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10772
10773         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10774         and CMP + SUB-immediate -> SUBS.
10775
10776 2018-01-26  Martin Sebor  <msebor@redhat.com>
10777
10778         PR tree-optimization/83896
10779         * tree-ssa-strlen.c (get_string_len): Rename...
10780         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
10781         Avoid assuming length is constant.
10782         (handle_char_store): Use HOST_WIDE_INT for string length.
10783
10784 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
10785
10786         PR target/81763
10787         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10788         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10789
10790 2018-01-26  Richard Biener  <rguenther@suse.de>
10791
10792         PR rtl-optimization/84003
10793         * dse.c (record_store): Only record redundant stores when
10794         the earlier store aliases at least all accesses the later one does.
10795
10796 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
10797
10798         PR rtl-optimization/83985
10799         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10800         REG_CFA_RESTORE insns.
10801         (delete_unmarked_insns): Don't ignore separate shrink wrapping
10802         REG_CFA_RESTORE insns here.
10803
10804         PR c/83989
10805         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10806         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10807
10808 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10809
10810         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10811         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10812         (arc_init): Likewise.
10813         (arc_override_options): Likewise.
10814         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10815         value.
10816         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10817         support.
10818         * config/arc/arc.h (TARGET_DBNZ): Define.
10819         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10820         properly set the tune attribute.
10821         (dbnz): Use TARGET_DBNZ guard.
10822         * config/arc/arc.opt (mtune): Add core3 option.
10823
10824 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10825
10826         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10827         recognize new pic like addresses.
10828         (arc_delegitimize_address): Clean up.
10829
10830 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10831
10832         * config/arc/arc-arches.def: Option mrf16 valid for all
10833         architectures.
10834         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10835         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10836         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10837         * config/arc/arc-tables.opt: Regenerate.
10838         * config/arc/arc.c (arc_conditional_register_usage): Handle
10839         reduced register file case.
10840         (arc_file_start): Set must have build attributes.
10841         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10842         mrf16 option value.
10843         * config/arc/arc.opt (mrf16): Add new option.
10844         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10845         * config/arc/genmultilib.awk: Handle new mrf16 option.
10846         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10847         * config/arc/t-multilib: Regenerate.
10848         * doc/invoke.texi (ARC Options): Document mrf16 option.
10849
10850 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10851
10852         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10853         * config/arc/arc.c (arc_handle_secure_attribute): New function.
10854         (arc_attribute_table): Add 'secure_call' attribute.
10855         (arc_print_operand): Print secure call operand.
10856         (arc_function_ok_for_sibcall): Don't optimize tail calls when
10857         secure.
10858         (arc_is_secure_call_p): New function.  * config/arc/arc.md
10859         (call_i): Add support for sjli instruction.
10860         (call_value_i): Likewise.
10861         * config/arc/constraints.md (Csc): New constraint.
10862
10863 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
10864             John Eric Martin  <John.Martin@emmicro-us.com>
10865
10866         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10867         * config/arc/arc.c (_arc_jli_section): New struct.
10868         (arc_jli_section): New type.
10869         (rc_jli_sections): New static variable.
10870         (arc_handle_jli_attribute): New function.
10871         (arc_attribute_table): Add jli_always and jli_fixed attribute.
10872         (arc_file_end): New function.
10873         (TARGET_ASM_FILE_END): Define.
10874         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10875         (arc_add_jli_section): New function.
10876         (jli_call_scan): Likewise.
10877         (arc_reorg): Call jli_call_scan.
10878         (arc_output_addsi): Remove 'S' from printing asm operand.
10879         (arc_is_jli_call_p): New function.
10880         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10881         operand.
10882         (movhi_insn): Likewise.
10883         (movsi_insn): Likewise.
10884         (movsi_set_cc_insn): Likewise.
10885         (loadqi_update): Likewise.
10886         (load_zeroextendqisi_update): Likewise.
10887         (load_signextendqisi_update): Likewise.
10888         (loadhi_update): Likewise.
10889         (load_zeroextendhisi_update): Likewise.
10890         (load_signextendhisi_update): Likewise.
10891         (loadsi_update): Likewise.
10892         (loadsf_update): Likewise.
10893         (movsicc_insn): Likewise.
10894         (bset_insn): Likewise.
10895         (bxor_insn): Likewise.
10896         (bclr_insn): Likewise.
10897         (bmsk_insn): Likewise.
10898         (bicsi3_insn): Likewise.
10899         (cmpsi_cc_c_insn): Likewise.
10900         (movsi_ne): Likewise.
10901         (movsi_cond_exec): Likewise.
10902         (clrsbsi2): Likewise.
10903         (norm_f): Likewise.
10904         (normw): Likewise.
10905         (swap): Likewise.
10906         (divaw): Likewise.
10907         (flag): Likewise.
10908         (sr): Likewise.
10909         (kflag): Likewise.
10910         (ffs): Likewise.
10911         (ffs_f): Likewise.
10912         (fls): Likewise.
10913         (call_i): Remove 'S' asm letter, add jli instruction.
10914         (call_value_i): Likewise.
10915         * config/arc/arc.op (mjli-always): New option.
10916         * config/arc/constraints.md (Cji): New constraint.
10917         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10918         operand.
10919         (subsf3_fpx): Likewise.
10920         (mulsf3_fpx): Likewise.
10921         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10922         asm operand.
10923         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10924         function attrbutes.
10925         * doc/invoke.texi (ARC): Document mjli-always option.
10926
10927 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
10928
10929         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10930         avoid addition with 0 and use incw and decw where possible.
10931
10932 2018-01-26  Richard Biener  <rguenther@suse.de>
10933
10934         PR tree-optimization/81082
10935         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10936         association if it requires casting to unsigned.
10937         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10938         from fold_plusminus_mult_expr to catch important cases late when
10939         range info is available.
10940
10941 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10942
10943         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10944         * configure.ac (hidden_linkonce): New test.
10945         * configure: Regenerate.
10946         * config.in: Regenerate.
10947
10948 2018-01-26  Julia Koval  <julia.koval@intel.com>
10949
10950         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10951         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10952         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10953         _mm_mask_bitshuffle_epi64_mask): Fix type.
10954         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10955         USI_FTYPE_V4DI_V4DI_USI): Remove.
10956         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10957         __builtin_ia32_vpshufbitqmb256_mask,
10958         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10959         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10960         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10961
10962 2018-01-26  Alan Modra  <amodra@gmail.com>
10963
10964         PR target/84033
10965         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10966         UNSPEC_VBPERMQ.  Sort other unspecs.
10967
10968 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
10969
10970         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10971
10972 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
10973
10974         PR middle-end/83055
10975         * predict.c (drop_profile): Do not push/pop cfun; update also
10976         node->count.
10977         (handle_missing_profiles): Fix logic looking for zero profiles.
10978
10979 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
10980
10981         PR middle-end/83977
10982         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10983         on functions with #pragma omp declare simd or functions with simd
10984         attribute.
10985         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10986         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10987         Remove trailing \n from warning_at calls.
10988
10989 2018-01-25  Tom de Vries  <tom@codesourcery.com>
10990
10991         PR target/84028
10992         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10993         for neutered workers.
10994
10995 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
10996
10997         PR target/68467
10998         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10999         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11000
11001 2018-01-24  Jeff Law  <law@redhat.com>
11002
11003         PR target/83994
11004         * i386.c (get_probe_interval): Move to earlier point.
11005         (ix86_compute_frame_layout): If -fstack-clash-protection and
11006         the frame is larger than the probe interval, then use pushes
11007         to save registers rather than reg->mem moves.
11008         (ix86_expand_prologue): Remove conditional for int_registers_saved
11009         assertion.
11010
11011 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
11012
11013         PR target/84014
11014         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11015         min/max for never referenced object.
11016
11017 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
11018
11019         PR middle-end/83977
11020         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11021         here.
11022         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11023         attributes from DECL_ATTRIBUTES (decl) without affecting
11024         DECL_ATTRIBUTES (current_function_decl).
11025         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11026         functions with non-NULL DECL_ABSTRACT_ORIGIN.
11027
11028 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
11029
11030         PR tree-optimization/83979
11031         * fold-const.c (fold_comparison): Use constant_boolean_node
11032         instead of boolean_{true,false}_node.
11033
11034 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
11035
11036         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11037         with zero counts.
11038
11039 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11040
11041         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11042         Simplify the clause that sets the length attribute.
11043         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11044         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11045         clause that sets the length attribute.
11046         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11047
11048 2018-01-24  Tom de Vries  <tom@codesourcery.com>
11049
11050         PR target/83589
11051         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11052         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11053         Add strict parameter.
11054         (prevent_branch_around_nothing): Insert dummy insn between branch to
11055         label and label with no ptx insn inbetween.
11056         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11057
11058 2018-01-24  Tom de Vries  <tom@codesourcery.com>
11059
11060         PR target/81352
11061         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11062         for neutered threads in warp.
11063         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11064
11065 2018-01-24  Richard Biener  <rguenther@suse.de>
11066
11067         PR tree-optimization/83176
11068         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11069         operands.
11070
11071 2018-01-24  Richard Biener  <rguenther@suse.de>
11072
11073         PR tree-optimization/82819
11074         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11075         code generating pluses that are no-ops in the target precision.
11076
11077 2018-01-24  Richard Biener  <rguenther@suse.de>
11078
11079         PR middle-end/84000
11080         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11081
11082 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
11083
11084         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11085         to merge probabilities.
11086         * predict.c (probably_never_executed): Also mark as cold functions
11087         with global 0 profile and guessed local profile.
11088         * profile-count.c (profile_probability::combine_with_count): New
11089         member function.
11090         * profile-count.h (profile_probability::operator*,
11091         profile_probability::operator*=, profile_probability::operator/,
11092         profile_probability::operator/=): Reduce precision to adjusted
11093         and set value to guessed on contradictory divisions.
11094         (profile_probability::combine_with_freq): Remove.
11095         (profile_probability::combine_wiht_count): Declare.
11096         (profile_count::force_nonzero):: Set to adjusted.
11097         (profile_count::probability_in):: Set quality to adjusted.
11098         * tree-ssa-tail-merge.c (replace_block_by): Use
11099         combine_with_count.
11100
11101 2018-01-23  Andrew Waterman  <andrew@sifive.com>
11102             Jim Wilson  <jimw@sifive.com>
11103
11104         * config/riscv/riscv.c (riscv_stack_boundary): New.
11105         (riscv_option_override): Set riscv_stack_boundary.  Handle
11106         riscv_preferred_stack_boundary_arg.
11107         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11108         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11109         (STACK_BOUNDARY): Set to riscv_stack_boundary.
11110         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11111         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11112         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11113
11114 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
11115
11116         PR target/83905
11117         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11118         of struct ix86_frame.
11119         (ix86_expand_epilogue): Likewise.  Add a local variable for
11120         the reg_save_offset field in struct ix86_frame.
11121
11122 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
11123
11124         PR tree-optimization/82604
11125         * tree-loop-distribution.c (enum partition_kind): New enum item
11126         PKIND_PARTIAL_MEMSET.
11127         (partition_builtin_p): Support above new enum item.
11128         (generate_code_for_partition): Ditto.
11129         (compute_access_range): Differentiate cases that equality can be
11130         proven at all loops, the innermost loops or no loops.
11131         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11132         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
11133         (finalize_partitions, distribute_loop): Don't fuse partition of
11134         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11135         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11136         parloop is enabled.
11137
11138 2018-01-23  Martin Liska  <mliska@suse.cz>
11139
11140         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11141         order to ignore the predictor.
11142         (PRED_POLYMORPHIC_CALL): Likewise.
11143         (PRED_RECURSIVE_CALL): Likewise.
11144
11145 2018-01-23  Martin Liska  <mliska@suse.cz>
11146
11147         * tree-profile.c (tree_profiling): Print function header to
11148         aware reader which function we are working on.
11149         * value-prof.c (gimple_find_values_to_profile): Do not print
11150         not interesting value histograms.
11151
11152 2018-01-23  Martin Liska  <mliska@suse.cz>
11153
11154         * profile-count.h (enum profile_quality): Add
11155         profile_uninitialized as the first value. Do not number values
11156         as they are zero based.
11157         (profile_count::verify): Update sanity check.
11158         (profile_probability::verify): Likewise.
11159
11160 2018-01-23  Nathan Sidwell  <nathan@acm.org>
11161
11162         * doc/invoke.texi (ffor-scope): Deprecate.
11163
11164 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
11165
11166         PR tree-optimization/83510
11167         * domwalk.c (set_all_edges_as_executable): New function.
11168         (dom_walker::dom_walker): Convert bool param
11169         "skip_unreachable_blocks" to enum reachability.  Move setup of
11170         edge flags to set_all_edges_as_executable and only do it when
11171         reachability is REACHABLE_BLOCKS.
11172         * domwalk.h (enum dom_walker::reachability): New enum.
11173         (dom_walker::dom_walker): Convert bool param
11174         "skip_unreachable_blocks" to enum reachability.
11175         (set_all_edges_as_executable): New decl.
11176         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
11177         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11178         "reachability".
11179         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11180         but converting true to REACHABLE_BLOCKS.
11181         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11182         * tree-vrp.c
11183         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11184         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11185         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11186         REACHABLE_BLOCKS.
11187         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11188         if check_all_array_refs will be called.
11189
11190 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
11191
11192         * tree.c (selftest::test_location_wrappers): Add more test
11193         coverage.
11194
11195 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
11196
11197         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11198         (selftest::test_bit_in_range): Likewise.
11199
11200 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
11201
11202         PR testsuite/83888
11203         * doc/sourcebuild.texi (vect_float): Say that the selector
11204         only describes the situation when -funsafe-math-optimizations is on.
11205         (vect_float_strict): Document.
11206
11207 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
11208
11209         PR tree-optimization/83965
11210         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11211         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11212         instead of checking only for a reduction.
11213         (vect_recog_widen_sum_pattern): Likewise.
11214
11215 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
11216
11217         * predict.c (probably_never_executed): Only use precise profile info.
11218         (compute_function_frequency): Skip after inlining hack since we now
11219         have quality checking.
11220
11221 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
11222
11223         * profile-count.h (profile_probability::very_unlikely,
11224         profile_probability::unlikely, profile_probability::even): Set
11225         precision to guessed.
11226
11227 2018-01-23  Richard Biener  <rguenther@suse.de>
11228
11229         PR tree-optimization/83963
11230         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11231         Properly terminate dominator walk when crossing the exit edge not
11232         when visiting its source block.
11233
11234 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
11235
11236         PR c++/83918
11237         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11238         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11239
11240 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
11241
11242         PR tree-optimization/83957
11243         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
11244         semicolon after for body surrounded by braces.
11245
11246         PR tree-optimization/83081
11247         * profile-count.h (profile_probability::split): New method.
11248         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11249         Use profile_probability::split.
11250         (do_compare_rtx_and_jump): Fix adjustment of probabilities
11251         when splitting a single conditional jump into 2.
11252
11253 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
11254
11255         PR tree-optimization/69452
11256         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11257         decl.
11258
11259 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11260
11261         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11262         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11263         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11264
11265 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11266
11267         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11268         declaration.
11269         * config/rl78/rl78.md (movdi): New define_expand.
11270         * config/rl78/rl78.c (rl78_split_movdi): New function.
11271
11272 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
11273
11274         PR target/83862
11275         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11276         no longer used.
11277         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11278         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11279         128-bit to produce an UNSPEC move to get the double word with the
11280         signbit and then a shift directly to do signbit.
11281         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11282         implementation with a new version that just does either a direct
11283         move or a regular move.  Move memory interface to separate insns.
11284         Move insns so they are next to the expander.
11285         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11286         with signbit move.  Split big and little endian case.
11287         (signbit<mode>2_dm_mem_le): Likewise.
11288         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11289         (signbit<mode>2_dm2): Likewise.
11290
11291 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11292
11293         * config/rl78/rl78.md (anddi3): New define_expand.
11294
11295 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11296
11297         * config/rl78/rl78.md (umindi3): New define_expand.
11298
11299 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11300
11301         * config/rl78/rl78.md (smindi3): New define_expand.
11302
11303 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11304
11305         * config/rl78/rl78.md (smaxdi3): New define_expand.
11306
11307 2018-01-22  Carl Love  <cel@us.ibm.com>
11308
11309         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11310         LVX_V1TI): Add macro expansion.
11311         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11312         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11313         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11314         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11315         Change check to determine if the instruction is a byte reversing
11316         entry.  Fix typo in comment.
11317         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11318         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11319         Add def_builtin calls for new builtins.
11320         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11321         Add define_insn expansion.
11322
11323 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11324
11325         * config/rl78/rl78.md (umaxdi3): New define_expand.
11326
11327 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
11328
11329         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11330         for non-QImode registers.
11331
11332 2018-01-22  Richard Biener  <rguenther@suse.de>
11333
11334         PR tree-optimization/83963
11335         * graphite-scop-detection.c (scop_detection::get_sese): Delay
11336         including the loop exit block.
11337         (scop_detection::merge_sese): Likewise.
11338         (scop_detection::add_scop): Do it here instead.
11339
11340 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11341
11342         * doc/sourcebuild.texi (arm_softfloat): Document.
11343
11344 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
11345
11346         PR gcc/77734
11347         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11348         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11349         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11350
11351 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11352             David Edelsohn  <dje.gcc@gmail.com>
11353
11354         PR target/83946
11355         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11356         Change "crset eq" to "crset 2".
11357         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11358         (*call_indirect_aix<mode>_nospec): Likewise.
11359         (*call_value_indirect_aix<mode>_nospec): Likewise.
11360         (*call_indirect_elfv2<mode>_nospec): Likewise.
11361         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11362         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11363         change assembly output from . to $.
11364         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11365         (indirect_jump<mode>_nospec): Change assembly output from . to $.
11366         (*tablejump<mode>_internal1_nospec): Likewise.
11367
11368 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
11369
11370         PR target/80870
11371         * config/sh/sh_optimize_sett_clrt.cc:
11372         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11373
11374 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
11375
11376         PR tree-optimization/83940
11377         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11378         offset_dt to vect_constant_def rather than vect_unknown_def_type.
11379         (vect_check_load_store_mask): Add a mask_dt_out parameter and
11380         use it to pass back the definition type.
11381         (vect_check_store_rhs): Likewise rhs_dt_out.
11382         (vect_build_gather_load_calls): Add a mask_dt argument and use
11383         it instead of a call to vect_is_simple_use.
11384         (vectorizable_store): Update calls to vect_check_load_store_mask
11385         and vect_check_store_rhs.  Use the dt returned by the latter instead
11386         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
11387         instead of calls to vect_is_simple_use.  Pass the scalar rather
11388         than the vector operand to vect_is_simple_use when handling
11389         second and subsequent copies of an rhs value.
11390         (vectorizable_load): Update calls to vect_check_load_store_mask
11391         and vect_build_gather_load_calls.  Use the cached mask_dt and
11392         gs_info.offset_dt instead of calls to vect_is_simple_use.
11393
11394 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
11395
11396         PR middle-end/83945
11397         * tree-emutls.c: Include gimplify.h.
11398         (lower_emutls_2): New function.
11399         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11400         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11401         it before further processing.
11402
11403         PR target/83930
11404         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11405         UINTVAL (trueop1) instead of INTVAL (op1).
11406
11407 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
11408
11409         PR debug/81570
11410         PR debug/83728
11411         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11412         INCOMING_FRAME_SP_OFFSET if not defined.
11413         (scan_trace): Add ENTRY argument.  If true and
11414         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11415         emit a note to adjust the CFA offset.
11416         (create_cfi_notes): Adjust scan_trace callers.
11417         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11418         INCOMING_FRAME_SP_OFFSET in the CIE.
11419         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11420         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11421         Likewise.
11422         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11423         * doc/tm.texi: Regenerated.
11424
11425 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
11426
11427         PR rtl-optimization/83147
11428         * lra-constraints.c (remove_inheritance_pseudos): Use
11429         lra_substitute_pseudo_within_insn.
11430
11431 2018-01-19  Tom de Vries  <tom@codesourcery.com>
11432             Cesar Philippidis  <cesar@codesourcery.com>
11433
11434         PR target/83920
11435         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11436
11437 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
11438
11439         PR target/83790
11440         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11441         spaces for function labels.
11442
11443 2018-01-19  Martin Liska  <mliska@suse.cz>
11444
11445         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11446         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11447         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11448         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11449         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11450         (PRED_CONST_RETURN): Change from 69 to 65.
11451         (PRED_NULL_RETURN): Change from 91 to 71.
11452         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11453         (PRED_LOOP_GUARD): Change from 66 to 73.
11454
11455 2018-01-19  Martin Liska  <mliska@suse.cz>
11456
11457         * predict.c (predict_insn_def): Add new assert.
11458         (struct branch_predictor): Change type to signed integer.
11459         (test_prediction_value_range): Amend test to cover
11460         PROB_UNINITIALIZED.
11461         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11462         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11463         (PRED_LOOP_ITERATIONS_MAX): Likewise.
11464         (PRED_LOOP_IV_COMPARE): Likewise.
11465         * predict.h (PROB_UNINITIALIZED): Define new constant.
11466
11467 2018-01-19  Martin Liska  <mliska@suse.cz>
11468
11469         * predict.c (dump_prediction): Add new format for
11470         analyze_brprob.py script which is enabled with -details
11471         suboption.
11472         * profile-count.h (precise_p): New function.
11473
11474 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
11475
11476         PR tree-optimization/83922
11477         * tree-vect-loop.c (vect_verify_full_masking): Return false if
11478         there are no statements that need masking.
11479         (vect_active_double_reduction_p): New function.
11480         (vect_analyze_loop_operations): Use it when handling phis that
11481         are not in the loop header.
11482
11483 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
11484
11485         PR tree-optimization/83914
11486         * tree-vect-loop.c (vectorizable_induction): Don't convert
11487         init_expr or apply the peeling adjustment for inductions
11488         that are nested within the vectorized loop.
11489
11490 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11491
11492         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11493         instead of NEG.
11494
11495 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
11496
11497         PR sanitizer/81715
11498         PR testsuite/83882
11499         * function.h (gimplify_parameters): Add gimple_seq * argument.
11500         * function.c: Include gimple.h and options.h.
11501         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11502         for the added local temporaries if needed.
11503         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11504         if there are any parameter cleanups, wrap whole body into a
11505         try/finally with the cleanups.
11506
11507 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
11508
11509         PR target/82964
11510         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11511         Use GET_MODE_CLASS for scalar floating point.
11512
11513 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
11514
11515         PR ipa/82256
11516         patch by PaX Team
11517         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11518         Fix call of call_cgraph_insertion_hooks.
11519
11520 2018-01-18  Martin Sebor  <msebor@redhat.com>
11521
11522         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11523
11524 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
11525
11526         PR ipa/83619
11527         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11528         frequencies.
11529
11530 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
11531
11532         PR other/70268
11533         * common.opt: (-ffile-prefix-map): New option.
11534         * opts.c (common_handle_option): Defer it.
11535         * opts-global.c (handle_common_deferred_options): Handle it.
11536         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11537         * file-prefix-map.h: New file.
11538         (remap_debug_filename, add_debug_prefix_map): ...here.
11539         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11540         * final.c (debug_prefix_map, add_debug_prefix_map
11541         remap_debug_filename): Move to...
11542         * file-prefix-map.c: New file.
11543         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11544         generalize, get rid of alloca(), use strrchr() instead of strchr().
11545         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11546         Implement in terms of add_prefix_map().
11547         (remap_macro_filename, remap_debug_filename): Implement in term of
11548         remap_filename().
