gcc/ChangeLog

   1 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
   2
   3         * internal-fn.h (lookup_internal_fn): Declare
   4         * internal-fn.c (lookup_internal_fn): New function.
   5         * gimple.c (gimple_build_call_from_tree): Handle calls to
   6         internal functions.
   7         * gimple-pretty-print.c (dump_gimple_call): Print "." before
   8         internal function names.
   9         * tree-pretty-print.c (dump_generic_node): Likewise.
  10         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
  11
  12 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
  13
  14         * gimple-fold.h (gimple_build): Make the function forms take
  15         combined_fn rather than built_in_function.
  16         (gimple_simplify): Likewise.
  17         * gimple-match-head.c (gimple_simplify): Likewise.
  18         * gimple-fold.c (gimple_build): Likewise.
  19         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
  20         rather than gimple_build_call_internal.
  21         (get_initial_defs_for_reduction): Likewise.
  22         (vect_create_epilog_for_reduction): Likewise.
  23         (vectorizable_live_operation): Likewise.
  24
  25 2018-05-17  Martin Liska  <mliska@suse.cz>
  26
  27         * gimple-ssa-sprintf.c (format_directive): Do not use
  28         space in between 'G_' and '('.
  29
  30 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
  31
  32         PR target/85323
  33         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
  34         even if the mask is not all ones.
  35
  36         PR target/85323
  37         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
  38         vector.
  39         (ix86_gimple_fold_builtin): Likewise.
  40
  41         PR target/85323
  42         * config/i386/i386.c: Include tree-vector-builder.h.
  43         (ix86_vector_shift_count): New function.
  44         (ix86_fold_builtin): Fold shift builtins by scalar count.
  45         (ix86_gimple_fold_builtin): Likewise.
  46
  47         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
  48         _mm512_setzero): New intrinsics.
  49
  50 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
  51             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
  52
  53         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
  54         code generation for cases where splatting a value is not useful.
  55         * simplify-rtx.c (simplify_ternary_operation): Simplify
  56         vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
  57         mode to a vec_concat.
  58
  59 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
  60
  61         * config.gcc: Support "goldmont-plus".
  62         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
  63         "goldmont-plus".
  64         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
  65         PROCESSOR_GOLDMONT_PLUS.
  66         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
  67         (processor_target_table): Add "goldmont-plus".
  68         (PTA_GOLDMONT_PLUS): Define.
  69         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
  70         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
  71         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
  72         (fold_builtin_cpu): Add "goldmont-plus".
  73         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
  74         (ix86_option_override_internal): Add "goldmont-plus".
  75         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
  76         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
  77         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
  78         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
  79
  80 2018-05-17  Richard Biener  <rguenther@suse.de>
  81
  82         PR tree-optimization/85757
  83         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
  84         remove defs that only feed that PHI from further processing.
  85
  86 2018-05-16  Jim Wilson  <jimw@sifive.com>
  87
  88         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
  89         asterisk to name.
  90         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
  91
  92 2018-05-16  Mark Wielaard  <mark@klomp.org>
  93
  94         * dwarf2out.c (count_index_strings): New function.
  95         (output_indirect_strings): Call count_index_strings and generate
  96         header for dwarf_version >= 5.
  97
  98 2018-05-16  Mark Wielaard  <mark@klomp.org>
  99
 100         * dwarf2out.c (dwarf_FORM): New function.
 101         (set_indirect_string): Use dwarf_FORM.
 102         (reset_indirect_string): Likewise.
 103         (size_of_die): Likewise.
 104         (value_format): Likewise.
 105         (output_die): Likewise.
 106         (add_skeleton_AT_string): Likewise.
 107         (output_macinfo_op): Likewise.
 108         (index_string): Likewise.
 109         (output_index_string_offset): Likewise.
 110         (output_index_string): Likewise.
 111         (count_index_strings): Likewise.
 112
 113 2018-05-16  Carl Love  <cel@us.ibm.com>
 114
 115         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
 116         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
 117
 118 2018-05-16  Martin Jambor  <mjambor@suse.cz>
 119
 120         * ipa-prop.c (ipa_free_all_edge_args): Remove.
 121         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
 122
 123 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
 124
 125         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
 126         (fnma<mode>4): Likewise.
 127         (fms<mode>4): Likewise.
 128         (fnms<mode>4): Likewise.
 129         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
 130         (aarch64_fnma<mode>4): Likewise.
 131         (aarch64_fms<mode>4): Likewise.
 132         (aarch64_fnms<mode>4): Likewise.
 133         (aarch64_fnmadd<mode>4): Likewise.
 134
 135 2018-05-16  Jason Merrill  <jason@redhat.com>
 136
 137         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
 138
 139 2018-05-16  Richard Biener  <rguenther@suse.de>
 140
 141         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
 142         (dump_stmt_cost): Declare.
 143         (add_stmt_cost): Dump cost we add.
 144         (add_stmt_costs): New function.
 145         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
 146         No longer exported.
 147         (vect_analyze_stmt): Adjust prototype.
 148         (vectorizable_condition): Likewise.
 149         (vectorizable_live_operation): Likewise.
 150         (vectorizable_reduction): Likewise.
 151         (vectorizable_induction): Likewise.
 152         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
 153         cost vector to pass to vectorizable_ and record afterwards.
 154         (vect_model_reduction_cost): Take cost vector argument and adjust.
 155         (vect_model_induction_cost): Likewise.
 156         (vectorizable_reduction): Likewise.
 157         (vectorizable_induction): Likewise.
 158         (vectorizable_live_operation): Likewise.
 159         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
 160         SLP_TREE_NUMBER_OF_VEC_STMTS.
 161         (vect_analyze_slp_cost_1): Remove.
 162         (vect_analyze_slp_cost): Likewise.
 163         (vect_slp_analyze_node_operations): Take visited args and
 164         a target cost vector.  Avoid processing already visited stmt sets.
 165         (vect_slp_analyze_operations): Use a local cost vector to gather
 166         costs and register those of non-discarded instances.
 167         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
 168         (vect_schedule_slp_instance): Remove copying of
 169         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
 170         zero.
 171         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
 172         adding cost.  Record cost entry location.
 173         (vect_prologue_cost_for_slp_op): Function to compute cost of
 174         a constant or invariant generated for SLP vect in the prologue,
 175         split out from vect_analyze_slp_cost_1.
 176         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
 177         (vect_model_promotion_demotion_cost): Likewise.
 178         (vect_model_store_cost): Likewise, make static.
 179         (vect_model_load_cost): Likewise.
 180         (vectorizable_bswap): Add cost vector arg and adjust.
 181         (vectorizable_call): Likewise.
 182         (vectorizable_simd_clone_call): Likewise.
 183         (vectorizable_conversion): Likewise.
 184         (vectorizable_assignment): Likewise.
 185         (vectorizable_shift): Likewise.
 186         (vectorizable_operation): Likewise.
 187         (vectorizable_store): Likewise.
 188         (vectorizable_load): Likewise.
 189         (vectorizable_condition): Likewise.
 190         (vectorizable_comparison): Likewise.
 191         (can_vectorize_live_stmts): Likewise.
 192         (vect_analyze_stmt): Likewise.
 193         (vect_transform_stmt): Adjust calls to vectorizable_*.
 194         * tree-vectorizer.c: Include gimple-pretty-print.h.
 195         (dump_stmt_cost): New function.
 196
 197 2018-05-16  Richard Biener  <rguenther@suse.de>
 198
 199         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
 200         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
 201         * tree-ssa-dse.c: Include tree-ssa-loop.h.
 202         (check_name): New callback.
 203         (dse_classify_store): Track cycles via a visited bitmap of PHI
 204         defs and simplify handling of in-loop and across loop dead stores
 205         and properly fail for loop-variant refs.  Handle byte-tracking with
 206         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
 207         limiting the walk.
 208
 209 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
 210
 211         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
 212         (vect_get_mask_type_for_stmt): Likewise.
 213         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
 214         split out from...
 215         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
 216         to determine the statement's vector type and the vector type that
 217         should be used for calculating nunits.  Deal with cases in which
 218         the type has to be deferred.
 219         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
 220         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
 221         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
 222         (vect_determine_vf_for_stmt): New functions, split out from...
 223         (vect_determine_vectorization_factor): ...here.
 224         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
 225         (vect_get_mask_type_for_stmt): New functions, split out from
 226         vect_determine_vectorization_factor.
 227
 228 2018-05-16  Richard Biener  <rguenther@suse.de>
 229
 230         * tree-cfg.c (verify_gimple_assign_ternary): Properly
 231         verify the [VEC_]COND_EXPR embedded comparison.
 232
 233 2018-05-15  Martin Sebor  <msebor@redhat.com>
 234
 235         PR tree-optimization/85753
 236         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
 237         RECORD_TYPE in addition to ARRAY_TYPE.
 238
 239 2018-05-15  Martin Sebor  <msebor@redhat.com>
 240
 241         PR middle-end/85643
 242         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
 243
 244 2018-05-15  Richard Biener  <rguenther@suse.de>
 245
 246         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
 247         add by_clobber_p one.  Change algorithm to collect all defs
 248         representing uses we need to walk and try reducing them to
 249         a single one before failing.
 250         (dse_dom_walker::dse_optimize_stmt): Adjust.
 251
 252 2018-05-13  Mark Wielaard  <mark@klomp.org>
 253
 254         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
 255         (size_of_loc_descr): Likewise.
 256         (output_loc_operands): Likewise.
 257         (output_loc_operands_raw): Likewise.
 258         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
 259         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
 260         (hash_loc_operands): Likewise.
 261         (compare_loc_operands): Likewise.
 262
 263 2018-05-14  Mark Wielaard  <mark@klomp.org>
 264
 265         * dwarf2out.c (count_index_addrs): New function.
 266         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
 267
 268 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
 269
 270         PR tree-optimization/83648
 271         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
 272         return value as malloc candidate.
 273
 274 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
 275
 276         PR ipa/85734
 277         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
 278         as true in call to suggest_attribute.
 279
 280 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
 281
 282         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
 283         -mreadonly-in-sdata.
 284
 285 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 286
 287         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
 288         New pattern.
 289         (aarch64_crypto_aesd_fused): Likewise.
 290
 291 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
 292
 293         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
 294         (movsi_aarch64): Likewise.
 295         (load_pairsi): Likewise.
 296         (load_pairdi): Likewise.
 297         (store_pairsi): Likewise.
 298         (store_pairdi): Likewise.
 299         (load_pairsf): Likewise.
 300         (load_pairdf): Likewise.
 301         (store_pairsf): Likewise.
 302         (store_pairdf): Likewise.
 303         (zero_extend): Likewise.
 304         (trunc): Swap alternatives.
 305         (fcvt_target): Add '?' to prefer w over r.
 306
 307 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
 308
 309         PR target/85756
 310         * config/i386/i386.md: Disallow non-commutative arithmetics in
 311         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
 312         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
 313         in the peephole2 before it.
 314
 315 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
 316
 317         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
 318         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
 319         (ix86_handle_option): Handle -mcldemote.
 320         * config.gcc: New header.
 321         * config/i386/cldemoteintrin.h: New file.
 322         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
 323         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
 324         -mcldemote.
 325         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
 326         OPTION_MASK_ISA_CLDEMOTE.
 327         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
 328         (ix86_valid_target_attribute_inner_p): Ditto.
 329         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
 330         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
 331         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
 332         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
 333         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
 334         (cldemote): New.
 335         * config/i386/i386.opt: Add -mcldemote.
 336         * config/i386/x86intrin.h: New header.
 337         * doc/invoke.texi: Add -mcldemote.
 338
 339 2018-05-14  Richard Biener  <rguenther@suse.de>
 340
 341         * doc/match-and-simplify.texi: Adjust :s documentation.
 342
 343 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
 344
 345         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
 346         intended memcpy size.
 347         (REORDER_45): Likewise.
 348
 349 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
 350
 351         * sort.cc: New file.
 352         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
 353         * vec.c (qsort_chk): Use gcc_qsort.
 354         * Makefile.in (OBJS-libcommon): Add sort.o.
 355         (build/sort.o): New target.  Use it...
 356         (BUILD_RTL): ... here, and...
 357         (build/gencfn-macros): ... here, and...
 358         (build/genmatch): ... here.
 359
 360 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
 361             Chung-Ju Wu  <jasonwucj@gmail.com>
 362
 363         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
 364         * config/nds32/nds32-graywolf.md: New file.
 365         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
 366         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
 367         pipeline.
 368         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
 369         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
 370         * config/nds32/nds32.md (pipeline_model): Add graywolf.
 371         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
 372         * config/nds32/pipelines.md: Include n15 settings.
 373
 374 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
 375             Chung-Ju Wu  <jasonwucj@gmail.com>
 376
 377         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
 378         * config/nds32/nds32-n13.md: New file.
 379         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
 380         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
 381         pipeline.
 382         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
 383         * config/nds32/nds32.md (pipeline_model): Add n13.
 384         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
 385         * config/nds32/pipelines.md: Include n13 settings.
 386
 387 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
 388             Chung-Ju Wu  <jasonwucj@gmail.com>
 389
 390         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
 391         * config/nds32/nds32-n10.md: New file.
 392         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
 393         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
 394         pipeline.
 395         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
 396         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
 397         * config/nds32/nds32.md (pipeline_model): Add n10.
 398         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
 399         * config/nds32/pipelines.md: Include n10 settings.
 400
 401 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
 402             Kito Cheng  <kito.cheng@gmail.com>
 403             Chung-Ju Wu  <jasonwucj@gmail.com>
 404
 405         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
 406         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
 407         Add enum values for DSP extension instructions.
 408         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
 409         New constraints.
 410         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
 411         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
 412         New code iterators.
 413         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
 414         * config/nds32/nds32-dspext.md: New file for DSP implementation.
 415         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
 416         * config/nds32/nds32-intrinsic.md: Likewise.
 417         * config/nds32/nds32_intrinsic.h: Likewise.
 418         * config/nds32/nds32-md-auxiliary.c: Likewise.
 419         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
 420         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
 421         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
 422         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
 423         * config/nds32/nds32-protos.h: New declarations for DSP extension.
 424         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
 425         TYPE_DMAC in switch statement.
 426         * config/nds32/nds32.c: New checking and implementation for DSP
 427         extension instructions.
 428         * config/nds32/nds32.h: Likewise.
 429         * config/nds32/nds32.md: Likewise.
 430         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
 431         * config/nds32/predicates.md: Implement new predicates for DSP
 432         extension.
 433
 434 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
 435
 436         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
 437         Reformat alternatives and attributes so it is easier to identify
 438         which constraints/attributes go with which instruction.
 439         (mov<mode>_hardfloat32, FMOVE64): Likewise.
 440         (mov<mode>_softfloat32, FMOVE64): Likewise.
 441         (mov<mode>_hardfloat64, FMOVE64): Likewise.
 442         (mov<mode>_softfloat64, FMOVE64): Likewise.
 443
 444 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 445
 446         * doc/extend.texi (PowerPC Built-in Functions): Rename this
 447         subsection.
 448         (Basic PowerPC Built-in Functions): The new name of the
 449         subsection previously known as "PowerPC Built-in Functions".
 450         (Basic PowerPC Built-in Functions Available on all Configurations):
 451         New subsubsection.
 452         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
 453         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
 454         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
 455         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
 456
 457 2018-05-11  Martin Jambor  <mjambor@suse.cz>
 458
 459         PR ipa/85655
 460         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
 461         single const.
 462
 463 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
 464
 465         PR target/85733
 466         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
 467
 468 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
 469
 470         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
 471         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
 472         (ix86_handle_option): Handle -mwaitpkg.
 473         * config.gcc: New header.
 474         * config/i386/cpuid.h (bit_WAITPKG): New bit.
 475         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
 476         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
 477         function type.
 478         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
 479         OPTION_MASK_ISA_WAITPKG.
 480         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
 481         (ix86_option_override_internal): Add PTA_WAITPKG.
 482         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
 483         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
 484         IX86_BUILTIN_TPAUSE.
 485         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
 486         __builtin_ia32_umwait and __builtin_ia32_tpause.
 487         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
 488         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
 489         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
 490         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
 491         UNSPECV_TPAUSE): New.
 492         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
 493         * config/i386/i386.opt: Add -mwaitpkg.
 494         * config/i386/waitpkgintrin.h: New file.
 495         * config/i386/x86intrin.h: New header.
 496         * doc/invoke.texi: Add -mwaitpkg.
 497
 498 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
 499
 500         PR target/85606
 501         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
 502         equivalent.
 503         (cortex-m0): Use armv6s-m isa.
 504         (cortex-m0plus): Likewise.
 505         (cortex-m1): Likewise.
 506         (cortex-m0.small-multiply): Likewise.
 507         (cortex-m0plus.small-multiply): Likewise.
 508         (cortex-m1.small-multiply): Likewise.
 509
 510 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
 511             Jakub Jelinek  <jakub@redhat.com>
 512
 513         PR tree-optimization/85692
 514         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
 515         source permute as well.
 516
 517 2018-05-11  Martin Liska  <mliska@suse.cz>
 518
 519         PR sanitizer/85556
 520         * doc/extend.texi: Document LLVM style format for no_sanitize
 521         attribute.
 522
 523 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
 524
 525         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
 526         mode_supports_vsx_dform_quad to mode_supports_dq_form.
 527         (mode_supports_vsx_dform_quad): Likewise.
 528         (mode_supports_vmx_dform): Move these functions to be next to the
 529         other mode_supports functions.
 530         (mode_supports_dq_form): Likewise.
 531         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
 532         mode_supports_dq_form.
 533         (reg_offset_addressing_ok_p): Likewise.
 534         (offsettable_ok_by_alignment): Likewise.
 535         (rs6000_legitimate_offset_address_p): Likewise.
 536         (legitimate_lo_sum_address_p): Likewise.
 537         (rs6000_legitimize_address): Likewise.
 538         (rs6000_legitimize_reload_address): Likewise.
 539         (rs6000_secondary_reload_inner): Likewise.
 540         (rs6000_preferred_reload_class): Likewise.
 541         (rs6000_output_move_128bit): Likewise.
 542
 543 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
 544
 545         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
 546         Generate SImode target register for null target.
 547         <case IX86_BUILTIN_XGETBV>: Ditto.
 548         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
 549         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
 550
 551 2018-05-10  Carl Love  <cel@us.ibm.com>
 552
 553         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
 554         dcbtt and dcbtstt if operands[2] is 0.
 555
 556 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
 557
 558         PR target/85693
 559         * config/i386/sse.md (usadv64qi): New expander.
 560
 561 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
 562
 563         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
 564         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
 565         -maltivec=be support.
 566         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
 567         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
 568         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
 569         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
 570         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
 571         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
 572         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
 573         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
 574         altivec_vsumsws): Adjust.
 575         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
 576         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
 577         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
 578         support.
 579         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
 580         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
 581         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
 582         (altivec_lve<VI_char>x): Delete expand.
 583         (*altivec_lve<VI_char>x_internal): Rename to...
 584         (altivec_lve<VI_char>x): ... this.
 585         (altivec_lvxl_<mode>): Delete expand.
 586         (*altivec_lvxl_<mode>_internal): Rename to ...
 587         (altivec_lvxl_<mode>): ... this.
 588         (altivec_stvxl_<mode>): Delete expand.
 589         (*altivec_stvxl_<mode>_internal): Rename to ...
 590         (altivec_stvxl_<mode>): ... this.
 591         (altivec_stve<VI_char>x): Delete expand.
 592         (*altivec_stve<VI_char>x_internal): Rename to ...
 593         (altivec_stve<VI_char>x): ... this.
 594         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
 595         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
 596         reduc_plus_scal_<mode>): Adjust.
 597         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
 598         comment.
 599         (rs6000_cpu_cpp_builtins): Adjust.
 600         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
 601         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
 602         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
 603         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
 604         -maltivec=be support.
 605         (rs6000_split_vec_extract_var): Adjust.
 606         (rs6000_split_v4si_init): Adjust.
 607         (swap_selector_for_mode): Delete.
 608         (altivec_expand_lvx_be, altivec_expand_stvx_be,
 609         altivec_expand_stvex_be): Delete.
 610         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
 611         -maltivec=be support.
 612         (rs6000_gimple_fold_builtin): Ditto.
 613         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
 614         Adjust.
 615         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
 616         (TARGET_DIRECT_MOVE_64BIT): Adjust.
 617         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
 618         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
 619         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
 620         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
 621         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
 622         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
 623         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
 624         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
 625         anonymous split): Adjust.
 626         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
 627         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
 628
 629 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
 630
 631         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
 632         when --with-gxx-include-dir is also specified.
 633         * configure: Regenerate.
 634
 635 2018-05-09  Jim Wilson  <jimw@sifive.com>
 636
 637         PR target/84797
 638         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
 639         * config/riscv/t-withmultilib: New.
 640         * config/riscv/withmultilib.h: New.
 641         * doc/install.texi: Document RISC-V --with-multilib-list support.
 642
 643 2018-05-09  Richard Biener  <rguenther@suse.de>
 644
 645         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
 646         vector.
 647         (vect_bb_vectorization_profitable_p): Adjust.  Compute
 648         actual scalar cost using the cost vector and the add_stmt_cost
 649         machinery.
 650
 651 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
 652
 653         PR rtl-optimization/85645
 654         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
 655         in the REG_CFA_REGISTER note for LR, don't leave it empty.
 656
 657 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
 658
 659         PR rtl-optimization/85645
 660         * shrink-wrap.c (spread_components): Return a boolean saying if
 661         anything was changed.
 662         (try_shrink_wrapping_separate): Iterate spread_components until
 663         nothing changes anymore.
 664
 665 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
 666
 667         PR rtl-optimization/85645
 668         * regrename.c (build_def_use): Also kill the chains that include the
 669         destination of a REG_CFA_REGISTER note.
 670
 671 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
 672
 673         PR rtl-optimization/85645
 674         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
 675         insn that has a REG_CFA_REGISTER note.
 676
 677 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
 678
 679         * cfgexpand.c (expand_clobber): New function.
 680         (expand_gimple_stmt_1): Use it.
 681         * tree-vect-stmts.c (vect_clobber_variable): New function,
 682         split out from...
 683         (vectorizable_simd_clone_call): ...here.
 684         (vectorizable_store): Emit a clobber either side of an
 685         IFN_STORE_LANES sequence.
 686         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
 687
 688 2018-05-09  Tom de Vries  <tom@codesourcery.com>
 689
 690         PR target/85626
 691         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
 692         (define_insn "trap_if_false"): Add exit after trap.
 693
 694 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
 695
 696         PR rtl-optimization/85638
 697         * bb-reorder.c: Include common/common-target.h.
 698         (create_forwarder_block): New function extracted from...
 699         (fix_up_crossing_landing_pad): ...here.  Rename into...
 700         (dw2_fix_up_crossing_landing_pad): ...this.
 701         (sjlj_fix_up_crossing_landing_pad): New function.
 702         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
 703         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
 704         from both partitions and exit the loop after one iteration.
 705
 706 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 707
 708         Revert:
 709         * doc/extend.texi (PowerPC Built-in Functions): Rename this
 710         subsection.
 711         (Basic PowerPC Built-in Functions): The new name of the
 712         subsection previously known as "PowerPC Built-in Functions".
 713         (Basic PowerPC Built-in Functions Available on all Configurations):
 714         New subsubsection.
 715         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
 716         subsubsection.
 717         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
 718         subsubsection.
 719         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
 720         subsubsection.
 721         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
 722         subsubsection.
 723
 724 2018-05-08  Jim Wilson  <jimw@sifive.com>
 725
 726         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
 727         (LD_EMUL_SUFFIX): New.
 728         (LINK_SPEC): Use it.
 729
 730 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 731
 732         * doc/extend.texi (PowerPC Built-in Functions): Rename this
 733         subsection.
 734         (Basic PowerPC Built-in Functions): The new name of the
 735         subsection previously known as "PowerPC Built-in Functions".
 736         (Basic PowerPC Built-in Functions Available on all Configurations):
 737         New subsubsection.
 738         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
 739         subsubsection.
 740         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
 741         subsubsection.
 742         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
 743         subsubsection.
 744         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
 745         subsubsection.
 746
 747 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
 748
 749         PR target/85683
 750         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
 751         after cmpelim optimization.
 752
 753 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
 754
 755         * config.gcc: Support "goldmont".
 756         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
 757         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
 758         PROCESSOR_GOLDMONT.
 759         * config/i386/i386.c (m_GOLDMONT): Define.
 760         (processor_target_table): Add "goldmont".
 761         (PTA_GOLDMONT): Define.
 762         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
 763         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
 764         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
 765         (fold_builtin_cpu): Add "goldmont".
 766         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
 767         (ix86_option_override_internal): Add "goldmont".
 768         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
 769         (processor_type): Add PROCESSOR_GOLDMONT.
 770         * config/i386/i386.md: Add CPU "glm".
 771         * config/i386/glm.md: New file.
 772         * config/i386/x86-tune.def: Add m_GOLDMONT.
 773         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
 774
 775 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
 776
 777         PR target/85572
 778         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
 779         E_V4DImode.
 780         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
 781         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
 782         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
 783
 784         PR target/85317
 785         * config/i386/i386.c (ix86_fold_builtin): Handle
 786         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
 787
 788         PR target/85480
 789         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
 790         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
 791
 792 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
 793
 794         PR target/85658
 795         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
 796         (check_arch): Likewise.
 797         (check_fpu): Return the result rather than printing it.
 798         (end arch): Fix operator precedence.
 799         (end cpu): Likewise.
 800         (END): Print the result from check_fpu.
 801
 802 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
 803             Alan Hayward  <alan.hayward@arm.com>
 804             David Sherwood  <david.sherwood@arm.com>
 805
 806         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
 807         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
 808         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
 809         (*fcmuo<mode>_and): New patterns.
 810
 811 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
 812
 813         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
 814         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
 815         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
 816         (cmp_op, sve_imm_con): New code attributes.
 817         (SVE_COND_INT_CMP, imm_con): Delete.
 818         (cmp_op): Remove above unspecs from int attribute.
 819         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
 820         to...
 821         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
 822         comparison-specific unspecs.
 823         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
 824         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
 825         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
 826         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
 827         (*vec_fcm<cmp_op><mode>): Rename to...
 828         (*fcm<cmp_op><mode>): ...this and adjust likewise.
 829         (*vec_fcmuo<mode>): Rename to...
 830         (*fcmuo<mode>): ...this and adjust likewise.
 831         (*pred_fcm<cmp_op><mode>): New pattern.
 832         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
 833         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
 834         functions.
 835         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
 836         and UNORDERED.
 837         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
 838         (aarch64_emit_sve_predicated_cond): New function.
 839         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
 840         (aarch64_emit_unspec_cond_or): Replace with...
 841         (aarch64_emit_sve_or_conds): ...this new function.  Use
 842         aarch64_emit_sve_ptrue_op for the individual comparisons and
 843         aarch64_emit_binop to OR them together.
 844         (aarch64_emit_inverted_unspec_cond): Replace with...
 845         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
 846         aarch64_emit_sve_ptrue_op for the comparison and
 847         aarch64_emit_unop to invert the result.
 848         (aarch64_expand_sve_vec_cmp_float): Update after the above
 849         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
 850
 851 2018-05-07  Nathan Sidwell  <nathan@acm.org>
 852
 853         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
 854         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
 855         (Backwards Compatibility): Likewise.
 856
 857 2018-05-07  Luis Machado  <luis.machado@linaro.org>
 858
 859         PR bootstrap/85681
 860         Revert:
 861         2018-05-07  Luis Machado  <luis.machado@linaro.org>
 862
 863         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
 864         <prefetch_dynamic_strides>: New const bool field.
 865         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
 866         prefetch_dynamic_strides.
 867         (exynosm1_prefetch_tune): Likewise.
 868         (thunderxt88_prefetch_tune): Likewise.
 869         (thunderx_prefetch_tune): Likewise.
 870         (thunderx2t99_prefetch_tune): Likewise.
 871         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
 872         (aarch64_override_options_internal): Update to set
 873         PARAM_PREFETCH_DYNAMIC_STRIDES.
 874         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
 875         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
 876         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
 877         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
 878         prefetch-dynamic-strides setting.
 879
 880         2018-05-07  Luis Machado  <luis.machado@linaro.org>
 881
 882         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
 883         <minimum_stride>: New const int field.
 884         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
 885         minimum_stride field.
 886         (exynosm1_prefetch_tune): Likewise.
 887         (thunderxt88_prefetch_tune): Likewise.
 888         (thunderx_prefetch_tune): Likewise.
 889         (thunderx2t99_prefetch_tune): Likewise.
 890         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
 891         (aarch64_override_options_internal): Update to set
 892         PARAM_PREFETCH_MINIMUM_STRIDE.
 893         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
 894         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
 895         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
 896         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
 897         stride is constant and is below the minimum stride threshold.
 898
 899 2018-05-07  Luis Machado  <luis.machado@linaro.org>
 900
 901         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
 902         to 512.
 903
 904 2018-05-07  Luis Machado  <luis.machado@linaro.org>
 905
 906         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
 907         <prefetch_dynamic_strides>: New const bool field.
 908         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
 909         prefetch_dynamic_strides.
 910         (exynosm1_prefetch_tune): Likewise.
 911         (thunderxt88_prefetch_tune): Likewise.
 912         (thunderx_prefetch_tune): Likewise.
 913         (thunderx2t99_prefetch_tune): Likewise.
 914         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
 915         (aarch64_override_options_internal): Update to set
 916         PARAM_PREFETCH_DYNAMIC_STRIDES.
 917         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
 918         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
 919         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
 920         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
 921         prefetch-dynamic-strides setting.
 922
 923 2018-05-07  Luis Machado  <luis.machado@linaro.org>
 924
 925         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
 926         <minimum_stride>: New const int field.
 927         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
 928         minimum_stride field.
 929         (exynosm1_prefetch_tune): Likewise.
 930         (thunderxt88_prefetch_tune): Likewise.
 931         (thunderx_prefetch_tune): Likewise.
 932         (thunderx2t99_prefetch_tune): Likewise.
 933         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
 934         (aarch64_override_options_internal): Update to set
 935         PARAM_PREFETCH_MINIMUM_STRIDE.
 936         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
 937         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
 938         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
 939         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
 940         stride is constant and is below the minimum stride threshold.
 941
 942 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
 943
 944         PR c++/85659
 945         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
 946         the type is addressable.  Don't force op into register if it has
 947         BLKmode.
 948
 949 2018-05-05  Roland McGrath  <mcgrathr@google.com>
 950
 951         PR other/77609
 952         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
 953         any section for which we don't know a specific type it should have,
 954         regardless of name.  Previously this was done only for the exact
 955         names ".init_array", ".fini_array", and ".preinit_array".
 956         (default_elf_asm_named_section): Add comment about
 957         relationship with default_section_type_flags and SECTION_NOTYPE.
 958         (get_section): Don't consider it a type conflict if one side has
 959         SECTION_NOTYPE and the other doesn't, as long as neither has the
 960         SECTION_BSS et al used in the default_section_type_flags logic.
 961
 962 2018-05-05  Tom de Vries  <tom@codesourcery.com>
 963
 964         PR target/85653
 965         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
 966         (workaround_barsyncs): New function.
 967         (nvptx_reorg): Use workaround_barsyncs.
 968         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
 969         (define_expand "nvptx_membar_cta"): New define_expand.
 970         (define_insn "*nvptx_membar_cta"): New insn.
 971
 972 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
 973
 974         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
 975         To improve optimization opportunities.
 976         * builtin-types.def: The new needed builtin types for the above.
 977
 978 2018-05-04  Richard Biener  <rguenther@suse.de>
 979
 980         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
 981         * gimple-ssa-store-merging.c
 982         (imm_store_chain_info::output_merged_store): Remove redundant create,
 983         release split_store vector contents on failure.
 984         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
 985         scalar stmt vector on cache hit.
 986
 987 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
 988
 989         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
 990         Xilinx FP support.
 991         * config.gcc (powerpc-xilinx-eabi*): Remove.
 992         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
 993         support.
 994         (fusion_addis_mem_combo_load): Ditto.
 995         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
 996         FP support.
 997         (rs6000_cpu_cpp_builtins): Ditto.
 998         * config/rs6000/rs6000-linux.c
 999         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1000         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1001         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1002         support.
1003         (rs6000_setup_reg_addr_masks): Ditto.
1004         (rs6000_init_hard_regno_mode_ok): Ditto.
1005         (rs6000_option_override_internal): Ditto.
1006         (legitimate_lo_sum_address_p): Ditto.
1007         (rs6000_legitimize_address): Ditto.
1008         (rs6000_legitimize_reload_address): Ditto.
1009         (rs6000_legitimate_address_p): Ditto.
1010         (abi_v4_pass_in_fpr): Ditto.
1011         (setup_incoming_varargs): Ditto.
1012         (rs6000_gimplify_va_arg): Ditto.
1013         (rs6000_split_multireg_move): Ditto.
1014         (rs6000_savres_strategy): Ditto.
1015         (rs6000_emit_prologue_components): Ditto.
1016         (rs6000_emit_epilogue_components): Ditto.
1017         (rs6000_emit_prologue): Ditto.
1018         (rs6000_emit_epilogue): Ditto.
1019         (rs6000_elf_file_end): Ditto.
1020         (rs6000_function_value): Ditto.
1021         (rs6000_libcall_value): Ditto.
1022         * config/rs6000/rs6000.h: Ditto.
1023         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1024         (TARGET_MINMAX): ... this.  New.
1025         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1026         * config/rs6000/rs6000.md: Remove Xilinx FP support.
1027         (*movsi_internal1_single): Delete.
1028         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1029         mfpu=, mxilinx-fpu): Delete.
1030         * config/rs6000/singlefp.h: Delete.
1031         * config/rs6000/sysv4.h: Remove Xilinx FP support.
1032         * config/rs6000/t-rs6000: Ditto.
1033         * config/rs6000/t-xilinx: Delete.
1034         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1035         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1036         (VStype_simple): Delete.
1037         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1038         * config/rs6000/xfpu.h: Delete.
1039         * config/rs6000/xfpu.md: Delete.
1040         * config/rs6000/xilinx.h: Delete.
1041         * config/rs6000/xilinx.opt: Delete.
1042         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1043         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1044
1045 2018-05-04  Tom de Vries  <tom@codesourcery.com>
1046
1047         PR libgomp/85639
1048         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1049         if ignore == 0.
1050
1051 2018-05-04  Richard Biener  <rguenther@suse.de>
1052
1053         PR middle-end/85627
1054         * tree-complex.c (update_complex_assignment): We are always in SSA form.
1055         (expand_complex_div_wide): Likewise.
1056         (expand_complex_operations_1): Likewise.
1057         (expand_complex_libcall): Preserve EH info of the original stmt.
1058         (tree_lower_complex): Handle removed blocks.
1059         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1060         on complex multiplication and division libcall builtins.
1061
1062 2018-05-04  Richard Biener  <rguenther@suse.de>
1063
1064         PR middle-end/85574
1065         * fold-const.c (negate_expr_p): Restrict negation of operand
1066         zero of a division to when we know that can happen without
1067         overflow.
1068         (fold_negate_expr_1): Likewise.
1069
1070 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
1071
1072         PR libstdc++/85466
1073         * real.h (real_nextafter): Declare.
1074         * real.c (real_nextafter): New function.
1075         * fold-const-call.c (fold_const_nextafter): New function.
1076         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1077         CASE_CFN_NEXTTOWARD.
1078         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1079         even when arg1_mode is different from arg0_mode.
1080
1081 2018-05-03  Nathan Sidwell  <nathan@acm.org>
1082
1083         * doc/extend.texi (Deprecated Features): Remove
1084         -ffriend-injection.
1085         (Backwards Compatibility): Likewise.
1086         * doc/invoke.texi (C++ Language Options): Likewise.
1087         (C++ Dialect Options): Likewise.
1088
1089 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
1090
1091         PR target/85530
1092         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1093         _mm512_mask_mullox_epi64): New intrinsics.
1094
1095 2018-05-03  Tom de Vries  <tom@codesourcery.com>
1096
1097         PR testsuite/85106
1098         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1099         dump files): Add offload-tree.
1100
1101 2018-05-03  Richard Biener  <rguenther@suse.de>
1102
1103         PR tree-optimization/85615
1104         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1105         to loops not nested in BBs loop father to avoid creating multi-entry
1106         loops.
1107
1108 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1109
1110         PR tree-optimization/70291
1111         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1112         arguments.  Change return type to tree.  Emit libcall as a new
1113         statement rather than replacing existing one when inplace_p is true.
1114         (expand_complex_multiplication_components): New function.
1115         (expand_complex_multiplication): Expand floating-point complex
1116         multiplication using the above.
1117         (expand_complex_division): Rename inner_type parameter to type.
1118         Update expand_complex_libcall call-site.
1119         (expand_complex_operations_1): Update expand_complex_multiplication
1120         and expand_complex_division call-sites.
1121
1122 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
1123
1124         PR target/85582
1125         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1126         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1127         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1128         the highest significant bit of the shift count mask is clear.  In
1129         check whether and[sq]i3 is needed verify that all significant bits
1130         of the shift count other than the highest are set.
1131
1132 2018-05-02  Tom de Vries  <tom@codesourcery.com>
1133
1134         PR libgomp/82428
1135         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1136         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1137         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1138         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1139         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1140         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1141         __builtin_goacc_parlevel_size.
1142
1143 2018-05-02  Richard Biener  <rguenther@suse.de>
1144
1145         PR tree-optimization/85597
1146         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1147         do not use split vect_get_vec_defs call but call vect_get_slp_defs
1148         directly.
1149
1150 2018-05-02  Tom de Vries  <tom@codesourcery.com>
1151
1152         PR testsuite/85106
1153         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1154         dump files): Add ltrans-tree.
1155
1156 2018-05-02  Tom de Vries  <tom@codesourcery.com>
1157
1158         PR testsuite/85106
1159         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1160         dump files): Add wpa-ipa.
1161
1162 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
1163
1164         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
1165         powerpc*-*-linux*paired* target.
1166         * config/rs6000/750cl.h: Delete.
1167         * config/rs6000/paired.h: Delete.
1168         * config/rs6000/paired.md: Delete.
1169         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1170         float support.
1171         * config/rs6000/rs6000-builtin.def: Remove paired float support.
1172         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1173         comment.  Remove paired float support.
1174         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1175         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1176         VECTOR_PAIRED.
1177         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1178         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1179         declarations.
1180         * config/rs6000/rs6000.c: Remove paired float support.
1181         (paired_expand_vector_init, paired_expand_vector_move,
1182         paired_emit_vector_compare, paired_emit_vector_cond_expr,
1183         (paired_expand_lv_builtin, paired_expand_stv_builtin,
1184         paired_expand_builtin, paired_expand_predicate_builtin,
1185         paired_init_builtins): Delete.
1186         * config/rs6000/rs6000.h: Remove paired float support.
1187         * config/rs6000/rs6000.md: Remove paired float support.
1188         (move_from_CR_ov_bit): Delete.
1189         * config/rs6000/rs6000.opt (mpaired): Delete.
1190         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1191         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1192
1193 2018-05-02  Richard Biener  <rguenther@suse.de>
1194
1195         PR middle-end/85567
1196         * gimplify.c (gimplify_save_expr): When in SSA form allow
1197         SAVE_EXPRs to compute to SSA vars.
1198
1199 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
1200
1201         PR target/85582
1202         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1203         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1204         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1205         clobber operands[2], instead use a new pseudo.  Formatting fixes.
1206
1207 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
1208
1209         PR tree-optimization/85586
1210         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1211         exit early for statements in the same group if the accesses are
1212         not strided.
1213
1214 2018-05-02  Tom de Vries  <tom@codesourcery.com>
1215
1216         PR lto/85451
1217         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1218         error message.
1219
1220 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
1221
1222         PR tree-optimization/85143
1223         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1224
1225 2018-05-01  Tom de Vries  <tom@codesourcery.com>
1226
1227         PR lto/85451
1228         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1229         not found" error message.
1230
1231 2018-05-01  Tom de Vries  <tom@codesourcery.com>
1232
1233         PR other/83786
1234         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1235         * vec.c (test_ordered_remove_if): New function.
1236         (vec_c_tests): Call test_ordered_remove_if.
1237         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1238         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1239         * tree-vect-patterns.c (vect_pattern_recog_1): Use
1240         VEC_ORDERED_REMOVE_IF.
1241
1242 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1243
1244         PR tree-optimization/82665
1245         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1246         pointer subtraction where arguments come from a memchr call.
1247
1248 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
1249
1250         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1251         --push-state --as-needed and --pop-state instead of --as-needed and
1252         --no-as-needed if ld supports it.
1253         * configure: Regenerated.
1254
1255         PR web/85578
1256         * doc/install.texi2html: Replace _002d with - and _002a with * in
1257         generated html files using sed.
1258
1259 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1260
1261         PR c++/85523
1262         * gcc-rich-location.c (blank_line_before_p): New function.
1263         (use_new_line): New function.
1264         (gcc_rich_location::add_fixit_insert_formatted): New function.
1265         * gcc-rich-location.h
1266         (gcc_rich_location::add_fixit_insert_formatted): New function.
1267
1268 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1269
1270         * selftest.c (assert_streq): Rename "expected" and "actual" to
1271         "val1" and "val2".  Extend NULL-handling to cover both inputs
1272         symmetrically, while still requiring both to be non-NULL for a pass.
1273         * selftest.h (assert_streq): Rename "expected" and "actual" to
1274         "val1" and "val2".
1275         (ASSERT_EQ): Likewise.
1276         (ASSERT_EQ_AT): Likewise.
1277         (ASSERT_KNOWN_EQ): Likewise.
1278         (ASSERT_KNOWN_EQ_AT): Likewise.
1279         (ASSERT_NE): Likewise.
1280         (ASSERT_MAYBE_NE): Likewise.
1281         (ASSERT_MAYBE_NE_AT): Likewise.
1282         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
1283         the assertion to pass.
1284         (ASSERT_STREQ_AT): Likewise.
1285
1286 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
1287
1288         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1289         interaction with -pie.
1290
1291 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1292
1293         * selftest.h: Fix alphabetization of per-source-file selftest
1294         declarations.
1295
1296 2018-04-30  Jason Merrill  <jason@redhat.com>
1297
1298         PR c++/61982 - dead stores to destroyed objects.
1299         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1300         of clobber.
1301
1302 2018-04-30  Jason Merrill  <jason@redhat.com>
1303
1304         * tree.c (build_clobber): New.
1305         * tree.h: Declare it.
1306         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1307
1308 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1309
1310         * diagnostic-show-locus.c (layout::layout): Update for
1311         location_get_source_line returning a char_span.