11549         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11550         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11551         * dbxout.c: Include file-prefix-map.h.
11552         * varasm.c: Likewise.
11553         * vmsdbgout.c: Likewise.
11554         * xcoffout.c: Likewise.
11555         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11556         * doc/cppopts.texi (-fmacro-prefix-map): Document.
11557         * doc/invoke.texi (-ffile-prefix-map): Document.
11558         (-fdebug-prefix-map): Update description.
11559
11560 2018-01-18  Martin Liska  <mliska@suse.cz>
11561
11562         * config/i386/i386.c (indirect_thunk_name): Document that also
11563         lfence is emitted.
11564         (output_indirect_thunk): Document why both instructions
11565         (pause and lfence) are generated.
11566
11567 2018-01-18  Richard Biener  <rguenther@suse.de>
11568
11569         PR tree-optimization/83887
11570         * graphite-scop-detection.c
11571         (scop_detection::get_nearest_dom_with_single_entry): Remove.
11572         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11573         (scop_detection::merge_sese): Re-implement with a flood-fill
11574         algorithm that properly finds a SESE region if it exists.
11575
11576 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
11577
11578         PR c/61240
11579         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11580         pointer_diff optimizations use view_convert instead of convert.
11581
11582 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11583
11584         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11585         Generate different code for -mno-speculate-indirect-jumps.
11586         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11587         (*call_indirect_aix<mode>): Disable for
11588         -mno-speculate-indirect-jumps.
11589         (*call_indirect_aix<mode>_nospec): New define_insn.
11590         (*call_value_indirect_aix<mode>): Disable for
11591         -mno-speculate-indirect-jumps.
11592         (*call_value_indirect_aix<mode>_nospec): New define_insn.
11593         (*sibcall_nonlocal_sysv<mode>): Generate different code for
11594         -mno-speculate-indirect-jumps.
11595         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11596
11597 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
11598
11599         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11600         long double type, set the flags for noting the default long double
11601         type, even if we don't pass or return a long double type.
11602
11603 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
11604
11605         PR ipa/83051
11606         * ipa-inline.c (flatten_function): Do not overwrite final inlining
11607         failure.
11608
11609 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
11610
11611         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11612         support for merge[hl].
11613         (fold_mergehl_helper): New helper function.
11614         (tree-vector-builder.h): New #include for tree_vector_builder usage.
11615         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11616         (altivec_vmrglw_direct): Add xxmrglw insn.
11617
11618 2018-01-17  Andrew Waterman  <andrew@sifive.com>
11619
11620         * config/riscv/riscv.c (riscv_conditional_register_usage): If
11621         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11622
11623 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
11624
11625         PR lto/83121
11626         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11627         call the lto_location_cache before reading the
11628         DECL_SOURCE_LOCATION of the types.
11629
11630 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
11631             Richard Sandiford  <richard.sandiford@linaro.org>
11632
11633         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11634         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11635         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11636         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11637         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11638         Add declaration.
11639         * config/aarch64/constraints.md (aarch64_movti_operand):
11640         Limit immediates.
11641         * config/aarch64/predicates.md (Uti): Add new constraint.
11642
11643 2018-01-17 Carl Love  <cel@us.ibm.com>
11644
11645         * config/rs6000/vsx.md (define_expand xl_len_r,
11646         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11647         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11648         lxvll.
11649         (define_expand, define_insn): Move the shift left from  the
11650         define_insn to the define_expand for lxvl and stxvl instructions.
11651         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11652         and XL_LEN_R definitions to PURE.
11653
11654 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
11655
11656         * config/i386/i386.c (indirect_thunk_name): Declare regno
11657         as unsigned int.  Compare regno with INVALID_REGNUM.
11658         (output_indirect_thunk): Ditto.
11659         (output_indirect_thunk_function): Ditto.
11660         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
11661         in the call to output_indirect_thunk_function.
11662
11663 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
11664
11665         PR middle-end/83884
11666         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11667         rather than the size of inner_type to determine the stack slot size
11668         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11669
11670 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
11671
11672         PR target/83546
11673         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11674         to PTA_SILVERMONT.
11675
11676 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
11677
11678         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11679         endian Linux systems to optionally enable multilibs for selecting
11680         the long double type if the user configured an explicit type.
11681         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11682         have no long double multilibs if not defined.
11683         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11684         warn if the user used -mabi={ieee,ibm}longdouble and we built
11685         multilibs for long double.
11686         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11687         appropriate multilib option.
11688         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11689         multilib options.
11690         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11691         for building long double multilibs.
11692         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11693
11694 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
11695
11696         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11697         copies.
11698
11699         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11700         64 bits.
11701         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11702         128 bits.
11703
11704         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11705         variables.
11706
11707         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11708         return value.
11709
11710 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
11711
11712         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11713         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11714
11715 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11716
11717         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11718         different rtl trees depending on TARGET_64BIT.
11719         (rs6000_gen_lvx): Likewise.
11720
11721 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
11722
11723         * config/visium/visium.md (nop): Tweak comment.
11724         (hazard_nop): Likewise.
11725
11726 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11727
11728         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11729         -mspeculate-indirect-jumps.
11730         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11731         for -mno-speculate-indirect-jumps.
11732         (*call_indirect_elfv2<mode>_nospec): New define_insn.
11733         (*call_value_indirect_elfv2<mode>): Disable for
11734         -mno-speculate-indirect-jumps.
11735         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11736         (indirect_jump): Emit different RTL for
11737         -mno-speculate-indirect-jumps.
11738         (*indirect_jump<mode>): Disable for
11739         -mno-speculate-indirect-jumps.
11740         (*indirect_jump<mode>_nospec): New define_insn.
11741         (tablejump): Emit different RTL for
11742         -mno-speculate-indirect-jumps.
11743         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11744         (tablejumpsi_nospec): New define_expand.
11745         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11746         (tablejumpdi_nospec): New define_expand.
11747         (*tablejump<mode>_internal1): Disable for
11748         -mno-speculate-indirect-jumps.
11749         (*tablejump<mode>_internal1_nospec): New define_insn.
11750         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11751         option.
11752
11753 2018-01-16  Artyom Skrobov tyomitch@gmail.com
11754
11755         * caller-save.c (insert_save): Drop unnecessary parameter.  All
11756         callers updated.
11757
11758 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
11759             Richard Biener  <rguenth@suse.de>
11760
11761         PR libgomp/83590
11762         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11763         return early, inline manually is_gimple_sizepos.  Make sure if we
11764         call gimplify_expr we don't end up with a gimple constant.
11765         * tree.c (variably_modified_type_p): Don't return true for
11766         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
11767         * gimplify.h (is_gimple_sizepos): Remove.
11768
11769 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
11770
11771         PR tree-optimization/83857
11772         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11773         vectorizable_live_operation for pure SLP statements.
11774         (vectorizable_live_operation): Handle PHIs.
11775
11776 2018-01-16  Richard Biener  <rguenther@suse.de>
11777
11778         PR tree-optimization/83867
11779         * tree-vect-stmts.c (vect_transform_stmt): Precompute
11780         nested_in_vect_loop_p since the scalar stmt may get invalidated.
11781
11782 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
11783
11784         PR c/83844
11785         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11786         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11787         If off is not INTEGER_CST, issue a may not be aligned warning
11788         rather than isn't aligned.  Use isn%'t rather than isn't.
11789         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11790         into MULT_EXPR.
11791         <case MULT_EXPR>: Improve the case when bottom and one of the
11792         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11793         operand, in that case check if the other operand is multiple of
11794         bottom divided by the INTEGER_CST operand.
11795
11796 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
11797
11798         PR target/83858
11799         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11800         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11801         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11802         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11803         * config/pa/pa.c (pa_function_arg_advance): Likewise.
11804         (pa_function_arg, pa_arg_partial_bytes): Likewise.
11805         (pa_function_arg_size): New function.
11806
11807 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
11808
11809         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11810         in a separate statement.
11811
11812 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
11813
11814         PR tree-optimization/83847
11815         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11816         group gathers and scatters.
11817
11818 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
11819
11820         PR rtl-optimization/86620
11821         * params.def (max-sched-ready-insns): Bump minimum value to 1.
11822
11823         PR rtl-optimization/83213
11824         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11825         to last if both are JUMP_INSNs.
11826
11827         PR tree-optimization/83843
11828         * gimple-ssa-store-merging.c
11829         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11830         store_immediate_info for bswap/nop orig_stores.
11831
11832 2018-01-15  Andrew Waterman  <andrew@sifive.com>
11833
11834         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11835         !TARGET_MUL.
11836         <UDIV>: Increase cost if !TARGET_DIV.
11837
11838 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
11839
11840         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11841         (define_attr "cr_logical_3op"): New.
11842         (cceq_ior_compare): Adjust.
11843         (cceq_ior_compare_complement): Adjust.
11844         (*cceq_rev_compare): Adjust.
11845         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11846         (is_cracked_insn): Adjust.
11847         (insn_must_be_first_in_group): Adjust.
11848         * config/rs6000/40x.md: Adjust.
11849         * config/rs6000/440.md: Adjust.
11850         * config/rs6000/476.md: Adjust.
11851         * config/rs6000/601.md: Adjust.
11852         * config/rs6000/603.md: Adjust.
11853         * config/rs6000/6xx.md: Adjust.
11854         * config/rs6000/7450.md: Adjust.
11855         * config/rs6000/7xx.md: Adjust.
11856         * config/rs6000/8540.md: Adjust.
11857         * config/rs6000/cell.md: Adjust.
11858         * config/rs6000/e300c2c3.md: Adjust.
11859         * config/rs6000/e500mc.md: Adjust.
11860         * config/rs6000/e500mc64.md: Adjust.
11861         * config/rs6000/e5500.md: Adjust.
11862         * config/rs6000/e6500.md: Adjust.
11863         * config/rs6000/mpc.md: Adjust.
11864         * config/rs6000/power4.md: Adjust.
11865         * config/rs6000/power5.md: Adjust.
11866         * config/rs6000/power6.md: Adjust.
11867         * config/rs6000/power7.md: Adjust.
11868         * config/rs6000/power8.md: Adjust.
11869         * config/rs6000/power9.md: Adjust.
11870         * config/rs6000/rs64.md: Adjust.
11871         * config/rs6000/titan.md: Adjust.
11872
11873 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11874
11875         * config/i386/predicates.md (indirect_branch_operand): Rewrite
11876         ix86_indirect_branch_register logic.
11877
11878 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11879
11880         * config/i386/constraints.md (Bs): Update
11881         ix86_indirect_branch_register check.  Don't check
11882         ix86_indirect_branch_register with GOT_memory_operand.
11883         (Bw): Likewise.
11884         * config/i386/predicates.md (GOT_memory_operand): Don't check
11885         ix86_indirect_branch_register here.
11886         (GOT32_symbol_operand): Likewise.
11887
11888 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11889
11890         * config/i386/predicates.md (constant_call_address_operand):
11891         Rewrite ix86_indirect_branch_register logic.
11892         (sibcall_insn_operand): Likewise.
11893
11894 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11895
11896         * config/i386/constraints.md (Bs): Replace
11897         ix86_indirect_branch_thunk_register with
11898         ix86_indirect_branch_register.
11899         (Bw): Likewise.
11900         * config/i386/i386.md (indirect_jump): Likewise.
11901         (tablejump): Likewise.
11902         (*sibcall_memory): Likewise.
11903         (*sibcall_value_memory): Likewise.
11904         Peepholes of indirect call and jump via memory: Likewise.
11905         * config/i386/i386.opt: Likewise.
11906         * config/i386/predicates.md (indirect_branch_operand): Likewise.
11907         (GOT_memory_operand): Likewise.
11908         (call_insn_operand): Likewise.
11909         (sibcall_insn_operand): Likewise.
11910         (GOT32_symbol_operand): Likewise.
11911
11912 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
11913
11914         PR middle-end/83837
11915         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11916         type rather than type addr's type points to.
11917         (expand_omp_atomic_mutex): Likewise.
11918         (expand_omp_atomic): Likewise.
11919
11920 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
11921
11922         PR target/83839
11923         * config/i386/i386.c (output_indirect_thunk_function): Use
11924         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11925         for  __x86_return_thunk.
11926
11927 2018-01-15  Richard Biener  <rguenther@suse.de>
11928
11929         PR middle-end/83850
11930         * expmed.c (extract_bit_field_1): Fix typo.
11931
11932 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11933
11934         PR target/83687
11935         * config/arm/iterators.md (VF): New mode iterator.
11936         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11937         Remove integer-related logic from pattern.
11938         (neon_vabd<mode>_3): Likewise.
11939
11940 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
11941
11942         PR middle-end/82694
11943         * common.opt (fstrict-overflow): No longer an alias.
11944         (fwrapv-pointer): New option.
11945         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11946         also for pointer types based on flag_wrapv_pointer.
11947         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11948         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11949         opts->x_flag_wrapv got set.
11950         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11951         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11952         POINTER_TYPE_OVERFLOW_UNDEFINED.
11953         * match.pd: Likewise in address comparison pattern.
11954         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11955
11956 2018-01-15  Richard Biener  <rguenther@suse.de>
11957
11958         PR lto/83804
11959         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11960         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
11961         Reset type names to their identifier if their TYPE_DECL doesn't
11962         have linkage (and thus is used for ODR and devirt).
11963         (save_debug_info_for_decl): Remove.
11964         (save_debug_info_for_type): Likewise.
11965         (add_tree_to_fld_list): Adjust.
11966         * tree-pretty-print.c (dump_generic_node): Make dumping of
11967         type names more robust.
11968
11969 2018-01-15  Richard Biener  <rguenther@suse.de>
11970
11971         * BASE-VER: Bump to 8.0.1.
11972
11973 2018-01-14  Martin Sebor  <msebor@redhat.com>
11974
11975         PR other/83508
11976         * builtins.c (check_access): Avoid warning when the no-warning bit
11977         is set.
11978
11979 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
11980
11981         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11982         * ira-color (allocno_hard_regs_compare): Likewise.
11983
11984 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
11985
11986         PR target/83013
11987         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11988         Use .pushsection/.popsection.
11989
11990 2018-01-14  Martin Sebor  <msebor@redhat.com>
11991
11992         PR c++/81327
11993         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11994
11995 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
11996
11997         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11998         entry from extra_headers.
11999         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12000         extra_headers, make the list bitwise identical to the i?86-*-* one.
12001
12002 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12003
12004         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12005         -mcmodel=large with -mindirect-branch=thunk,
12006         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12007         -mfunction-return=thunk-extern.
12008         * doc/invoke.texi: Document -mcmodel=large is incompatible with
12009         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12010         -mfunction-return=thunk and -mfunction-return=thunk-extern.
12011
12012 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12013
12014         * config/i386/i386.c (print_reg): Print the name of the full
12015         integer register without '%'.
12016         (ix86_print_operand): Handle 'V'.
12017         * doc/extend.texi: Document 'V' modifier.
12018
12019 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12020
12021         * config/i386/constraints.md (Bs): Disallow memory operand for
12022         -mindirect-branch-register.
12023         (Bw): Likewise.
12024         * config/i386/predicates.md (indirect_branch_operand): Likewise.
12025         (GOT_memory_operand): Likewise.
12026         (call_insn_operand): Likewise.
12027         (sibcall_insn_operand): Likewise.
12028         (GOT32_symbol_operand): Likewise.
12029         * config/i386/i386.md (indirect_jump): Call convert_memory_address
12030         for -mindirect-branch-register.
12031         (tablejump): Likewise.
12032         (*sibcall_memory): Likewise.
12033         (*sibcall_value_memory): Likewise.
12034         Disallow peepholes of indirect call and jump via memory for
12035         -mindirect-branch-register.
12036         (*call_pop): Replace m with Bw.
12037         (*call_value_pop): Likewise.
12038         (*sibcall_pop_memory): Replace m with Bs.
12039         * config/i386/i386.opt (mindirect-branch-register): New option.
12040         * doc/invoke.texi: Document -mindirect-branch-register option.
12041
12042 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12043
12044         * config/i386/i386-protos.h (ix86_output_function_return): New.
12045         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12046         set function_return_type.
12047         (indirect_thunk_name): Add ret_p to indicate thunk for function
12048         return.
12049         (output_indirect_thunk_function): Pass false to
12050         indirect_thunk_name.
12051         (ix86_output_indirect_branch_via_reg): Likewise.
12052         (ix86_output_indirect_branch_via_push): Likewise.
12053         (output_indirect_thunk_function): Create alias for function
12054         return thunk if regno < 0.
12055         (ix86_output_function_return): New function.
12056         (ix86_handle_fndecl_attribute): Handle function_return.
12057         (ix86_attribute_table): Add function_return.
12058         * config/i386/i386.h (machine_function): Add
12059         function_return_type.
12060         * config/i386/i386.md (simple_return_internal): Use
12061         ix86_output_function_return.
12062         (simple_return_internal_long): Likewise.
12063         * config/i386/i386.opt (mfunction-return=): New option.
12064         (indirect_branch): Mention -mfunction-return=.
12065         * doc/extend.texi: Document function_return function attribute.
12066         * doc/invoke.texi: Document -mfunction-return= option.
12067
12068 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
12069
12070         * config/i386/i386-opts.h (indirect_branch): New.
12071         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12072         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12073         with local indirect jump when converting indirect call and jump.
12074         (ix86_set_indirect_branch_type): New.
12075         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12076         (indirectlabelno): New.
12077         (indirect_thunk_needed): Likewise.
12078         (indirect_thunk_bnd_needed): Likewise.
12079         (indirect_thunks_used): Likewise.
12080         (indirect_thunks_bnd_used): Likewise.
12081         (INDIRECT_LABEL): Likewise.
12082         (indirect_thunk_name): Likewise.
12083         (output_indirect_thunk): Likewise.
12084         (output_indirect_thunk_function): Likewise.
12085         (ix86_output_indirect_branch_via_reg): Likewise.
12086         (ix86_output_indirect_branch_via_push): Likewise.
12087         (ix86_output_indirect_branch): Likewise.
12088         (ix86_output_indirect_jmp): Likewise.
12089         (ix86_code_end): Call output_indirect_thunk_function if needed.
12090         (ix86_output_call_insn): Call ix86_output_indirect_branch if
12091         needed.
12092         (ix86_handle_fndecl_attribute): Handle indirect_branch.
12093         (ix86_attribute_table): Add indirect_branch.
12094         * config/i386/i386.h (machine_function): Add indirect_branch_type
12095         and has_local_indirect_jump.
12096         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12097         to true.
12098         (tablejump): Likewise.
12099         (*indirect_jump): Use ix86_output_indirect_jmp.
12100         (*tablejump_1): Likewise.
12101         (simple_return_indirect_internal): Likewise.
12102         * config/i386/i386.opt (mindirect-branch=): New option.
12103         (indirect_branch): New.
12104         (keep): Likewise.
12105         (thunk): Likewise.
12106         (thunk-inline): Likewise.
12107         (thunk-extern): Likewise.
12108         * doc/extend.texi: Document indirect_branch function attribute.
12109         * doc/invoke.texi: Document -mindirect-branch= option.
12110
12111 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
12112
12113         PR ipa/83051
12114         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12115
12116 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
12117
12118         * ipa-inline.c (want_inline_small_function_p): Return false if
12119         inlining has already failed with CIF_FINAL_ERROR.
12120         (update_caller_keys): Call want_inline_small_function_p before
12121         can_inline_edge_p.
12122         (update_callee_keys): Likewise.
12123
12124 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12125
12126         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12127         New function.
12128         (rs6000_quadword_masked_address_p): Likewise.
12129         (quad_aligned_load_p): Likewise.
12130         (quad_aligned_store_p): Likewise.
12131         (const_load_sequence_p): Add comment to describe the outer-most loop.
12132         (mimic_memory_attributes_and_flags): New function.
12133         (rs6000_gen_stvx): Likewise.
12134         (replace_swapped_aligned_store): Likewise.
12135         (rs6000_gen_lvx): Likewise.
12136         (replace_swapped_aligned_load): Likewise.
12137         (replace_swapped_load_constant): Capitalize argument name in
12138         comment describing this function.
12139         (rs6000_analyze_swaps): Add a third pass to search for vector loads
12140         and stores that access quad-word aligned addresses and replace
12141         with stvx or lvx instructions when appropriate.
12142         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12143         New function prototype.
12144         (rs6000_quadword_masked_address_p): Likewise.
12145         (rs6000_gen_lvx): Likewise.
12146         (rs6000_gen_stvx): Likewise.
12147         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12148         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12149         when memory address is aligned.
12150         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12151         this split to select lvx instruction when memory address is aligned.
12152         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12153         instruction when memory address is aligned.
12154         (*vsx_le_perm_load_v16qi): Likewise.
12155         (four unnamed splitters): Modify to select the stvx instruction
12156         when memory is aligned.
12157
12158 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
12159
12160         * predict.c (determine_unlikely_bbs): Handle correctly BBs
12161         which appears in the queue multiple times.
12162
12163 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12164             Alan Hayward  <alan.hayward@arm.com>
12165             David Sherwood  <david.sherwood@arm.com>
12166
12167         * tree-vectorizer.h (vec_lower_bound): New structure.
12168         (_loop_vec_info): Add check_nonzero and lower_bounds.
12169         (LOOP_VINFO_CHECK_NONZERO): New macro.
12170         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12171         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12172         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12173         fields.  Make seg_len the distance travelled, not including the
12174         access size.
12175         (dr_direction_indicator): Declare.
12176         (dr_zero_step_indicator): Likewise.
12177         (dr_known_forward_stride_p): Likewise.
12178         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12179         tree-ssanames.h.
12180         (runtime_alias_check_p): Allow runtime alias checks with
12181         variable strides.
12182         (operator ==): Compare access_size and align.
12183         (prune_runtime_alias_test_list): Rework for new distinction between
12184         the access_size and seg_len.
12185         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
12186         segment lengths.
12187         (get_segment_min_max): New function.
12188         (create_intersect_range_checks): Use it.
12189         (dr_step_indicator): New function.
12190         (dr_direction_indicator): Likewise.
12191         (dr_zero_step_indicator): Likewise.
12192         (dr_known_forward_stride_p): Likewise.
12193         * tree-loop-distribution.c (data_ref_segment_size): Return
12194         DR_STEP * (niters - 1).
12195         (compute_alias_check_pairs): Update call to the dr_with_seg_len
12196         constructor.
12197         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12198         (vect_preserves_scalar_order_p): New function, split out from...
12199         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
12200         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12201         (vect_vfa_access_size): New function.
12202         (vect_vfa_align): Likewise.
12203         (vect_compile_time_alias): Take access_size_a and access_b arguments.
12204         (dump_lower_bound): New function.
12205         (vect_check_lower_bound): Likewise.
12206         (vect_small_gap_p): Likewise.
12207         (vectorizable_with_step_bound_p): Likewise.
12208         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12209         depencies if the vectorization factor is 1.  Convert the checks
12210         for nonzero steps into checks on the bounds of DR_STEP.  Try using
12211         a bunds check for variable steps if the minimum required step is
12212         relatively small. Update calls to the dr_with_seg_len
12213         constructor and to vect_compile_time_alias.
12214         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12215         function.
12216         (vect_loop_versioning): Call it.
12217         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12218         when retrying.
12219         (vect_estimate_min_profitable_iters): Account for any bounds checks.
12220
12221 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12222             Alan Hayward  <alan.hayward@arm.com>
12223             David Sherwood  <david.sherwood@arm.com>
12224
12225         * doc/sourcebuild.texi (vect_scatter_store): Document.
12226         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12227         optabs.
12228         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12229         Document.
12230         * genopinit.c (main): Add supports_vec_scatter_store and
12231         supports_vec_scatter_store_cached to target_optabs.
12232         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12233         IFN_MASK_SCATTER_STORE.
12234         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12235         functions.
12236         * internal-fn.h (internal_store_fn_p): Declare.
12237         (internal_fn_stored_value_index): Likewise.
12238         * internal-fn.c (scatter_store_direct): New macro.
12239         (expand_scatter_store_optab_fn): New function.
12240         (direct_scatter_store_optab_supported_p): New macro.
12241         (internal_store_fn_p): New function.
12242         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12243         IFN_MASK_SCATTER_STORE.
12244         (internal_fn_mask_index): Likewise.
12245         (internal_fn_stored_value_index): New function.
12246         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12247         for scatter stores.
12248         * optabs-query.h (supports_vec_scatter_store_p): Declare.
12249         * optabs-query.c (supports_vec_scatter_store_p): New function.
12250         * tree-vectorizer.h (vect_get_store_rhs): Declare.
12251         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12252         true for scatter stores.
12253         (vect_gather_scatter_fn_p): Handle scatter stores too.
12254         (vect_check_gather_scatter): Consider using scatter stores if
12255         supports_vec_scatter_store_p.
12256         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12257         scatter stores too.
12258         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12259         internal_fn_stored_value_index.
12260         (check_load_store_masking): Handle scatter stores too.
12261         (vect_get_store_rhs): Make public.
12262         (vectorizable_call): Use internal_store_fn_p.
12263         (vectorizable_store): Handle scatter store internal functions.
12264         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12265         when deciding whether the end of the group has been reached.
12266         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12267         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12268         (mask_scatter_store<mode>): New insns.
12269
12270 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12271             Alan Hayward  <alan.hayward@arm.com>
12272             David Sherwood  <david.sherwood@arm.com>
12273
12274         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12275         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12276         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12277         function.
12278         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12279         Use vect_truncate_gather_scatter_offset if we can't treat the
12280         operation as a normal gather load or scatter store.
12281         (get_group_load_store_type): Take the gather_scatter_info
12282         as argument.  Try using a gather load or scatter store for
12283         single-element groups.
12284         (get_load_store_type): Update calls to get_group_load_store_type
12285         and vect_use_strided_gather_scatters_p.
12286
12287 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12288             Alan Hayward  <alan.hayward@arm.com>
12289             David Sherwood  <david.sherwood@arm.com>
12290
12291         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12292         optional tree argument.
12293         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12294         null target hooks.
12295         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12296         but continue to use the current value as a fallback.
12297         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12298         to compare the updates.
12299         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12300         (get_load_store_type): Use it when handling a strided access.
12301         (vect_get_strided_load_store_ops): New function.
12302         (vect_get_data_ptr_increment): Likewise.
12303         (vectorizable_load): Handle strided gather loads.  Always pass
12304         a step to vect_create_data_ref_ptr and bump_vector_ptr.
12305
12306 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12307             Alan Hayward  <alan.hayward@arm.com>
12308             David Sherwood  <david.sherwood@arm.com>
12309
12310         * doc/md.texi (gather_load@var{m}): Document.
12311         (mask_gather_load@var{m}): Likewise.
12312         * genopinit.c (main): Add supports_vec_gather_load and
12313         supports_vec_gather_load_cached to target_optabs.
12314         * optabs-tree.c (init_tree_optimization_optabs): Use
12315         ggc_cleared_alloc to allocate target_optabs.
12316         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12317         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12318         functions.
12319         * internal-fn.h (internal_load_fn_p): Declare.
12320         (internal_gather_scatter_fn_p): Likewise.
12321         (internal_fn_mask_index): Likewise.
12322         (internal_gather_scatter_fn_supported_p): Likewise.
12323         * internal-fn.c (gather_load_direct): New macro.
12324         (expand_gather_load_optab_fn): New function.
12325         (direct_gather_load_optab_supported_p): New macro.
12326         (direct_internal_fn_optab): New function.
12327         (internal_load_fn_p): Likewise.
12328         (internal_gather_scatter_fn_p): Likewise.
12329         (internal_fn_mask_index): Likewise.
12330         (internal_gather_scatter_fn_supported_p): Likewise.
12331         * optabs-query.c (supports_at_least_one_mode_p): New function.
12332         (supports_vec_gather_load_p): Likewise.
12333         * optabs-query.h (supports_vec_gather_load_p): Declare.
12334         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12335         and memory_type field.
12336         (NUM_PATTERNS): Bump to 15.
12337         * tree-vect-data-refs.c: Include internal-fn.h.
12338         (vect_gather_scatter_fn_p): New function.
12339         (vect_describe_gather_scatter_call): Likewise.
12340         (vect_check_gather_scatter): Try using internal functions for
12341         gather loads.  Recognize existing calls to a gather load function.
12342         (vect_analyze_data_refs): Consider using gather loads if
12343         supports_vec_gather_load_p.
12344         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12345         (vect_get_gather_scatter_offset_type): Likewise.
12346         (vect_convert_mask_for_vectype): Likewise.
12347         (vect_add_conversion_to_patterm): Likewise.
12348         (vect_try_gather_scatter_pattern): Likewise.
12349         (vect_recog_gather_scatter_pattern): New pattern recognizer.
12350         (vect_vect_recog_func_ptrs): Add it.
12351         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12352         internal_fn_mask_index and internal_gather_scatter_fn_p.
12353         (check_load_store_masking): Take the gather_scatter_info as an
12354         argument and handle gather loads.
12355         (vect_get_gather_scatter_ops): New function.
12356         (vectorizable_call): Check internal_load_fn_p.
12357         (vectorizable_load): Likewise.  Handle gather load internal
12358         functions.
12359         (vectorizable_store): Update call to check_load_store_masking.
12360         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12361         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12362         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12363         (aarch64_gather_scale_operand_d): New predicates.
12364         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12365         (mask_gather_load<mode>): New insns.
12366
12367 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12368             Alan Hayward  <alan.hayward@arm.com>
12369             David Sherwood  <david.sherwood@arm.com>
12370
12371         * optabs.def (fold_left_plus_optab): New optab.
12372         * doc/md.texi (fold_left_plus_@var{m}): Document.
12373         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12374         * internal-fn.c (fold_left_direct): Define.
12375         (expand_fold_left_optab_fn): Likewise.
12376         (direct_fold_left_optab_supported_p): Likewise.
12377         * fold-const-call.c (fold_const_fold_left): New function.
12378         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12379         * tree-parloops.c (valid_reduction_p): New function.
12380         (gather_scalar_reductions): Use it.
12381         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12382         (vect_finish_replace_stmt): Declare.
12383         * tree-vect-loop.c (fold_left_reduction_fn): New function.
12384         (needs_fold_left_reduction_p): New function, split out from...
12385         (vect_is_simple_reduction): ...here.  Accept reductions that
12386         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12387         (vect_force_simple_reduction): Also store the reduction type in
12388         the assignment's STMT_VINFO_REDUC_TYPE.
12389         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12390         (merge_with_identity): New function.
12391         (vect_expand_fold_left): Likewise.
12392         (vectorize_fold_left_reduction): Likewise.
12393         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
12394         scalar phi in place for it.  Check for target support and reject
12395         cases that would reassociate the operation.  Defer the transform
12396         phase to vectorize_fold_left_reduction.
12397         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12398         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12399         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12400
12401 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12402
12403         * tree-if-conv.c (predicate_mem_writes): Remove redundant
12404         call to ifc_temp_var.
12405
12406 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12407             Alan Hayward  <alan.hayward@arm.com>
12408             David Sherwood  <david.sherwood@arm.com>
12409
12410         * target.def (legitimize_address_displacement): Take the original
12411         offset as a poly_int.
12412         * targhooks.h (default_legitimize_address_displacement): Update
12413         accordingly.
12414         * targhooks.c (default_legitimize_address_displacement): Likewise.
12415         * doc/tm.texi: Regenerate.
12416         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12417         as an argument, moving assert of ad->disp == ad->disp_term to...
12418         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
12419         Try calling targetm.legitimize_address_displacement before expanding
12420         the address rather than afterwards, and adjust for the new interface.
12421         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12422         Match the new hook interface.  Handle SVE addresses.
12423         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12424         new hook interface.
12425
12426 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12427
12428         * Makefile.in (OBJS): Add early-remat.o.
12429         * target.def (select_early_remat_modes): New hook.
12430         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12431         * doc/tm.texi: Regenerate.
12432         * targhooks.h (default_select_early_remat_modes): Declare.
12433         * targhooks.c (default_select_early_remat_modes): New function.
12434         * timevar.def (TV_EARLY_REMAT): New timevar.
12435         * passes.def (pass_early_remat): New pass.
12436         * tree-pass.h (make_pass_early_remat): Declare.
12437         * early-remat.c: New file.
12438         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12439         function.
12440         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12441
12442 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12443             Alan Hayward  <alan.hayward@arm.com>
12444             David Sherwood  <david.sherwood@arm.com>
12445
12446         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12447         vfm1 with a bound_epilog parameter.
12448         (vect_do_peeling): Update calls accordingly, and move the prologue
12449         call earlier in the function.  Treat the base bound_epilog as 0 for
12450         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
12451         this base when peeling for gaps.
12452         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12453         with fully-masked loops.
12454         (vect_estimate_min_profitable_iters): Handle the single peeled
12455         iteration in that case.
12456
12457 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12458             Alan Hayward  <alan.hayward@arm.com>
12459             David Sherwood  <david.sherwood@arm.com>
12460
12461         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12462         single-element interleaving even if the size is not a power of 2.
12463         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12464         accesses for single-element interleaving if the group size is
12465         not a power of 2.
12466
12467 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12468             Alan Hayward  <alan.hayward@arm.com>
12469             David Sherwood  <david.sherwood@arm.com>
12470
12471         * doc/md.texi (fold_extract_last_@var{m}): Document.
12472         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12473         * optabs.def (fold_extract_last_optab): New optab.
12474         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12475         * internal-fn.c (fold_extract_direct): New macro.
12476         (expand_fold_extract_optab_fn): Likewise.
12477         (direct_fold_extract_optab_supported_p): Likewise.
12478         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12479         * tree-vect-loop.c (vect_model_reduction_cost): Handle
12480         EXTRACT_LAST_REDUCTION.
12481         (get_initial_def_for_reduction): Do not create an initial vector
12482         for EXTRACT_LAST_REDUCTION reductions.
12483         (vectorizable_reduction): Leave the scalar phi in place for
12484         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
12485         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
12486         epilogue code for EXTRACT_LAST_REDUCTION and defer the
12487         transform phase to vectorizable_condition.
12488         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12489         split out from...
12490         (vect_finish_stmt_generation): ...here.
12491         (vect_finish_replace_stmt): New function.
12492         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12493         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12494         pattern.
12495         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12496
12497 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12498             Alan Hayward  <alan.hayward@arm.com>
12499             David Sherwood  <david.sherwood@arm.com>
12500
12501         * doc/md.texi (extract_last_@var{m}): Document.
12502         * optabs.def (extract_last_optab): New optab.
12503         * internal-fn.def (EXTRACT_LAST): New internal function.
12504         * internal-fn.c (cond_unary_direct): New macro.
12505         (expand_cond_unary_optab_fn): Likewise.
12506         (direct_cond_unary_optab_supported_p): Likewise.
12507         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12508         loops using EXTRACT_LAST.
12509         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12510         (extract_last_<mode>): ...this optab.
12511         (vec_extract<mode><Vel>): Update accordingly.
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         * target.def (empty_mask_is_expensive): New hook.
12518         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12519         * doc/tm.texi: Regenerate.
12520         * targhooks.h (default_empty_mask_is_expensive): Declare.
12521         * targhooks.c (default_empty_mask_is_expensive): New function.
12522         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12523         if the target says that empty masks are expensive.
12524         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12525         New function.
12526         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12527
12528 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12529             Alan Hayward  <alan.hayward@arm.com>
12530             David Sherwood  <david.sherwood@arm.com>
12531
12532         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12533         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12534         (vect_use_loop_mask_for_alignment_p): New function.
12535         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12536         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12537         niters_skip argument.  Make sure that the first niters_skip elements
12538         of the first iteration are inactive.
12539         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12540         Update call to vect_set_loop_masks_directly.
12541         (get_misalign_in_elems): New function, split out from...
12542         (vect_gen_prolog_loop_niters): ...here.
12543         (vect_update_init_of_dr): Take a code argument that specifies whether
12544         the adjustment should be added or subtracted.
12545         (vect_update_init_of_drs): Likewise.
12546         (vect_prepare_for_masked_peels): New function.
12547         (vect_do_peeling): Skip prologue peeling if we're using a mask
12548         instead.  Update call to vect_update_inits_of_drs.
12549         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12550         mask_skip_niters.
12551         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12552         alignment.  Do not include the number of peeled iterations in
12553         the minimum threshold in that case.
12554         (vectorizable_induction): Adjust the start value down by
12555         LOOP_VINFO_MASK_SKIP_NITERS iterations.
12556         (vect_transform_loop): Call vect_prepare_for_masked_peels.
12557         Take the number of skipped iterations into account when calculating
12558         the loop bounds.
12559         * tree-vect-stmts.c (vect_gen_while_not): New function.
12560
12561 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12562             Alan Hayward  <alan.hayward@arm.com>
12563             David Sherwood  <david.sherwood@arm.com>
12564
12565         * doc/sourcebuild.texi (vect_fully_masked): Document.
12566         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12567         default value to 0.
12568         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12569         split out from...
12570         (vect_analyze_loop_2): ...here. Don't check the vectorization
12571         factor against the number of loop iterations if the loop is
12572         fully-masked.
12573
12574 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12575             Alan Hayward  <alan.hayward@arm.com>
12576             David Sherwood  <david.sherwood@arm.com>
12577
12578         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12579         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12580         (dump_groups): Update accordingly.
12581         (iv_use::mem_type): New member variable.
12582         (address_p): New function.
12583         (record_use): Add a mem_type argument and initialize the new
12584         mem_type field.
12585         (record_group_use): Add a mem_type argument.  Use address_p.
12586         Remove obsolete null checks of base_object.  Update call to record_use.
12587         (find_interesting_uses_op): Update call to record_group_use.
12588         (find_interesting_uses_cond): Likewise.
12589         (find_interesting_uses_address): Likewise.
12590         (get_mem_type_for_internal_fn): New function.
12591         (find_address_like_use): Likewise.
12592         (find_interesting_uses_stmt): Try find_address_like_use before
12593         calling find_interesting_uses_op.
12594         (addr_offset_valid_p): Use the iv mem_type field as the type
12595         of the addressed memory.
12596         (add_autoinc_candidates): Likewise.
12597         (get_address_cost): Likewise.
12598         (split_small_address_groups_p): Use address_p.
12599         (split_address_groups): Likewise.
12600         (add_iv_candidate_for_use): Likewise.
12601         (autoinc_possible_for_pair): Likewise.
12602         (rewrite_groups): Likewise.
12603         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12604         (determine_group_iv_cost): Update after split of USE_ADDRESS.
12605         (get_alias_ptr_type_for_ptr_address): New function.
12606         (rewrite_use_address): Rewrite address uses in calls that were
12607         identified by find_address_like_use.
12608
12609 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12610             Alan Hayward  <alan.hayward@arm.com>
12611             David Sherwood  <david.sherwood@arm.com>
12612
12613         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12614         TARGET_MEM_REFs.
12615         * gimple-expr.h (is_gimple_addressable: Likewise.
12616         * gimple-expr.c (is_gimple_address): Likewise.
12617         * internal-fn.c (expand_call_mem_ref): New function.
12618         (expand_mask_load_optab_fn): Use it.
12619         (expand_mask_store_optab_fn): Likewise.
12620
12621 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12622             Alan Hayward  <alan.hayward@arm.com>
12623             David Sherwood  <david.sherwood@arm.com>
12624
12625         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12626         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12627         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12628         (cond_umax@var{mode}): Document.
12629         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12630         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12631         (cond_umin_optab, cond_umax_optab): New optabs.
12632         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12633         (COND_IOR, COND_XOR): New internal functions.
12634         * internal-fn.h (get_conditional_internal_fn): Declare.
12635         * internal-fn.c (cond_binary_direct): New macro.
12636         (expand_cond_binary_optab_fn): Likewise.
12637         (direct_cond_binary_optab_supported_p): Likewise.
12638         (get_conditional_internal_fn): New function.
12639         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12640         Cope with reduction statements that are vectorized as calls rather
12641         than assignments.
12642         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12643         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12644         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12645         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12646         (UNSPEC_COND_EOR): New unspecs.
12647         (optab): Add mappings for them.
12648         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12649         (sve_int_op, sve_fp_op): New int attributes.
12650
12651 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12652             Alan Hayward  <alan.hayward@arm.com>
12653             David Sherwood  <david.sherwood@arm.com>
12654
12655         * optabs.def (while_ult_optab): New optab.
12656         * doc/md.texi (while_ult@var{m}@var{n}): Document.
12657         * internal-fn.def (WHILE_ULT): New internal function.
12658         * internal-fn.h (direct_internal_fn_supported_p): New override
12659         that takes two types as argument.
12660         * internal-fn.c (while_direct): New macro.
12661         (expand_while_optab_fn): New function.
12662         (convert_optab_supported_p): Likewise.
12663         (direct_while_optab_supported_p): New macro.
12664         * wide-int.h (wi::udiv_ceil): New function.
12665         * tree-vectorizer.h (rgroup_masks): New structure.
12666         (vec_loop_masks): New typedef.
12667         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12668         and fully_masked_p.
12669         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12670         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12671         (vect_max_vf): New function.
12672         (slpeel_make_loop_iterate_ntimes): Delete.
12673         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12674         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12675         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12676         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12677         internal-fn.h, stor-layout.h and optabs-query.h.
12678         (vect_set_loop_mask): New function.
12679         (add_preheader_seq): Likewise.
12680         (add_header_seq): Likewise.
12681         (interleave_supported_p): Likewise.
12682         (vect_maybe_permute_loop_masks): Likewise.
12683         (vect_set_loop_masks_directly): Likewise.
12684         (vect_set_loop_condition_masked): Likewise.
12685         (vect_set_loop_condition_unmasked): New function, split out from
12686         slpeel_make_loop_iterate_ntimes.
12687         (slpeel_make_loop_iterate_ntimes): Rename to..
12688         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
12689         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12690         (vect_do_peeling): Update call accordingly.
12691         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12692         loops.
12693         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12694         mask_compare_type, can_fully_mask_p and fully_masked_p.
12695         (release_vec_loop_masks): New function.
12696         (_loop_vec_info): Use it to free the loop masks.
12697         (can_produce_all_loop_masks_p): New function.
12698         (vect_get_max_nscalars_per_iter): Likewise.
12699         (vect_verify_full_masking): Likewise.
12700         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12701         retries, and free the mask rgroups before retrying.  Check loop-wide
12702         reasons for disallowing fully-masked loops.  Make the final decision
12703         about whether use a fully-masked loop or not.
12704         (vect_estimate_min_profitable_iters): Do not assume that peeling
12705         for the number of iterations will be needed for fully-masked loops.
12706         (vectorizable_reduction): Disable fully-masked loops.
12707         (vectorizable_live_operation): Likewise.
12708         (vect_halve_mask_nunits): New function.
12709         (vect_double_mask_nunits): Likewise.
12710         (vect_record_loop_mask): Likewise.
12711         (vect_get_loop_mask): Likewise.
12712         (vect_transform_loop): Handle the case in which the final loop
12713         iteration might handle a partial vector.  Call vect_set_loop_condition
12714         instead of slpeel_make_loop_iterate_ntimes.
12715         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12716         (check_load_store_masking): New function.
12717         (prepare_load_store_mask): Likewise.
12718         (vectorizable_store): Handle fully-masked loops.
12719         (vectorizable_load): Likewise.
12720         (supportable_widening_operation): Use vect_halve_mask_nunits for
12721         booleans.
12722         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12723         (vect_gen_while): New function.
12724         * config/aarch64/aarch64.md (umax<mode>3): New expander.
12725         (aarch64_uqdec<mode>): New insn.
12726
12727 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12728             Alan Hayward  <alan.hayward@arm.com>
12729             David Sherwood  <david.sherwood@arm.com>
12730
12731         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12732         (reduc_xor_scal_optab): New optabs.
12733         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12734         (reduc_xor_scal_@var{m}): Document.
12735         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12736         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12737         internal functions.
12738         * fold-const-call.c (fold_const_call): Handle them.
12739         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12740         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12741         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12742         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12743         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12744         (UNSPEC_XORV): New unspecs.
12745         (optab): Add entries for them.
12746         (BITWISEV): New int iterator.
12747         (bit_reduc_op): New int attributes.
12748
12749 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12750             Alan Hayward  <alan.hayward@arm.com>
12751             David Sherwood  <david.sherwood@arm.com>
12752
12753         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12754         * internal-fn.def (VEC_SHL_INSERT): New internal function.
12755         * optabs.def (vec_shl_insert_optab): New optab.
12756         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12757         (duplicate_and_interleave): Likewise.
12758         * tree-vect-loop.c: Include internal-fn.h.
12759         (neutral_op_for_slp_reduction): New function, split out from
12760         get_initial_defs_for_reduction.
12761         (get_initial_def_for_reduction): Handle option 2 for variable-length
12762         vectors by loading the neutral value into a vector and then shifting
12763         the initial value into element 0.
12764         (get_initial_defs_for_reduction): Replace the code argument with
12765         the neutral value calculated by neutral_op_for_slp_reduction.
12766         Use gimple_build_vector for constant-length vectors.
12767         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12768         but the first group_size elements have a neutral value.
12769         Use duplicate_and_interleave otherwise.
12770         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12771         Update call to get_initial_defs_for_reduction.  Handle SLP
12772         reductions for variable-length vectors by creating one vector
12773         result for each scalar result, with the elements associated
12774         with other scalar results stubbed out with the neutral value.
12775         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12776         Require IFN_VEC_SHL_INSERT for double reductions on
12777         variable-length vectors, or SLP reductions that have
12778         a neutral value.  Require can_duplicate_and_interleave_p
12779         support for variable-length unchained SLP reductions if there
12780         is no neutral value, such as for MIN/MAX reductions.  Also require
12781         the number of vector elements to be a multiple of the number of
12782         SLP statements when doing variable-length unchained SLP reductions.
12783         Update call to vect_create_epilog_for_reduction.
12784         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12785         and remove initial values.
12786         (duplicate_and_interleave): Make public.
12787         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12788         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12789
12790 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12791             Alan Hayward  <alan.hayward@arm.com>
12792             David Sherwood  <david.sherwood@arm.com>
12793
12794         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12795         (can_duplicate_and_interleave_p): New function.
12796         (vect_get_and_check_slp_defs): Take the vector of statements
12797         rather than just the current one.  Remove excess parentheses.
12798         Restriction rejectinon of vect_constant_def and vect_external_def
12799         for variable-length vectors to boolean types, or types for which
12800         can_duplicate_and_interleave_p is false.
12801         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12802         (duplicate_and_interleave): New function.
12803         (vect_get_constant_vectors): Use gimple_build_vector for
12804         constant-length vectors and suitable variable-length constant
12805         vectors.  Use duplicate_and_interleave for other variable-length
12806         vectors.  Don't defer the update when inserting new statements.