1312         (struct char_span): Move to input.h.
1313         (struct correction): Update for fields in char_span becoming
1314         private.
1315         (struct source_line): Update for location_get_source_line
1316         returning a char_span.
1317         (layout::print_line): Likewise.
1318         * edit-context.c (edited_file::print_content): Likewise.
1319         (edited_file::print_diff_hunk): Likewise.
1320         (edited_file::print_run_of_changed_lines): Likewise.
1321         (edited_file::get_num_lines): Likewise.
1322         (edited_line::edited_line): Likewise.
1323         * final.c (asm_show_source): Likewise.
1324         * input.c (location_get_source_line): Convert return type
1325         from const char * to char_span, losing the final "line_len"
1326         param.
1327         (dump_location_info): Update for the above.
1328         (get_substring_ranges_for_loc): Likewise.  Use a char_span
1329         when handling the literal within the line.
1330         (test_reading_source_line): Update for location_get_source_line
1331         returning a char_span.
1332         * input.h (class char_span): Move here from
1333         diagnostic-show-locus.c, converting from a struct to a class.
1334         Make data members private.
1335         (char_span::operator bool): New.
1336         (char_span::length): New.
1337         (char_span::get_buffer): New.
1338         (char_span::operator[]): New.
1339         (char_span::subspan): Make const.
1340         (char_span::xstrdup): New.
1341         (location_get_source_line): Convert return type from const char *
1342         to char_span, losing the final "line_size" param.
1343
1344 2018-04-30  Jan Hubicka  <jh@suse.cz>
1345
1346         * lto-wrapper.c (ltrans_priorities): New static var.
1347         (cmp_priority): New.
1348         (run_gcc): Read priorities and if doing parallel build order
1349         the Makefile by them.
1350
1351 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1352
1353         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1354
1355 2018-04-30  Richard Biener  <rguenther@suse.de>
1356
1357         * tree-cfg.c (verify_address): Remove base argument, add
1358         flag whether to check TREE_ADDRESSABLE and do that.
1359         (verify_expr): Remove.
1360         (verify_types_in_gimple_reference): Add pieces from verify_expr.
1361         (verify_gimple_assign_single): Likewise.
1362         (verify_gimple_switch): Likewise.
1363         (verify_expr_location_1): Dereference tp once.  Add (disabled)
1364         piece from verify_expr.
1365         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1366
1367 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
1368
1369         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1370
1371 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
1372
1373         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1374         (small_data_pattern): Likewise.
1375         (arc_rewrite_small_data): Likewise.
1376         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1377         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1378         (get_symbol_alignment): New function.
1379         (legitimate_small_data_address_p): Likewise.
1380         (legitimate_scaled_address): Update, call
1381         legitimate_small_data_address_p.
1382         (output_sdata): New static variable.
1383         (arc_print_operand): Update how we handle small data operands.
1384         (arc_print_operand_address): Likewise.
1385         (arc_legitimate_address_p): Update, use
1386         legitimate_small_data_address_p.
1387         (arc_rewrite_small_data_p): Remove.
1388         (arc_rewrite_small_data_1): Likewise.
1389         (arc_rewrite_small_data): Likewise.
1390         (small_data_pattern): Likewise.
1391         (compact_sda_memory_operand): Update to use
1392         legitimate_small_data_address_p and get_symbol_alignment.
1393         (prepare_move_operands): Don't rewite sdata pattern.
1394         (prepare_extend_operands): Remove.
1395         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1396         pattern.
1397         (zero_extendqisi2): Likewise.
1398         (zero_extendhisi2): Likewise.
1399         (extendqihi2): Likewise.
1400         (extendqisi2): Likewise.
1401         (extendhisi2): Likewise.
1402         (addsi3): Likewise.
1403         (subsi3): Likewise.
1404         (andsi3): Likewise.
1405         * config/arc/constraints.md (Usd): Change it to memory constraint.
1406
1407 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
1408
1409         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1410         as source of std instructions.
1411         * config/arc/arc.md (movsi_insn): Update pattern predicate to
1412         allow 6-bit constants as source for store instructions.
1413         (movdi_insn): Update instruction pattern to allow 6-bit constants
1414         as source for store instructions.
1415
1416 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
1417
1418         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1419
1420 2018-04-30  Nathan Sidwell  <nathan@acm.org>
1421             Sandra Loosemore <sandra@codesourcery.com>
1422
1423         * dumpfile.c (dump_open): Allow '-' for stdout.
1424         * doc/invoke.texi (Developer Options): Document dump filename
1425         determination early.  Document stdin/stdout selection.
1426
1427 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
1428
1429         Microblaze Target: PIC data text relative
1430
1431         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1432         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1433         Add declaration.
1434         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1435         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1436         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1437         New addressing mode for data-text relative position indepenedent code.
1438         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1439         'ADDRESS_SYMBOLIC_TXT_REL'.
1440         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1441         (microblaze_legitimate_pic_operand): Exclude function calls from
1442         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1443         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1444         addresses cases.
1445         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1446         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1447         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1448         for 'address + offset'.
1449         (microblaze_expand_prologue): Add new function prologue call for
1450         'r20' assignation.
1451         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1452         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1453         table in case of TARGET_PIC_DATA_TEXT_REL.
1454         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1455         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1456         Add new macros 'UNSPEC_TEXT',
1457         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1458         + exclude function calls from 'UNSPEC_PLT' in case of data text
1459         relative mode.
1460         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1461         new target hook for generating address diff vector tables in case of
1462         flag_pic.
1463         * doc/tm.texi : Regenerate.
1464         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1465         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1466         of addr diff vector generation.
1467         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1468         target hook definition.
1469         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1470         Add default function for generate_pic_addr_diff_vec -> flag_pic.
1471         * doc/invoke.texi (Add new pic option): Add new microblaze pic
1472         option for data text relative.
1473
1474 2018-04-30  Richard Biener  <rguenther@suse.de>
1475
1476         * tree-chrec.h (evolution_function_is_constant_p): Remove
1477         redundant check.
1478         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1479
1480 2018-04-30  Richard Biener  <rguenther@suse.de>
1481
1482         PR bootstrap/85571
1483         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1484
1485 2018-04-30  Richard Biener  <rguenther@suse.de>
1486
1487         PR tree-optimization/28364
1488         PR tree-optimization/85275
1489         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1490         copying first exit test.
1491
1492 2018-04-28  Mark Wielaard  <mark@klomp.org>
1493
1494         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1495         dwarf_version >= 5.
1496         (dwarf_AT): Handle DW_AT_addr_base.
1497         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1498
1499 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
1500
1501         PR target/84431
1502         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1503         (*ashl<dwi>3_doubleword_mask_1): Ditto.
1504         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1505         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1506
1507 2018-04-28  Richard Biener  <rguenther@suse.de>
1508
1509         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1510         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1511         to reflect use.  Only add interesting stmts.
1512
1513 2018-04-27  Martin Jambor  <mjambor@suse.cz>
1514
1515         PR ipa/85549
1516         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1517         the jump function allows for passing through aggregate values.
1518
1519 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
1520
1521         * input.h (in_system_header_at): Convert from macro to inline
1522         function.
1523         (from_macro_expansion_at): Likewise.
1524         (from_macro_definition_at): Likewise.
1525
1526 2018-04-27  Jeff Law  <law@redhat.com>
1527
1528         * config.gcc: Mark tile* targets as deprecated/obsolete.
1529
1530 2018-04-27  Richard Biener  <rguenther@suse.de>
1531
1532         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1533         fix for ILP32.
1534
1535 2018-04-27  Richard Biener  <rguenther@suse.de>
1536
1537         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1538
1539 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
1540
1541         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1542         with Yd constraint. Set "preferred_for_speed" attribute from
1543         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1544         with Yd constraint.
1545         (*movdi_internal): Ditto.
1546         (movti_interunit splitters): Remove
1547         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1548         (movdi_interunit splitters): Ditto.
1549         * config/i386/constraints.md (Ye): Remove.
1550         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1551
1552 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1553
1554         PR target/85512
1555         * config/aarch64/constraints.md (Usg): Limit to 31.
1556         (Usj): Limit to 63.
1557
1558 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
1559
1560         PR tree-optimization/85529
1561         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1562         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1563         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1564         zero extension or masking of the MSB bit.
1565         (optimize_range_tests): Add FIRST_BB argument, pass it through
1566         to optimize_range_tests_var_bound.
1567         (maybe_optimize_range_tests, reassociate_bb): Adjust
1568         optimize_range_tests callers.
1569
1570 2018-04-26  Richard Biener  <rguenther@suse.de>
1571             Jakub Jelinek  <jakub@redhat.com>
1572
1573         * cgraph.h (symbol_table): Just declare debug method here.
1574         * symtab.c (symbol_table::debug): Define.
1575
1576 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
1577
1578         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1579
1580 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
1581
1582         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1583         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1584         (*movdi_internal): Substitute Yi and Yj constraint with x
1585         and Ym and Yn constraint with y constraint.  Update "isa"
1586         attribute and set "preferred_for_speed" attribute from
1587         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1588         (*movsi_internal): Ditto.
1589         (*movdf_internal): Ditto.
1590         (*movsf_internal): Ditto.
1591         (*zero_extendsidi2): Ditto.
1592         * config/i386/sse.md (vec_set<mode>_0): Ditto.
1593         (sse2_loadld): Ditto.
1594         (*vec_extract<ssevecmodelower>_0): Ditto.
1595         (*vec_extractv4si_0_zext_sse4): Ditto.
1596         (vec_concatv2di): Ditto.
1597         (*vec_dup<mode>): Ditto.
1598         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1599         * config/i386/constraints.md (Yi): Remove.
1600         (Yj): Remove.
1601         (Ym): Remove.
1602         (Yn): Remove.
1603
1604 2018-04-26  Nathan Sidwell  <nathan@acm.org>
1605
1606         * dumpfile.c (dump_open): New.
1607         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1608         (dump_finish): Detect stdio/stderr by value not name.
1609
1610 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
1611
1612         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1613
1614 2018-04-26  Tom de Vries  <tom@codesourcery.com>
1615
1616         PR target/84952
1617         * config/nvptx/nvptx.c (verify_neutering_jumps)
1618         (verify_neutering_labels): New function
1619         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1620
1621 2018-04-26  Tom de Vries  <tom@codesourcery.com>
1622
1623         PR target/84025
1624         * config/nvptx/nvptx.c (needs_neutering_p): New function.
1625         (nvptx_single): Use needs_neutering_p to skip over insns that do not
1626         need neutering.
1627
1628 2018-04-26  Richard Biener <rguenther@suse.de>
1629             Tom de Vries  <tom@codesourcery.com>
1630
1631         PR lto/85422
1632         * lto-streamer-out.c (output_function): Fixup loops if required to match
1633         discovery done in the reader.
1634
1635 2018-04-26  Richard Biener  <rguenther@suse.de>
1636
1637         PR tree-optimization/85116
1638         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1639         have a loop exit from the single latch predecessor.  Remove
1640         case of header with just condition.
1641         (ch_base::copy_headers): Exclude infinite loops from any
1642         processing.
1643         (pass_ch::execute): Record exits.
1644
1645 2018-04-26  Richard Biener  <rguenther@suse.de>
1646
1647         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1648         prologue cost vector and pass it to vect_get_load_cost.
1649         (vect_get_peeling_costs_all_drs): Likewise.
1650         (vect_peeling_hash_get_lowest_cost): Likewise.
1651         (vect_enhance_data_refs_alignment): Likewise.
1652
1653 2018-04-26  Richard Biener  <rguenther@suse.de>
1654
1655         PR middle-end/85450
1656         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1657         checking of integer<->pointer conversions.
1658         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1659         sign-/zero-extending pointer types.
1660         (expand_omp_for_static_chunk): Likewise.
1661
1662 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
1663             Jean Lee  <xiaoyur347@gmail.com>
1664
1665         * config/mips/mips.c (mips_asan_shadow_offset): New function.
1666         (TARGET_ASAN_SHADOW_OFFSET): Define.
1667         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1668         true for -fsanitize=address.
1669
1670 2018-04-25  Mark Wielaard  <mark@klomp.org>
1671
1672         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1673         shorter ones.
1674
1675 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
1676
1677         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1678         than "alu", remove explicit "memory" and "imm_disp" attributes.
1679         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1680
1681         PR middle-end/85414
1682         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1683         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1684         gen_lowpart_no_emit.
1685
1686 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
1687
1688         PR target/85473
1689         * config/i386/i386.c (ix86_expand_builtin): Change memory
1690         operand to XI, extend p0 to Pmode.
1691         * config/i386/i386.md: Change unspec volatile and operand
1692         1 mode to XI, change operand 0 mode to P.
1693
1694 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
1695
1696         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1697         GET_MODE_MASK before any checking.
1698         (nds32_can_use_bset_p): Likewise.
1699         (nds32_can_use_btgl_p): Likewise.
1700
1701 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
1702
1703         * config/nds32/nds32-doubleword.md: New define_split pattern for
1704         illegal register number.
1705
1706 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
1707
1708         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1709
1710 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
1711
1712         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1713
1714 2018-04-25  Richard Biener  <rguenther@suse.de>
1715
1716         * lto-streamer.h (LTO_major_version): Bump to 8.
1717
1718 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
1719
1720         * BASE-VER: Set to 9.0.0.
1721
1722 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
1723
1724         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1725         in __abskf2 and __powikf2.
1726
1727 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1728
1729         PR target/85512
1730         * config/aarch64/constraints.md (Usg, Usj): New constraints.
1731         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1732         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1733         Use the above on operand 2.  Reindent.
1734         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1735
1736 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
1737
1738         PR target/85485
1739         * common/config/i386/i386-common.c (ix86_handle_option): Don't
1740         handle OPT_mcet.
1741         * config/i386/i386.opt (mcet): Removed.
1742         * doc/install.texi: Remove -mcet documentation.
1743         * doc/invoke.texi: Likewise.
1744
1745 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
1746
1747         PR target/85485
1748         * doc/install.texi: Remove -mcet from bootstrap-cet.
1749
1750 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
1751
1752         PR target/85511
1753         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1754         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1755         if TARGET_64BIT.
1756
1757         PR target/85503
1758         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1759         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1760         containing a CONST_VECTOR.
1761
1762 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
1763
1764         * doc/install.texi: Update newlib dependency for nvptx.
1765
1766 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
1767
1768         PR target/85508
1769         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1770         instead of INTVAL when shifting x left.
1771
1772 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
1773
1774         PR tree-optimization/85478
1775         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1776         vect_grouped_store_supported for single element vectors.
1777
1778 2018-04-24  Richard Biener  <rguenther@suse.de>
1779
1780         PR target/85491
1781         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1782         load cost increase to the case of non-constant step.
1783
1784 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
1785
1786         PR target/84828
1787         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1788         destination if any_malformed_asm.
1789
1790 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
1791
1792         PR middle-end/85496
1793         * expr.c (store_field): In the bitfield case, if the value comes from
1794         a function call and is returned in registers by means of a PARALLEL,
1795         do not change the mode of the temporary unless BLKmode and VOIDmode.
1796
1797 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
1798
1799         PR rtl-optimization/85423
1800         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1801         dependencies to debug insns when the previous insn is non-debug.
1802
1803 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
1804
1805         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1806         enums into a single definition.
1807         (fls): Fix predicates and printing.
1808         (seti): Likewise.
1809
1810 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
1811
1812         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1813         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1814         and short u6 immediate.
1815         (check_if_valid_sleep_operand): Remove.
1816         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1817
1818 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1819
1820         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1821         flag_always_save_lp condition.
1822         * config/nds32/nds32.opt (malways-save-lp): New option.
1823
1824 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
1825
1826         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1827         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1828         * config/nds32/nds32.h
1829         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1830         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1831
1832 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
1833
1834         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1835         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1836
1837 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
1838             Chung-Ju Wu  <jasonwucj@gmail.com>
1839
1840         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1841         Declare.
1842         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1843         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1844
1845 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1846
1847         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1848
1849 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1850
1851         * config/nds32/nds32-protos.h (nds32_data_alignment,
1852         nds32_local_alignment): Declare.
1853         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1854         nds32_local_alignment): New functions.
1855         (TARGET_CONSTANT_ALIGNMENT): Define.
1856         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1857
1858 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1859
1860         * config/nds32/nds32.c
1861         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1862         (TARGET_MODES_TIEABLE_P): Likewise.
1863
1864 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1865
1866         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1867         level Ofast and Og.
1868
1869 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
1870             Chung-Ju Wu  <jasonwucj@gmail.com>
1871
1872         * config/nds32/constants.md (unspec_volatile_element): Add enum values
1873         for unaligned access.
1874         * config/nds32/nds32-intrinsic.c: Implementation of expanding
1875         unaligned access.
1876         * config/nds32/nds32-intrinsic.md: Likewise.
1877         * config/nds32/nds32_intrinsic.h: Likewise.
1878         * config/nds32/nds32.h (nds32_builtins): Likewise.
1879         * config/nds32/nds32.opt (munaligned-access): New option.
1880         * config/nds32/nds32.c (nds32_asm_file_start): Display
1881         flag_unaligned_access status.
1882
1883 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
1884
1885         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1886         -mno-relax is present.
1887         * config/riscv/linux.h (LINK_SPEC): Ditto.
1888
1889 2018-04-20  Martin Sebor  <msebor@redhat.com>
1890
1891         PR c/85365
1892         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1893         for null pointers.
1894         (gimple_fold_builtin_stxcpy_chk): Same.
1895         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1896
1897 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
1898
1899         PR target/85456
1900         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1901         __powikf2 when long double is IEEE 128-bit.
1902
1903 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
1904
1905         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1906         step to make sure stack always aligned.
1907
1908 2018-04-20  Carl Love  <cel@us.ibm.com>
1909
1910         PR target/83402
1911         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1912         size check for arg0.
1913
1914 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
1915             Tom de Vries  <tom@codesourcery.com>
1916
1917         PR target/85445
1918         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1919         Emit insns for calls too.
1920         (nvptx_find_par): Always look for worker-level predecessor insn.
1921         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
1922         calls.
1923         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1924         (nvptx_process_pars): Propagate frames for calls.
1925
1926 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
1927
1928         PR target/85469
1929         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1930         Removed.
1931         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1932         (ix86_handle_option): Don't handle OPT_mibt.
1933         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1934         __SHSTK__.
1935         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1936         has_ibt and ibt.
1937         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1938         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1939         (ix86_target_macros): Define __CET__ with flag_cf_protection
1940         for -fcf-protection.
1941         * config/i386/i386.c (isa2_opts): Remove -mibt.
1942         * config/i386/i386.h (TARGET_IBT): Removed.
1943         (TARGET_IBT_P): Likewise.
1944         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1945         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1946         * config/i386/i386.opt (mcet): Update help message.
1947         (mshstk): Likewise.
1948         (mibt): Removed.
1949         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
1950         -mcet as an alias for -mshstk.
1951
1952 2018-04-20  Richard Biener <rguenther@suse.de>
1953
1954         PR middle-end/85475
1955         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1956         complexity by forcing a single use of the multiply operand.
1957
1958 2018-04-20  Martin Jambor  <mjambor@suse.cz>
1959
1960         ipa/85449
1961         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1962         recursion dependency to only apply to non-clones.
1963
1964 2018-04-20  Martin Jambor  <mjambor@suse.cz>
1965
1966         ipa/85447
1967         * ipa-cp.c (create_specialized_node): Check that clones of
1968         self-recursive edges exist during IPA-CP.
1969
1970 2018-04-19  Toon Moene  <toon@moene.org>
1971
1972         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1973         by -O3.
1974
1975 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
1976
1977         PR tree-optimization/85467
1978         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1979         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
1980         VECTOR_CST element to type.
1981
1982 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
1983
1984         PR target/85397
1985         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1986         * config/i386/i386.md (builtin_setjmp_setup): Removed.
1987         (builtin_longjmp): Likewise.
1988         (save_stack_nonlocal): New pattern.
1989         (restore_stack_nonlocal): Likewise.
1990
1991 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
1992
1993         PR target/85404
1994         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1995         Replace ASM_OUTPUT_LABEL with fprintf.
1996
1997 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
1998
1999         PR target/85417
2000         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2001         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2002         * config/i386/i386-c.c (ix86_target_macros_internal): Also
2003         define __IBT__ and __SHSTK__ for -fcf-protection.
2004         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2005         TARGET_IBT.
2006         (ix86_trampoline_init): Likewise.
2007         (x86_output_mi_thunk): Likewise.
2008         (ix86_notrack_prefixed_insn_p): Likewise.
2009         (ix86_option_override_internal): Don't disallow -fcf-protection.
2010         * config/i386/i386.md (rdssp<mode>): Also enable for
2011         -fcf-protection.
2012         (incssp<mode>): Likewise.
2013         (nop_endbr): Likewise.
2014         * config/i386/i386.opt (mcet): Change help message to built-in
2015         functions only.
2016         (mibt): Likewise.
2017         (mshstk): Likewise.
2018         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2019         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
2020         enable CET built-in functions.
2021
2022 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
2023
2024         * common/config/i386/i386-common.c
2025         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2026         OPTION_MASK_ISA_MOVDIRI_UNSET,
2027         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2028         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2029         * config.gcc (movdirintrin.h): New header.
2030         * config/i386/cpuid.h (bit_MOVDIRI,
2031         bit_MOVDIR64B): New bits.
2032         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2033         and -mmvodir64b.
2034         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2035         (VOID, PVOID, PCVOID)): New function types.
2036         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2037         __builtin_ia32_directstoreu_u64,
2038          __builtin_ia32_movdir64b): New builtins.
2039         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2040         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2041         and -mmovdiri.
2042         (ix86_valid_target_attribute_inner_p): Ditto.
2043         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2044         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2045         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2046         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2047         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2048         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2049         (movdiri<mode>, movdir64b_<mode>): New.
2050         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2051         * config/i386/immintrin.h: Include movdirintrin.h.
2052         * config/i386/movdirintrin.h: New file.
2053         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2054
2055 2018-04-19  Richard Biener  <rguenther@suse.de>
2056
2057         PR middle-end/85455
2058         * cfg.c (clear_bb_flags): When loop state says we have
2059         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2060
2061 2018-04-19  Richard Biener  <rguenther@suse.de>
2062
2063         PR tree-optimization/84737
2064         * tree-vect-data-refs.c (vect_copy_ref_info): New function
2065         copying restrict info.
2066         (vect_setup_realignment): Use it.
2067         * tree-vectorizer.h (vect_copy_ref_info): Declare.
2068         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2069         the first DR to all generated stores.
2070         (vectorizable_load): Likewise for loads.
2071
2072 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
2073
2074         PR tree-optimization/85446
2075         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2076         the integral and pointer types to have the same precision.
2077
2078         * doc/install.texi: Document --disable-cet being the default and
2079         --enable-cet=auto.
2080
2081 2018-04-18  Martin Liska  <mliska@suse.cz>
2082
2083         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2084         style.
2085
2086 2018-04-18  Martin Liska  <mliska@suse.cz>
2087
2088         Revert
2089         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
2090
2091         PR ipa/83983
2092         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2093         arguments if they are comparable.
2094
2095 2018-04-18  Martin Liska  <mliska@suse.cz>
2096
2097         Revert
2098         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
2099
2100         PR lto/84805
2101         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2102         incomplete types.
2103
2104 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
2105
2106         PR target/85388
2107         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2108         ENDBR after calling __morestack.
2109
2110 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
2111
2112         PR jit/85384
2113         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2114         by using gcc_base_ver to generate a gcc_driver_version, and use
2115         it when generating GCC_DRIVER_NAME.
2116         * configure: Regenerate.
2117
2118 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
2119
2120         PR target/81084
2121         * config.gcc: Obsolete powerpc*-*-*spe*.
2122
2123 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
2124
2125         PR debug/84637
2126         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2127         (stabstr_D): Change type of unum from unsigned int to
2128         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
2129         type.
2130
2131 2018-04-17  Jim Wilson  <jimw@sifive.com>
2132
2133         PR 84856
2134         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2135         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2136         Set arg_pointer_offset after using pretend_args_size.
2137
2138 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
2139
2140         PR rtl-optimization/85431
2141         * dse.c (record_store): Ignore zero width stores.
2142
2143         PR sanitizer/85230
2144         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
2145         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2146         __builtin_stack_restore rather than after it.
2147         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2148         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2149         argument instead of virtual_dynamic_stack_rtx.
2150
2151 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2152
2153         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2154         New prototype.
2155         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2156         Add note to error message to explain internal mapping of overloaded
2157         built-in function name to non-overloaded built-in function name.
2158         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2159         function.
2160
2161 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
2162
2163         PR target/85424
2164         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2165         where the inputs overlap with the output.
2166
2167 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
2168
2169         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2170         (=v, v) alternative and explicit "memory" attribute.
2171         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
2172         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2173         attributes.
2174         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2175         "sselog1" type instead of "sselog".
2176         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2177         "sselog".  Remove explicit "memory" attribute.
2178         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2179         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2180         attributes.
2181         (vec_extract_hi_v32hi): Merge all alternatives into one, use
2182         "sselog1" type instead of "sselog".  Remove explicit "memory"
2183         attribute.
2184         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2185         use "sselog1" type instead of "sselog".  Remove explicit "memory"
2186         attribute.
2187         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2188         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2189         attributes.
2190         (vec_extract_hi_v64qi): Merge all alternatives into one, use
2191         "sselog1" type instead of "sselog".  Remove explicit "memory"
2192         attribute.
2193         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2194         use "sselog1" type instead of "sselog".  Remove explicit "memory"
2195         attribute.
2196
2197         PR target/85430
2198         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2199
2200         PR middle-end/85414
2201         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2202         on a SUBREG.
2203
2204 2018-04-17  Martin Jambor  <mjambor@suse.cz>
2205
2206         PR ipa/85421
2207         * ipa-cp.c (create_specialized_node): Call
2208         expand_all_artificial_thunks if necessary.
2209
2210 2018-04-17  Martin Liska  <mliska@suse.cz>
2211
2212         PR lto/85405
2213         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2214         in message, remote space in between '_G' and '('.
2215
2216 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
2217
2218         PR target/85281
2219         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2220         avx512f_vmcmp<mode>3<round_saeonly_name>,
2221         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2222         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2223         avx512f_rndscale<mode><round_saeonly_name>,
2224         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2225         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2226         Use %<iptr>2 instead of %2 for -masm=intel.
2227         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2228         avx512f_vcvttss2usi<round_saeonly_name>,
2229         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2230         -masm=intel.
2231         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2232         avx512f_vcvttsd2usi<round_saeonly_name>,
2233         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2234         Use %q1 instead of %1 for -masm=intel.
2235         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2236         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2237         of %3 for -masm=intel.
2238         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2239         -masm=intel.
2240         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2241         -masm=intel.
2242         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2243         -masm=intel.
2244         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2245         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2246         %g1.
2247         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2248         -masm=intel.
2249         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2250         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2251         %g1 and one with %0 and %1.
2252         (avx512er_vmrcp28<mode><round_saeonly_name>,
2253         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2254         %1 for -masm=intel.
2255         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2256         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2257         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2258         of %0 and %{%4%} for -masm=intel.
2259         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2260         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2261         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2262         order of %0 and %{%5%}%{z%} for -masm=intel.
2263
2264 2018-04-17  Jan Hubicka  <jh@suse.cz>
2265
2266         PR lto/85405
2267         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2268
2269 2018-04-17  Martin Liska  <mliska@suse.cz>
2270
2271         PR ipa/85329
2272         * multiple_target.c (create_dispatcher_calls): Set apostrophes
2273         for target_clone error message.  Make default implementation
2274         clone to be a local declaration.
2275         (separate_attrs): Add new argument and check for an empty
2276         string.
2277         (expand_target_clones): Handle it.
2278         (ipa_target_clone): Make redirection just for target_clones
2279         functions.
2280
2281 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
2282             Tom de Vries  <tom@codesourcery.com>
2283
2284         PR middle-end/84955
2285         * omp-expand.c (expand_oacc_for): Add dummy false branch for
2286         tiled basic blocks without omp continue statements.
2287
2288 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
2289
2290         PR target/83660
2291         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2292         vec_extract expression as having side effects to make sure it gets
2293         a cleanup point.
2294
2295 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
2296
2297         PR target/85403
2298         * config/i386/i386.c (get_builtin_code_for_version): Check
2299         error_mark_node.
2300
2301 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
2302
2303         PR target/84331
2304         * gcc/config.gcc: Support "skylake".
2305         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2306         PROCESSOR_SKYLAKE.
2307         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2308         (processor_target_table): Add "skylake".
2309         (ix86_option_override_internal): Add "skylake".
2310         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2311         PROCESSOR_CANNONLAKE.
2312         (get_builtin_code_for_version): Fix priority for
2313         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2314         PROCESSOR_SKYLAKE-AVX512.
2315         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2316         (processor_type): Add PROCESSOR_SKYLAKE.
2317
2318 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
2319             Jason Merrill  <jason@redhat.com>
2320
2321         PR c++/85112
2322         * convert.c (convert_to_integer_1): Use direct recursion for
2323         enumeral types and types with a precision less than the number
2324         of bits in their mode.
2325
2326 2018-04-16  Julia Koval  <julia.koval@intel.com>
2327
2328         PR target/84413
2329         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2330         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2331
2332 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
2333
2334         PR target/85293
2335         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2336         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2337         and -mno-direct-move.
2338
2339 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
2340
2341         PR target/83402
2342         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2343         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2344         Ensure negative shifts result in {0}.
2345
2346 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
2347
2348         PR rtl-optimization/79916
2349         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2350         regs (if any) to define how to gnerate SD moves when LRA is in
2351         progress.
2352
2353 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
2354
2355         PR rtl-optimization/85393
2356         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2357         * except.c (expand_dw2_landing_pad_for_region): Make static.
2358         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2359         a label and unconditional jump to old_bb, rather than
2360         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2361         basic block.
2362
2363         PR rtl-optimization/85376
2364         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2365         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2366         instead of a specific value.
2367
2368 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
2369             Bin Cheng  <bin.cheng@arm.com>
2370
2371         PR tree-optimization/82965
2372         PR tree-optimization/83991
2373         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2374         by_profile_only parameter.
2375         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2376         information if the loop was predicted to iterate too many times.
2377         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2378
2379 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
2380
2381         PR lto/71991
2382         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2383         always inline.
2384
2385 2018-04-13  Martin Liska  <mliska@suse.cz>
2386             Jakub Jelinek  <jakub@redhat.com>
2387
2388         PR middle-end/81657
2389         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2390         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2391         * builtins.c (expand_builtin_memory_copy_args): Use
2392         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2393         handle dest_addr == pc_rtx.
2394
2395 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
2396
2397         PR target/85291
2398         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2399         asked to not generate direct moves.
2400         (fix_trunc<mode>si2_stfiwx): Similar.
2401         (fix_trunc<mode>si2_internal): Similar.
2402
2403 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
2404
2405         PR debug/83157
2406         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2407         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2408         lookup if dest in some wider mode is known to be const0_rtx and
2409         if so, record permanent equivalence for it to be ZERO_EXTEND of
2410         the narrower mode destination.
2411
2412 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
2413
2414         * lto-streamer-out.c (output_function): Revert 259346.
2415         * omp-expand.c (expand_oacc_for): Likewise.
2416
2417 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
2418
2419         PR rtl-optimization/85354
2420         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2421         * sel-sched.c (sel_global_init): ... here.
2422
2423 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
2424
2425         PR target/85238
2426         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2427         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2428         mode for PE-COFF targets.
2429         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2430         (i386_pe_asm_lto_end): Likewise.
2431         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2432         (TARGET_ASM_LTO_END): Likewise.
2433         * config/i386/winnt.c (saved_debug_info_level): New static variable.
2434         (i386_pe_asm_lto_start): New function.
2435         (i386_pe_asm_lto_end): Likewise.
2436
2437 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
2438             Richard Biener  <rguenther@suse.de>
2439
2440         PR middle-end/84955
2441         * lto-streamer-out.c (output_function): Fix CFG loop state before
2442         streaming out.
2443         * omp-expand.c (expand_oacc_for): Handle calls to internal
2444         functions like regular functions.
2445
2446 2018-04-12  Richard Biener  <rguenther@suse.de>
2447
2448         PR lto/85371
2449         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2450         for the early LTO debug to properly generate references to it
2451         during DIE emission.  Do not re-use that for the skeleton for
2452         split-dwarf.
2453         (dwarf2out_early_finish): Likewise.
2454
2455 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
2456
2457         PR target/85328
2458         * config/i386/sse.md
2459         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2460         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2461         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2462         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2463         and output is a reg, avoid creating invalid lowpart subreg, but
2464         instead split into a 512-bit move.  Don't split if not AVX512VL,
2465         input is xmm16+ reg and output is a mem.
2466         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2467         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2468         xmm16+ reg and output is a mem.
2469
2470 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
2471
2472         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2473         also for flag_dwarf2_cfi_asm.
2474
2475 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
2476
2477         PR rtl-optimization/85342
2478         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2479         a bool scalar var inside of the loop instead.  Don't try to update
2480         recog_data.operand after failed apply_change_group.
2481
2482 2018-04-12  Tom de Vries  <tom@codesourcery.com>
2483
2484         PR target/85296
2485         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2486         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
2487         array with flexible array member as array without given dimension.
2488         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2489         argument for undefined param to true.
2490
2491 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
2492
2493         PR target/85321
2494         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2495         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2496         from PowerPC section.
2497         * config/rs6000/sysv4.opt (mcall-): Improve help text.
2498         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2499         help text that is too long.
2500         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2501         help text that is too long.
2502         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2503         help text that is too long.
2504
2505 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
2506
2507         * config/alpha/alpha.md (stack_probe_internal): Rename
2508         from "probe_stack".  Update all callers.
2509
2510 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
2511
2512         PR rtl-optimization/84566
2513         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2514         sched_macro_fuse_insns.
2515
2516 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
2517
2518         PR target/84301
2519         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2520         (compute_block_dependences): ... from here.
2521
2522 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
2523
2524         PR tree-optimization/85331
2525         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2526         from int to HOST_WIDE_INT.
2527
2528 2018-04-11  Martin Jambor  <mjambor@suse.cz>
2529
2530         PR ipa/84149
2531         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2532         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2533         not the same as the source val.
2534         (cgraph_edge_brings_value_p): New parameter.
2535         (gather_edges_for_value): Pass destination value to
2536         cgraph_edge_brings_value_p.
2537         (perhaps_add_new_callers): Likewise.
2538         (get_info_about_necessary_edges): Likewise and exclude values brought
2539         only by self-recursive edges.
2540         (create_specialized_node): Redirect only clones of self-calling edges.
2541         (+self_recursive_pass_through_p): New function.
2542         (find_more_scalar_values_for_callers_subset): Use it.
2543         (find_aggregate_values_for_callers_subset): Likewise.
2544         (known_aggs_to_agg_replacement_list): Removed.
2545         (decide_whether_version_node): Re-calculate known constants for all
2546         remaining context clones.
2547
2548 2018-04-11  Richard Biener  <rguenther@suse.de>
2549
2550         PR lto/85339
2551         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2552         from early DWARF output.
2553         (dwarf2out_early_finish): Output line info unconditionally into
2554         early DWARF and add reference to it.
2555
2556 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
2557
2558         PR target/85281
2559         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2560         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2561         other than V2DFmode using iptr mode attribute.
2562         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2563
2564 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
2565
2566         PR rtl-optimization/84659
2567         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2568
2569 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
2570
2571         PR debug/85302
2572         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2573         SIZEP is NULL.
2574         (output_loc_list): Pass address of a dummy size variable even in the
2575         locview handling loop.
2576         (index_location_lists): Add comment on why skip_loc_list_entry can't
2577         call size_of_locs.
2578
2579 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
2580
2581         PR target/85261
2582         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2583         into register.
2584
2585 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
2586
2587         PR target/85321
2588         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2589         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2590         and -mstring-compare-inline-limit.
2591
2592 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
2593
2594         PR target/85287
2595         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2596         for stack clash protection in a register whenever we need it to be in
2597         a register.
2598
2599 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
2600
2601         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2602         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2603
2604 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
2605
2606         PR target/85321
2607         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2608         the help text.
2609         (mlong-double-): Ditto.
2610         * config/rs6000/sysv4.opt (msdata=): Ditto.
2611         (mtls-size=): Ditto.
2612
2613 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2614
2615         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2616         erroneous entries for
2617         "vector int vec_ldl (int, long int *)", and
2618         "vector unsigned int vec_ldl (int, unsigned long int *)".
2619         Add comments and entries for
2620         "vector bool char vec_ldl (int, bool char *)",
2621         "vector bool short vec_ldl (int, bool short *)",
2622         "vector bool int vec_ldl (int, bool int *)",
2623         "vector bool long long vec_ldl (int, bool long long *)",
2624         "vector pixel vec_ldl (int, pixel *)",
2625         "vector long long vec_ldl (int, long long *)",
2626         "vector unsigned long long vec_ldl (int, unsigned long long *)".
2627         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2628         type tree bool_long_long_type_node and correct definition of
2629         bool_V2DI_type_node to make reference to this new type tree.
2630         (rs6000_mangle_type): Replace erroneous reference to
2631         bool_long_type_node with bool_long_long_type_node.
2632         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2633         comments to emphasize sign distinctions for char and int types and
2634         replace RS6000_BTI_bool_long constant with
2635         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
2636         use of RS6000_BTI_pixel.
2637         (bool_long_type_node): Remove this macro definition.
2638         (bool_long_long_type_node): New macro definition
2639
2640 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
2641
2642         PR rtl-optimization/85300
2643         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2644         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2645         simplify_unary_operation fails.
2646
2647 2018-04-10  Martin Liska  <mliska@suse.cz>
2648
2649         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2650         cgraph_edge and ipa_ref.
2651
2652 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
2653
2654         PR target/85177
2655         PR target/85255
2656         * config/i386/sse.md
2657         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2658         computation of the VEC_MERGE selector from mask.
2659         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2660         Fix decoding of the VEC_MERGE selector into mask.
2661
2662 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
2663
2664         PR tree-optimization/85286
2665         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2666
2667 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
2668
2669         * final.c (final_1): Set insn_last_address as well as
2670         insn_current_address.
2671
2672 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2673
2674         PR target/85173
2675         * explow.c (emit_stack_probe): Call validize_mem on memory location
2676         before passing it to gen_probe_stack.  Create address operand and
2677         legitimize it for the probe_stack_address case.
2678
2679 2018-04-09  Jan Hubicka  <jh@suse.cz>
2680
2681         PR lto/85078
2682         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2683         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2684         * tree.c (free_lang_data_in_type): Fix handling of binfos;
2685         walk basetypes.
2686         (free_lang_data): Rebuild type inheritance graph.
2687
2688 2018-04-09  Martin Sebor  <msebor@redhat.com>
2689
2690         * invoke.texi (-finline-small-functions): Mention other optimization
2691         options.
2692         (-findirect-inlining, -fpartial-inlining): Same.
2693         (-finline-functions-called-once): Same.
2694         (-freorder-blocks-and-partition): Same.
2695
2696 2018-04-09  Jan Hubicka  <jh@suse.cz>
2697
2698         PR rtl/84058
2699         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2700         jumps; choose last target that matches the criteria (i.e.
2701         no partition changes for non-crossing jumps).
2702         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2703         support for redirecting crossing jumps to non-crossing.
2704
2705 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
2706
2707         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2708         also for naked functions.
2709
2710 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
2711
2712         * config/arc/arc.md (add_shift): New pattern.
2713         (add_shift2): Likewise.
2714         (sub_shift): Likewise.
2715         (sub_shift_cmp0_noout): Likewise.
2716         (compare_si_ashiftsi): Likewise.
2717         (xbfu_cmp0_noout): New combine pattern.
2718         (xbfu_cmp0"): Likewise.
2719         (movsi_set_cc_insn): Place the predicable variant first.
2720         (commutative_binary_cmp0_noout): Remove clobber.
2721         (commutative_binary_cmp0): New pattern.
2722         (noncommutative_binary_cmp0): Likewise.
2723         (noncommutative_binary_cmp0_noout): Likewise.
2724         (noncommutative_binary_comparison_result_used): Removed.
2725         (rsub_cmp0): New pattern.
2726         (rsub_cmp0_noout): Likewise.
2727         (extzvsi): Changed, keep only meaningful variants.
2728         (SQH, SEZ): New iterators.
2729         (SQH_postfix): New mode attribute.
2730         (SEZ_prefix): New code attribute.
2731         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2732         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2733         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2734         of numerical value.
2735         (noncommutative_operator): Check the availability of barrel
2736         shifter option.
2737
2738 2018-04-09  Richard Biener  <rguenther@suse.de>
2739
2740         PR tree-optimization/85284
2741         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2742         Only use the niter constraining form of simple_iv when the exit
2743         is always executed.
2744
2745 2018-04-09  Tom de Vries  <tom@codesourcery.com>
2746
2747         PR target/84041
2748         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2749         (define_expand "*memory_barrier"): New define_expand.
2750         (define_insn "memory_barrier"): New insn.
2751
2752 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
2753
2754         PR rtl-optimization/80463
2755         PR rtl-optimization/83972
2756         PR rtl-optimization/83480
2757
2758         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2759         correct producer for the insn.
2760         (tidy_control_flow): Fixup seqnos in case of debug insns.
2761
2762 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
2763
2764         PR rtl-optimization/83913
2765
2766         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2767         different sched-times when merging exprs.
2768
2769 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
2770
2771         PR rtl-optimization/83962
2772
2773         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2774         tidy_fallthru_edge and tidy_control_flow.
2775
2776 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
2777
2778         PR rtl-optimization/83530
2779
2780         * sel-sched.c (force_next_insn): New global variable.
2781         (remove_insn_for_debug): When force_next_insn is true, also leave only
2782         next insn in the ready list.
2783         (sel_sched_region): When the region wasn't scheduled, make another pass
2784         over it with force_next_insn set to 1.
2785
2786 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
2787
2788         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2789         into tm_file.
2790         * config/nds32/constants.md (unspec_volatile_element): Add enum values
2791         for interrupt control.
2792         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2793         functions for interrupt control.
2794         * config/nds32/nds32-intrinsic.md: Likewise.
2795         * config/nds32/nds32_intrinsic.h: Likewise.
2796         * config/nds32/nds32.h (nds32_builtins): Likewise.
2797
2798 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
2799
2800         * config/nds32/nds32.c (nds32_init_machine_status,
2801         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2802         strict_aligned_p field.
2803         (nds32_expand_to_rtl_hook): New function.
2804         (TARGET_EXPAND_TO_RTL_HOOK): Define.