12807
12808 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12809             Alan Hayward  <alan.hayward@arm.com>
12810             David Sherwood  <david.sherwood@arm.com>
12811
12812         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12813         min_profitable_iters doesn't go negative.
12814
12815 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12816             Alan Hayward  <alan.hayward@arm.com>
12817             David Sherwood  <david.sherwood@arm.com>
12818
12819         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12820         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12821         * optabs.def (vec_mask_load_lanes_optab): New optab.
12822         (vec_mask_store_lanes_optab): Likewise.
12823         * internal-fn.def (MASK_LOAD_LANES): New internal function.
12824         (MASK_STORE_LANES): Likewise.
12825         * internal-fn.c (mask_load_lanes_direct): New macro.
12826         (mask_store_lanes_direct): Likewise.
12827         (expand_mask_load_optab_fn): Handle masked operations.
12828         (expand_mask_load_lanes_optab_fn): New macro.
12829         (expand_mask_store_optab_fn): Handle masked operations.
12830         (expand_mask_store_lanes_optab_fn): New macro.
12831         (direct_mask_load_lanes_optab_supported_p): Likewise.
12832         (direct_mask_store_lanes_optab_supported_p): Likewise.
12833         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12834         parameter.
12835         (vect_load_lanes_supported): Likewise.
12836         * tree-vect-data-refs.c (strip_conversion): New function.
12837         (can_group_stmts_p): Likewise.
12838         (vect_analyze_data_ref_accesses): Use it instead of checking
12839         for a pair of assignments.
12840         (vect_store_lanes_supported): Take a masked_p parameter.
12841         (vect_load_lanes_supported): Likewise.
12842         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12843         vect_store_lanes_supported and vect_load_lanes_supported.
12844         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12845         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12846         parameter.  Don't allow gaps for masked accesses.
12847         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
12848         and vect_load_lanes_supported.
12849         (get_load_store_type): Take a masked_p parameter and update
12850         call to get_group_load_store_type.
12851         (vectorizable_store): Update call to get_load_store_type.
12852         Handle IFN_MASK_STORE_LANES.
12853         (vectorizable_load): Update call to get_load_store_type.
12854         Handle IFN_MASK_LOAD_LANES.
12855
12856 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12857             Alan Hayward  <alan.hayward@arm.com>
12858             David Sherwood  <david.sherwood@arm.com>
12859
12860         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12861         modes for SVE.
12862         * config/aarch64/aarch64-protos.h
12863         (aarch64_sve_struct_memory_operand_p): Declare.
12864         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12865         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12866         (VPRED, vpred): Handle SVE structure modes.
12867         * config/aarch64/constraints.md (Utx): New constraint.
12868         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12869         (aarch64_sve_struct_nonimmediate_operand): New predicates.
12870         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12871         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12872         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12873         structure modes.  Split into pieces after RA.
12874         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12875         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12876         New patterns.
12877         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12878         SVE structure modes.
12879         (aarch64_classify_address): Likewise.
12880         (sizetochar): Move earlier in file.
12881         (aarch64_print_operand): Handle SVE register lists.
12882         (aarch64_array_mode): New function.
12883         (aarch64_sve_struct_memory_operand_p): Likewise.
12884         (TARGET_ARRAY_MODE): Redefine.
12885
12886 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12887             Alan Hayward  <alan.hayward@arm.com>
12888             David Sherwood  <david.sherwood@arm.com>
12889
12890         * target.def (array_mode): New target hook.
12891         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12892         * doc/tm.texi: Regenerate.
12893         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12894         * hooks.c (hook_optmode_mode_uhwi_none): New function.
12895         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12896         targetm.array_mode.
12897         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
12898         type sizes.
12899
12900 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12901             Alan Hayward  <alan.hayward@arm.com>
12902             David Sherwood  <david.sherwood@arm.com>
12903
12904         * fold-const.c (fold_binary_loc): Check the argument types
12905         rather than the result type when testing for a vector operation.
12906
12907 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12908
12909         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12910         * doc/tm.texi: Regenerate.
12911
12912 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12913             Alan Hayward  <alan.hayward@arm.com>
12914             David Sherwood  <david.sherwood@arm.com>
12915
12916         * doc/invoke.texi (-msve-vector-bits=): Document new option.
12917         (sve): Document new AArch64 extension.
12918         * doc/md.texi (w): Extend the description of the AArch64
12919         constraint to include SVE vectors.
12920         (Upl, Upa): Document new AArch64 predicate constraints.
12921         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12922         enum.
12923         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12924         (msve-vector-bits=): New option.
12925         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12926         SVE when these are disabled.
12927         (sve): New extension.
12928         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12929         modes.  Adjust their number of units based on aarch64_sve_vg.
12930         (MAX_BITSIZE_MODE_ANY_MODE): Define.
12931         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12932         aarch64_addr_query_type.
12933         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12934         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12935         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12936         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12937         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12938         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12939         (aarch64_simd_imm_zero_p): Delete.
12940         (aarch64_check_zero_based_sve_index_immediate): Declare.
12941         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12942         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12943         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12944         (aarch64_sve_float_mul_immediate_p): Likewise.
12945         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12946         rather than an rtx.
12947         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12948         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12949         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12950         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12951         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12952         (aarch64_regmode_natural_size): Likewise.
12953         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12954         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12955         left one place.
12956         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12957         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12958         for VG and the SVE predicate registers.
12959         (V_ALIASES): Add a "z"-prefixed alias.
12960         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12961         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12962         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12963         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12964         (REG_CLASS_NAMES): Add entries for them.
12965         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
12966         and the predicate registers.
12967         (aarch64_sve_vg): Declare.
12968         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12969         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12970         (REGMODE_NATURAL_SIZE): Define.
12971         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12972         SVE macros.
12973         * config/aarch64/aarch64.c: Include cfgrtl.h.
12974         (simd_immediate_info): Add a constructor for series vectors,
12975         and an associated step field.
12976         (aarch64_sve_vg): New variable.
12977         (aarch64_dbx_register_number): Handle VG and the predicate registers.
12978         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12979         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12980         (VEC_ANY_DATA, VEC_STRUCT): New constants.
12981         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12982         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12983         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12984         (aarch64_get_mask_mode): New functions.
12985         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12986         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12987         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
12988         predicate modes and predicate registers.  Explicitly restrict
12989         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
12990         to store a vector mode if it is recognized by
12991         aarch64_classify_vector_mode.
12992         (aarch64_regmode_natural_size): New function.
12993         (aarch64_hard_regno_caller_save_mode): Return the original mode
12994         for predicates.
12995         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12996         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12997         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12998         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12999         functions.
13000         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
13001         does not overlap dest if the function is frame-related.  Handle
13002         SVE constants.
13003         (aarch64_split_add_offset): New function.
13004         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13005         them aarch64_add_offset.
13006         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13007         and update call to aarch64_sub_sp.
13008         (aarch64_add_cfa_expression): New function.
13009         (aarch64_expand_prologue): Pass extra temporary registers to the
13010         functions above.  Handle the case in which we need to emit new
13011         DW_CFA_expressions for registers that were originally saved
13012         relative to the stack pointer, but now have to be expressed
13013         relative to the frame pointer.
13014         (aarch64_output_mi_thunk): Pass extra temporary registers to the
13015         functions above.
13016         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
13017         IP0 and IP1 values for SVE frames.
13018         (aarch64_expand_vec_series): New function.
13019         (aarch64_expand_sve_widened_duplicate): Likewise.
13020         (aarch64_expand_sve_const_vector): Likewise.
13021         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13022         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
13023         into the register, rather than emitting a SET directly.
13024         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13025         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13026         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13027         (offset_9bit_signed_scaled_p): New functions.
13028         (aarch64_replicate_bitmask_imm): New function.
13029         (aarch64_bitmask_imm): Use it.
13030         (aarch64_cannot_force_const_mem): Reject expressions involving
13031         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
13032         (aarch64_classify_index): Handle SVE indices, by requiring
13033         a plain register index with a scale that matches the element size.
13034         (aarch64_classify_address): Handle SVE addresses.  Assert that
13035         the mode of the address is VOIDmode or an integer mode.
13036         Update call to aarch64_classify_symbol.
13037         (aarch64_classify_symbolic_expression): Update call to
13038         aarch64_classify_symbol.
13039         (aarch64_const_vec_all_in_range_p): New function.
13040         (aarch64_print_vector_float_operand): Likewise.
13041         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
13042         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
13043         and the FP immediates 1.0 and 0.5.
13044         (aarch64_print_address_internal): Handle SVE addresses.
13045         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13046         (aarch64_regno_regclass): Handle predicate registers.
13047         (aarch64_secondary_reload): Handle big-endian reloads of SVE
13048         data modes.
13049         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13050         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13051         (aarch64_convert_sve_vector_bits): New function.
13052         (aarch64_override_options): Use it to handle -msve-vector-bits=.
13053         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13054         rather than an rtx.
13055         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13056         Handle SVE vector and predicate modes.  Accept VL-based constants
13057         that need only one temporary register, and VL offsets that require
13058         no temporary registers.
13059         (aarch64_conditional_register_usage): Mark the predicate registers
13060         as fixed if SVE isn't available.
13061         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13062         Return true for SVE vector and predicate modes.
13063         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13064         rather than an unsigned int.  Handle SVE modes.
13065         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
13066         SVE modes.
13067         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13068         if SVE is enabled.
13069         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13070         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13071         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13072         (aarch64_sve_float_mul_immediate_p): New functions.
13073         (aarch64_sve_valid_immediate): New function.
13074         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13075         Explicitly reject structure modes.  Check for INDEX constants.
13076         Handle PTRUE and PFALSE constants.
13077         (aarch64_check_zero_based_sve_index_immediate): New function.
13078         (aarch64_simd_imm_zero_p): Delete.
13079         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13080         vector modes.  Accept constants in the range of CNT[BHWD].
13081         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13082         ask for an Advanced SIMD mode.
13083         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13084         (aarch64_simd_vector_alignment): Handle SVE predicates.
13085         (aarch64_vectorize_preferred_vector_alignment): New function.
13086         (aarch64_simd_vector_alignment_reachable): Use it instead of
13087         the vector size.
13088         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13089         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13090         functions.
13091         (MAX_VECT_LEN): Delete.
13092         (expand_vec_perm_d): Add a vec_flags field.
13093         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13094         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13095         (aarch64_evpc_ext): Don't apply a big-endian lane correction
13096         for SVE modes.
13097         (aarch64_evpc_rev): Rename to...
13098         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
13099         (aarch64_evpc_rev_global): New function.
13100         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13101         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13102         MAX_VECT_LEN.
13103         (aarch64_evpc_sve_tbl): New function.
13104         (aarch64_expand_vec_perm_const_1): Update after rename of
13105         aarch64_evpc_rev.  Handle SVE permutes too, trying
13106         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13107         than aarch64_evpc_tbl.
13108         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13109         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13110         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13111         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13112         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13113         (aarch64_expand_sve_vcond): New functions.
13114         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13115         of aarch64_vector_mode_p.
13116         (aarch64_dwarf_poly_indeterminate_value): New function.
13117         (aarch64_compute_pressure_classes): Likewise.
13118         (aarch64_can_change_mode_class): Likewise.
13119         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13120         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13121         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13122         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13123         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13124         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13125         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13126         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13127         constraints.
13128         (Dn, Dl, Dr): Accept const as well as const_vector.
13129         (Dz): Likewise.  Compare against CONST0_RTX.
13130         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13131         of "vector" where appropriate.
13132         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13133         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13134         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13135         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13136         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13137         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13138         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13139         (v_int_equiv): Extend to SVE modes.
13140         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13141         mode attributes.
13142         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13143         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13144         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13145         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13146         (SVE_COND_FP_CMP): New int iterators.
13147         (perm_hilo): Handle the new unpack unspecs.
13148         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13149         attributes.
13150         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13151         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13152         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13153         (aarch64_equality_operator, aarch64_constant_vector_operand)
13154         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13155         (aarch64_sve_nonimmediate_operand): Likewise.
13156         (aarch64_sve_general_operand): Likewise.
13157         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13158         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13159         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13160         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13161         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13162         (aarch64_sve_float_arith_immediate): Likewise.
13163         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13164         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13165         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13166         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13167         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13168         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13169         (aarch64_sve_float_arith_operand): Likewise.
13170         (aarch64_sve_float_arith_with_sub_operand): Likewise.
13171         (aarch64_sve_float_mul_operand): Likewise.
13172         (aarch64_sve_vec_perm_operand): Likewise.
13173         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13174         (aarch64_mov_operand): Accept const_poly_int and const_vector.
13175         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13176         as well as const_vector.
13177         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13178         in file.  Use CONST0_RTX and CONSTM1_RTX.
13179         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
13180         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13181         Use aarch64_simd_imm_zero.
13182         * config/aarch64/aarch64-sve.md: New file.
13183         * config/aarch64/aarch64.md: Include it.
13184         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13185         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13186         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13187         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13188         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13189         (sve): New attribute.
13190         (enabled): Disable instructions with the sve attribute unless
13191         TARGET_SVE.
13192         (movqi, movhi): Pass CONST_POLY_INT operaneds through
13193         aarch64_expand_mov_immediate.
13194         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13195         CNT[BHSD] immediates.
13196         (movti): Split CONST_POLY_INT moves into two halves.
13197         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13198         Split additions that need a temporary here if the destination
13199         is the stack pointer.
13200         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13201         (*add<mode>3_poly_1): New instruction.
13202         (set_clobber_cc): New expander.
13203
13204 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
13205
13206         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13207         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
13208         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13209         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13210         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
13211         Change innermode from fixed_mode_size to machine_mode.
13212         (simplify_subreg): Update call accordingly.  Handle a constant-sized
13213         subreg of a variable-length CONST_VECTOR.
13214
13215 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
13216             Alan Hayward  <alan.hayward@arm.com>
13217             David Sherwood  <david.sherwood@arm.com>
13218
13219         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13220         (add_offset_to_base): New function, split out from...
13221         (create_mem_ref): ...here.  When handling a scale other than 1,
13222         check first whether the address is valid without the offset.
13223         Add it into the base if so, leaving the index and scale as-is.
13224
13225 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
13226
13227         PR c++/83778
13228         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13229         fold_for_warn before checking if arg2 is INTEGER_CST.
13230
13231 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
13232
13233         * config/rs6000/predicates.md (load_multiple_operation): Delete.
13234         (store_multiple_operation): Delete.
13235         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13236         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13237         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13238         guarded by TARGET_STRING.
13239         (rs6000_output_load_multiple): Delete.
13240         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13241         OPTION_MASK_STRING / TARGET_STRING handling.
13242         (print_operand) <'N', 'O'>: Add comment that these are unused now.
13243         (const rs6000_opt_masks) <"string">: Change mask to 0.
13244         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13245         (MASK_STRING): Delete.
13246         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13247         parts.  Simplify.
13248         (load_multiple): Delete.
13249         (*ldmsi8): Delete.
13250         (*ldmsi7): Delete.
13251         (*ldmsi6): Delete.
13252         (*ldmsi5): Delete.
13253         (*ldmsi4): Delete.
13254         (*ldmsi3): Delete.
13255         (store_multiple): Delete.
13256         (*stmsi8): Delete.
13257         (*stmsi7): Delete.
13258         (*stmsi6): Delete.
13259         (*stmsi5): Delete.
13260         (*stmsi4): Delete.
13261         (*stmsi3): Delete.
13262         (movmemsi_8reg): Delete.
13263         (corresponding unnamed define_insn): Delete.
13264         (movmemsi_6reg): Delete.
13265         (corresponding unnamed define_insn): Delete.
13266         (movmemsi_4reg): Delete.
13267         (corresponding unnamed define_insn): Delete.
13268         (movmemsi_2reg): Delete.
13269         (corresponding unnamed define_insn): Delete.
13270         (movmemsi_1reg): Delete.
13271         (corresponding unnamed define_insn): Delete.
13272         * config/rs6000/rs6000.opt (mno-string): New.
13273         (mstring): Replace by deprecation warning stub.
13274         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13275
13276 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
13277
13278         * regrename.c (regrename_do_replace): If replacing the same
13279         reg multiple times, try to reuse last created gen_raw_REG.
13280
13281         PR debug/81155
13282         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13283         main to workaround a bug in GDB.
13284
13285 2018-01-12  Tom de Vries  <tom@codesourcery.com>
13286
13287         PR target/83737
13288         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13289
13290 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
13291
13292         PR rtl-optimization/80481
13293         * ira-color.c (get_cap_member): New function.
13294         (allocnos_conflict_by_live_ranges_p): Use it.
13295         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13296         (setup_slot_coalesced_allocno_live_ranges): Ditto.
13297
13298 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
13299
13300         PR target/83628
13301         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13302         (*saddl_se_1): Ditto.
13303         (*ssubsi_1): Ditto.
13304         (*ssubl_se_1): Ditto.
13305
13306 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
13307
13308         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13309         rather than wi::to_widest for DR_INITs.
13310         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13311         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13312         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13313         INTEGER_CSTs.
13314         (vect_analyze_group_access_1): Note that here.
13315
13316 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
13317
13318         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13319         polynomial type sizes.
13320
13321 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
13322
13323         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13324         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13325         (gimple_add_tmp_var): Likewise.
13326
13327 2018-01-12  Martin Liska  <mliska@suse.cz>
13328
13329         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13330         (gimple_alloc_sizes): Likewise.
13331         (dump_gimple_statistics): Use PRIu64 in printf format.
13332         * gimple.h: Change uint64_t to int.
13333
13334 2018-01-12  Martin Liska  <mliska@suse.cz>
13335
13336         * tree-core.h: Use uint64_t instead of int.
13337         * tree.c (tree_node_counts): Likewise.
13338         (tree_node_sizes): Likewise.
13339         (dump_tree_statistics): Use PRIu64 in printf format.
13340
13341 2018-01-12  Martin Liska  <mliska@suse.cz>
13342
13343         * Makefile.in: As qsort_chk is implemented in vec.c, add
13344         vec.o to linkage of gencfn-macros.
13345         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13346         passing the info to record_node_allocation_statistics.
13347         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13348         and pass the info.
13349         * ggc-common.c (struct ggc_usage): Add operator== and use
13350         it in operator< and compare function.
13351         * mem-stats.h (struct mem_usage): Likewise.
13352         * vec.c (struct vec_usage): Remove operator< and compare
13353         function. Can be simply inherited.
13354
13355 2018-01-12  Martin Jambor  <mjambor@suse.cz>
13356
13357         PR target/81616
13358         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13359         * tree-ssa-math-opts.c: Include domwalk.h.
13360         (convert_mult_to_fma_1): New function.
13361         (fma_transformation_info): New type.
13362         (fma_deferring_state): Likewise.
13363         (cancel_fma_deferring): New function.
13364         (result_of_phi): Likewise.
13365         (last_fma_candidate_feeds_initial_phi): Likewise.
13366         (convert_mult_to_fma): Added deferring logic, split actual
13367         transformation to convert_mult_to_fma_1.
13368         (math_opts_dom_walker): New type.
13369         (math_opts_dom_walker::after_dom_children): New method, body moved
13370         here from pass_optimize_widening_mul::execute, added deferring logic
13371         bits.
13372         (pass_optimize_widening_mul::execute): Moved most of code to
13373         math_opts_dom_walker::after_dom_children.
13374         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13375         * config/i386/i386.c (ix86_option_override_internal): Added
13376         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13377
13378 2018-01-12  Richard Biener  <rguenther@suse.de>
13379
13380         PR debug/83157
13381         * dwarf2out.c (gen_variable_die): Do not reset old_die for
13382         inline instance vars.
13383
13384 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
13385
13386         PR target/81819
13387         * config/rx/rx.c (rx_is_restricted_memory_address):
13388         Handle SUBREG case.
13389
13390 2018-01-12  Richard Biener  <rguenther@suse.de>
13391
13392         PR tree-optimization/80846
13393         * target.def (split_reduction): New target hook.
13394         * targhooks.c (default_split_reduction): New function.
13395         * targhooks.h (default_split_reduction): Declare.
13396         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13397         target requests first reduce vectors by combining low and high
13398         parts.
13399         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13400         (get_vectype_for_scalar_type_and_size): Export.
13401         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13402         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13403         * doc/tm.texi: Regenerate.
13404         * config/i386/i386.c (ix86_split_reduction): Implement
13405         TARGET_VECTORIZE_SPLIT_REDUCTION.
13406
13407 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
13408
13409         PR target/83368
13410         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13411         in PIC mode except for TARGET_VXWORKS_RTP.
13412         * config/sparc/sparc.c: Include cfgrtl.h.
13413         (TARGET_INIT_PIC_REG): Define.
13414         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13415         (sparc_pic_register_p): New predicate.
13416         (sparc_legitimate_address_p): Use it.
13417         (sparc_legitimize_pic_address): Likewise.
13418         (sparc_delegitimize_address): Likewise.
13419         (sparc_mode_dependent_address_p): Likewise.
13420         (gen_load_pcrel_sym): Remove 4th parameter.
13421         (load_got_register): Adjust call to above.  Remove obsolete stuff.
13422         (sparc_expand_prologue): Do not call load_got_register here.
13423         (sparc_flat_expand_prologue): Likewise.
13424         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13425         (sparc_use_pseudo_pic_reg): New function.
13426         (sparc_init_pic_reg): Likewise.
13427         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13428         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13429
13430 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
13431
13432         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13433         Add item for branch_cost.
13434
13435 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
13436
13437         PR rtl-optimization/83565
13438         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13439         not extend the result to a larger mode for rotate operations.
13440         (num_sign_bit_copies1): Likewise.
13441
13442 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13443
13444         PR target/40411
13445         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13446         -symbolic.
13447         Use values-Xc.o for -pedantic.
13448         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13449
13450 2018-01-12  Martin Liska  <mliska@suse.cz>
13451
13452         PR ipa/83054
13453         * ipa-devirt.c (final_warning_record::grow_type_warnings):
13454         New function.
13455         (possible_polymorphic_call_targets): Use it.
13456         (ipa_devirt): Likewise.
13457
13458 2018-01-12  Martin Liska  <mliska@suse.cz>
13459
13460         * profile-count.h (enum profile_quality): Use 0 as invalid
13461         enum value of profile_quality.
13462
13463 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
13464
13465         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13466         -mext-string options.
13467
13468 2018-01-12  Richard Biener  <rguenther@suse.de>
13469
13470         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13471         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13472         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13473         Likewise.
13474         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13475
13476 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
13477
13478         * configure.ac (--with-long-double-format): Add support for the
13479         configuration option to change the default long double format on
13480         PowerPC systems.
13481         * config.gcc (powerpc*-linux*-*): Likewise.
13482         * configure: Regenerate.
13483         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13484         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13485         used without modification.
13486
13487 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13488
13489         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13490         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13491         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13492         MISC_BUILTIN_SPEC_BARRIER.
13493         (rs6000_init_builtins): Likewise.
13494         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13495         enum value.
13496         (speculation_barrier): New define_insn.
13497         * doc/extend.texi: Document __builtin_speculation_barrier.
13498
13499 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
13500
13501         PR target/83203
13502         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13503         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13504         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13505         iterators.
13506         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
13507         integral modes instead of "ss" and "sd".
13508         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13509         vectors with 32-bit and 64-bit elements.
13510         (vecdupssescalarmodesuffix): New mode attribute.
13511         (vec_dup<mode>): Use it.
13512
13513 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
13514
13515         PR target/83330
13516         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13517         frame if argument is passed on stack.
13518
13519 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
13520
13521         PR target/82682
13522         * ree.c (combine_reaching_defs): Optimize also
13523         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13524         reg2=any_extend(exp); reg1=reg2;, formatting fix.