2805         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2806
2807 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
2808             Chung-Ju Wu  <jasonwucj@gmail.com>
2809
2810         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2811         * config/nds32/nds32-n7.md: New file.
2812         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2813         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2814         pipeline.
2815         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2816         * config/nds32/nds32.md (pipeline_model): Add n7.
2817         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2818         * config/nds32/pipelines.md: Include n7 settings.
2819
2820 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
2821             Chung-Ju Wu  <jasonwucj@gmail.com>
2822
2823         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2824         * config/nds32/nds32-e8.md: New file.
2825         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2826         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2827         pipeline.
2828         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2829         * config/nds32/nds32.md (pipeline_model): Add e8.
2830         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2831         * config/nds32/pipelines.md: Include e8 settings.
2832
2833 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
2834             Chung-Ju Wu  <jasonwucj@gmail.com>
2835
2836         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2837         * config/nds32/nds32-n8.md: New file.
2838         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2839         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2840         pipeline.
2841         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2842         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2843         * config/nds32/nds32.md (pipeline_model): Add n8.
2844         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2845         * config/nds32/pipelines.md: Include n8 settings.
2846
2847 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
2848             Chung-Ju Wu  <jasonwucj@gmail.com>
2849
2850         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2851         * config/nds32/nds32-n9-2r1w.md: New file.
2852         * config/nds32/nds32-n9-3r2w.md: New file.
2853         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2854         nds32_register_ports): New or modify for cpu n9.
2855         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2856         pipeline.
2857         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2858         * config/nds32/nds32-utils.c: New file.
2859         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2860         TARGET_MUL_SLOW): Define.
2861         * config/nds32/nds32.md (pipeline_model): New attribute.
2862         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2863         New options that support cpu n9.
2864         * config/nds32/pipelines.md: Include n9 settings.
2865         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2866
2867 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
2868
2869         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2870         information if necessary.
2871         (output_cond_branch_compare_zero): Likewise.
2872         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2873         (nds32_target_alignment): Refine for alignment.
2874         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2875         (FUNCTION_BOUNDARY): Modify.
2876         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2877         align case.
2878         * config/nds32/nds32.opt (malways-align, malign-functions): New.
2879
2880 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
2881
2882         * config/nds32/constants.md (unspec_volatile_element): Add values for
2883         TLB operation and data prefetch.
2884         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2885         functions for TLB operation and data prefetch.
2886         * config/nds32/nds32-intrinsic.md: Likewise.
2887         * config/nds32/nds32_intrinsic.h: Likewise.
2888         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2889         (nds32_print_operand): Likewise.
2890         * config/nds32/nds32.h (nds32_builtins): Likewise.
2891
2892 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
2893         Andrew Pinski <pinsika@gcc.gnu.org>
2894
2895         PR middle-end/82976
2896         * match.pd: Use constant_boolean_node of correct type instead of
2897         boolean_true_node or boolean_false_node for simplifying
2898         pointer comparisons to zero.
2899
2900 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
2901
2902         PR tree-optimization/80021
2903         * tree.c (verify_type_variant): Make error call in verify_variant_match
2904         translatable and remove final full stop.
2905
2906 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
2907
2908         * config/nds32/constants.md (unspec_volatile_element): Add
2909         UNSPEC_VOLATILE_EH_RETURN.
2910         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2911         nds32_output_stack_pop): Support dwarf exception handling process.
2912         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2913         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2914         exception handling process.
2915         (nds32_compute_stack_frame): Likewise.
2916         (nds32_return_addr_rtx): Likewise.
2917         (nds32_initial_elimination_offset): Likewise.
2918         (nds32_expand_prologue): Likewise.
2919         (nds32_expand_epilogue): Likewise.
2920         (nds32_dynamic_chain_address): New function.
2921         * config/nds32/nds32.h (machine_function): Add fields for dwarf
2922         exception handling.
2923         (DYNAMIC_CHAIN_ADDRESS): Define.
2924         (EH_RETURN_DATA_REGNO): Define.
2925         (EH_RETURN_STACKADJ_RTX): Define.
2926         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2927         patterns for dwarf exception handling.
2928
2929 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
2930
2931         * config/nds32/nds32.h: Clean up obsolete macros.
2932
2933 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
2934
2935         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2936         Add enum values for particular instructions.
2937         * config/nds32/nds32-intrinsic.c: Implementation of expanding
2938         particular intrinsic functions.
2939         * config/nds32/nds32-intrinsic.md: Likewise.
2940         * config/nds32/nds32_intrinsic.h: Likewise.
2941         * config/nds32/nds32.h (nds32_builtins): Likewise.
2942         * config/nds32/nds32.md (type): Add pbsad and pbsada.
2943         (btst, ave): New patterns for particular instructions.
2944
2945 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
2946
2947         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2948         Add enum values for atomic load/store and memory sync.
2949         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2950         and memory sync.
2951         * config/nds32/nds32-intrinsic.md: Likewise.
2952         * config/nds32/nds32_intrinsic.h: Likewise.
2953         * config/nds32/nds32.h (nds32_builtins): Likewise.
2954
2955 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
2956
2957         PR tree-optimization/85257
2958         * fold-const.c (native_encode_vector): If not all elts could fit
2959         and off is -1, return 0 rather than offset.
2960         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2961         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2962         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
2963         adjust buffer in native_interpret_expr call.
2964
2965 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
2966
2967         * config/nds32/constants.md (unspec_volatile_element): Add cache
2968         control enum values.
2969         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2970         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2971         * config/nds32/nds32.c (nds32_cctl_names): New.
2972         (nds32_print_operand): Handle cache control register names.
2973         * config/nds32/nds32.h (nds32_builtins): New enum values.
2974         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2975         macros.
2976         * config/nds32/nds32.md (type): Add mmu.
2977         * config/nds32/pipelines.md (simple_insn): Add mmu.
2978
2979 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
2980
2981         * config/nds32/nds32.md (type): Remove call.
2982         * config/nds32/pipelines.md (simple_insn): Likewise.
2983
2984 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
2985
2986         * config/nds32/constants.md (unspec_volatile_element): Add
2987         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2988         UNSPEC_VOLATILE_FMFCFG.
2989         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2990         description for fmfcfg and fmfcsr.
2991         (bdesc_1arg): Add fmtcsr.
2992         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2993         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2994         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2995         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2996         unspec_fmfcfg): New patterns.
2997         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2998         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2999         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3000         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3001         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3002         __nds32__fmfcfg): Define.
3003
3004 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
3005
3006         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3007         intrinsic register names.
3008         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3009         intrinsic register enum values and macros.
3010
3011 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
3012
3013         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3014         for load/store addressing form.
3015         (nds32_print_operand_address): Likewise.
3016
3017 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
3018
3019         PR target/85196
3020         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3021         based on LABEL_REF.  Remove useless assertion.
3022         (pic_address_needs_scratch): Fix formatting.
3023         (sparc_legitimize_pic_address): Minor tweaks.
3024         (sparc_delegitimize_address): Adjust assertion accordingly.
3025         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3026         into symbolic_operand.
3027         (movsi_high_pic_label_ref): Likewise.
3028         (movsi_lo_sum_pic_label_ref): Likewise.
3029         (movdi_pic_label_ref): Likewise.
3030         (movdi_high_pic_label_ref): Likewise.
3031         (movdi_lo_sum_pic_label_ref): Likewise.
3032
3033 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
3034
3035         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3036         custom LIB_SPEC setup.
3037
3038 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
3039             Kito Cheng  <kito.cheng@gmail.com>
3040
3041         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3042         * config/riscv/freebsd.h: New.
3043
3044 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
3045
3046         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3047         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3048         file.
3049
3050 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
3051             Kito Cheng  <kito.cheng@gmail.com>
3052
3053         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3054         nds32_output_call, nds32_symbol_binds_local_p): New functions.
3055         * config/nds32/nds32-protos.h (nds32_output_call,
3056         nds32_output_return): Declare.
3057         * config/nds32/nds32.md: Refine all the call and return patterns.
3058
3059 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
3060
3061         PR debug/85252
3062         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3063         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3064
3065         PR rtl-optimization/84872
3066         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3067         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3068         EDGE_CROSSING edge.
3069
3070 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
3071
3072         * expr.c (copy_blkmode_to_reg): Revert 254862.
3073         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3074
3075 2018-04-06  Richard Biener  <rguenther@suse.de>
3076
3077         PR middle-end/85244
3078         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3079         after seeing a component reference with an adjacent field.  Treat
3080         refs to arrays at struct end of external decls similar to
3081         refs to unconstrained commons.
3082
3083 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
3084
3085         PR sanitizer/85213
3086         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3087         look through SAVE_EXPRs with non-side-effects argument.  Adjust
3088         recursive calls.
3089         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3090         save_p here.
3091
3092 2018-04-06  Richard Biener  <rguenther@suse.de>
3093
3094         PR middle-end/85180
3095         * alias.c (find_base_term): New wrapper around find_base_term
3096         unwinding CSELIB_VAL_PTR changes.
3097         (find_base_term): Do not restore CSELIB_VAL_PTR during the
3098         recursion.
3099
3100 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3101
3102         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3103         instructions.
3104         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3105         constant definitions.
3106         ("nop"): lr 0,0 -> nopr r0
3107         ("nop_lr0", "nop_lr1"): New insn definitions.
3108
3109 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
3110
3111         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3112         NDS32_V3PUSH_AVAILABLE_P macro.
3113
3114 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
3115             Chung-Ju Wu  <jasonwucj@gmail.com>
3116
3117         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3118         (nds32*-*-*): Add float and fpu_config into supported_defaults.
3119         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3120         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3121         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3122         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3123         * config/nds32/constraints.md: New constraints and checking for hard
3124         float configuration.
3125         * config/nds32/iterators.md: New mode iterator and attribute for hard
3126         float configuration.
3127         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3128         patterns.
3129         * config/nds32/nds32-fpu.md: New file.
3130         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3131         deal with hard float code generation.
3132         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3133         ARCH_V3S.
3134         (abi_type, float_reg_number): New enum type.
3135         * config/nds32/nds32-predicates.c: New predicates for hard float.
3136         * config/nds32/nds32-protos.h: Declare functions for hard float.
3137         * config/nds32/nds32.c: Implementation for hard float configuration.
3138         * config/nds32/nds32.h: Definitions for hard float configuration.
3139         * config/nds32/nds32.md: Include hard float machine description and
3140         modify patterns for hard float configuration.
3141         * config/nds32/nds32.opt: New options for hard float configuration.
3142         * config/nds32/predicates.md: New predicates for hard float
3143         configuration.
3144
3145 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
3146
3147         * common/config/nds32/nds32-common.c
3148         (nds32_option_optimization_table): Enable -mreleax-hint by default.
3149
3150 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
3151
3152         PR middle-end/85195
3153         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3154         CONSTRUCTOR_ELT (ctor, ...)->value.
3155
3156 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
3157
3158         PR target/85193
3159         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3160
3161 2018-04-05  Tom de Vries  <tom@codesourcery.com>
3162
3163         PR target/85204
3164         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3165         cond jump.
3166
3167 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
3168             Kito Cheng  <kito.cheng@gmail.com>
3169
3170         * config/nds32/constraints.md (U33): Fine-tune checking condition.
3171         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3172         * config/nds32/nds32.h (nds32_16bit_address_type): Add
3173         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3174
3175 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
3176             Kito Cheng  <kito.cheng@gmail.com>
3177
3178         * config/nds32/constraints.md (Ufe): New memory constraint.
3179         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3180         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3181         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3182         operands.
3183         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3184         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3185
3186 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3187
3188         * config/nds32/nds32.md: Use optimize_size in the condition for
3189         alu-shift instructions.
3190
3191 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3192
3193         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3194
3195 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3196
3197         * config/nds32/nds32.md (negsi2): Refine pattern.
3198
3199 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
3200             Chung-Ju Wu  <jasonwucj@gmail.com>
3201
3202         * config/nds32/iterators.md (shift_rotate): New code iterator.
3203         (shift): New code attribute.
3204         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3205         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3206         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3207         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3208         bit-wise operations.
3209         (andsi3, *andsi3): Ditto.
3210         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3211         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3212         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3213         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3214         nds32_ior_operand, nds32_xor_operand): New predicates.
3215
3216 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3217
3218         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3219         (addsi3, subsi3): ... this.
3220
3221 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3222
3223         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3224
3225 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3226
3227         * config/nds32/nds32.md: Adjust indention.
3228
3229 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
3230
3231         * config/nds32/nds32.md (feature): New attribute.
3232
3233 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3234
3235         * config/nds32/nds32.md (subtype): New attribute.
3236
3237 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
3238
3239         PR target/85203
3240         * config/arm/arm-builtins.c (arm_expand_builtin): Change
3241         expansion to perform a bitwise AND of the argument followed by a
3242         boolean negation of the result.
3243
3244 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
3245
3246         PR rtl-optimization/84878
3247         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3248         the basic block.  Assert the use reference is not artificial and that
3249         it has an associated insn.
3250
3251 2018-04-04  Michael Matz  <matz@suse.de>
3252
3253         * builtins.c (compute_objsize): Pass correct operand
3254         to array_at_struct_end_p.
3255
3256 2018-04-04  Richard Biener  <rguenther@suse.de>
3257
3258         PR lto/85176
3259         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3260         from contexts for DINFO_LEVEL_TERSE and below.
3261
3262 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
3263
3264         * config/nds32/nds32-doubleword.md (move_<mode>): Require
3265         resiter_operand condition.
3266         * config/nds32/nds32.md (*move<mode>): Ditto.
3267
3268 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
3269             Monk Chiang  <sh.chiang04@gmail.com>
3270
3271         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3272
3273 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
3274
3275         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3276
3277 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
3278             Kito Cheng  <kito.cheng@gmail.com>
3279
3280         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3281         nds32_cond_code_str, output_cond_branch,
3282         output_cond_branch_compare_zero, nds32_expand_cbranch,
3283         nds32_expand_cstore, nds32_expand_movcc,
3284         nds32_output_cbranchsi4_equality_zero,
3285         nds32_output_cbranchsi4_equality_reg,
3286         nds32_output_cbranchsi4_equality_reg_or_const_int,
3287         nds32_output_cbranchsi4_greater_less_zero: New functions.
3288         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3289         nds32_expand_cstore, nds32_expand_movcc,
3290         nds32_output_cbranchsi4_equality_zero,
3291         nds32_output_cbranchsi4_equality_reg,
3292         nds32_output_cbranchsi4_equality_reg_or_const_int,
3293         nds32_output_cbranchsi4_greater_less_zero): Declare.
3294         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3295         nds32_rimm11s_operand): New predicates.
3296         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3297         * config/nds32/nds32.md: Rewrite all the branch and conditional move
3298         patterns.
3299
3300 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
3301
3302         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3303         * config/nds32/nds32.md: Ditto.
3304         * config/nds32/pipelines.md: Ditto.
3305
3306 2018-04-04  Richard Biener  <rguenther@suse.de>
3307
3308         PR tree-optimization/85168
3309         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3310         propagating abnormals.
3311
3312 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
3313
3314         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3315
3316 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
3317             Kito Cheng  <kito.cheng@gmail.com>
3318
3319         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3320         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3321         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3322         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3323         * config/nds32/nds32.md (sibcall_internal): New.
3324         (sibcall_register): Remove.
3325         (sibcall_immediate): Remove.
3326         (sibcall_value_internal): New.
3327         (sibcall_value_register): Remove.
3328         (sibcall_value_immediate): Remove.
3329         * config/nds32/predicates.md (nds32_general_register_operand): New.
3330         (nds32_call_address_operand): New.
3331
3332 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
3333
3334         PR rtl-optimization/85167
3335         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3336         bb_defs if *split_p, instead preinitialize it to NULL.
3337
3338         PR tree-optimization/85156
3339         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3340         evaluating the argument multiple times.
3341
3342 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
3343
3344         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3345         than vector.
3346         (_mm_cvtpd_ps): Likewise.
3347         (_mm_cvttpd_epi32): Likewise.
3348         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3349         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3350         vector, pixel, and bool following altivec.h include.
3351
3352 2018-04-03  Martin Sebor  <msebor@redhat.com>
3353
3354         * doc/extend.texi (Common Function Attributes): Clarify.
3355         (const attribute): Likewise.
3356         (pure attribute): Likewise.
3357
3358 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
3359
3360         PR target/85169
3361         * config/i386/i386.c (ix86_expand_vector_set): Use
3362         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
3363
3364 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
3365
3366         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3367         instructions when changing rounding bits to preserve precision bits
3368         in the x87 control word.
3369
3370 2018-04-03  Martin Liska  <mliska@suse.cz>
3371
3372         PR tree-optimization/82491
3373         * rtl.h (strip_offset_and_add): Replace += suboffset with
3374         poly_uint64 () + suboffset.
3375
3376 2018-03-29  Martin Liska  <mliska@suse.cz>
3377             Martin Jambor  <mjambor@suse.cz>
3378
3379         PR ipa/84947
3380         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3381         param_type is not an integral or pointer type.
3382
3383 2018-04-03  Richard Biener  <rguenther@suse.de>
3384
3385         * sese.h (recompute_all_dominators): Remove.
3386
3387 2018-04-02  Martin Sebor  <msebor@redhat.com>
3388
3389         * doc/invoke.texi (-Wrestrict): Fix typos.
3390
3391 2018-04-02  Jim Wilson  <jimw@sifive.com>
3392
3393         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3394         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3395         (<optab>di3, <optab>si3_extend): Likewise.
3396         (<optab>si3_mask, <optab>si3_mask_1): New.
3397         (<optab>di3_mask, <optab>di3_mask_1): New.
3398         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3399         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3400         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3401
3402 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
3403
3404         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3405         example.
3406
3407 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
3408
3409         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3410         (nds32_canonicalize_comparison): New function.
3411
3412 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
3413             Kito Cheng  <kito.cheng@gmail.com>
3414             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
3415
3416         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3417         * config/nds32/constants.md (unspec_volatile_element): Add
3418         UNSPEC_VOLATILE_RELAX_GROUP.
3419         * config/nds32/nds32-relax-opt.c: New file.
3420         * config/nds32/nds32-predicates.c
3421         (nds32_symbol_load_store_p): New function.
3422         * config/nds32/nds32-protos.h
3423         (nds32_symbol_load_store_p): Declare function.
3424         (make_pass_nds32_relax_opt): Declare new rtl pass function.
3425         * config/nds32/nds32.c
3426         (nds32_register_pass): New function to register pass.
3427         (nds32_register_passes): New function to register passes.
3428         * config/nds32/nds32.md (relax_group): New pattern.
3429         * config/nds32/nds32.opt (mrelax-hint): New option.
3430         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3431
3432 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
3433
3434         * config/nds32/t-nds32: Modify files dependency.
3435
3436 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
3437
3438         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3439         (PROFILE_HOOK): Define its implementation.
3440
3441 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
3442
3443         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3444         type and 32-bit size.
3445
3446 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
3447
3448         PR middle-end/85090
3449         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3450         (V_128_256): New mode iterator.
3451         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3452         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3453         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3454         of V.
3455         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3456         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3457
3458 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
3459
3460         PR target/83315
3461         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3462         NaN inputs correctly.
3463
3464 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
3465
3466         PR target/80546
3467         * config/rs6000/vsx.md (??r): New mode attribute.
3468         (*vsx_mov<mode>_64bit): Use it.
3469         (*vsx_mov<mode>_32bit): Likewise.
3470
3471 2018-03-30  Martin Sebor  <msebor@redhat.com>
3472
3473         PR tree-optimization/84818
3474         * builtins.c (check_access): Use warning_n.
3475
3476 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3477
3478         PR target/83822
3479         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3480         condition.
3481         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3482         condition.
3483
3484 2018-03-30  Julia Koval  <julia.koval@intel.com>
3485
3486         PR target/84413
3487         * x86-tune.def (movx, partial_reg_dependency): Enable for
3488         m_SKYLAKE_AVX512.
3489
3490 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
3491
3492         PR inline-asm/84985
3493         * lra-constraints.c (process_alt_operands): Move setting
3494         this_alternative_matches below.
3495
3496 2018-03-29  Martin Liska  <mliska@suse.cz>
3497
3498         PR lto/84995.
3499         * doc/invoke.texi: Document how LTO works with debug info.
3500         Describe auto-load support of binutils.  Mention 'x86-64'
3501         as valid option value of -march option.
3502
3503 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
3504
3505         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3506
3507         PR c/85094
3508         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3509         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3510         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3511         checking.
3512
3513 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
3514
3515         PR target/84912
3516         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3517         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3518         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3519         for RS6000_BTM_POWERPC64.
3520         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3521         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3522         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3523         definition.
3524         (DIVDE): Use it.
3525         (DIVDEU): Likewise.
3526
3527 2018-03-28 Carl Love  <cel@us.ibm.com>
3528
3529         Revert
3530         2017-09-27  Carl Love  <cel@us.ibm.com>
3531
3532         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3533         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3534         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3535         fctiw instruction.
3536
3537 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3538
3539         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3540         instead of __vector bool.
3541         (_mm_max_pu8): Likewise.
3542         (_mm_min_pi16): Likewise.
3543
3544 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
3545
3546         PR target/84912
3547         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3548         (DIVWEUO): Likewise.
3549         (DIVDEO): Likewise.
3550         (DIVDEUO): Likewise.
3551         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3552         DIVWEUO and DIVDEUO.
3553         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3554         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3555         (div_extend): Likewise.
3556         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3557         builtin function.
3558         (__builtin_divweuo): Likewise.
3559         (__builtin_divdeo): Likewise.
3560         (__builtin_divdeuo): Likewise.
3561
3562 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
3563
3564         PR target/85095
3565         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3566         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3567
3568         PR tree-optimization/82004
3569         * gimple-match-head.c (optimize_pow_to_exp): New function.
3570         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3571         Don't fold to exp if optimize_pow_to_exp is false.
3572
3573 2018-03-28  Martin Liska  <mliska@suse.cz>
3574
3575         PR other/84819
3576         * calls.c (initialize_argument_information): Fix trailing space.
3577         * common.opt: Fix typo and provide better explanation for
3578         -fsanitize-coverage option.
3579         * config/i386/i386.opt: Fix typo.
3580
3581 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
3582             Martin Liska  <mliska@suse.cz>
3583
3584         PR sanitizer/85081
3585         * gimplify.c (asan_poison_variable): Don't do the check for
3586         gimplify_omp_ctxp here.
3587         (gimplify_decl_expr): Do it here.
3588         (gimplify_target_expr): Likewise.
3589
3590 2018-03-28  Martin Liska  <mliska@suse.cz>
3591
3592         PR target/84988
3593         * config/i386/i386.c (ix86_function_arg_advance): Do not call
3594         chkp_type_bounds_count if MPX is not enabled.
3595
3596 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
3597
3598         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3599
3600 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
3601
3602         PR target/84914
3603         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3604         function to create the function decl for complex long double
3605         multiply and divide for -mabi=ieeelongdouble.
3606         (init_float128_ieee): Call it.
3607
3608 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
3609
3610         PR target/85044
3611         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3612         -fcf-protection=branch -mibt.
3613         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3614
3615 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
3616
3617         PR target/81863
3618         * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3619
3620 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
3621
3622         PR target/85056
3623         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3624         extern array declarations.
3625
3626 2018-03-27  Richard Biener  <rguenther@suse.de>
3627
3628         PR middle-end/84067
3629         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3630         explicit single_use checks.
3631
3632 2018-03-27  Richard Biener  <rguenther@suse.de>
3633
3634         PR tree-optimization/85082
3635         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3636         Valueize the VUSE.
3637
3638 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
3639
3640         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3641         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3642         Turn on fasynchronous-unwind-tables and funwind-tables.
3643
3644 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
3645
3646         PR target/85073
3647         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3648         (*bmi_blsr_<mode>_ccz): Ditto.
3649
3650 2018-03-26  Tom de Vries  <tom@codesourcery.com>
3651
3652         PR tree-optimization/85063
3653         * omp-general.c (offloading_function_p): New function.  Factor out
3654         of ...
3655         * omp-offload.c (pass_omp_target_link::gate): ... here.
3656         * omp-general.h (offloading_function_p): Declare.
3657         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3658         with attribute omp declare target for offloading functions.
3659
3660 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
3661
3662         PR tree-optimization/84005
3663         * tree-data-ref.h (get_base_for_alignment): Declare.
3664         * tree-data-ref.c (get_base_for_alignment_1): New function.
3665         (get_base_for_alignment): Likewise.
3666         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3667         get_base_for_alignment to find a suitable base object, instead
3668         of always using drb->base_address.
3669
3670 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
3671
3672         PR inline-asm/85022
3673         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3674         known size by default.
3675
3676 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
3677
3678         PR inline-asm/85030
3679         * lra-constraints.c (process_alt_operands): Don't match BLKmode
3680         and non BLKmode operands.
3681
3682 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3683
3684         PR target/85026
3685         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3686         Clean up attributes.
3687
3688 2018-03-23  Richard Biener  <rguenther@suse.de>
3689
3690         PR debug/85020
3691         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3692         we are going to emit early debug for LTO.
3693
3694 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
3695
3696         PR inline-asm/85034
3697         * function.c (match_asm_constraints_1): Don't optimize if input
3698         doesn't satisfy general_operand predicate for output's mode.
3699
3700         PR inline-asm/85022
3701         * alias.c (write_dependence_p): Don't require for x_canonicalized
3702         non-VOIDmode if x has VOIDmode.
3703
3704         PR sanitizer/85029
3705         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3706         just don't try to optimize it rather than assert it never happens.
3707
3708 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3709
3710         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3711         macro expansions for definition of ST_INTERNAL_<mode> and
3712         LD_INTERNAL_<mode> builtins.
3713         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3714         Remove prototype.
3715         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3716         function.
3717         (altivec_expand_st_builtin): Likewise.
3718         (altivec_expand_builtin): Remove calls to deleted functions.
3719         (rs6000_address_for_altivec): Delete this function.
3720         * config/rs6000/vector.md: Remove expands for
3721         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3722
3723 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
3724
3725         PR target/84826
3726         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3727         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3728         re-computing once computed.
3729         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3730         (arm_init_machine_status): Initialize
3731         machine->static_chain_stack_bytes.
3732
3733 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3734
3735         PR target/84760
3736         * doc/extend.texi: Add four new prototypes for vec_ld.
3737         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3738         definitions for more logical presentation.
3739         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3740         entries for V1TI variants of __builtin_altivec_ld builtin.
3741         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3742         handling of V1TI variant of LVX icode pattern.
3743         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3744         (rs6000_gimple_fold_builtin): Likewise.
3745         (altivec_init_builtins): Add code to define
3746         __builtin_altivec_lvx_v1ti function.
3747
3748 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
3749
3750         PR inline-asm/84941
3751         * function.c (match_asm_constraints_1): Don't do the optimization
3752         if input isn't a REG, SUBREG, MEM or constant.
3753
3754 2018-03-22  Tom de Vries  <tom@codesourcery.com>
3755
3756         PR tree-optimization/84956
3757         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3758         bb_has_abnormal_pred.
3759
3760 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
3761
3762         PR sanitizer/85018
3763         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3764         DECL_INITIAL (decl) to decl at the end.
3765         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3766         adjust the comment.
3767
3768 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
3769
3770         * doc/extend.texi (__builtin_tgmath): Document when complex
3771         integer types are treated as _Complex _Float64.
3772
3773 2018-03-21  Tom de Vries  <tom@codesourcery.com>
3774
3775         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3776
3777 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
3778
3779         PR tree-optimization/84960
3780         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3781         if it is ENTRY block, move them into single succ of ENTRY in that case.
3782
3783 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
3784
3785         PR tree-optimization/84811
3786         * poly-int.h (poly_span_traits): Remove the T3 parameter and
3787         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3788         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3789         (known_subrange_p): Update accordingly.  Cast each value involved
3790         in the size comparison, rather than casting the result of the
3791         subtraction.
3792
3793 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
3794
3795         PR tree-optimization/84982
3796         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3797         by flipping the least significant bit rather than all bits from
3798         bitpos to bitpos + bitsize - 1.
3799
3800 2018-03-21  Nathan Sidwell  <nathan@acm.org>
3801
3802         * doc/extend.texi (Deprecated Features): Remove mention of
3803         long-deleted deprecations.
3804
3805 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3806
3807         PR jit/84288
3808         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3809         * configure: Regenerate.
3810
3811 2018-03-21  Tom de Vries  <tom@codesourcery.com>
3812
3813         PR tree-optimization/83126
3814         * tree-parloops.c (num_phis): New function.
3815         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3816
3817 2018-03-21  Nathan Sidwell  <nathan@acm.org>
3818
3819         * doc/extend.texi (Deprecated Features): Update deprecated flags,
3820         mention anon-struct/union members and trailing attributes.
3821
3822 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
3823
3824         PR tree-optimization/84969
3825         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3826         builtin memset partitions if they set different rhs values.
3827
3828 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
3829
3830         PR rtl-optimization/84989
3831         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3832         VEC_DUPLICATE with scalar result mode.
3833
3834 2018-03-21  Martin Liska  <mliska@suse.cz>
3835
3836         PR ipa/84963
3837         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3838         not intended return statement.
3839
3840 2018-03-21  Martin Liska  <mliska@suse.cz>
3841
3842         PR target/84988
3843         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3844         (chkp_find_bound_slots_1): Limit number of iterations.
3845
3846 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
3847
3848         PR target/84838
3849         * Minor grammar fixes for x86 options.
3850
3851 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
3852
3853         PR debug/84875
3854         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3855         holding REG_CFA_RESTORE notes, instead turn them into a USE.
3856
3857 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
3858
3859         PR target/83789
3860         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3861         (altivec_lvx_<mode>_1op): Likewise.
3862         (altivec_stvx_<mode>_2op): Likewise.
3863         (altivec_stvx_<mode>_1op): Likewise.
3864         (altivec_lvx_<VM2:mode>): New define_expand.
3865         (altivec_stvx_<VM2:mode>): Likewise.
3866         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3867         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3868         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3869         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3870         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3871         (rs6000_gen_lvx): Likewise.
3872         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3873         (altivec_expand_stv_builtin): Likewise.
3874         (altivec_expand_builtin): Likewise.
3875         * config/rs6000/vector.md: Likewise.
3876
3877 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3878
3879         PR target/82518
3880         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3881         BYTES_BIG_ENDIAN.
3882
3883 2018-03-20  Richard Biener  <rguenther@suse.de>
3884
3885         PR target/84986
3886         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3887         sign-conversions as zero, fall back to standard scalar_stmt
3888         cost for the rest.
3889
3890 2018-03-20  Martin Liska  <mliska@suse.cz>
3891
3892         PR ipa/84825
3893         * predict.c (rebuild_frequencies): Handle case when we have
3894         PROFILE_ABSENT, but flag_guess_branch_prob is false.
3895
3896 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
3897
3898         PR target/84990
3899         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3900         flag_section_anchors.
3901         * varasm.c (use_blocks_for_decl_p): Remove hack for
3902         dw2_force_const_mem.
3903
3904         PR target/84845
3905         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3906         to ...
3907         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
3908         be created, use lowpart_subreg of operands[0] rather than operands[0]
3909         itself.
3910         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3911         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3912         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3913         and n constraint instead of aarch64_shift_imm_di and Usd.
3914         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3915         (*aarch64_<optab>_reg_minus<mode>3): ... this.
3916
3917 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
3918
3919         PR target/82989
3920         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3921         to favor GPR over NEON registers.
3922         (<shift>di3_neon): Likewise.
3923
3924 2018-03-20  Tom de Vries  <tom@codesourcery.com>
3925
3926         PR target/84952
3927         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3928         (nvptx_process_pars): Emit bar.sync asap and alap.
3929
3930 2018-03-20  Tom de Vries  <tom@codesourcery.com>
3931
3932         PR target/84954
3933         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3934         seen_label if seen_label is already set.
3935
3936 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
3937
3938         PR target/84945
3939         * config/i386/i386.c (fold_builtin_cpu): For features above 31
3940         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3941         Use 1U instead of 1.  Formatting fixes.
3942
3943         PR c/84953
3944         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3945         instead of TREE_TYPE (s1) for the return value.
3946
3947 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
3948
3949         PR tree-optimization/84946
3950         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3951         bitsize + bitsize in poly_uint64 rather than poly_int64.
3952
3953         PR sanitizer/78651
3954         * dwarf2asm.c: Include fold-const.c.
3955         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3956         of decl rather than decl itself.
3957
3958         PR rtl-optimization/84643
3959         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3960
3961 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
3962
3963         PR sanitizer/78651
3964         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3965         calling assemble_variable.
3966
3967 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
3968
3969         PR target/81647
3970         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3971         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3972
3973 2018-03-19  Jim Wilson  <jimw@sifive.com>
3974
3975         PR bootstrap/84856
3976         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3977         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3978         (riscv_first_stack_step): Likewise.
3979         (riscv_option_override): Use STACK_BOUNDARY instead of
3980         MIN_STACK_BOUNDARY.
3981         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3982         MIN_STACK_BOUNDARY.
3983         (BIGGEST_ALIGNMENT): Set to 128.
3984         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3985         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3986         STACK_BOUNDARY.
3987
3988 2018-03-19  Richard Biener  <rguenther@suse.de>
3989
3990         PR tree-optimization/84933
3991         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3992         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3993
3994 2018-03-19  Richard Biener  <rguenther@suse.de>
3995
3996         PR tree-optimization/84859
3997         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3998         (cond_if_else_store_replacement): Perform sinking operation on
3999         single-store BBs regardless of MAX_STORES_TO_SINK setting.
4000         Generalize what a BB with a single eligible store is.
4001
4002 2018-03-19  Richard Biener  <rguenther@suse.de>
4003
4004         PR tree-optimization/84929
4005         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4006         chrec_is_positive against non-chrec arg.
4007
4008 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
4009
4010         PR target/84711
4011         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4012
4013 2018-03-18  Martin Liska  <mliska@suse.cz>
4014
4015         PR rtl-optimization/84635
4016         * regrename.c (build_def_use): Use matches_mode only when
4017         matches >= 0.
4018
4019 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
4020
4021         PR tree-optimization/84913
4022         * tree-vect-loop.c (vectorizable_reduction): Don't try to
4023         vectorize chains of COND_EXPRs.
4024
4025 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
4026
4027         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4028
4029 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
4030
4031         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4032
4033 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
4034
4035         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4036
4037 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
4038             Kito Cheng  <kito.cheng@gmail.com>
4039
4040         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4041         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4042         (nds32_adjust_reg_alloc_order): New function.
4043         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4044
4045 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
4046
4047         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4048         nds32_print_operand, nds32_print_operand_address): Use
4049         HOST_WIDE_INT_PRINT_DEC instead.
4050
4051 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
4052
4053         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4054
4055 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
4056
4057         PR target/84902
4058         * config/i386/i386.c (initial_ix86_tune_features,
4059         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4060         unsigned long long.
4061         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4062         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4063         rather than 1u << ix86_tune.  Formatting fix.
4064         (ix86_option_override_internal): Change ix86_arch_mask from
4065         unsigned int to unsigned HOST_WIDE_INT, initialize to
4066         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4067         (ix86_function_specific_restore): Likewise.
4068
4069 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
4070
4071         PR target/84899
4072         * postreload.c (reload_combine_recognize_pattern): Perform
4073         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4074         truncate_int_for_mode the result for the destination's mode.
4075
4076         PR c/84909
4077         * hsa-gen.c (mem_type_for_type): Fix comment typo.
4078         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4079         Likewise.
4080         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4081         Likewise.
4082
4083 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
4084
4085         PR target/84876
4086         * lra-assigns.c (lra_split_hard_reg_for): Don't use
4087         regno_allocno_class_array and sorted_pseudos.
4088         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4089         insns where regno is used.
4090
4091 2018-03-16  Martin Liska  <mliska@suse.cz>
4092
4093         PR ipa/84833
4094         * multiple_target.c (create_dispatcher_calls): Redirect
4095         reference in the symbol table.
4096
4097 2018-03-16  Martin Liska  <mliska@suse.cz>
4098
4099         PR ipa/84722
4100         * multiple_target.c (create_dispatcher_calls): Redirect also
4101         an alias.
4102
4103 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
4104
4105         PR c++/79937
4106         PR c++/82410
4107         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4108         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4109         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4110
4111 2018-03-16  Julia Koval  <julia.koval@intel.com>
4112
4113         * doc/invoke.texi (Skylake Server): Add CLWB.
4114         Cannonlake): Remove CLWB.
4115
4116 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
4117
4118         PR tree-optimization/84841
4119         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4120         1 << 3.
4121         (FLOAT_ONE_CONST_TYPE): Define.
4122         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4123         (sort_by_operand_rank): Put entries with higher constant_type last
4124         rather than first to match comments.
4125
4126 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
4127
4128         * config/nios2/nios2.md (movsi_internal): Fix thinko in
4129         split predicate.
4130
4131 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
4132
4133         PR c++/79085
4134         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4135         check and use address of target always.
4136
4137 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
4138
4139         PR target/84574
4140         * config/i386/i386.c (indirect_thunk_needed): Update comments.
4141         (indirect_thunk_bnd_needed): Likewise.
4142         (indirect_thunks_used): Likewise.
4143         (indirect_thunks_bnd_used): Likewise.
4144         (indirect_return_needed): New.
4145         (indirect_return_bnd_needed): Likewise.
4146         (output_indirect_thunk_function): Add a bool argument for
4147         function return.
4148         (output_indirect_thunk_function): Don't generate alias for
4149         function return thunk.
4150         (ix86_code_end): Call output_indirect_thunk_function to generate
4151         function return thunks.
4152         (ix86_output_function_return): Set indirect_return_bnd_needed
4153         and indirect_return_needed instead of indirect_thunk_bnd_needed
4154         and indirect_thunk_needed.
4155
4156 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
4157
4158         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4159         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4160         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4161
4162 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
4163             Paul Hua <paul.hua.gm@gmail.com>
4164
4165         PR c/84852
4166         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4167
4168 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
4169
4170         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4171         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4172         resp. SFmode cases.
4173
4174 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
4175
4176         PR target/84711
4177         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4178         instead of GET_MODE_SIZE when comparing Units.
4179
4180 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
4181
4182         PR target/68256
4183         * varasm.c (hash_section): Return an unchangeble hash value
4184         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4185         Return !aarch64_can_use_per_function_literal_pools_p ().
4186
4187 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
4188
4189         PR target/84860
4190         * optabs.c (emit_conditional_move): Pass address of cmode's copy
4191         rather than address of cmode as last argument to prepare_cmp_insn.
4192
4193 2018-03-15  Julia Koval  <julia.koval@intel.com>
4194
4195         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4196         F_AVX512VNNI, F_AVX512BITALG): New.
4197
4198 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
4199
4200         PR target/83451
4201         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
4202         insn for floating-point loads and stores.
4203
4204 2018-03-14  Carl Love  <cel@us.ibm.com>
4205
4206         * config/rs6000/rs6000-c.c: Add macro definitions for
4207         ALTIVEC_BUILTIN_VEC_PERMXOR.
4208         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4209         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4210         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4211         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4212         UNSPEC_VPERMXOR.
4213         * config/doc/extend.texi: Add prototypes for vec_permxor.
4214
4215 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
4216
4217         PR c/84852
4218         * diagnostic-show-locus.c (class layout_point): Convert m_line
4219         from int to linenum_type.
4220         (line_span::comparator): Use linenum "compare" function when
4221         comparing line numbers.
4222         (test_line_span): New function.
4223         (layout_range::contains_point): Convert param "row" from int to
4224         linenum_type.
4225         (layout_range::intersects_line_p): Likewise.
4226         (layout::will_show_line_p): Likewise.
4227         (layout::print_source_line): Likewise.
4228         (layout::should_print_annotation_line_p): Likewise.
4229         (layout::print_annotation_line): Likewise.
4230         (layout::print_leading_fixits): Likewise.
4231         (layout::annotation_line_showed_range_p): Likewise.
4232         (struct line_corrections): Likewise for field m_row.
4233         (line_corrections::line_corrections): Likewise for param "row".
4234         (layout::print_trailing_fixits): Likewise.
4235         (layout::get_state_at_point): Likewise.
4236         (layout::get_x_bound_for_row): Likewise.
4237         (layout::print_line): Likewise.
4238         (diagnostic_show_locus): Likewise for locals "last_line" and
4239         "row".
4240         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4241         * input.c (selftest::test_linenum_comparisons): New function.
4242         (selftest::input_c_tests): Call it.
4243         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4244         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4245         * selftest.h (ASSERT_GT): New macro.
4246         (ASSERT_GT_AT): New macro.
4247         (ASSERT_LT): New macro.
4248         (ASSERT_LT_AT): New macro.
4249
4250 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
4251
4252         PR rtl-optimization/84780
4253         * combine.c (distribute_links): Don't make a link based on pc_rtx.
4254
4255 2018-03-14  Martin Liska  <mliska@suse.cz>
4256
4257         * tree.c (record_node_allocation_statistics): Use
4258         get_stats_node_kind.
4259         (get_stats_node_kind): New function extracted from
4260         record_node_allocation_statistics.
4261         (free_node): Use get_stats_node_kind.
4262
4263 2018-03-14  Richard Biener  <rguenther@suse.de>
4264
4265         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4266         that the value-set of ANTIC_IN doesn't grow.
4267
4268         Revert
4269         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4270         member.
4271         (BB_VISITED_WITH_VISITED_SUCCS): New define.
4272         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4273
4274 2018-03-14  Julia Koval  <julia.koval@intel.com>
4275
4276         * config.gcc (icelake-client, icelake-server): New.
4277         (icelake): Remove.
4278         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4279         (initial_ix86_arch_features): Ditto.
4280         (PTA_SKYLAKE): Add SGX.
4281         (PTA_ICELAKE): Remove.
4282         (PTA_ICELAKE_CLIENT): New.
4283         (PTA_ICELAKE_SERVER): New.
4284         (ix86_option_override_internal): Split up icelake on icelake client and
4285         icelake server.
4286         (get_builtin_code_for_version): Ditto.
4287         (fold_builtin_cpu): Ditto.
4288         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4289         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4290         * config/i386/i386.h (processor_type): Ditto.
4291         * doc/invoke.texi: Ditto.
4292
4293 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
4294
4295         PR sanitizer/83392
4296         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4297         INTEGER_CST offset, add it together with bitpos / 8 and
4298         sign extend based on POINTER_SIZE.
4299
4300         PR target/84844
4301         Revert
4302         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
4303
4304         PR target/78090
4305         * config/i386/constraints.md (Yc): New register constraint.
4306         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4307         Use Yc constraint for alternative 2 of operand 0.  Remove
4308         preferred_for_speed attribute.