13525
13526 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
13527
13528         PR middle-end/83189
13529         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13530
13531 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
13532
13533         PR middle-end/83718
13534         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13535         after they are computed.
13536
13537 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
13538
13539         PR tree-optimization/83695
13540         * gimple-loop-linterchange.cc
13541         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13542         reset cached scev information after interchange.
13543         (pass_linterchange::execute): Remove call to scev_reset_htab.
13544
13545 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13546
13547         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13548         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13549         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13550         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13551         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13552         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13553         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13554         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13555         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13556         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13557         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13558         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13559         (V_lane_reg): Likewise.
13560         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13561         New define_expand.
13562         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13563         (vfmal_lane_low<mode>_intrinsic,
13564         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13565         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13566         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13567         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13568         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13569         vfmsl_lane_high<mode>_intrinsic): New define_insns.
13570
13571 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13572
13573         * config/arm/arm-cpus.in (fp16fml): New feature.
13574         (ALL_SIMD): Add fp16fml.
13575         (armv8.2-a): Add fp16fml as an option.
13576         (armv8.3-a): Likewise.
13577         (armv8.4-a): Add fp16fml as part of fp16.
13578         * config/arm/arm.h (TARGET_FP16FML): Define.
13579         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13580         when appropriate.
13581         * config/arm/arm-modes.def (V2HF): Define.
13582         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13583         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13584         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13585         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13586         vfmsl_low, vfmsl_high): New set of builtins.
13587         * config/arm/iterators.md (PLUSMINUS): New code iterator.
13588         (vfml_op): New code attribute.
13589         (VFMLHALVES): New int iterator.
13590         (VFML, VFMLSEL): New mode attributes.
13591         (V_reg): Define mapping for V2HF.
13592         (V_hi, V_lo): New mode attributes.
13593         (VF_constraint): Likewise.
13594         (vfml_half, vfml_half_selector): New int attributes.
13595         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13596         define_expand.
13597         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13598         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13599         New define_insn.
13600         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13601         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13602         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13603         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
13604         documentation.
13605         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13606         Document new effective target and option set.
13607
13608 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13609
13610         * config/arm/arm-cpus.in (armv8_4): New feature.
13611         (ARMv8_4a): New fgroup.
13612         (armv8.4-a): New arch.
13613         * config/arm/arm-tables.opt: Regenerate.
13614         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13615         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13616         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13617         Add matching rules for -march=armv8.4-a and extensions.
13618         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13619
13620 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
13621
13622         PR target/81821
13623         * config/rx/rx.md (BW): New mode attribute.
13624         (sync_lock_test_and_setsi): Add mode suffix to insn output.
13625
13626 2018-01-11  Richard Biener  <rguenther@suse.de>
13627
13628         PR tree-optimization/83435
13629         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13630         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13631         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13632
13633 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13634             Alan Hayward  <alan.hayward@arm.com>
13635             David Sherwood  <david.sherwood@arm.com>
13636
13637         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13638         field.
13639         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
13640         (aarch64_print_address_internal): Use it to check for a zero offset.
13641
13642 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13643             Alan Hayward  <alan.hayward@arm.com>
13644             David Sherwood  <david.sherwood@arm.com>
13645
13646         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13647         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13648         Return a poly_int64 rather than a HOST_WIDE_INT.
13649         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13650         rather than a HOST_WIDE_INT.
13651         * config/aarch64/aarch64.h (aarch64_frame): Protect with
13652         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
13653         hard_fp_offset, frame_size, initial_adjust, callee_offset and
13654         final_offset from HOST_WIDE_INT to poly_int64.
13655         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13656         to_constant when getting the number of units in an Advanced SIMD
13657         mode.
13658         (aarch64_builtin_vectorized_function): Check for a constant number
13659         of units.
13660         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13661         GET_MODE_SIZE.
13662         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13663         attribute instead of GET_MODE_NUNITS.
13664         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13665         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13666         GET_MODE_SIZE for fixed-size registers.
13667         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13668         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13669         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13670         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13671         (aarch64_print_operand, aarch64_print_address_internal)
13672         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13673         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13674         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13675         Handle polynomial GET_MODE_SIZE.
13676         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
13677         wider than SImode without modification.
13678         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13679         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13680         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13681         passing and returning SVE modes.
13682         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13683         rather than GEN_INT.
13684         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13685         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13686         (aarch64_allocate_and_probe_stack_space): Likewise.
13687         (aarch64_layout_frame): Cope with polynomial offsets.
13688         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13689         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
13690         polynomial offsets.
13691         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13692         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13693         poly_int64 rather than a HOST_WIDE_INT.
13694         (aarch64_get_separate_components, aarch64_process_components)
13695         (aarch64_expand_prologue, aarch64_expand_epilogue)
13696         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13697         (aarch64_anchor_offset): New function, split out from...
13698         (aarch64_legitimize_address): ...here.
13699         (aarch64_builtin_vectorization_cost): Handle polynomial
13700         TYPE_VECTOR_SUBPARTS.
13701         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13702         GET_MODE_NUNITS.
13703         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13704         number of elements from the PARALLEL rather than the mode.
13705         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13706         rather than GET_MODE_BITSIZE.
13707         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13708         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13709         (aarch64_expand_vec_perm_const_1): Handle polynomial
13710         d->perm.length () and d->perm elements.
13711         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
13712         Apply to_constant to d->perm elements.
13713         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13714         polynomial CONST_VECTOR_NUNITS.
13715         (aarch64_move_pointer): Take amount as a poly_int64 rather
13716         than an int.
13717         (aarch64_progress_pointer): Avoid temporary variable.
13718         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13719         the mode attribute instead of GET_MODE.
13720
13721 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13722             Alan Hayward  <alan.hayward@arm.com>
13723             David Sherwood  <david.sherwood@arm.com>
13724
13725         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13726         x exists before using it.
13727         (aarch64_add_constant_internal): Rename to...
13728         (aarch64_add_offset_1): ...this.  Replace regnum with separate
13729         src and dest rtxes.  Handle the case in which they're different,
13730         including when the offset is zero.  Replace scratchreg with an rtx.
13731         Use 2 additions if there is no spare register into which we can
13732         move a 16-bit constant.
13733         (aarch64_add_constant): Delete.
13734         (aarch64_add_offset): Replace reg with separate src and dest
13735         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
13736         Use aarch64_add_offset_1.
13737         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13738         an rtx rather than an int.  Take the delta as a poly_int64
13739         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
13740         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13741         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13742         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13743         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13744         and aarch64_add_sp.
13745         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13746         aarch64_add_constant.
13747
13748 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13749
13750         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13751         Use scalar_float_mode.
13752
13753 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13754
13755         * config/aarch64/aarch64-simd.md
13756         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13757         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13758         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13759         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13760         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13761         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13762         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13763         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13764         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13765         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13766
13767 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13768
13769         PR target/83514
13770         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13771         targ_options->x_arm_arch_string is non NULL.
13772
13773 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
13774
13775         * config/aarch64/aarch64.h
13776         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
13777
13778 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
13779
13780         PR target/82096
13781         * expmed.c (emit_store_flag_force): Swap if const op0
13782         and change VOIDmode to mode of op0.
13783
13784 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
13785
13786         PR rtl-optimization/83761
13787         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13788         than bytes to mode_for_size.
13789
13790 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
13791
13792         PR middle-end/83189
13793         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13794         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13795         profile.
13796
13797 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
13798
13799         PR middle-end/83575
13800         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13801         when in layout mode.
13802         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13803         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13804         partition fixup.
13805
13806 2018-01-10  Michael Collison  <michael.collison@arm.com>
13807
13808         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13809         * config/aarch64/aarch64-option-extension.def: Add
13810         AARCH64_OPT_EXTENSION of 'fp16fml'.
13811         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13812         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13813         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13814         * config/aarch64/constraints.md (Ui7): New constraint.
13815         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13816         (VFMLA_SEL_W): Ditto.
13817         (f16quad): Ditto.
13818         (f16mac1): Ditto.
13819         (VFMLA16_LOW): New int iterator.
13820         (VFMLA16_HIGH): Ditto.
13821         (UNSPEC_FMLAL): New unspec.
13822         (UNSPEC_FMLSL): Ditto.
13823         (UNSPEC_FMLAL2): Ditto.
13824         (UNSPEC_FMLSL2): Ditto.
13825         (f16mac): New code attribute.
13826         * config/aarch64/aarch64-simd-builtins.def
13827         (aarch64_fmlal_lowv2sf): Ditto.
13828         (aarch64_fmlsl_lowv2sf): Ditto.
13829         (aarch64_fmlalq_lowv4sf): Ditto.
13830         (aarch64_fmlslq_lowv4sf): Ditto.
13831         (aarch64_fmlal_highv2sf): Ditto.
13832         (aarch64_fmlsl_highv2sf): Ditto.
13833         (aarch64_fmlalq_highv4sf): Ditto.
13834         (aarch64_fmlslq_highv4sf): Ditto.
13835         (aarch64_fmlal_lane_lowv2sf): Ditto.
13836         (aarch64_fmlsl_lane_lowv2sf): Ditto.
13837         (aarch64_fmlal_laneq_lowv2sf): Ditto.
13838         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13839         (aarch64_fmlalq_lane_lowv4sf): Ditto.
13840         (aarch64_fmlsl_lane_lowv4sf): Ditto.
13841         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13842         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13843         (aarch64_fmlal_lane_highv2sf): Ditto.
13844         (aarch64_fmlsl_lane_highv2sf): Ditto.
13845         (aarch64_fmlal_laneq_highv2sf): Ditto.
13846         (aarch64_fmlsl_laneq_highv2sf): Ditto.
13847         (aarch64_fmlalq_lane_highv4sf): Ditto.
13848         (aarch64_fmlsl_lane_highv4sf): Ditto.
13849         (aarch64_fmlalq_laneq_highv4sf): Ditto.
13850         (aarch64_fmlsl_laneq_highv4sf): Ditto.
13851         * config/aarch64/aarch64-simd.md:
13852         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13853         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13854         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13855         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13856         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13857         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13858         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13859         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13860         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13861         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13862         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13863         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13864         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13865         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13866         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13867         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13868         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13869         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13870         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13871         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13872         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13873         (vfmlsl_low_u32): Ditto.
13874         (vfmlalq_low_u32): Ditto.
13875         (vfmlslq_low_u32): Ditto.
13876         (vfmlal_high_u32): Ditto.
13877         (vfmlsl_high_u32): Ditto.
13878         (vfmlalq_high_u32): Ditto.
13879         (vfmlslq_high_u32): Ditto.
13880         (vfmlal_lane_low_u32): Ditto.
13881         (vfmlsl_lane_low_u32): Ditto.
13882         (vfmlal_laneq_low_u32): Ditto.
13883         (vfmlsl_laneq_low_u32): Ditto.
13884         (vfmlalq_lane_low_u32): Ditto.
13885         (vfmlslq_lane_low_u32): Ditto.
13886         (vfmlalq_laneq_low_u32): Ditto.
13887         (vfmlslq_laneq_low_u32): Ditto.
13888         (vfmlal_lane_high_u32): Ditto.
13889         (vfmlsl_lane_high_u32): Ditto.
13890         (vfmlal_laneq_high_u32): Ditto.
13891         (vfmlsl_laneq_high_u32): Ditto.
13892         (vfmlalq_lane_high_u32): Ditto.
13893         (vfmlslq_lane_high_u32): Ditto.
13894         (vfmlalq_laneq_high_u32): Ditto.
13895         (vfmlslq_laneq_high_u32): Ditto.
13896         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13897         (AARCH64_FL_FOR_ARCH8_4): New.
13898         (AARCH64_ISA_F16FML): New ISA flag.
13899         (TARGET_F16FML): New feature flag for fp16fml.
13900         (doc/invoke.texi): Document new fp16fml option.
13901
13902 2018-01-10  Michael Collison  <michael.collison@arm.com>
13903
13904         * config/aarch64/aarch64-builtins.c:
13905         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13906         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13907         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13908         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13909         (AARCH64_ISA_SHA3): New ISA flag.
13910         (TARGET_SHA3): New feature flag for sha3.
13911         * config/aarch64/iterators.md (sha512_op): New int attribute.
13912         (CRYPTO_SHA512): New int iterator.
13913         (UNSPEC_SHA512H): New unspec.
13914         (UNSPEC_SHA512H2): Ditto.
13915         (UNSPEC_SHA512SU0): Ditto.
13916         (UNSPEC_SHA512SU1): Ditto.
13917         * config/aarch64/aarch64-simd-builtins.def
13918         (aarch64_crypto_sha512hqv2di): New builtin.
13919         (aarch64_crypto_sha512h2qv2di): Ditto.
13920         (aarch64_crypto_sha512su0qv2di): Ditto.
13921         (aarch64_crypto_sha512su1qv2di): Ditto.
13922         (aarch64_eor3qv8hi): Ditto.
13923         (aarch64_rax1qv2di): Ditto.
13924         (aarch64_xarqv2di): Ditto.
13925         (aarch64_bcaxqv8hi): Ditto.
13926         * config/aarch64/aarch64-simd.md:
13927         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13928         (aarch64_crypto_sha512su0qv2di): Ditto.
13929         (aarch64_crypto_sha512su1qv2di): Ditto.
13930         (aarch64_eor3qv8hi): Ditto.
13931         (aarch64_rax1qv2di): Ditto.
13932         (aarch64_xarqv2di): Ditto.
13933         (aarch64_bcaxqv8hi): Ditto.
13934         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13935         (vsha512h2q_u64): Ditto.
13936         (vsha512su0q_u64): Ditto.
13937         (vsha512su1q_u64): Ditto.
13938         (veor3q_u16): Ditto.
13939         (vrax1q_u64): Ditto.
13940         (vxarq_u64): Ditto.
13941         (vbcaxq_u16): Ditto.
13942         * config/arm/types.md (crypto_sha512): New type attribute.
13943         (crypto_sha3): Ditto.
13944         (doc/invoke.texi): Document new sha3 option.
13945
13946 2018-01-10  Michael Collison  <michael.collison@arm.com>
13947
13948         * config/aarch64/aarch64-builtins.c:
13949         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13950         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13951         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13952         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13953         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13954         (AARCH64_ISA_SM4): New ISA flag.
13955         (TARGET_SM4): New feature flag for sm4.
13956         * config/aarch64/aarch64-simd-builtins.def
13957         (aarch64_sm3ss1qv4si): Ditto.
13958         (aarch64_sm3tt1aq4si): Ditto.
13959         (aarch64_sm3tt1bq4si): Ditto.
13960         (aarch64_sm3tt2aq4si): Ditto.
13961         (aarch64_sm3tt2bq4si): Ditto.
13962         (aarch64_sm3partw1qv4si): Ditto.
13963         (aarch64_sm3partw2qv4si): Ditto.
13964         (aarch64_sm4eqv4si): Ditto.
13965         (aarch64_sm4ekeyqv4si): Ditto.
13966         * config/aarch64/aarch64-simd.md:
13967         (aarch64_sm3ss1qv4si): Ditto.
13968         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13969         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13970         (aarch64_sm4eqv4si): Ditto.
13971         (aarch64_sm4ekeyqv4si): Ditto.
13972         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13973         (sm3part_op): Ditto.
13974         (CRYPTO_SM3TT): Ditto.
13975         (CRYPTO_SM3PART): Ditto.
13976         (UNSPEC_SM3SS1): New unspec.
13977         (UNSPEC_SM3TT1A): Ditto.
13978         (UNSPEC_SM3TT1B): Ditto.
13979         (UNSPEC_SM3TT2A): Ditto.
13980         (UNSPEC_SM3TT2B): Ditto.
13981         (UNSPEC_SM3PARTW1): Ditto.
13982         (UNSPEC_SM3PARTW2): Ditto.
13983         (UNSPEC_SM4E): Ditto.
13984         (UNSPEC_SM4EKEY): Ditto.
13985         * config/aarch64/constraints.md (Ui2): New constraint.
13986         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13987         * config/arm/types.md (crypto_sm3): New type attribute.
13988         (crypto_sm4): Ditto.
13989         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13990         (vsm3tt1aq_u32): Ditto.
13991         (vsm3tt1bq_u32): Ditto.
13992         (vsm3tt2aq_u32): Ditto.
13993         (vsm3tt2bq_u32): Ditto.
13994         (vsm3partw1q_u32): Ditto.
13995         (vsm3partw2q_u32): Ditto.
13996         (vsm4eq_u32): Ditto.
13997         (vsm4ekeyq_u32): Ditto.
13998         (doc/invoke.texi): Document new sm4 option.
13999
14000 2018-01-10  Michael Collison  <michael.collison@arm.com>
14001
14002         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14003         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14004         (AARCH64_FL_FOR_ARCH8_4): New.
14005         (AARCH64_FL_V8_4): New flag.
14006         (doc/invoke.texi): Document new armv8.4-a option.
14007
14008 2018-01-10  Michael Collison  <michael.collison@arm.com>
14009
14010         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14011         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14012         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14013         * config/aarch64/aarch64-option-extension.def: Add
14014         AARCH64_OPT_EXTENSION of 'sha2'.
14015         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14016         (crypto): Disable sha2 and aes if crypto disabled.
14017         (crypto): Enable aes and sha2 if enabled.
14018         (simd): Disable sha2 and aes if simd disabled.
14019         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14020         New flags.
14021         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14022         (TARGET_SHA2): New feature flag for sha2.
14023         (TARGET_AES): New feature flag for aes.
14024         * config/aarch64/aarch64-simd.md:
14025         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14026         conditional on TARGET_AES.
14027         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14028         (aarch64_crypto_sha1hsi): Make pattern conditional
14029         on TARGET_SHA2.
14030         (aarch64_crypto_sha1hv4si): Ditto.
14031         (aarch64_be_crypto_sha1hv4si): Ditto.
14032         (aarch64_crypto_sha1su1v4si): Ditto.
14033         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14034         (aarch64_crypto_sha1su0v4si): Ditto.
14035         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14036         (aarch64_crypto_sha256su0v4si): Ditto.
14037         (aarch64_crypto_sha256su1v4si): Ditto.
14038         (doc/invoke.texi): Document new aes and sha2 options.
14039
14040 2018-01-10  Martin Sebor  <msebor@redhat.com>
14041
14042         PR tree-optimization/83781
14043         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14044         as string arrays.
14045
14046 2018-01-11  Martin Sebor  <msebor@gmail.com>
14047             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14048
14049         PR tree-optimization/83501
14050         PR tree-optimization/81703
14051
14052         * tree-ssa-strlen.c (get_string_cst): Rename...
14053         (get_string_len): ...to this.  Handle global constants.
14054         (handle_char_store): Adjust.
14055
14056 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
14057             Jim Wilson  <jimw@sifive.com>
14058
14059         * config/riscv/riscv-protos.h (riscv_output_return): New.
14060         * config/riscv/riscv.c (struct machine_function): New naked_p field.
14061         (riscv_attribute_table, riscv_output_return),
14062         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14063         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14064         (riscv_compute_frame_info): Only compute frame->mask if not a naked
14065         function.
14066         (riscv_expand_prologue): Add early return for naked function.
14067         (riscv_expand_epilogue): Likewise.
14068         (riscv_function_ok_for_sibcall): Return false for naked function.
14069         (riscv_set_current_function): New.
14070         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14071         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14072         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14073         * doc/extend.texi (RISC-V Function Attributes): New.
14074
14075 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
14076
14077         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14078         check for 128-bit long double before checking TCmode.
14079         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14080         128-bit long doubles before checking TFmode or TCmode.
14081         (FLOAT128_IBM_P): Likewise.
14082
14083 2018-01-10  Martin Sebor  <msebor@redhat.com>
14084
14085         PR tree-optimization/83671
14086         * builtins.c (c_strlen): Unconditionally return zero for the empty
14087         string.
14088         Use -Warray-bounds for warnings.
14089         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14090         for non-constant array indices with COMPONENT_REF, arrays of
14091         arrays, and pointers to arrays.
14092         (gimple_fold_builtin_strlen): Determine and set length range for
14093         non-constant character arrays.
14094
14095 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
14096
14097         PR middle-end/81897
14098         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14099         empty blocks.
14100
14101 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
14102
14103         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14104
14105 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
14106
14107         PR target/83399
14108         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14109         VECTOR_MEM_ALTIVEC_OR_VSX_P.
14110         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14111         indexed_or_indirect_operand predicate.
14112         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14113         (*vsx_le_perm_load_v8hi): Likewise.
14114         (*vsx_le_perm_load_v16qi): Likewise.
14115         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14116         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14117         (*vsx_le_perm_store_v8hi): Likewise.
14118         (*vsx_le_perm_store_v16qi): Likewise.
14119         (eight unnamed splitters): Likewise.
14120
14121 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
14122
14123         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14124         * config/rs6000/emmintrin.h: Likewise.
14125         * config/rs6000/mmintrin.h: Likewise.
14126         * config/rs6000/xmmintrin.h: Likewise.
14127
14128 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
14129
14130         PR c++/43486
14131         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14132         "public_flag".
14133         * tree.c (tree_nop_conversion): Return true for location wrapper
14134         nodes.
14135         (maybe_wrap_with_location): New function.
14136         (selftest::check_strip_nops): New function.
14137         (selftest::test_location_wrappers): New function.
14138         (selftest::tree_c_tests): Call it.
14139         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14140         (maybe_wrap_with_location): New decl.
14141         (EXPR_LOCATION_WRAPPER_P): New macro.
14142         (location_wrapper_p): New inline function.
14143         (tree_strip_any_location_wrapper): New inline function.
14144
14145 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
14146
14147         PR target/83735
14148         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14149         stack_realign_offset for the largest alignment of stack slot
14150         actually used.
14151         (ix86_find_max_used_stack_alignment): New function.
14152         (ix86_finalize_stack_frame_flags): Use it.  Set
14153         max_used_stack_alignment if we don't realign stack.
14154         * config/i386/i386.h (machine_function): Add
14155         max_used_stack_alignment.
14156
14157 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
14158
14159         * config/arm/arm.opt (-mbranch-cost): New option.
14160         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14161         account.
14162
14163 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
14164
14165         PR target/83629
14166         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14167         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14168
14169 2018-01-10  Richard Biener  <rguenther@suse.de>
14170
14171         PR debug/83765
14172         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14173         early out so it also covers the case where we have a non-NULL
14174         origin.
14175
14176 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
14177
14178         PR tree-optimization/83753
14179         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14180         for non-strided grouped accesses if the number of elements is 1.
14181
14182 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
14183
14184         PR target/81616
14185         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14186         * i386.h (TARGET_USE_GATHER): Define.
14187         * x86-tune.def (X86_TUNE_USE_GATHER): New.
14188
14189 2018-01-10  Martin Liska  <mliska@suse.cz>
14190
14191         PR bootstrap/82831
14192         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14193         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14194         partitioning.
14195         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14196         CLEANUP_NO_PARTITIONING is not set.
14197
14198 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
14199
14200         * doc/rtl.texi: Remove documentation of (const ...) wrappers
14201         for vectors, as a partial revert of r254296.
14202         * rtl.h (const_vec_p): Delete.
14203         (const_vec_duplicate_p): Don't test for vector CONSTs.
14204         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14205         * expmed.c (make_tree): Likewise.
14206
14207         Revert:
14208         * common.md (E, F): Use CONSTANT_P instead of checking for
14209         CONST_VECTOR.
14210         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14211         checking for CONST_VECTOR.