4309
4310 2018-03-14  Richard Biener  <rguenther@suse.de>
4311
4312         PR tree-optimization/84830
4313         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4314         with the old one to avoid oscillations.
4315
4316 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
4317
4318         PR target/83712
4319         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4320         pseudos.
4321         (assign_by_spills): Return a flag of reload assignment failure.
4322         Do not process the reload assignment failures.  Do not spill other
4323         reload pseudos if they has the same reg class.  Update n if
4324         necessary.
4325         (lra_assign): Add a return arg.  Set up from the result of
4326         assign_by_spills call.
4327         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4328         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
4329         usage_insns if it is not NULL.
4330         (spill_hard_reg_in_range): New function.
4331         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4332         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4333         function prototypes.
4334         (lra_assign): Change prototype.
4335         * lra.c (lra): Add code to deal with fails by splitting hard reg
4336         live ranges.
4337
4338 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
4339
4340         * config/riscv/riscv.opt (mrelax): New option.
4341         * config/riscv/riscv.c (riscv_file_start): Emit ".option
4342         "norelax" when riscv_mrelax is disabled.
4343         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4344
4345 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
4346
4347         PR target/84743
4348         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4349         reassociation for int modes.
4350
4351 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
4352
4353         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4354         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4355         for big-endian.
4356         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4357         * config/aarch64/aarch64-sve.md
4358         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4359         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4360         (*extend<mode><Vwide>2): Rename to...
4361         (aarch64_sve_extend<mode><Vwide>2): ...this.
4362         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4363         renaming the old pattern to...
4364         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
4365         unsigned packs.
4366         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4367         define_expand, renaming the old pattern to...
4368         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4369         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4370         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4371         account when deciding which SVE instruction the optab should use.
4372         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4373
4374 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
4375
4376         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4377         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4378         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4379         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4380         (tlsdesc_small_<mode>): Turn a define_expand and use
4381         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
4382         (tlsdesc_small_advsimd_<mode>): ...this.
4383         (tlsdesc_small_sve_<mode>): New pattern.
4384
4385 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
4386
4387         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4388         (UNSPEC_UMUL_HIGHPART): New constants.
4389         (MUL_HIGHPART): New int iteraor.
4390         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4391         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4392         define_expand.
4393         (*<su>mul<mode>3_highpart): New define_insn.
4394
4395 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
4396
4397         PR lto/84805
4398         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4399         incomplete types.
4400
4401 2018-03-13  Martin Liska  <mliska@suse.cz>
4402
4403         PR ipa/84658.
4404         * (sem_item_optimizer::sem_item_optimizer): Initialize new
4405         vector.
4406         (sem_item_optimizer::~sem_item_optimizer): Release it.
4407         (sem_item_optimizer::merge_classes): Register variable aliases.
4408         (sem_item_optimizer::fixup_pt_set): New function.
4409         (sem_item_optimizer::fixup_points_to_sets): Likewise.
4410         * ipa-icf.h: Declare new variables and functions.
4411
4412 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
4413
4414         PR middle-end/84834
4415         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4416         integer_pow2p@2 and test integer_pow2p in condition.
4417         (A < 0 ? C : 0): Similarly for @1.
4418
4419         PR middle-end/84831
4420         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4421         characters starting at p contain '\0' character, don't look beyond
4422         that.
4423
4424         PR target/84827
4425         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4426         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4427
4428         PR target/84828
4429         * reg-stack.c (change_stack): Change update_end var from int to
4430         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4431         also call set_block_for_insn on the newly added insns and rescan.
4432
4433         PR target/84786
4434         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4435         on the last operand.
4436
4437         PR c++/84704
4438         * tree.c (stabilize_reference_1): Return save_expr (e) for
4439         STATEMENT_LIST even if it doesn't have side-effects.
4440
4441 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
4442
4443         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4444
4445 2018-03-12  Renlin Li  <renlin.li@arm.com>
4446
4447         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4448         aarch64_output_scalar_simd_mov_immediate.
4449
4450 2018-03-12  Martin Sebor  <msebor@redhat.com>
4451
4452         PR tree-optimization/83456
4453         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4454         for perfectly overlapping calls to memcpy.
4455         (gimple_fold_builtin_memory_chk): Same.
4456         (gimple_fold_builtin_strcpy): Handle no-warning.
4457         (gimple_fold_builtin_stxcpy_chk): Same.
4458         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4459
4460 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
4461
4462         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4463         parameter.  Use it for SFmode.
4464         (rs6000_function_arg_advance_1): Adjust.
4465         (rs6000_function_arg): Adjust.
4466         (rs6000_gimplify_va_arg): Pass false for that new parameter.
4467
4468 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
4469
4470         PR rtl-optimization/84169
4471         PR rtl-optimization/84780
4472         * combine.c (can_combine_p): Check for a 2-insn combination whether
4473         the destination register is used between the two insns, too.
4474
4475 2018-03-12  Richard Biener  <rguenther@suse.de>
4476
4477         PR tree-optimization/84803
4478         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4479         for refs DR analysis didn't process.
4480
4481 2018-03-12  Richard Biener  <rguenther@suse.de>
4482
4483         PR tree-optimization/84777
4484         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4485         force-vectorize loops ignore whether we are optimizing for size.
4486
4487 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4488
4489         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4490         (TARGET_MD_ASM_ADJUST): Define.
4491
4492 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
4493             Kito Cheng  <kito.cheng@gmail.com>
4494             Chung-Ju Wu  <jasonwucj@gmail.com>
4495
4496         * config/nds32/nds32.c (nds32_compute_stack_frame,
4497         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4498         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4499         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4500         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4501         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4502         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4503         * config/nds32/nds32.md (prologue, epilogue): Use macro
4504         NDS32_V3PUSH_AVAILABLE_P to do checking.
4505
4506 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
4507
4508         PR debug/58150
4509         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4510         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4511         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
4512         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
4513         addition of most attributes on !orig_type_die or the attribute not
4514         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4515
4516 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
4517             Chung-Ju Wu  <jasonwucj@gmail.com>
4518
4519         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4520         __NDS32_VH__ macro.
4521         * config/nds32/nds32.opt (mvh): New option.
4522
4523 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
4524             Chung-Ju Wu  <jasonwucj@gmail.com>
4525
4526         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4527         function.
4528         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4529         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4530         definition.
4531
4532 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
4533             Chung-Ju Wu  <jasonwucj@gmail.com>
4534
4535         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4536         function.
4537         * config/nds32/nds32-multiple.md (strlensi): New pattern.
4538         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4539
4540 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
4541             Kito Cheng  <kito.cheng@gmail.com>
4542             Chung-Ju Wu  <jasonwucj@gmail.com>
4543
4544         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4545         UNSPEC_FFMISM and UNSPEC_FLMISM.
4546         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4547         for ffb, ffmism and flmism.
4548         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4549         (unspec_ffmism): Ditto.
4550         (unspec_flmism): Ditto.
4551         (nds32_expand_builtin_impl): Check if string extension is available.
4552         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4553         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4554
4555 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
4556
4557         Reverting patch:
4558         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
4559
4560         PR target/83712
4561         * lra-assigns.c (assign_by_spills): Return a flag of reload
4562         assignment failure.  Do not process the reload assignment
4563         failures.  Do not spill other reload pseudos if they has the same
4564         reg class.
4565         (lra_assign): Add a return arg.  Set up from the result of
4566         assign_by_spills call.
4567         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4568         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
4569         usage_insns if it is not NULL.
4570         (spill_hard_reg_in_range): New function.
4571         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4572         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4573         function prototypes.
4574         (lra_assign): Change prototype.
4575         * lra.c (lra): Add code to deal with fails by splitting hard reg
4576         live ranges.
4577
4578 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
4579
4580         PR target/84807
4581         * config/i386/i386.opt: Replace Enforcment with Enforcement.
4582
4583 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
4584
4585         PR debug/84620
4586         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4587         (dw_val_node): Add val_symbolic_view.
4588         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4589         (symview_upper_bound): New.
4590         (new_line_info_table): Initialize symviews_since_reset.
4591         (dwarf2out_source_line): Count symviews_since_reset and set
4592         symview_upper_bound.
4593         (dw_val_equal_p): Handle symview.
4594         (add_AT_symview): New.
4595         (print_dw_val): Handle symview.
4596         (attr_checksum, attr_checksum_ordered): Likewise.
4597         (same_dw_val_p, size_of_die): Likewise.
4598         (value_format, output_die): Likewise.
4599         (add_high_low_attributes): Use add_AT_symview for entry_view.
4600         (dwarf2out_finish): Reset symview_upper_bound, clear
4601         zero_view_p.
4602
4603 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
4604
4605         PR target/83969
4606         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4607         Add strict argument and use it.
4608         (rs6000_split_multireg_move): Update for new strict argument.
4609         (mem_operand_gpr): Disallow all non-offsettable addresses.
4610         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4611
4612 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
4613
4614         PR target/84772
4615         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4616         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4617         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4618
4619         PR c++/84767
4620         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4621         decl, use remap_type if we want to use the type.
4622
4623 2018-03-09  Martin Sebor  <msebor@redhat.com>
4624
4625         PR tree-optimization/84526
4626         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4627         Remove dead code.
4628         (builtin_access::generic_overlap): Be prepared to handle non-array
4629         base objects.
4630
4631 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
4632
4633         PR rtl-optimization/84682
4634         * lra-constraints.c (process_address_1): Check is_address flag
4635         for address constraints.
4636         (process_alt_operands): Likewise.
4637         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4638         preprocess_constraints.
4639         * recog.h (preprocess_constraints): Add oploc parameter.
4640         Adjust callers.
4641         * recog.c (preprocess_constraints): Test address_operand for
4642         CT_ADDRESS constraints.
4643
4644 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
4645
4646         PR target/83712
4647         * lra-assigns.c (assign_by_spills): Return a flag of reload
4648         assignment failure.  Do not process the reload assignment
4649         failures.  Do not spill other reload pseudos if they has the same
4650         reg class.
4651         (lra_assign): Add a return arg.  Set up from the result of
4652         assign_by_spills call.
4653         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4654         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
4655         usage_insns if it is not NULL.
4656         (spill_hard_reg_in_range): New function.
4657         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4658         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4659         function prototypes.
4660         (lra_assign): Change prototype.
4661         * lra.c (lra): Add code to deal with fails by splitting hard reg
4662         live ranges.
4663
4664 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4665
4666         PR target/83193
4667         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4668         Accept complain bool parameter.  Only emit errors if it is true.
4669         (arm_parse_cpu_option_name): Likewise.
4670         (arm_target_thumb_only): Adjust callers of the above.
4671         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4672         prototype to take a default true bool parameter.
4673         (arm_parse_arch_option_name): Likewise.
4674
4675 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
4676             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
4677
4678         PR jit/64089
4679         PR jit/84288
4680         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4681         * configure: Regenerate.
4682         * configure.ac ("linker --version-script option"): New.
4683         ("linker soname option"): New.
4684
4685 2018-03-09  Richard Biener  <rguenther@suse.de>
4686
4687         PR tree-optimization/84775
4688         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4689         immediate uses of predicate stmts and mark them modified.
4690
4691         Revert
4692         PR tree-optimization/84178
4693         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4694         to caller.
4695         (version_loop_for_if_conversion): Delay update_ssa call.
4696         (tree_if_conversion): Delay update_ssa until after predicate
4697         insertion.
4698
4699 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
4700
4701         PR target/84763
4702         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4703         when the function accesses prior frames.
4704
4705 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
4706
4707         PR debug/84456
4708         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4709         gen_llsym, otherwise call maybe_gen_llsym.
4710
4711         PR inline-asm/84742
4712         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4713         has ',' character inside of it.
4714
4715 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4716
4717         PR target/84748
4718         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4719         as clobbering CC_REGNUM.
4720
4721 2018-03-08  Richard Biener  <rguenther@suse.de>
4722
4723         PR middle-end/84552
4724         * tree-scalar-evolution.c: Include tree-into-ssa.h.
4725         (follow_copies_to_constant): Do not follow SSA names registered
4726         for update.
4727
4728 2018-03-08  Richard Biener  <rguenther@suse.de>
4729
4730         PR tree-optimization/84178
4731         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4732         to caller.
4733         (version_loop_for_if_conversion): Delay update_ssa call.
4734         (tree_if_conversion): Delay update_ssa until after predicate
4735         insertion.
4736
4737 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
4738
4739         PR tree-optimization/84178
4740         * tree-if-conv.c (release_bb_predicate): Remove the
4741         the assertion that the stmts have NULL use_ops.
4742         Discard the statements, asserting that they haven't
4743         yet been added to a BB.
4744
4745 2018-03-08  Richard Biener  <rguenther@suse.de>
4746
4747         PR tree-optimization/84746
4748         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4749         (phi_translate): Pass in destination ANTIC_OUT set.
4750         (phi_translate_1): Likewise.  For a simplified result lookup
4751         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4752         (phi_translate_set): Adjust.
4753         (do_pre_regular_insertion): Likewise.
4754         (do_pre_partial_partial_insertion): Likewise.
4755
4756 2018-03-08  Martin Liska  <mliska@suse.cz>
4757
4758         PR gcov-profile/84735
4759         * doc/gcov.texi: Document usage of profile files.
4760         * gcov-io.h: Document changes in the format.
4761
4762 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
4763
4764         PR debug/84404
4765         PR debug/84408
4766         * dwarf2out.c (struct dw_line_info_table): Update comments for
4767         view == -1.
4768         (FORCE_RESET_NEXT_VIEW): New.
4769         (FORCE_RESETTING_VIEW_P): New.
4770         (RESETTING_VIEW_P): Check for -1 too.
4771         (ZERO_VIEW_P): Likewise.
4772         (new_line_info_table): Force-reset next view.
4773         (dwarf2out_begin_function): Likewise.
4774         (dwarf2out_source_line): Simplify zero_view_p initialization.
4775         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4776         view directly.  Omit view when omitting .loc at line 0.
4777
4778 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
4779
4780         PR tree-optimization/84740
4781         * tree-switch-conversion.c (process_switch): Call build_constructors
4782         only if info.phi_count is non-zero.
4783
4784         PR tree-optimization/84739
4785         * tree-tailcall.c (find_tail_calls): Check call arguments against
4786         DECL_ARGUMENTS (current_function_decl) rather than
4787         DECL_ARGUMENTS (func) when checking for tail recursion.
4788
4789 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
4790
4791         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4792         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4793         Volker Reichelt's entry and add entries for people that perform
4794         GCC fuzzy testing and report numerous bugs.
4795
4796 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
4797
4798         PR target/82411
4799         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4800         readonly data in sdata, if that is disabled.
4801         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4802         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4803         -mreadonly-in-sdata option.
4804
4805 2018-03-07  Martin Sebor  <msebor@redhat.com>
4806
4807         PR tree-optimization/84468
4808         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4809         basic block when looking for nul assignment.
4810
4811 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
4812
4813         PR target/84277
4814         * except.h (output_function_exception_table): Adjust prototype.
4815         * except.c (output_function_exception_table): Remove FNNAME parameter
4816         and add SECTION parameter.  Ouput one part of the table at a time.
4817         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4818         the first part of the exception table and emit unwind directives.
4819         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4820         (i386_pe_seh_cold_init): Likewise.
4821         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4822         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4823         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4824         (ix86_output_call_insn): Emit a nop in one more case for SEH.
4825         * config/i386/winnt.c: Include except.h.
4826         (struct seh_frame_state): Add reg_offset, after_prologue and
4827         in_cold_section fields.
4828         (i386_pe_seh_end_prologue): Set seh->after_prologue.
4829         (i386_pe_seh_cold_init): New function.
4830         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4831         to seh->in_cold_section.
4832         (seh_emit_push): Record the offset of the push.
4833         (seh_emit_save): Record the offet of the save.
4834         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4835         Test seh->after_prologue to disregard the epilogue.
4836         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4837         (i386_pe_end_cold_function): New function.
4838
4839 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
4840
4841         PR fortran/84565
4842         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4843         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4844
4845         PR c++/84704
4846         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4847         on tmp_var.
4848         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4849         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4850
4851         PR middle-end/84723
4852         * multiple_target.c: Include tree-inline.h and intl.h.
4853         (expand_target_clones): Diagnose and fail if node->definition and
4854         !tree_versionable_function_p (node->decl).
4855
4856 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
4857
4858         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4859         sprint_ul.
4860         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4861         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4862         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4863
4864 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
4865
4866         PR target/84710
4867         * combine.c (try_combine): Use reg_or_subregno instead of handling
4868         just paradoxical SUBREGs and REGs.
4869
4870 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
4871
4872          * config/arc/arc.c (arc_finalize_pic): Remove function.
4873          (arc_must_save_register): We use single base PIC register, remove
4874          checks to save/restore the PIC register.
4875          (arc_expand_prologue): Likewise.
4876          * config/arc/arc-protos.h (arc_set_default_type_attributes):
4877          Remove.
4878          (arc_verify_short): Likewise.
4879          (arc_attr_type): Likewise.
4880          * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4881          (walk_stores): Likewise.
4882          (arc_address_cost): Make it static.
4883          (arc_verify_short): Likewise.
4884          (branch_dest): Likewise.
4885          (arc_attr_type): Likewise.
4886          * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4887          (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4888          (arc_final_prescan_insn): Remove inserting the nops due to
4889          hardware hazards.  It is done in reorg step.
4890          (insn_length_variant_t): Remove.
4891          (insn_length_parameters_t): Likewise.
4892          (arc_insn_length_parameters): Likewise.
4893          (arc_get_insn_variants): Likewise.
4894          * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4895
4896 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
4897
4898         PR inline-asm/84683
4899         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4900         assertion failure.
4901
4902         PR tree-optimization/84687
4903         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4904         on new_node->decl.
4905         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4906
4907 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4908
4909         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4910         Rename to ppc_speculation_barrier.
4911         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4912         __builtin_ppc_speculation_barrier.
4913
4914 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
4915
4916         PR target/84700
4917         * combine.c (combine_simplify_rtx): Don't try to simplify if
4918         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4919         are equal to x.
4920
4921 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
4922
4923         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4924         to 32 bytes when compiling for POWER9.
4925
4926 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
4927
4928         PR target/84564
4929         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4930         regparm >= 3 with no arg reg available also for calls with
4931         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
4932
4933         PR target/84524
4934         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4935         orig,vex.
4936         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
4937
4938 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
4939
4940         PR target/84264
4941         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4942
4943 2018-03-05  Richard Biener  <rguenther@suse.de>
4944
4945         PR tree-optimization/84486
4946         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4947         When inserting a __builtin_assume_aligned call set the LHS
4948         SSA name alignment info accordingly.
4949
4950 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
4951
4952         PR tree-optimization/84114
4953         * config/aarch64/aarch64.c (aarch64_reassociation_width)
4954         Avoid reassociation of FLOAT_MODE addition.
4955
4956 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
4957
4958         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4959         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4960         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4961         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4962         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4963         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4964         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4965         and -mwbnoinvd.
4966         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4967         __builtin_ia32_wbinvd): New builtins.
4968         (SPECIAL_ARGS2): New.
4969         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4970         (SPECIAL_ARGS2): New.
4971         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4972         (ix86_valid_target_attribute_inner_p): Ditto.
4973         (ix86_init_mmx_sse_builtins): Add special_args2.
4974         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4975         TARGET_WBNOINVD_P): New.
4976         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4977         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4978         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4979         * config/i386/immintrin.h (_wbinvd): New intrinsic.
4980         * config/i386/pconfigintrin.h: New file.
4981         * config/i386/wbnoinvdintrin.h: Ditto.
4982         * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4983         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4984
4985 2018-03-05  Richard Biener  <rguenther@suse.de>
4986
4987         PR tree-optimization/84670
4988         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4989         member.
4990         (BB_VISITED_WITH_VISITED_SUCCS): New define.
4991         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4992         (compute_antic_aux): Only assert the number of values in ANTIC_IN
4993         doesn't grow if all successors (recursively) were visited at least
4994         once.
4995
4996 2018-03-05  Richard Biener  <rguenther@suse.de>
4997
4998         PR tree-optimization/84650
4999         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5000         if executed in the loop pipeline.
5001
5002 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
5003
5004         * doc/configfiles.texi (Configuration Files): Move info about
5005         conditionalizing $target-protos.h to...
5006         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
5007         differs from $target-protos.h.
5008
5009 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
5010             Chung-Ju Wu  <jasonwucj@gmail.com>
5011
5012         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5013         * config/nds32/nds32-multiple.md (setmemsi): Define.
5014         * config/nds32/nds32-memory-manipulation.c
5015         (nds32_gen_dup_4_byte_to_word_value): New.
5016         (emit_setmem_word_loop): New.
5017         (emit_setmem_byte_loop): New.
5018         (nds32_expand_setmem_loop): New.
5019         (nds32_expand_setmem_loop_v3m): New.
5020         (nds32_expand_setmem_unroll): New.
5021         (nds32_expand_setmem): New.
5022
5023 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
5024             Chung-Ju Wu  <jasonwucj@gmail.com>
5025
5026         * config/nds32/nds32-memory-manipulation.c
5027         (nds32_emit_load_store): New.
5028         (nds32_emit_post_inc_load_store): New.
5029         (nds32_emit_mem_move): New.
5030         (nds32_emit_mem_move_block): New.
5031         (nds32_expand_movmemsi_loop_unknown_size): New.
5032         (nds32_expand_movmemsi_loop_known_size): New.
5033         (nds32_expand_movmemsi_loop): New.