14212
14213 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
14214
14215         PR middle-end/83575
14216         * predict.c (force_edge_cold): Handle in more sane way edges
14217         with no prediction.
14218
14219 2018-01-09  Carl Love  <cel@us.ibm.com>
14220
14221         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14222         V4SI, V4SF types.
14223         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14224         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14225         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14226         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
14227         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14228         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
14229         * config/rs6000/rs6000-protos.h: Add extern defition for
14230         rs6000_generate_float2_double_code.
14231         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14232         function.
14233         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14234         (float2_v2df): Add define_expand.
14235
14236 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
14237
14238         PR target/83628
14239         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14240         op_mode in the force_to_mode call.
14241
14242 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
14243
14244         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14245         instead of checking each element individually.
14246         (aarch64_evpc_uzp): Likewise.
14247         (aarch64_evpc_zip): Likewise.
14248         (aarch64_evpc_ext): Likewise.
14249         (aarch64_evpc_rev): Likewise.
14250         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14251         instead of checking each element individually.  Return true without
14252         generating rtl if
14253         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14254         whether all selected elements come from the same input, instead of
14255         checking each element individually.  Remove calls to gen_rtx_REG,
14256         start_sequence and end_sequence and instead assert that no rtl is
14257         generated.
14258
14259 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
14260
14261         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14262         order of HIGH and CONST checks.
14263
14264 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
14265
14266         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14267         if the destination isn't an SSA_NAME.
14268
14269 2018-01-09  Richard Biener  <rguenther@suse.de>
14270
14271         PR tree-optimization/83668
14272         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14273         move prologue...
14274         (canonicalize_loop_form): ... here, renamed from ...
14275         (canonicalize_loop_closed_ssa_form): ... this and amended to
14276         swap successor edges for loop exit blocks to make us use
14277         the RPO order we need for initial schedule generation.
14278
14279 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
14280
14281         PR tree-optimization/64811
14282         * match.pd: When optimizing comparisons with Inf, avoid
14283         introducing or losing exceptions from comparisons with NaN.
14284
14285 2018-01-09  Martin Liska  <mliska@suse.cz>
14286
14287         PR sanitizer/82517
14288         * asan.c (shadow_mem_size): Add gcc_assert.
14289
14290 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
14291
14292         Don't save registers in main().
14293
14294         PR target/83738
14295         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14296         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14297         * config/avr/avr.c (avr_set_current_function): Don't error if
14298         naked, OS_task or OS_main are specified at the same time.
14299         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14300         OS_main.
14301         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14302         attribute.
14303         * common/config/avr/avr-common.c (avr_option_optimization_table):
14304         Switch on -mmain-is-OS_task for optimizing compilations.
14305
14306 2018-01-09  Richard Biener  <rguenther@suse.de>
14307
14308         PR tree-optimization/83572
14309         * graphite.c: Include cfganal.h.
14310         (graphite_transform_loops): Connect infinite loops to exit
14311         and remove fake edges at the end.
14312
14313 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
14314
14315         * ipa-inline.c (edge_badness): Revert accidental checkin.
14316
14317 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
14318
14319         PR ipa/80763
14320         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14321         symbols; not inline clones.
14322
14323 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
14324
14325         PR target/83507
14326         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14327         hard registers.  Formatting fixes.
14328
14329         PR preprocessor/83722
14330         * gcc.c (try_generate_repro): Pass
14331         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14332         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14333         do_report_bug.
14334
14335 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
14336             Kito Cheng  <kito.cheng@gmail.com>
14337
14338         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14339         (riscv_leaf_function_p): Delete.
14340         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14341
14342 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14343
14344         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14345         function.
14346         (do_ifelse): New function.
14347         (do_isel): New function.
14348         (do_sub3): New function.
14349         (do_add3): New function.
14350         (do_load_mask_compare): New function.
14351         (do_overlap_load_compare): New function.
14352         (expand_compare_loop): New function.
14353         (expand_block_compare): Call expand_compare_loop() when appropriate.
14354         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14355         option description.
14356         (-mblock-compare-inline-loop-limit): New option.
14357
14358 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14359
14360         PR target/83677
14361         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14362         Reverse order of second and third operands in first alternative.
14363         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14364         of first and second elements in UNSPEC_VPERMR vector.
14365         (altivec_expand_vec_perm_le): Likewise.
14366
14367 2018-01-08  Jeff Law  <law@redhat.com>
14368
14369         PR rtl-optimizatin/81308
14370         * tree-switch-conversion.c (cfg_altered): New file scoped static.
14371         (process_switch): If group_case_labels makes a change, then set
14372         cfg_altered.
14373         (pass_convert_switch::execute): If a switch is converted, then
14374         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
14375
14376         PR rtl-optimization/81308
14377         * recog.c (split_all_insns): Conditionally cleanup the CFG after
14378         splitting insns.
14379
14380 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
14381
14382         PR target/83663 - Revert r255946
14383         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14384         generation for cases where splatting a value is not useful.
14385         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14386         across a vec_duplicate and a paradoxical subreg forming a vector
14387         mode to a vec_concat.
14388
14389 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14390
14391         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14392         -march=armv8.3-a variants.
14393         * config/arm/t-multilib: Likewise.
14394         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
14395
14396 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14397
14398         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14399         to generate rtl.
14400         (cceq_ior_compare_complement): Give it a name so I can use it, and
14401         change boolean_or_operator predicate to boolean_operator so it can
14402         be used to generate a crand.
14403         (eqne): New code iterator.
14404         (bd/bd_neg): New code_attrs.
14405         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14406         a single define_insn.
14407         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14408         decrement (bdnzt/bdnzf/bdzt/bdzf).
14409         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14410         with the new names of the branch decrement patterns, and added the
14411         names of the branch decrement conditional patterns.
14412
14413 2018-01-08  Richard Biener  <rguenther@suse.de>
14414
14415         PR tree-optimization/83563
14416         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14417         cache.
14418
14419 2018-01-08  Richard Biener  <rguenther@suse.de>
14420
14421         PR middle-end/83713
14422         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14423
14424 2018-01-08  Richard Biener  <rguenther@suse.de>
14425
14426         PR tree-optimization/83685
14427         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14428         references to abnormals.
14429
14430 2018-01-08  Richard Biener  <rguenther@suse.de>
14431
14432         PR lto/83719
14433         * dwarf2out.c (output_indirect_strings): Handle empty
14434         skeleton_debug_str_hash.
14435         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14436
14437 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
14438
14439         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14440         (emit_store_direct): Likewise.
14441         (arc_trampoline_adjust_address): Likewise.
14442         (arc_asm_trampoline_template): New function.
14443         (arc_initialize_trampoline): Use asm_trampoline_template.
14444         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14445         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14446         * config/arc/arc.md (flush_icache): Delete pattern.
14447
14448 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
14449
14450         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14451         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14452         munaligned-access.
14453
14454 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14455
14456         PR target/83681
14457         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14458         by not USED_FOR_TARGET.
14459         (make_pass_resolve_sw_modes): Likewise.
14460
14461 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14462
14463         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14464         USED_FOR_TARGET.
14465
14466 2018-01-08  Richard Biener  <rguenther@suse.de>
14467
14468         PR middle-end/83580
14469         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14470
14471 2018-01-08  Richard Biener  <rguenther@suse.de>
14472
14473         PR middle-end/83517
14474         * match.pd ((t * 2) / 2) -> t): Add missing :c.
14475
14476 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
14477
14478         PR middle-end/81897
14479         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14480         basic blocks with a small number of successors.
14481         (convert_control_dep_chain_into_preds): Improve handling of
14482         forwarder blocks.
14483         (dump_predicates): Split apart into...
14484         (dump_pred_chain): ...here...
14485         (dump_pred_info): ...and here.
14486         (can_one_predicate_be_invalidated_p): Add debugging printfs.
14487         (can_chain_union_be_invalidated_p): Improve check for invalidation
14488         of paths.
14489         (uninit_uses_cannot_happen): Avoid unnecessary if
14490         convert_control_dep_chain_into_preds yielded nothing.
14491
14492 2018-01-06  Martin Sebor  <msebor@redhat.com>
14493
14494         PR tree-optimization/83640
14495         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14496         subtracting negative offset from size.
14497         (builtin_access::overlap): Adjust offset bounds of the access to fall
14498         within the size of the object if possible.
14499
14500 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
14501
14502         PR rtl-optimization/83699
14503         * expmed.c (extract_bit_field_1): Restrict the vector usage of
14504         extract_bit_field_as_subreg to cases in which the extracted
14505         value is also a vector.
14506
14507         * lra-constraints.c (process_alt_operands): Test for the equivalence
14508         substitutions when detecting a possible reload cycle.
14509
14510 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
14511
14512         PR debug/83480
14513         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14514         by default if flag_selective_schedling{,2}.  Formatting fixes.
14515
14516         PR rtl-optimization/83682
14517         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14518         if it has non-VECTOR_MODE element mode.
14519         (vec_duplicate_p): Likewise.
14520
14521         PR middle-end/83694
14522         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14523         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14524
14525 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
14526
14527         PR target/83604
14528         * config/i386/i386-builtin.def
14529         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14530         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14531         Require also OPTION_MASK_ISA_AVX512F in addition to
14532         OPTION_MASK_ISA_GFNI.
14533         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14534         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14535         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14536         to OPTION_MASK_ISA_GFNI.
14537         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14538         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14539         OPTION_MASK_ISA_AVX512BW.
14540         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14541         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14542         addition to OPTION_MASK_ISA_GFNI.
14543         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14544         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14545         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14546         to OPTION_MASK_ISA_GFNI.
14547         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14548         a requirement for all ISAs rather than any of them with a few
14549         exceptions.
14550         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14551         processing.
14552         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14553         bitmasks to be enabled with 3 exceptions, instead of requiring any
14554         enabled ISA with lots of exceptions.
14555         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14556         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14557         Change avx512bw in isa attribute to avx512f.
14558         * config/i386/sgxintrin.h: Add license boilerplate.
14559         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
14560         to __AVX512F__ and __AVX512VL to __AVX512VL__.
14561         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14562         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14563         defined.
14564         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14565         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14566         temporarily sse2 rather than sse if not enabled already.
14567
14568         PR target/83604
14569         * config/i386/sse.md (VI248_VLBW): Rename to ...
14570         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
14571         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14572         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14573         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14574         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14575         mode iterator instead of VI248_VLBW.
14576
14577 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
14578
14579         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14580         (record_modified): Skip clobbers; add debug output.
14581         (param_change_prob): Use sreal frequencies.
14582
14583 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
14584
14585         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14586         punt for user-aligned variables.
14587
14588 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
14589
14590         * tree-chrec.c (chrec_contains_symbols): Return true for
14591         POLY_INT_CST.
14592
14593 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
14594
14595         PR target/82439
14596         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14597         of (x|y) == x for BICS pattern.
14598
14599 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
14600
14601         PR tree-optimization/83605
14602         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14603         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14604         can throw.
14605
14606 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14607
14608         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14609         * config/epiphany/rtems.h: New file.
14610
14611 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
14612             Uros Bizjak  <ubizjak@gmail.com>
14613
14614         PR target/83554
14615         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14616         QIreg_operand instead of register_operand predicate.
14617         * config/i386/i386.c (ix86_rop_should_change_byte_p,
14618         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14619         comments instead of -fmitigate[-_]rop.
14620
14621 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14622
14623         PR bootstrap/81926
14624         * cgraphunit.c (symbol_table::compile): Switch to text_section
14625         before calling assembly_start debug hook.
14626         * run-rtl-passes.c (run_rtl_passes): Likewise.
14627         Include output.h.
14628
14629 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14630
14631         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14632         range_int_cst_p rather than !symbolic_range_p before calling
14633         extract_range_from_multiplicative_op_1.
14634
14635 2018-01-04  Jeff Law  <law@redhat.com>
14636
14637         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14638         redundant test in assertion.
14639
14640 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14641
14642         * doc/rtl.texi: Document machine_mode wrapper classes.
14643
14644 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14645
14646         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14647         using tree_to_uhwi.
14648
14649 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14650
14651         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14652         the VEC_PERM_EXPR fold to fail.
14653
14654 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
14655
14656         PR debug/83585
14657         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14658         to switched_sections.
14659
14660 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14661
14662         PR target/83680
14663         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14664         test for d.testing.
14665
14666 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
14667
14668         PR target/83387
14669         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14670         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14671
14672 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
14673
14674         PR debug/83666
14675         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14676         is BLKmode and bitpos not zero or mode change is needed.
14677
14678 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
14679
14680         PR target/83675
14681         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14682         TARGET_VIS2.
14683
14684 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
14685
14686         PR target/83628
14687         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14688         instead of MULT rtx.  Update all corresponding splitters.
14689         (*saddl_se): Ditto.
14690         (*ssub<modesuffix>): Ditto.
14691         (*ssubl_se): Ditto.
14692         (*cmp_sadd_di): Update split patterns.
14693         (*cmp_sadd_si): Ditto.
14694         (*cmp_sadd_sidi): Ditto.
14695         (*cmp_ssub_di): Ditto.
14696         (*cmp_ssub_si): Ditto.
14697         (*cmp_ssub_sidi): Ditto.
14698         * config/alpha/predicates.md (const23_operand): New predicate.
14699         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14700         Look for ASHIFT, not MULT inner operand.
14701         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14702
14703 2018-01-04  Martin Liska  <mliska@suse.cz>
14704
14705         PR gcov-profile/83669
14706         * gcov.c (output_intermediate_file): Add version to intermediate
14707         gcov file.
14708         * doc/gcov.texi: Document new field 'version' in intermediate
14709         file format. Fix location of '-k' option of gcov command.
14710
14711 2018-01-04  Martin Liska  <mliska@suse.cz>
14712
14713         PR ipa/82352
14714         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14715
14716 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
14717
14718         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14719
14720 2018-01-03  Martin Sebor  <msebor@redhat.com>
14721
14722         PR tree-optimization/83655
14723         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14724         checking calls with invalid arguments.
14725
14726 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14727
14728         * tree-vect-stmts.c (vect_get_store_rhs): New function.
14729         (vectorizable_mask_load_store): Delete.
14730         (vectorizable_call): Return false for masked loads and stores.
14731         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
14732         instead of gimple_assign_rhs1.
14733         (vectorizable_load): Handle IFN_MASK_LOAD.
14734         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14735
14736 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14737
14738         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14739         split out from..,
14740         (vectorizable_mask_load_store): ...here.
14741         (vectorizable_load): ...and here.
14742
14743 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14744
14745         * tree-vect-stmts.c (vect_build_all_ones_mask)
14746         (vect_build_zero_merge_argument): New functions, split out from...
14747         (vectorizable_load): ...here.
14748
14749 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14750
14751         * tree-vect-stmts.c (vect_check_store_rhs): New function,
14752         split out from...
14753         (vectorizable_mask_load_store): ...here.
14754         (vectorizable_store): ...and here.
14755
14756 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14757
14758         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14759         split out from...
14760         (vectorizable_mask_load_store): ...here.
14761
14762 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14763
14764         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14765         (vect_model_store_cost): Take a vec_load_store_type instead of a
14766         vect_def_type.
14767         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14768         (vect_model_store_cost): Take a vec_load_store_type instead of a
14769         vect_def_type.
14770         (vectorizable_mask_load_store): Update accordingly.
14771         (vectorizable_store): Likewise.
14772         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14773
14774 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14775
14776         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14777         IFN_MASK_LOAD calls here rather than...
14778         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14779
14780 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14781             Alan Hayward  <alan.hayward@arm.com>
14782             David Sherwood  <david.sherwood@arm.com>
14783
14784         * expmed.c (extract_bit_field_1): For vector extracts,
14785         fall back to extract_bit_field_as_subreg if vec_extract
14786         isn't available.
14787
14788 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14789             Alan Hayward  <alan.hayward@arm.com>
14790             David Sherwood  <david.sherwood@arm.com>
14791
14792         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14793         they are variable or constant sized.
14794         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14795         slots for constant-sized data.
14796
14797 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14798             Alan Hayward  <alan.hayward@arm.com>
14799             David Sherwood  <david.sherwood@arm.com>
14800
14801         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14802         handling COND_EXPRs with boolean comparisons, try to find a better
14803         basis for the mask type than the boolean itself.
14804
14805 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14806
14807         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14808         is calculated and how it can be overridden.
14809         * genmodes.c (max_bitsize_mode_any_mode): New variable.
14810         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14811         if defined.
14812         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14813         if nonzero.
14814
14815 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14816             Alan Hayward  <alan.hayward@arm.com>
14817             David Sherwood  <david.sherwood@arm.com>
14818
14819         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14820         Remove the mode argument.
14821         (aarch64_simd_valid_immediate): Remove the mode and inverse
14822         arguments.
14823         * config/aarch64/iterators.md (bitsize): New iterator.
14824         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14825         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14826         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14827         aarch64_simd_valid_immediate.
14828         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14829         (aarch64_reg_or_bic_imm): Likewise.
14830         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14831         with an insn_type enum and msl with a modifier_type enum.
14832         Replace element_width with a scalar_mode.  Change the shift
14833         to unsigned int.  Add constructors for scalar_float_mode and
14834         scalar_int_mode elements.
14835         (aarch64_vect_float_const_representable_p): Delete.
14836         (aarch64_can_const_movi_rtx_p)
14837         (aarch64_simd_scalar_immediate_valid_for_move)
14838         (aarch64_simd_make_constant): Update call to
14839         aarch64_simd_valid_immediate.
14840         (aarch64_advsimd_valid_immediate_hs): New function.
14841         (aarch64_advsimd_valid_immediate): Likewise.
14842         (aarch64_simd_valid_immediate): Remove mode and inverse
14843         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
14844         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14845         and aarch64_float_const_representable_p on the result.
14846         (aarch64_output_simd_mov_immediate): Remove mode argument.
14847         Update call to aarch64_simd_valid_immediate and use of
14848         simd_immediate_info.
14849         (aarch64_output_scalar_simd_mov_immediate): Update call
14850         accordingly.
14851
14852 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14853             Alan Hayward  <alan.hayward@arm.com>
14854             David Sherwood  <david.sherwood@arm.com>
14855
14856         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14857         (mode_nunits): Likewise CONST_MODE_NUNITS.
14858         * machmode.def (ADJUST_NUNITS): Document.
14859         * genmodes.c (mode_data::need_nunits_adj): New field.
14860         (blank_mode): Update accordingly.
14861         (adj_nunits): New variable.
14862         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14863         parameter.
14864         (emit_mode_size_inline): Set need_bytesize_adj for all modes
14865         listed in adj_nunits.
14866         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14867         listed in adj_nunits.  Don't emit case statements for such modes.
14868         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14869         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
14870         nothing if adj_nunits is nonnull.
14871         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14872         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14873         (emit_mode_fbit): Update use of print_maybe_const_decl.
14874         (emit_move_size): Likewise.  Treat the array as non-const
14875         if adj_nunits.
14876         (emit_mode_adjustments): Handle adj_nunits.
14877
14878 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14879
14880         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14881         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14882         (VECTOR_MODES): Use it.
14883         (make_vector_modes): Take the prefix as an argument.
14884
14885 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14886             Alan Hayward  <alan.hayward@arm.com>
14887             David Sherwood  <david.sherwood@arm.com>
14888
14889         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14890         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14891         for MODE_VECTOR_BOOL.
14892         * machmode.def (VECTOR_BOOL_MODE): Document.
14893         * genmodes.c (VECTOR_BOOL_MODE): New macro.
14894         (make_vector_bool_mode): New function.
14895         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14896         MODE_VECTOR_BOOL.
14897         * lto-streamer-in.c (lto_input_mode_table): Likewise.
14898         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14899         Likewise.
14900         * stor-layout.c (int_mode_for_mode): Likewise.
14901         * tree.c (build_vector_type_for_mode): Likewise.
14902         * varasm.c (output_constant_pool_2): Likewise.
14903         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14904         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
14905         for MODE_VECTOR_BOOL.
14906         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14907         of mode class checks.
14908         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14909         instead of a list of mode class checks.
14910         (expand_vector_scalar_condition): Likewise.
14911         (type_for_widest_vector_mode): Handle BImode as an inner mode.
14912
14913 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14914             Alan Hayward  <alan.hayward@arm.com>
14915             David Sherwood  <david.sherwood@arm.com>
14916
14917         * machmode.h (mode_size): Change from unsigned short to
14918         poly_uint16_pod.
14919         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14920         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14921         or if measurement_type is not polynomial.
14922         (fixed_size_mode::includes_p): Check for constant-sized modes.
14923         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14924         return a poly_uint16 rather than an unsigned short.
14925         (emit_mode_size): Change the type of mode_size from unsigned short
14926         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
14927         (emit_mode_adjustments): Cope with polynomial vector sizes.
14928         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14929         for GET_MODE_SIZE.
14930         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14931         for GET_MODE_SIZE.
14932         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14933         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14934         * caller-save.c (setup_save_areas): Likewise.
14935         (replace_reg_with_saved_mem): Likewise.
14936         * calls.c (emit_library_call_value_1): Likewise.
14937         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14938         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14939         (gen_lowpart_for_combine): Likewise.
14940         * convert.c (convert_to_integer_1): Likewise.
14941         * cse.c (equiv_constant, cse_insn): Likewise.
14942         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14943         (cselib_subst_to_values): Likewise.
14944         * dce.c (word_dce_process_block): Likewise.
14945         * df-problems.c (df_word_lr_mark_ref): Likewise.
14946         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14947         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14948         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14949         (rtl_for_decl_location): Likewise.
14950         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14951         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14952         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14953         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14954         (expand_expr_real_1): Likewise.
14955         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14956         (pad_below): Likewise.
14957         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14958         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14959         * ira.c (get_subreg_tracking_sizes): Likewise.
14960         * ira-build.c (ira_create_allocno_objects): Likewise.
14961         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14962         (ira_sort_regnos_for_alter_reg): Likewise.
14963         * ira-costs.c (record_operand_costs): Likewise.
14964         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14965         (resolve_simple_move): Likewise.
14966         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14967         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14968         (lra_constraints): Likewise.
14969         (CONST_POOL_OK_P): Reject variable-sized modes.
14970         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14971         (add_pseudo_to_slot, lra_spill): Likewise.
14972         * omp-low.c (omp_clause_aligned_alignment): Likewise.
14973         * optabs-query.c (get_best_extraction_insn): Likewise.
14974         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14975         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14976         (expand_mult_highpart, valid_multiword_target_p): Likewise.
14977         * recog.c (offsettable_address_addr_space_p): Likewise.
14978         * regcprop.c (maybe_mode_change): Likewise.
14979         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14980         * regrename.c (build_def_use): Likewise.
14981         * regstat.c (dump_reg_info): Likewise.
14982         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14983         (find_reloads, find_reloads_subreg_address): Likewise.
14984         * reload1.c (eliminate_regs_1): Likewise.
14985         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14986         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14987         (simplify_binary_operation_1, simplify_subreg): Likewise.
14988         * targhooks.c (default_function_arg_padding): Likewise.
14989         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14990         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14991         (verify_gimple_assign_ternary): Likewise.
14992         * tree-inline.c (estimate_move_cost): Likewise.
14993         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14994         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14995         (get_address_cost_ainc): Likewise.
14996         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14997         (vect_supportable_dr_alignment): Likewise.
14998         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14999         (vectorizable_reduction): Likewise.
15000         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15001         (vectorizable_operation, vectorizable_load): Likewise.
15002         * tree.c (build_same_sized_truth_vector_type): Likewise.
15003         * valtrack.c (cleanup_auto_inc_dec): Likewise.
15004         * var-tracking.c (emit_note_insn_var_location): Likewise.
15005         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15006         (ADDR_VEC_ALIGN): Likewise.
15007
15008 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15009             Alan Hayward  <alan.hayward@arm.com>
15010             David Sherwood  <david.sherwood@arm.com>
15011
15012         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15013         unsigned short.
15014         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15015         or if measurement_type is polynomial.
15016         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15017         * combine.c (make_extraction): Likewise.
15018         * dse.c (find_shift_sequence): Likewise.
15019         * dwarf2out.c (mem_loc_descriptor): Likewise.
15020         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15021         (extract_bit_field, extract_low_bits): Likewise.
15022         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15023         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15024         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15025         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15026         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15027         * reload.c (find_reloads): Likewise.
15028         * reload1.c (alter_reg): Likewise.
15029         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15030         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15031         * tree-if-conv.c (predicate_mem_writes): Likewise.
15032         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15033         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15034         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15035         * valtrack.c (dead_debug_insert_temp): Likewise.
15036         * varasm.c (mergeable_constant_section): Likewise.
15037         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15038
15039 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15040             Alan Hayward  <alan.hayward@arm.com>
15041             David Sherwood  <david.sherwood@arm.com>
15042
15043         * expr.c (expand_assignment): Cope with polynomial mode sizes
15044         when assigning to a CONCAT.
15045
15046 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15047             Alan Hayward  <alan.hayward@arm.com>
15048             David Sherwood  <david.sherwood@arm.com>
15049
15050         * machmode.h (mode_precision): Change from unsigned short to
15051         poly_uint16_pod.
15052         (mode_to_precision): Return a poly_uint16 rather than an unsigned
15053         short.
15054         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15055         or if measurement_type is not polynomial.
15056         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
15057         in which the mode is already known to be a scalar_int_mode.
15058         * genmodes.c (emit_mode_precision): Change the type of mode_precision
15059         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
15060         initializer.
15061         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15062         for GET_MODE_PRECISION.
15063         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15064         for GET_MODE_PRECISION.
15065         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15066         as polynomial.
15067         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15068         (expand_field_assignment, make_extraction): Likewise.
15069         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15070         (get_last_value): Likewise.
15071         * convert.c (convert_to_integer_1): Likewise.
15072         * cse.c (cse_insn): Likewise.
15073         * expr.c (expand_expr_real_1): Likewise.
15074         * lra-constraints.c (simplify_operand_subreg): Likewise.
15075         * optabs-query.c (can_atomic_load_p): Likewise.
15076         * optabs.c (expand_atomic_load): Likewise.
15077         (expand_atomic_store): Likewise.
15078         * ree.c (combine_reaching_defs): Likewise.
15079         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15080         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15081         * tree.h (type_has_mode_precision_p): Likewise.
15082         * ubsan.c (instrument_si_overflow): Likewise.
15083
15084 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15085             Alan Hayward  <alan.hayward@arm.com>
15086             David Sherwood  <david.sherwood@arm.com>
15087
15088         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15089         polynomial numbers of units.
15090         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15091         (valid_vector_subparts_p): New function.
15092         (build_vector_type): Remove temporary shim and take the number
15093         of units as a poly_uint64 rather than an int.
15094         (build_opaque_vector_type): Take the number of units as a
15095         poly_uint64 rather than an int.
15096         * tree.c (build_vector_from_ctor): Handle polynomial
15097         TYPE_VECTOR_SUBPARTS.
15098         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15099         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15100         (build_vector_from_val): If the number of units is variable,
15101         use build_vec_duplicate_cst for constant operands and
15102         VEC_DUPLICATE_EXPR otherwise.
15103         (make_vector_type): Remove temporary is_constant ().
15104         (build_vector_type, build_opaque_vector_type): Take the number of
15105         units as a poly_uint64 rather than an int.
15106         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15107         VECTOR_CST_NELTS.
15108         * cfgexpand.c (expand_debug_expr): Likewise.
15109         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15110         (store_constructor, expand_expr_real_1): Likewise.
15111         (const_scalar_mask_from_tree): Likewise.
15112         * fold-const-call.c (fold_const_reduction): Likewise.
15113         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15114         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15115         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15116         (fold_relational_const): Likewise.
15117         (native_interpret_vector): Likewise.  Change the size from an
15118         int to an unsigned int.
15119         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15120         TYPE_VECTOR_SUBPARTS.
15121         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15122         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15123         duplicating a non-constant operand into a variable-length vector.
15124         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15125         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15126         * ipa-icf.c (sem_variable::equals): Likewise.
15127         * match.pd: Likewise.
15128         * omp-simd-clone.c (simd_clone_subparts): Likewise.
15129         * print-tree.c (print_node): Likewise.
15130         * stor-layout.c (layout_type): Likewise.
15131         * targhooks.c (default_builtin_vectorization_cost): Likewise.
15132         * tree-cfg.c (verify_gimple_comparison): Likewise.
15133         (verify_gimple_assign_binary): Likewise.
15134         (verify_gimple_assign_ternary): Likewise.
15135         (verify_gimple_assign_single): Likewise.
15136         * tree-pretty-print.c (dump_generic_node): Likewise.
15137         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15138         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15139         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15140         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15141         (vect_shift_permute_load_chain): Likewise.
15142         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15143         (expand_vector_condition, optimize_vector_constructor): Likewise.
15144         (lower_vec_perm, get_compute_type): Likewise.
15145         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15146         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15147         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15148         (vect_recog_mask_conversion_pattern): Likewise.
15149         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15150         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15151         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15152         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15153         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15154         (vectorizable_shift, vectorizable_operation, vectorizable_store)
15155         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15156         (supportable_widening_operation): Likewise.
15157         (supportable_narrowing_operation): Likewise.
15158         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15159         Likewise.
15160         * varasm.c (output_constant): Likewise.
15161
15162 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15163             Alan Hayward  <alan.hayward@arm.com>
15164             David Sherwood  <david.sherwood@arm.com>
15165
15166         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15167         so that both the length == 3 and length != 3 cases set up their
15168         own permute vectors.  Add comments explaining why we know the
15169         number of elements is constant.
15170         (vect_permute_load_chain): Likewise.
15171
15172 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15173             Alan Hayward  <alan.hayward@arm.com>
15174             David Sherwood  <david.sherwood@arm.com>
15175
15176         * machmode.h (mode_nunits): Change from unsigned char to
15177         poly_uint16_pod.
15178         (ONLY_FIXED_SIZE_MODES): New macro.
15179         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15180         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15181         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15182         New typedefs.
15183         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15184         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15185         or if measurement_type is not polynomial.
15186         * genmodes.c (ZERO_COEFFS): New macro.
15187         (emit_mode_nunits_inline): Make mode_nunits_inline return a
15188         poly_uint16.
15189         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15190         Use ZERO_COEFFS when emitting initializers.
15191         * data-streamer.h (bp_pack_poly_value): New function.
15192         (bp_unpack_poly_value): Likewise.
15193         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15194         for GET_MODE_NUNITS.
15195         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15196         for GET_MODE_NUNITS.
15197         * tree.c (make_vector_type): Remove temporary shim and make
15198         the real function take the number of units as a poly_uint64
15199         rather than an int.
15200         (build_vector_type_for_mode): Handle polynomial nunits.
15201         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15202         * emit-rtl.c (const_vec_series_p_1): Likewise.
15203         (gen_rtx_CONST_VECTOR): Likewise.
15204         * fold-const.c (test_vec_duplicate_folding): Likewise.
15205         * genrecog.c (validate_pattern): Likewise.
15206         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15207         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15208         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15209         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15210         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15211         * rtlanal.c (subreg_get_info): Likewise.
15212         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15213         (vect_grouped_load_supported): Likewise.
15214         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15215         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15216         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15217         (simplify_const_unary_operation, simplify_binary_operation_1)
15218         (simplify_const_binary_operation, simplify_ternary_operation)
15219         (test_vector_ops_duplicate, test_vector_ops): Likewise.
15220         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15221         instead of CONST_VECTOR_NUNITS.
15222         * varasm.c (output_constant_pool_2): Likewise.
15223         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15224         explicit-encoded elements in the XVEC for variable-length vectors.
15225
15226 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15227
15228         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15229
15230 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15231             Alan Hayward  <alan.hayward@arm.com>
15232             David Sherwood  <david.sherwood@arm.com>
15233
15234         * coretypes.h (fixed_size_mode): Declare.
15235         (fixed_size_mode_pod): New typedef.
15236         * builtins.h (target_builtins::x_apply_args_mode)
15237         (target_builtins::x_apply_result_mode): Change type to
15238         fixed_size_mode_pod.
15239         * builtins.c (apply_args_size, apply_result_size, result_vector)
15240         (expand_builtin_apply_args_1, expand_builtin_apply)
15241         (expand_builtin_return): Update accordingly.
15242
15243 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15244
15245         * cse.c (hash_rtx_cb): Hash only the encoded elements.
15246         * cselib.c (cselib_hash_rtx): Likewise.
15247         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15248         CONST_VECTOR encoding.
15249
15250 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
15251             Jeff Law  <law@redhat.com>
15252
15253         PR target/83641
15254         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15255         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15256         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15257         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15258
15259         PR target/83641
15260         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15261         explicitly probe *sp in a noreturn function if there were any callee
15262         register saves or frame pointer is needed.
15263
15264 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
15265
15266         PR debug/83621
15267         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15268         BLKmode for ternary, binary or unary expressions.
15269
15270         PR debug/83645
15271         * var-tracking.c (delete_vta_debug_insn): New inline function.
15272         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15273         insns from get_insns () to NULL instead of each bb separately.
15274         Use delete_vta_debug_insn.  No longer static.
15275         (vt_debug_insns_local, variable_tracking_main_1): Adjust
15276         delete_vta_debug_insns callers.
15277         * rtl.h (delete_vta_debug_insns): Declare.
15278         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15279         instead of variable_tracking_main.
15280
15281 2018-01-03  Martin Sebor  <msebor@redhat.com>
15282
15283         PR tree-optimization/83603
15284         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15285         arguments past the endof the argument list in functions declared
15286         without a prototype.
15287         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15288         Avoid checking when arguments are null.
15289
15290 2018-01-03  Martin Sebor  <msebor@redhat.com>
15291
15292         PR c/83559
15293         * doc/extend.texi (attribute const): Fix a typo.
15294         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15295         issuing -Wsuggest-attribute for void functions.
15296
15297 2018-01-03  Martin Sebor  <msebor@redhat.com>
15298
15299         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15300         offset_int::from instead of wide_int::to_shwi.
15301         (maybe_diag_overlap): Remove assertion.
15302         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15303         * gimple-ssa-sprintf.c (format_directive): Same.
15304         (parse_directive): Same.
15305         (sprintf_dom_walker::compute_format_length): Same.
15306         (try_substitute_return_value): Same.
15307
15308 2018-01-03  Jeff Law  <law@redhat.com>
15309
15310         PR middle-end/83654
15311         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15312         non-constant residual for zero at runtime and avoid probing in
15313         that case.  Reorganize code for trailing problem to mirror handling
15314         of the residual.
15315
15316 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15317
15318         PR tree-optimization/83501
15319         * tree-ssa-strlen.c (get_string_cst): New.
15320         (handle_char_store): Call get_string_cst.
15321
15322 2018-01-03  Martin Liska  <mliska@suse.cz>
15323
15324         PR tree-optimization/83593
15325         * tree-ssa-strlen.c: Include tree-cfg.h.
15326         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15327         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15328         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15329         to false.
15330         (strlen_dom_walker::before_dom_children): Call
15331         gimple_purge_dead_eh_edges. Dump tranformation with details
15332         dump flags.
15333         (strlen_dom_walker::before_dom_children): Update call by adding
15334         new argument cleanup_eh.
15335         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15336
15337 2018-01-03  Martin Liska  <mliska@suse.cz>
15338
15339         PR ipa/83549
15340         * cif-code.def (VARIADIC_THUNK): New enum value.
15341         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15342         thunks.
15343
15344 2018-01-03  Jan Beulich  <jbeulich@suse.com>
15345
15346         * sse.md (mov<mode>_internal): Tighten condition for when to use
15347         vmovdqu<ssescalarsize> for TI and OI modes.
15348
15349 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
15350
15351         Update copyright years.
15352
15353 2018-01-03  Martin Liska  <mliska@suse.cz>
15354
15355         PR ipa/83594
15356         * ipa-visibility.c (function_and_variable_visibility): Skip
15357         functions with noipa attribure.
15358
15359 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
15360
15361         * gcc.c (process_command): Update copyright notice dates.
15362         * gcov-dump.c (print_version): Ditto.
15363         * gcov.c (print_version): Ditto.
15364         * gcov-tool.c (print_version): Ditto.
15365         * gengtype.c (create_file): Ditto.
15366         * doc/cpp.texi: Bump @copying's copyright year.
15367         * doc/cppinternals.texi: Ditto.
15368         * doc/gcc.texi: Ditto.
15369         * doc/gccint.texi: Ditto.
15370         * doc/gcov.texi: Ditto.
15371         * doc/install.texi: Ditto.
15372         * doc/invoke.texi: Ditto.
15373
15374 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15375
15376         * vector-builder.h (vector_builder::m_full_nelts): Change from
15377         unsigned int to poly_uint64.
15378         (vector_builder::full_nelts): Update prototype accordingly.
15379         (vector_builder::new_vector): Likewise.
15380         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15381         (vector_builder::operator ==): Likewise.
15382         (vector_builder::finalize): Likewise.
15383         * int-vector-builder.h (int_vector_builder::int_vector_builder):
15384         Take the number of elements as a poly_uint64 rather than an
15385         unsigned int.
15386         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15387         from unsigned int to poly_uint64.
15388         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15389         (vec_perm_indices::new_vector): Likewise.
15390         (vec_perm_indices::length): Likewise.
15391         (vec_perm_indices::nelts_per_input): Likewise.
15392         (vec_perm_indices::input_nelts): Likewise.
15393         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15394         number of elements per input as a poly_uint64 rather than an
15395         unsigned int.  Use the original encoding for variable-length
15396         vectors, rather than clamping each individual element.
15397         For the second and subsequent elements in each pattern,
15398         clamp the step and base before clamping their sum.
15399         (vec_perm_indices::series_p): Handle polynomial element counts.
15400         (vec_perm_indices::all_in_range_p): Likewise.
15401         (vec_perm_indices_to_tree): Likewise.
15402         (vec_perm_indices_to_rtx): Likewise.
15403         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15404         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15405         (tree_vector_builder::new_binary_operation): Handle polynomial
15406         element counts.  Return false if we need to know the number
15407         of elements at compile time.
15408         * fold-const.c (fold_vec_perm): Punt if the number of elements
15409         isn't known at compile time.
15410
15411 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15412
15413         * vec-perm-indices.h (vec_perm_builder): Change element type
15414         from HOST_WIDE_INT to poly_int64.
15415         (vec_perm_indices::element_type): Update accordingly.
15416         (vec_perm_indices::clamp): Handle polynomial element_types.
15417         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15418         (vec_perm_indices::all_in_range_p): Likewise.
15419         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15420         than shwi trees.
15421         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15422         polynomial vec_perm_indices element types.
15423         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15424         * fold-const.c (fold_vec_perm): Likewise.
15425         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15426         * tree-vect-generic.c (lower_vec_perm): Likewise.
15427         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15428         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15429         element type to HOST_WIDE_INT.
15430
15431 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15432             Alan Hayward  <alan.hayward@arm.com>
15433             David Sherwood  <david.sherwood@arm.com>
15434
15435         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15436         rather than an int.  Use plus_constant.
15437         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15438         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15439
15440 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15441             Alan Hayward  <alan.hayward@arm.com>
15442             David Sherwood  <david.sherwood@arm.com>
15443
15444         * calls.c (emit_call_1, expand_call): Change struct_value_size from
15445         a HOST_WIDE_INT to a poly_int64.
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         * calls.c (load_register_parameters): Cope with polynomial
15452         mode sizes.  Require a constant size for BLKmode parameters
15453         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
15454         forces a parameter to be padded at the lsb end in order to
15455         fill a complete number of words, require the parameter size
15456         to be ordered wrt UNITS_PER_WORD.
15457
15458 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15459             Alan Hayward  <alan.hayward@arm.com>
15460             David Sherwood  <david.sherwood@arm.com>
15461
15462         * reload1.c (spill_stack_slot_width): Change element type
15463         from unsigned int to poly_uint64_pod.
15464         (alter_reg): Treat mode sizes as polynomial.
15465
15466 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15467             Alan Hayward  <alan.hayward@arm.com>
15468             David Sherwood  <david.sherwood@arm.com>
15469
15470         * reload.c (complex_word_subreg_p): New function.
15471         (reload_inner_reg_of_subreg, push_reload): Use it.
15472
15473 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15474             Alan Hayward  <alan.hayward@arm.com>
15475             David Sherwood  <david.sherwood@arm.com>
15476
15477         * lra-constraints.c (process_alt_operands): Reject matched
15478         operands whose sizes aren't ordered.
15479         (match_reload): Refer to this check here.
15480
15481 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15482             Alan Hayward  <alan.hayward@arm.com>
15483             David Sherwood  <david.sherwood@arm.com>
15484
15485         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15486         that the mode size is in the set {1, 2, 4, 8, 16}.
15487
15488 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15489             Alan Hayward  <alan.hayward@arm.com>
15490             David Sherwood  <david.sherwood@arm.com>
15491
15492         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15493         Use plus_constant instead of gen_rtx_PLUS.
15494
15495 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15496             Alan Hayward  <alan.hayward@arm.com>
15497             David Sherwood  <david.sherwood@arm.com>
15498
15499         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15500         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15501         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15502         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15503         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15504         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15505         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15506         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15507         * config/i386/i386.c (ix86_push_rounding): ...this new function.
15508         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15509         a poly_int64.
15510         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15511         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15512         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15513         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15514         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15515         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15516         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15517         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15518         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15519         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15520         function.
15521         * expr.c (emit_move_resolve_push): Treat the input and result
15522         of PUSH_ROUNDING as a poly_int64.
15523         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15524         (emit_push_insn): Likewise.
15525         * lra-eliminations.c (mark_not_eliminable): Likewise.
15526         * recog.c (push_operand): Likewise.
15527         * reload1.c (elimination_effects): Likewise.
15528         * rtlanal.c (nonzero_bits1): Likewise.
15529         * calls.c (store_one_arg): Likewise.  Require the padding to be
15530         known at compile time.
15531
15532 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15533             Alan Hayward  <alan.hayward@arm.com>
15534             David Sherwood  <david.sherwood@arm.com>
15535
15536         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15537         Use plus_constant instead of gen_rtx_PLUS.
15538
15539 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15540             Alan Hayward  <alan.hayward@arm.com>
15541             David Sherwood  <david.sherwood@arm.com>
15542
15543         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15544         rather than an int.
15545
15546 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15547             Alan Hayward  <alan.hayward@arm.com>
15548             David Sherwood  <david.sherwood@arm.com>
15549
15550         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15551         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15552         via stack temporaries.  Treat the mode size as polynomial too.
15553
15554 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15555             Alan Hayward  <alan.hayward@arm.com>
15556             David Sherwood  <david.sherwood@arm.com>
15557
15558         * expr.c (expand_expr_real_2): When handling conversions involving
15559         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15560         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
15561         as a poly_uint64 too.
15562
15563 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15564             Alan Hayward  <alan.hayward@arm.com>
15565             David Sherwood  <david.sherwood@arm.com>
15566
15567         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15568
15569 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15570             Alan Hayward  <alan.hayward@arm.com>
15571             David Sherwood  <david.sherwood@arm.com>
15572
15573         * combine.c (can_change_dest_mode): Handle polynomial
15574         REGMODE_NATURAL_SIZE.
15575         * expmed.c (store_bit_field_1): Likewise.
15576         * expr.c (store_constructor): Likewise.
15577         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15578         and polynomial REGMODE_NATURAL_SIZE.
15579         (gen_lowpart_common): Likewise.
15580         * reginfo.c (record_subregs_of_mode): Likewise.
15581         * rtlanal.c (read_modify_subreg_p): Likewise.
15582
15583 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15584             Alan Hayward  <alan.hayward@arm.com>
15585             David Sherwood  <david.sherwood@arm.com>
15586
15587         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15588         numbers of elements.
15589
15590 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15591             Alan Hayward  <alan.hayward@arm.com>
15592             David Sherwood  <david.sherwood@arm.com>
15593
15594         * match.pd: Cope with polynomial numbers of vector elements.
15595
15596 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15597             Alan Hayward  <alan.hayward@arm.com>
15598             David Sherwood  <david.sherwood@arm.com>
15599
15600         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15601         in a POINTER_PLUS_EXPR.
15602
15603 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15604             Alan Hayward  <alan.hayward@arm.com>
15605             David Sherwood  <david.sherwood@arm.com>
15606
15607         * omp-simd-clone.c (simd_clone_subparts): New function.
15608         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15609         (ipa_simd_modify_function_body): Likewise.
15610
15611 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15612             Alan Hayward  <alan.hayward@arm.com>
15613             David Sherwood  <david.sherwood@arm.com>
15614
15615         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15616         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15617         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15618         (expand_vector_condition, vector_element): Likewise.
15619         (subparts_gt): New function.
15620         (get_compute_type): Use subparts_gt.
15621         (count_type_subparts): Delete.
15622         (expand_vector_operations_1): Use subparts_gt instead of
15623         count_type_subparts.
15624
15625 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15626             Alan Hayward  <alan.hayward@arm.com>
15627             David Sherwood  <david.sherwood@arm.com>
15628
15629         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15630         (vect_compile_time_alias): ...this new function.  Do the calculation
15631         on poly_ints rather than trees.
15632         (vect_prune_runtime_alias_test_list): Update call accordingly.
15633
15634 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15635             Alan Hayward  <alan.hayward@arm.com>
15636             David Sherwood  <david.sherwood@arm.com>
15637
15638         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15639         numbers of units.
15640         (vect_schedule_slp_instance): Likewise.
15641
15642 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15643             Alan Hayward  <alan.hayward@arm.com>
15644             David Sherwood  <david.sherwood@arm.com>
15645
15646         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15647         constant and extern definitions for variable-length vectors.
15648         (vect_get_constant_vectors): Note that the number of units
15649         is known to be constant.
15650
15651 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15652             Alan Hayward  <alan.hayward@arm.com>
15653             David Sherwood  <david.sherwood@arm.com>
15654
15655         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15656         of units as polynomial.  Choose between WIDE and NARROW based
15657         on multiple_p.
15658
15659 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15660             Alan Hayward  <alan.hayward@arm.com>
15661             David Sherwood  <david.sherwood@arm.com>
15662
15663         * tree-vect-stmts.c (simd_clone_subparts): New function.
15664         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15665
15666 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15667             Alan Hayward  <alan.hayward@arm.com>
15668             David Sherwood  <david.sherwood@arm.com>
15669
15670         * tree-vect-stmts.c (vectorizable_call): Treat the number of
15671         vectors as polynomial.  Use build_index_vector for
15672         IFN_GOMP_SIMD_LANE.
15673
15674 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15675             Alan Hayward  <alan.hayward@arm.com>
15676             David Sherwood  <david.sherwood@arm.com>
15677
15678         * tree-vect-stmts.c (get_load_store_type): Treat the number of
15679         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15680         for variable-length vectors.
15681         (vectorizable_mask_load_store): Treat the number of units as
15682         polynomial, asserting that it is constant if the condition has
15683         already been enforced.
15684         (vectorizable_store, vectorizable_load): Likewise.
15685
15686 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15687             Alan Hayward  <alan.hayward@arm.com>
15688             David Sherwood  <david.sherwood@arm.com>
15689
15690         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15691         of units as polynomial.  Punt if we can't tell at compile time
15692         which vector contains the final result.
15693
15694 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15695             Alan Hayward  <alan.hayward@arm.com>
15696             David Sherwood  <david.sherwood@arm.com>
15697
15698         * tree-vect-loop.c (vectorizable_induction): Treat the number
15699         of units as polynomial.  Punt on SLP inductions.  Use an integer
15700         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
15701         cast of such a series for variable-length floating-point
15702         reductions.
15703
15704 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15705             Alan Hayward  <alan.hayward@arm.com>
15706             David Sherwood  <david.sherwood@arm.com>
15707
15708         * tree.h (build_index_vector): Declare.
15709         * tree.c (build_index_vector): New function.
15710         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15711         of units as polynomial, forcibly converting it to a constant if
15712         vectorizable_reduction has already enforced the condition.
15713         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
15714         to create a {1,2,3,...} vector.
15715         (vectorizable_reduction): Treat the number of units as polynomial.
15716         Choose vectype_in based on the largest scalar element size rather
15717         than the smallest number of units.  Enforce the restrictions
15718         relied on above.
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-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15725         number of units as polynomial.
15726
15727 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15728             Alan Hayward  <alan.hayward@arm.com>
15729             David Sherwood  <david.sherwood@arm.com>
15730
15731         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15732         * target.def (autovectorize_vector_sizes): Return the vector sizes
15733         by pointer, using vector_sizes rather than a bitmask.
15734         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15735         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15736         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15737         Likewise.
15738         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15739         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15740         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15741         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15742         * omp-general.c (omp_max_vf): Likewise.
15743         * omp-low.c (omp_clause_aligned_alignment): Likewise.
15744         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15745         * tree-vect-loop.c (vect_analyze_loop): Likewise.
15746         * tree-vect-slp.c (vect_slp_bb): Likewise.
15747         * doc/tm.texi: Regenerate.
15748         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15749         to a poly_uint64.
15750         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15751         the vector size as a poly_uint64 rather than an unsigned int.
15752         (current_vector_size): Change from an unsigned int to a poly_uint64.
15753         (get_vectype_for_scalar_type): Update accordingly.
15754         * tree.h (build_truth_vector_type): Take the size and number of
15755         units as a poly_uint64 rather than an unsigned int.
15756         (build_vector_type): Add a temporary overload that takes
15757         the number of units as a poly_uint64 rather than an unsigned int.
15758         * tree.c (make_vector_type): Likewise.
15759         (build_truth_vector_type): Take the number of units as a poly_uint64
15760         rather than an unsigned int.
15761
15762 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15763             Alan Hayward  <alan.hayward@arm.com>
15764             David Sherwood  <david.sherwood@arm.com>
15765
15766         * target.def (get_mask_mode): Take the number of units and length
15767         as poly_uint64s rather than unsigned ints.
15768         * targhooks.h (default_get_mask_mode): Update accordingly.
15769         * targhooks.c (default_get_mask_mode): Likewise.
15770         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15771         * doc/tm.texi: Regenerate.
15772
15773 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15774             Alan Hayward  <alan.hayward@arm.com>
15775             David Sherwood  <david.sherwood@arm.com>
15776
15777         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15778         * omp-general.c (omp_max_vf): Likewise.
15779         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15780         (expand_omp_simd): Handle polynomial safelen.
15781         * omp-low.c (omplow_simd_context): Add a default constructor.
15782         (omplow_simd_context::max_vf): Change from int to poly_uint64.
15783         (lower_rec_simd_input_clauses): Update accordingly.
15784         (lower_rec_input_clauses): Likewise.
15785
15786 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15787             Alan Hayward  <alan.hayward@arm.com>
15788             David Sherwood  <david.sherwood@arm.com>
15789
15790         * tree-vectorizer.h (vect_nunits_for_cost): New function.
15791         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15792         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15793         (vect_analyze_slp_cost): Likewise.
15794         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15795         (vect_model_load_cost): Likewise.
15796
15797 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15798             Alan Hayward  <alan.hayward@arm.com>
15799             David Sherwood  <david.sherwood@arm.com>
15800
15801         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15802         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15803         from an unsigned int * to a poly_uint64_pod *.
15804         (calculate_unrolling_factor): New function.
15805         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
15806
15807 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15808             Alan Hayward  <alan.hayward@arm.com>
15809             David Sherwood  <david.sherwood@arm.com>
15810
15811         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15812         from an unsigned int to a poly_uint64.
15813         (_loop_vec_info::slp_unrolling_factor): Likewise.
15814         (_loop_vec_info::vectorization_factor): Change from an int
15815         to a poly_uint64.
15816         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15817         (vect_get_num_vectors): New function.
15818         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15819         (vect_get_num_copies): Use vect_get_num_vectors.
15820         (vect_analyze_data_ref_dependences): Change max_vf from an int *
15821         to an unsigned int *.
15822         (vect_analyze_data_refs): Change min_vf from an int * to a
15823         poly_uint64 *.
15824         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15825         than an unsigned HOST_WIDE_INT.
15826         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15827         (vect_analyze_data_ref_dependence): Change max_vf from an int *
15828         to an unsigned int *.
15829         (vect_analyze_data_ref_dependences): Likewise.
15830         (vect_compute_data_ref_alignment): Handle polynomial vf.
15831         (vect_enhance_data_refs_alignment): Likewise.
15832         (vect_prune_runtime_alias_test_list): Likewise.
15833         (vect_shift_permute_load_chain): Likewise.
15834         (vect_supportable_dr_alignment): Likewise.
15835         (dependence_distance_ge_vf): Take the vectorization factor as a
15836         poly_uint64 rather than an unsigned HOST_WIDE_INT.
15837         (vect_analyze_data_refs): Change min_vf from an int * to a
15838         poly_uint64 *.
15839         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15840         vfm1 as a poly_uint64 rather than an int.  Make the same change
15841         for the returned bound_scalar.
15842         (vect_gen_vector_loop_niters): Handle polynomial vf.
15843         (vect_do_peeling): Likewise.  Update call to
15844         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15845         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15846         be constant.
15847         * tree-vect-loop.c (vect_determine_vectorization_factor)
15848         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15849         (vect_get_known_peeling_cost): Likewise.
15850         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15851         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15852         (vect_transform_loop): Likewise.  Use the lowest possible VF when
15853         updating the upper bounds of the loop.
15854         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
15855         rather than an int.
15856         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15857         polynomial unroll factors.
15858         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15859         (vect_make_slp_decision): Likewise.
15860         (vect_supported_load_permutation_p): Likewise, and polynomial
15861         vf too.
15862         (vect_analyze_slp_cost): Handle polynomial vf.
15863         (vect_slp_analyze_node_operations): Likewise.
15864         (vect_slp_analyze_bb_1): Likewise.
15865         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15866         than an unsigned HOST_WIDE_INT.
15867         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15868         (vectorizable_load): Handle polynomial vf.
15869         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15870         a poly_uint64.
15871         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15872
15873 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15874             Alan Hayward  <alan.hayward@arm.com>
15875             David Sherwood  <david.sherwood@arm.com>
15876
15877         * match.pd: Handle bit operations involving three constants
15878         and try to fold one pair.
15879
15880 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15881
15882         * tree-vect-loop-manip.c: Include gimple-fold.h.
15883         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15884         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
15885         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15886         Add a path that uses a step of VF instead of 1, but disable it
15887         for now.
15888         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15889         and niters_no_overflow parameters.  Update calls to
15890         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15891         Create a new SSA name if the latter choses to use a ste other
15892         than zero, and return it via niters_vector_mult_vf_var.
15893         * tree-vect-loop.c (vect_transform_loop): Update calls to
15894         vect_do_peeling, vect_gen_vector_loop_niters and
15895         slpeel_make_loop_iterate_ntimes.
15896         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15897         (vect_gen_vector_loop_niters): Update declarations after above changes.
15898
15899 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
15900
15901         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15902         128-bit round to integer instructions.
15903         (ceil<mode>2): Likewise.
15904         (btrunc<mode>2): Likewise.
15905         (round<mode>2): Likewise.
15906
15907 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
15908
15909         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15910         unaligned VSX load/store on P8/P9.
15911         (expand_block_clear): Allow the use of unaligned VSX
15912         load/store on P8/P9.
15913
15914 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
15915
15916         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15917         New function.
15918         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15919         swap associated with both a load and a store.
15920
15921 2018-01-02  Andrew Waterman  <andrew@sifive.com>
15922
15923         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15924         * config/riscv/riscv.md (clear_cache): Use it.
15925
15926 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
15927
15928         * web.c: Remove out-of-date comment.
15929
15930 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15931
15932         * expr.c (fixup_args_size_notes): Check that any existing
15933         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15934         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15935         (emit_single_push_insn): ...here.
15936
15937 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15938
15939         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15940         (const_vector_encoded_nelts): New function.
15941         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15942         (const_vector_int_elt, const_vector_elt): Declare.
15943         * emit-rtl.c (const_vector_int_elt_1): New function.
15944         (const_vector_elt): Likewise.
15945         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15946         of CONST_VECTOR_ELT.
15947
15948 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15949
15950         * expr.c: Include rtx-vector-builder.h.
15951         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15952         directly on the tree encoding.
15953         (const_vector_from_tree): Likewise.
15954         * optabs.c: Include rtx-vector-builder.h.
15955         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15956         sequence of "u" values.
15957         * vec-perm-indices.c: Include rtx-vector-builder.h.
15958         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15959         directly on the vec_perm_indices encoding.
15960
15961 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15962
15963         * doc/rtl.texi (const_vector): Describe new encoding scheme.
15964         * Makefile.in (OBJS): Add rtx-vector-builder.o.
15965         * rtx-vector-builder.h: New file.
15966         * rtx-vector-builder.c: Likewise.
15967         * rtl.h (rtx_def::u2): Add a const_vector field.
15968         (CONST_VECTOR_NPATTERNS): New macro.
15969         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15970         (CONST_VECTOR_DUPLICATE_P): Likewise.
15971         (CONST_VECTOR_STEPPED_P): Likewise.
15972         (CONST_VECTOR_ENCODED_ELT): Likewise.
15973         (const_vec_duplicate_p): Check for a duplicated vector encoding.
15974         (unwrap_const_vec_duplicate): Likewise.
15975         (const_vec_series_p): Check for a non-duplicated vector encoding.
15976         Say that the function only returns true for integer vectors.
15977         * emit-rtl.c: Include rtx-vector-builder.h.
15978         (gen_const_vec_duplicate_1): Delete.
15979         (gen_const_vector): Call gen_const_vec_duplicate instead of
15980         gen_const_vec_duplicate_1.
15981         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15982         (gen_const_vec_duplicate): Use rtx_vector_builder.
15983         (gen_const_vec_series): Likewise.
15984         (gen_rtx_CONST_VECTOR): Likewise.
15985         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15986         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15987         Build a new vector rather than modifying a CONST_VECTOR in-place.
15988         (handle_special_swappables): Update call accordingly.
15989         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15990         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15991         Build a new vector rather than modifying a CONST_VECTOR in-place.
15992         (handle_special_swappables): Update call accordingly.
15993
15994 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15995
15996         * simplify-rtx.c (simplify_const_binary_operation): Use
15997         CONST_VECTOR_ELT instead of XVECEXP.
15998
15999 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16000
16001         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16002         the selector elements to be different from the data elements
16003         if the selector is a VECTOR_CST.
16004         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16005         ssizetype for the selector.
16006
16007 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16008
16009         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16010         before testing each element individually.
16011         * tree-vect-generic.c (lower_vec_perm): Likewise.
16012
16013 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16014
16015         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16016         * selftest-run-tests.c (selftest::run_tests): Call it.
16017         * vector-builder.h (vector_builder::operator ==): New function.
16018         (vector_builder::operator !=): Likewise.
16019         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16020         (vec_perm_indices::all_from_input_p): New function.
16021         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16022         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16023         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16024         instead of reading the VECTOR_CST directly.  Detect whether both
16025         vector inputs are the same before constructing the vec_perm_indices,
16026         and update the number of inputs argument accordingly.  Use the
16027         utility functions added above.  Only construct sel2 if we need to.
16028
16029 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16030
16031         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16032         the broadcast of the low byte.
16033         (expand_mult_highpart): Use an explicit encoding for the permutes.
16034         * optabs-query.c (can_mult_highpart_p): Likewise.
16035         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16036         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16037         (vectorizable_bswap): Likewise.
16038         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16039         explicit encoding for the power-of-2 permutes.
16040         (vect_permute_store_chain): Likewise.
16041         (vect_grouped_load_supported): Likewise.
16042         (vect_permute_load_chain): Likewise.
16043
16044 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16045
16046         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16047         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16048         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16049         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16050         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16051         (vect_gen_perm_mask_any): Likewise.
16052
16053 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16054
16055         * int-vector-builder.h: New file.
16056         * vec-perm-indices.h: Include int-vector-builder.h.
16057         (vec_perm_indices): Redefine as an int_vector_builder.
16058         (auto_vec_perm_indices): Delete.
16059         (vec_perm_builder): Redefine as a stand-alone class.
16060         (vec_perm_indices::vec_perm_indices): New function.
16061         (vec_perm_indices::clamp): Likewise.
16062         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16063         (vec_perm_indices::new_vector): New function.
16064         (vec_perm_indices::new_expanded_vector): Update for new
16065         vec_perm_indices class.
16066         (vec_perm_indices::rotate_inputs): New function.
16067         (vec_perm_indices::all_in_range_p): Operate directly on the
16068         encoded form, without computing elided elements.
16069         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16070         encoding.  Update for new vec_perm_indices class.
16071         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16072         the given vec_perm_builder.
16073         (expand_vec_perm_var): Update vec_perm_builder constructor.
16074         (expand_mult_highpart): Use vec_perm_builder instead of
16075         auto_vec_perm_indices.
16076         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16077         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
16078         or double series encoding as appropriate.
16079         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16080         vec_perm_indices instead of auto_vec_perm_indices.
16081         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16082         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16083         (vect_permute_store_chain): Likewise.
16084         (vect_grouped_load_supported): Likewise.
16085         (vect_permute_load_chain): Likewise.
16086         (vect_shift_permute_load_chain): Likewise.
16087         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16088         (vect_transform_slp_perm_load): Likewise.
16089         (vect_schedule_slp_instance): Likewise.
16090         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16091         (vectorizable_mask_load_store): Likewise.
16092         (vectorizable_bswap): Likewise.
16093         (vectorizable_store): Likewise.
16094         (vectorizable_load): Likewise.
16095         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16096         vec_perm_indices instead of auto_vec_perm_indices.  Use
16097         tree_to_vec_perm_builder to read the vector from a tree.
16098         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16099         vec_perm_builder instead of a vec_perm_indices.
16100         (have_whole_vector_shift): Use vec_perm_builder and
16101         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
16102         truncation to calc_vec_perm_mask_for_shift.
16103         (vect_create_epilog_for_reduction): Likewise.
16104         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16105         from auto_vec_perm_indices to vec_perm_indices.
16106         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16107         instead of changing individual elements.
16108         (aarch64_vectorize_vec_perm_const): Use new_vector to install
16109         the vector in d.perm.
16110         * config/arm/arm.c (expand_vec_perm_d::perm): Change
16111         from auto_vec_perm_indices to vec_perm_indices.
16112         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16113         instead of changing individual elements.
16114         (arm_vectorize_vec_perm_const): Use new_vector to install
16115         the vector in d.perm.
16116         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16117         Update vec_perm_builder constructor.
16118         (rs6000_expand_interleave): Likewise.
16119         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16120         (rs6000_expand_interleave): Likewise.
16121
16122 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16123
16124         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16125         to qimode could truncate the indices.
16126         * optabs.c (expand_vec_perm_var): Likewise.
16127
16128 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16129
16130         * Makefile.in (OBJS): Add vec-perm-indices.o.
16131         * vec-perm-indices.h: New file.
16132         * vec-perm-indices.c: Likewise.
16133         * target.h (vec_perm_indices): Replace with a forward class
16134         declaration.
16135         (auto_vec_perm_indices): Move to vec-perm-indices.h.
16136         * optabs.h: Include vec-perm-indices.h.
16137         (expand_vec_perm): Delete.
16138         (selector_fits_mode_p, expand_vec_perm_var): Declare.
16139         (expand_vec_perm_const): Declare.
16140         * target.def (vec_perm_const_ok): Replace with...
16141         (vec_perm_const): ...this new hook.
16142         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16143         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16144         * doc/tm.texi: Regenerate.
16145         * optabs.def (vec_perm_const): Delete.
16146         * doc/md.texi (vec_perm_const): Likewise.
16147         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16148         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16149         expand_vec_perm for constant permutation vectors.  Assert that
16150         the mode of variable permutation vectors is the integer equivalent
16151         of the mode that is being permuted.
16152         * optabs-query.h (selector_fits_mode_p): Declare.
16153         * optabs-query.c: Include vec-perm-indices.h.
16154         (selector_fits_mode_p): New function.
16155         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16156         is defined, instead of checking whether the vec_perm_const_optab
16157         exists.  Use targetm.vectorize.vec_perm_const instead of
16158         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
16159         fit in the vector mode before using a variable permute.
16160         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16161         vec_perm_indices instead of an rtx.
16162         (expand_vec_perm): Replace with...
16163         (expand_vec_perm_const): ...this new function.  Take the selector
16164         as a vec_perm_indices rather than an rtx.  Also take the mode of
16165         the selector.  Update call to shift_amt_for_vec_perm_mask.
16166         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16167         Use vec_perm_indices::new_expanded_vector to expand the original
16168         selector into bytes.  Check whether the indices fit in the vector
16169         mode before using a variable permute.
16170         (expand_vec_perm_var): Make global.
16171         (expand_mult_highpart): Use expand_vec_perm_const.
16172         * fold-const.c: Includes vec-perm-indices.h.
16173         * tree-ssa-forwprop.c: Likewise.
16174         * tree-vect-data-refs.c: Likewise.
16175         * tree-vect-generic.c: Likewise.
16176         * tree-vect-loop.c: Likewise.
16177         * tree-vect-slp.c: Likewise.
16178         * tree-vect-stmts.c: Likewise.
16179         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16180         Delete.
16181         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16182         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16183         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16184         (aarch64_vectorize_vec_perm_const): ...this new function.
16185         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16186         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16187         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16188         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16189         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16190         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16191         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16192         into...
16193         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
16194         check for NEON modes.
16195         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16196         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16197         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16198         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16199         into...
16200         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
16201         the old VEC_PERM_CONST conditions.
16202         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16203         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16204         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16205         (ia64_vectorize_vec_perm_const_ok): Merge into...
16206         (ia64_vectorize_vec_perm_const): ...this new function.
16207         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16208         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16209         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16210         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16211         * config/mips/mips.c (mips_expand_vec_perm_const)
16212         (mips_vectorize_vec_perm_const_ok): Merge into...
16213         (mips_vectorize_vec_perm_const): ...this new function.
16214         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16215         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16216         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16217         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16218         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16219         (rs6000_expand_vec_perm_const): Delete.
16220         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16221         Delete.
16222         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16223         (altivec_expand_vec_perm_const_le): Take each operand individually.
16224         Operate on constant selectors rather than rtxes.
16225         (altivec_expand_vec_perm_const): Likewise.  Update call to
16226         altivec_expand_vec_perm_const_le.
16227         (rs6000_expand_vec_perm_const): Delete.
16228         (rs6000_vectorize_vec_perm_const_ok): Delete.
16229         (rs6000_vectorize_vec_perm_const): New function.
16230         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16231         an element count and rtx array.
16232         (rs6000_expand_extract_even): Update call accordingly.
16233         (rs6000_expand_interleave): Likewise.
16234         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16235         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16236         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16237         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16238         (rs6000_expand_vec_perm_const): Delete.
16239         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16240         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16241         (altivec_expand_vec_perm_const_le): Take each operand individually.
16242         Operate on constant selectors rather than rtxes.
16243         (altivec_expand_vec_perm_const): Likewise.  Update call to
16244         altivec_expand_vec_perm_const_le.
16245         (rs6000_expand_vec_perm_const): Delete.
16246         (rs6000_vectorize_vec_perm_const_ok): Delete.
16247         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
16248         reference to the SPE evmerge intructions.
16249         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16250         an element count and rtx array.
16251         (rs6000_expand_extract_even): Update call accordingly.
16252         (rs6000_expand_interleave): Likewise.
16253         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16254         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16255         new function.
16256         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16257
16258 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16259
16260         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16261         vector mode and that that mode matches the mode of the data
16262         being permuted.
16263         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16264         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
16265         directly using expand_vec_perm_1 when forcing selectors into
16266         registers.
16267         (expand_vec_perm_var): New function, split out from expand_vec_perm.
16268
16269 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16270
16271         * optabs-query.h (can_vec_perm_p): Delete.
16272         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16273         * optabs-query.c (can_vec_perm_p): Split into...
16274         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16275         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16276         particular selector is valid.
16277         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16278         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16279         (vect_grouped_load_supported): Likewise.
16280         (vect_shift_permute_load_chain): Likewise.
16281         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16282         (vect_transform_slp_perm_load): Likewise.
16283         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16284         (vectorizable_bswap): Likewise.
16285         (vect_gen_perm_mask_checked): Likewise.
16286         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
16287         implementations of variable permutation vectors into account
16288         when deciding which selector to use.
16289         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16290         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16291         with a false third argument.
16292         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16293         to test whether the constant selector is valid and can_vec_perm_var_p
16294         to test whether a variable selector is valid.
16295
16296 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16297
16298         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16299         * optabs-query.c (can_vec_perm_p): Likewise.
16300         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16301         instead of vec_perm_indices.
16302         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16303         (vect_gen_perm_mask_checked): Likewise,
16304         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16305         (vect_gen_perm_mask_checked): Likewise,
16306
16307 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
16308
16309         * optabs-query.h (qimode_for_vec_perm): Declare.
16310         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16311         (qimode_for_vec_perm): ...this new function.
16312         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16313
16314 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
16315
16316         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16317         does not have a conditional at the top.
16318
16319 2018-01-02  Richard Biener  <rguenther@suse.de>
16320
16321         * ipa-inline.c (big_speedup_p): Fix expression.
16322
16323 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
16324
16325         PR target/81616
16326         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16327         for generic 4->6.
16328
16329 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
16330
16331         PR target/81616
16332         Generic tuning.
16333         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16334         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16335         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16336         cond_taken_branch_cost 3->4.
16337
16338 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
16339
16340         PR tree-optimization/83581
16341         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16342         TODO_cleanup_cfg if any changes have been made.
16343
16344         PR middle-end/83608
16345         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16346         convert_modes if target mode has the right side, but different mode
16347         class.
16348
16349         PR middle-end/83609
16350         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16351         last argument when extracting from CONCAT.  If either from_real or
16352         from_imag is NULL, use expansion through memory.  If result is not
16353         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16354         the parts directly to inner mode, if even that fails, use expansion
16355         through memory.
16356
16357         PR middle-end/83623
16358         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16359         check for bswap in mode rather than HImode and use that in expand_unop
16360         too.
16361 \f
16362 Copyright (C) 2018 Free Software Foundation, Inc.
16363
16364 Copying and distribution of this file, with or without modification,
16365 are permitted in any medium without royalty provided the copyright
16366 notice and this notice are preserved.