5034         (nds32_expand_movmemsi_unroll): New.
5035         (nds32_expand_movmemqi): Rename ...
5036         (nds32_expand_movmemsi): ... to this.
5037         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5038         (movmemsi): ... to this.
5039         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5040         (nds32_expand_movmemsi): ... to this.
5041
5042 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
5043             Monk Chiang  <sh.chiang04@gmail.com>
5044             Chung-Ju Wu  <jasonwucj@gmail.com>
5045
5046         * config/nds32/nds32-protos.h
5047         (nds32_expand_load_multiple): New arguments.
5048         (nds32_expand_store_multiple): Ditto.
5049         (nds32_valid_multiple_load_store): Rename ...
5050         (nds32_valid_multiple_load_store_p): ... to this.
5051         * config/nds32/nds32-memory-manipulation.c
5052         (nds32_expand_load_multiple): Refine implementation.
5053         (nds32_expand_store_multiple): Ditto.
5054         * config/nds32/nds32-multiple.md
5055         (load_multiple): Update nds32_expand_load_multiple interface.
5056         (store_multiple): Update nds32_expand_store_multiple interface.
5057         * config/nds32/nds32-predicates.c
5058         (nds32_valid_multiple_load_store): Rename ...
5059         (nds32_valid_multiple_load_store_p): ... to this and refine
5060         implementation.
5061         * config/nds32/predicates.md
5062         (nds32_load_multiple_and_update_address_operation): New predicate.
5063         (nds32_store_multiple_and_update_address_operation): New predicate.
5064
5065 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
5066             Chung-Ju Wu  <jasonwucj@gmail.com>
5067
5068         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5069         (combo): New attribute.
5070         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5071
5072 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
5073
5074         * config/nds32/nds32.opt: Change -mcmodel= default value.
5075
5076 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
5077             Monk Chiang  <sh.chiang04@gmail.com>
5078             Chung-Ju Wu  <jasonwucj@gmail.com>
5079
5080         * config/nds32/constants.md (unspec_element): New enum.
5081         * config/nds32/constraints.md (Umw): New constraint.
5082         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5083         * config/nds32/nds32-intrinsic.md: Likewise.
5084         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5085         (nds32_valid_smw_lwm_base_p): New.
5086         (nds32_output_smw_single_word): New.
5087         (nds32_output_lmw_single_word): New.
5088         (nds32_expand_unaligned_load): New.
5089         (nds32_expand_unaligned_store): New.
5090         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5091         (nds32_output_smw_single_word): Declare.
5092         (nds32_output_lmw_single_word): Declare.
5093         (nds32_expand_unaligned_load): Declare.
5094         (nds32_expand_unaligned_store): Declare.
5095         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5096         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5097         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5098         NDS32_BUILTIN_UASTORE_DW.
5099         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5100         predicate.
5101
5102 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
5103             Kito Cheng  <kito.cheng@gmail.com>
5104             Chung-Ju Wu  <jasonwucj@gmail.com>
5105
5106         * config/nds32/nds32-intrinsic.c
5107         (nds32_expand_builtin_null_ftype_reg): Delete.
5108         (nds32_expand_builtin_reg_ftype_imm): Ditto.
5109         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5110         (nds32_read_argument): New.
5111         (nds32_legitimize_target): Ditto.
5112         (nds32_legitimize_argument): Ditto.
5113         (nds32_check_constant_argument): Ditto.
5114         (nds32_expand_unop_builtin): Ditto.
5115         (nds32_expand_unopimm_builtin): Ditto.
5116         (nds32_expand_binop_builtin): Ditto.
5117         (nds32_builtin_decl_impl): Ditto.
5118         (builtin_description): Ditto.
5119         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5120         (nds32_init_builtins_impl): Ditto.
5121         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5122         (nds32_builtin_decl): New.
5123         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5124         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5125
5126 2018-03-02  Jeff Law  <law@redhat.com>
5127
5128         * reorg.c (stop_search_p): Handle DEBUG_INSN.
5129         (redundant_insn, fill_simple_delay_slots): Likewise.
5130         (fill_slots_from_thread): Likewise.
5131         * resource.c (mark_referenced_resources): Likewise.
5132         (mark_set_resources, find_dead_or_set_registers): Likewise.
5133
5134 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
5135
5136         * substring-locations.h (format_warning_va): Formatting fix for
5137         ATTRIBUTE_GCC_DIAG.
5138         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5139         argument.
5140         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5141         * substring-locations.c: Include intl.h.
5142         (format_warning_va): Turned into small wrapper around
5143         format_warning_n_va, renamed to ...
5144         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5145         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5146         use ngettext.
5147         (format_warning_at_substring_n): New function.
5148         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5149         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
5150         format_warning_at_substring with just a shorter name instead of
5151         const function pointer.
5152         (fmtwarn_n): New function.
5153         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5154         appropriate, get rid of all the fmtstr temporaries, move conditionals
5155         with G_() wrapped string literals directly into fmtwarn arguments,
5156         cast dir.len to (int), formatting fixes.
5157
5158 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
5159
5160         * doc/invoke.texi: Remove "Cilk Plus" references.
5161
5162 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
5163             Richard Biener  <rguenther@suse.de>
5164
5165         PR ipa/84628
5166         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5167         for error or warning attributes if CALL_FROM_THUNK_P is set.
5168         Formatting fixes.
5169
5170 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
5171
5172         PR target/56540
5173         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5174         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5175
5176         PR target/56540
5177         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5178         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5179
5180         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5181         instead of -1U in last predictors element's probability member.
5182
5183 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
5184
5185         PR ipa/83983
5186         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5187         arguments if they are comparable.
5188
5189 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
5190
5191         PR tree-optimization/84634
5192         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5193         masks and masked_loop_p with a single loop_masks, making sure it's
5194         null for bb vectorization.
5195
5196 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
5197
5198         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5199         (vect_analyze_data_ref_access): Use loop->safe_len rather than
5200         loop->force_vectorize to check whether there is no alias.
5201
5202 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
5203
5204         PR target/84614
5205         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5206         prototypes.
5207         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5208         comments.
5209         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5210         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5211         instead of a loop around prev_real_insn.
5212         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5213         prev_real_insn.
5214
5215         PR inline-asm/84625
5216         * config/i386/i386.c (ix86_print_operand): Use conditional
5217         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5218         zero vector.
5219
5220 2018-03-02  Richard Biener  <rguenther@suse.de>
5221
5222         PR tree-optimization/84427
5223         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5224         (bitmap_set_subtract_values): Rewrite to handle multiple
5225         exprs per value.
5226         (clean): Likewise.
5227         (prune_clobbered_mems): Likewise.
5228         (phi_translate): Take edge instead of pred/phiblock.
5229         (phi_translate_1): Likewise.
5230         (phi_translate_set): Likewise.  Insert all translated
5231         exprs for a value into the set, keeping possibly multiple
5232         expressions per value.
5233         (compute_antic_aux): Adjust for phi_translate changes.
5234         When intersecting union the expressions and prune those
5235         not in the final value set, keeping possibly multiple
5236         expressions per value.  Do not use value-insertion
5237         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5238         all expressions.  Add verification that the value-sets
5239         only shrink during iteration.
5240         (compute_partial_antic_aux): Adjust for the phi_translate changes.
5241         (do_pre_regular_insertion): Likewise.
5242         (do_pre_partial_partial_insertion): Likewise.
5243
5244 2018-03-02  Richard Biener  <rguenther@suse.de>
5245
5246         PR target/82005
5247         * config/darwin.c (saved_debug_info_level): New static global.
5248         (darwin_asm_lto_start): Disable debug info generation for LTO out.
5249         (darwin_asm_lto_end): Restore debug info generation settings.
5250
5251 2018-03-01  Martin Liska  <mliska@suse.cz>
5252
5253         PR sanitizer/82484
5254         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5255         volatile arguments.
5256
5257 2018-03-01  Richard Biener  <rguenther@suse.de>
5258
5259         PR debug/84645
5260         * dwarf2out.c (gen_variable_die): Properly handle late VLA
5261         type annotation with LTO when debug was disabled at compile-time.
5262
5263 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
5264
5265         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5266         XINT with INTVAL.
5267         (mips_final_postscan_insn): Likewise.
5268
5269 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
5270
5271         PR rtl-optimization/84528
5272         * alias.c (init_alias_target): Add commentary.
5273         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5274         a unique base value if the frame pointer is not eliminated
5275         to the stack pointer.
5276
5277 2018-03-01  Tom de Vries  <tom@codesourcery.com>
5278
5279         PR rtl-optimization/83327
5280         * lra-int.h (hard_regs_spilled_into): Declare.
5281         * lra.c (hard_regs_spilled_into): Define.
5282         (init_reg_info): Init hard_regs_spilled_into.
5283         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5284         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5285         (process_bb_lives): Handle hard_regs_spilled_into.
5286         (lra_create_live_ranges_1): Before doing liveness propagation, clear
5287         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5288
5289 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
5290
5291         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5292         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5293         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5294         * config/rs6000/aix72.h: New file.
5295
5296 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
5297
5298         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5299         instead of warning_at with conditional singular and plural messages
5300         where possible.
5301
5302         PR target/52991
5303         * stor-layout.c (update_alignment_for_field): For
5304         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5305         && !DECL_PACKED (field), do the alignment update, just use
5306         only desired_align instead of MAX (type_align, desired_align)
5307         as the alignment.
5308         (place_field): Don't do known_align < desired_align handling
5309         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5310         is non-NULL, instead do it after rli->prev_field handling and
5311         only if not within a bitfield word.  For DECL_PACKED (field)
5312         use type_align of BITS_PER_UNIT.
5313
5314 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
5315
5316         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5317         superfluous parentheses and trailing spaces.
5318
5319 2018-02-28  Richard Biener  <rguenther@suse.de>
5320
5321         PR tree-optimization/84584
5322         * graphite-scop-detection.c (scop_detection::add_scop): Discard
5323         SCoPs with fake exit edge.
5324
5325 2018-02-28  Martin Liska  <mliska@suse.cz>
5326
5327         PR testsuite/84597
5328         * timevar.c (timer::print): Fix format to properly print 100%
5329         values.
5330
5331 2018-02-28  Richard Biener  <rguenther@suse.de>
5332
5333         PR middle-end/84607
5334         * genmatch.c (capture_info::walk_match): Do not mark
5335         captured expressions without operands as expr_p given
5336         they act more like predicates and should be subject to
5337         "lost tail" side-effect preserving.
5338
5339 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
5340
5341         PR rtl-optimization/81611
5342         * auto-inc-dec.c (attempt_change): Move dead note from
5343         mem_insn if it's the next use of regno
5344         (find_address): Take address use of reg holding
5345         non-incremented value.  Add parm to limit search to the named
5346         reg only.
5347         (merge_in_block): Attempt to use a mem insn that is the next
5348         use of the original regno.
5349
5350 2018-02-27  Martin Sebor  <msebor@redhat.com>
5351
5352         PR c++/83871
5353         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5354         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5355
5356 2018-02-27  Martin Sebor  <msebor@redhat.com>
5357
5358         PR translation/84207
5359         * diagnostic-core.h (warning_n, error_n, inform_n): Change
5360         n argument to unsigned HOST_WIDE_INT.
5361         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5362         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
5363         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5364         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5365
5366 2018-02-27  Richard Biener  <rguenther@suse.de>
5367
5368         PR tree-optimization/84512
5369         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5370         Do not use the estimate returned from record_stmt_cost for
5371         the scalar iteration cost but sum properly using add_stmt_cost.
5372
5373 2018-02-27  Richard Biener  <rguenther@suse.de>
5374
5375         PR tree-optimization/84466
5376         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5377         Adjust last change to less strictly validate use operands.
5378
5379 2018-02-27  Martin Liska  <mliska@suse.cz>
5380
5381         PR gcov-profile/84548
5382         * gcov.c (process_file): Allow partial overlap and consider it
5383         also as group functions.
5384         (output_lines): Properly calculate range of lines for a group.
5385
5386 2018-02-27  Martin Liska  <mliska@suse.cz>
5387
5388         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5389         'ggc' suffixes.  Change first column width.
5390         (timer::print): Fix formatting of the column.
5391
5392 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
5393
5394         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5395         preserve inline entry blocks for the sake of debug inline
5396         entry point markers alone.
5397         (remove_unused_locals): Suggest in comments a better place to
5398         force the preservation of inline entry blocks that are
5399         otherwise unused, but do not preserve them.
5400
5401 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
5402
5403         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5404
5405 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
5406
5407         PR target/84039
5408         * config/i386/constraints.md (Bs): Replace
5409         ix86_indirect_branch_register with
5410         TARGET_INDIRECT_BRANCH_REGISTER.
5411         (Bw): Likewise.
5412         * config/i386/i386.md (indirect_jump): Likewise.
5413         (tablejump): Likewise.
5414         (*sibcall_memory): Likewise.
5415         (*sibcall_value_memory): Likewise.
5416         Peepholes of indirect call and jump via memory: Likewise.
5417         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5418         (*sibcall_value_GOT_32): Likewise.
5419         * config/i386/predicates.md (indirect_branch_operand): Likewise.
5420         (GOT_memory_operand): Likewise.
5421         (call_insn_operand): Likewise.
5422         (sibcall_insn_operand): Likewise.
5423         (GOT32_symbol_operand): Likewise.
5424         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5425
5426 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
5427
5428         PR rtl-optimization/83496
5429         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5430         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
5431         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5432         redundant insn, if any.
5433         (relax_delay_slots): Likewise.
5434         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5435
5436 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
5437
5438         PR tree-optimization/83965
5439         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5440         that grouped statements are part of a reduction chain.  Return
5441         true if the statement is not marked as a reduction itself but
5442         is part of a group.
5443         (vect_recog_dot_prod_pattern): Don't check whether the statement
5444         is part of a group here.
5445         (vect_recog_sad_pattern): Likewise.
5446         (vect_recog_widen_sum_pattern): Likewise.
5447
5448 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
5449
5450         PR debug/84545
5451         * final.c (rest_of_clean_state): Also look for calls inside sequences.
5452
5453 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
5454
5455         PR target/84530
5456         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5457         the bool argument.
5458         (ix86_output_indirect_function_return): New prototype.
5459         (ix86_split_simple_return_pop_internal): Likewise.
5460         * config/i386/i386.c (indirect_return_via_cx): New.
5461         (indirect_return_via_cx_bnd): Likewise.
5462         (indirect_thunk_name): Handle return va CX_REG.
5463         (output_indirect_thunk_function): Create alias for
5464         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5465         (ix86_output_indirect_jmp): Remove the bool argument.
5466         (ix86_output_indirect_function_return): New function.
5467         (ix86_split_simple_return_pop_internal): Likewise.
5468         * config/i386/i386.md (*indirect_jump): Don't pass false
5469         to ix86_output_indirect_jmp.
5470         (*tablejump_1): Likewise.
5471         (simple_return_pop_internal): Change it to define_insn_and_split.
5472         Call ix86_split_simple_return_pop_internal to split it for
5473         -mfunction-return=.
5474         (simple_return_indirect_internal): Call
5475         ix86_output_indirect_function_return instead of
5476         ix86_output_indirect_jmp.
5477
5478 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
5479
5480         PR bootstrap/84405
5481         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5482         memset and value initialization afterwards.
5483
5484 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
5485
5486         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5487
5488 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5489
5490         PR target/84521
5491         * common/config/aarch64/aarch64-common.c
5492         (aarch_option_optimization_table[]): Switch
5493         off fomit-frame-pointer
5494
5495 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
5496             Chung-Ju Wu  <jasonwucj@gmail.com>
5497
5498         * config/nds32/nds32-multiple.md (load_multiple): Disallow
5499         volatile memory.
5500         (store_multiple): Ditto.
5501
5502 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
5503
5504         * config.gcc: Add --with-cpu support for nds32 target.
5505         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5506         * config/nds32/nds32.opt: Add -mcpu= option.
5507
5508 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
5509
5510         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5511         isel=yes): Warn for these deprecated options.
5512
5513 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
5514
5515         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5516         ISA_2_5_MASKS_EMBEDDED.
5517
5518 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
5519
5520         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5521         p->max as pointers rather than using iterative_hash_expr.
5522
5523 2018-02-23  Carl Love  <cel@us.ibm.com>
5524
5525         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5526         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5527         BU_P8V_OVERLOAD_2.
5528         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5529         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5530         P8V_BUILTIN_VEC_VUNSIGNED2.
5531
5532 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
5533
5534         PR target/81572
5535         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5536         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5537         LRA_UNKNOWN_ALT.
5538         * lra-constraints.c (curr_insn_transform): Set up
5539         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
5540         LRA_UNKNOWN_ALT.
5541         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5542         * lra-eliminations.c (spill_pseudos): Ditto.
5543         (process_insn_for_elimination): Ditto.
5544         * lra-lives.c (reg_early_clobber_p): Use the new macros.
5545         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5546         LRA_NON_CLOBBERED_ALT.
5547
5548 2018-02-22  Martin Sebor  <msebor@redhat.com>
5549
5550         PR tree-optimization/84480
5551         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5552         to maybe_diag_stxncpy_trunc.  Call it.
5553         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5554         from gimple_fold_builtin_strcpy.  Print inlining stack.
5555         (handle_builtin_stxncpy): Print inlining stack.
5556         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5557
5558 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
5559
5560         PR target/84176
5561         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5562         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5563         and -fcheck-pointer-bounds are used together.
5564         (indirect_thunk_prefix): New enum.
5565         (indirect_thunk_need_prefix): New function.
5566         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
5567         "_nt" instead of "_bnd" for NOTRACK prefix.
5568         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5569         (output_indirect_thunk_function): Likewise.
5570         (): Likewise.
5571         (ix86_code_end): Update output_indirect_thunk_function calls.
5572         (ix86_output_indirect_branch_via_reg): Replace
5573         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5574         (ix86_output_indirect_branch_via_push): Likewise.
5575         (ix86_output_function_return): Likewise.
5576         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5577         incompatible with -fcf-protection=branch and
5578         -fcheck-pointer-bounds.
5579
5580 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
5581
5582         PR target/83335
5583         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5584         Change gcc_assert call to output_operand_lossage.
5585
5586 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
5587
5588         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5589
5590 2018-02-22  DJ Delorie  <dj@redhat.com>
5591             Sebastian Perta  <sebastian.perta@renesas.com>
5592             Oleg Endo  <olegendo@gcc.gnu.org>
5593
5594         * config/rx/rx.c (rx_rtx_costs): New function.
5595         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5596
5597 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
5598
5599         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5600
5601 2018-02-22  Martin Liska  <mliska@suse.cz>
5602
5603         PR driver/83193
5604         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5605         Add "native" as a possible value.
5606
5607 2018-02-22  Martin Liska  <mliska@suse.cz>
5608
5609         PR driver/83193
5610         * config/i386/i386.c (ix86_option_override_internal):
5611         Add "native" as a possible value for -march and -mtune.
5612
5613 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
5614
5615         PR target/84502
5616         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5617         to all type variants.
5618
5619         PR tree-optimization/84503
5620         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5621         width as info->bitpos + info->bitsize - start.
5622         (merged_store_group::merge_overlapping): Simplify width computation.
5623         (check_no_overlap): New function.
5624         (imm_store_chain_info::try_coalesce_bswap): Compute expected
5625         start + width and last_order of the group, fail if check_no_overlap
5626         fails.
5627         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5628         to group if check_no_overlap fails.
5629
5630 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
5631
5632         * config/rs6000/altivec.md: Delete contraint arguments to
5633         define_expand, define_split, and define_peephole2, and in
5634         define_insn_and_split if always unused.
5635         * config/rs6000/darwin.md: Ditto.
5636         * config/rs6000/dfp.md: Ditto.
5637         * config/rs6000/rs6000.md: Ditto.
5638         * config/rs6000/sync.md: Ditto.
5639         * config/rs6000/vector.md: Ditto.
5640         * config/rs6000/vsx.md: Ditto.
5641
5642 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
5643
5644         * config/rs6000/altivec.md: Write output control strings as braced
5645         blocks instead of double-quoted strings.
5646         * config/rs6000/darwin.md: Ditto.
5647         * config/rs6000/rs6000.md: Ditto.
5648         * config/rs6000/vector.md: Ditto.
5649         * config/rs6000/vsx.md: Ditto.
5650
5651 2018-02-21  Jason Merrill  <jason@redhat.com>
5652
5653         PR c++/84314 - ICE with templates and fastcall attribute.
5654         * attribs.c (build_type_attribute_qual_variant): Remove assert.
5655
5656 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
5657
5658         * ipa-cp.c (determine_versionability): Fix comment typos.
5659
5660 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
5661
5662         PR c/84229
5663         * ipa-cp.c (determine_versionability): Do not version functions caling
5664         va_arg_pack.
5665
5666 2018-02-21  Martin Liska  <mliska@suse.cz>
5667
5668         PR driver/83193
5669         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5670         Add "native" as a possible value.
5671         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
5672         the macro when native cpu detection is available.
5673
5674 2018-02-21  Martin Liska  <mliska@suse.cz>
5675
5676         PR driver/83193
5677         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5678         Add "native" as a possible value.
5679         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5680         when native cpu detection is available.
5681
5682 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
5683             Martin Sebor  <msebor@redhat.com>
5684
5685         PR tree-optimization/84478
5686         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5687         false.
5688         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5689         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5690         support which is conservatively correct, for 2 only stay conservative
5691         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
5692         argument to the 2 argument get_range_strlen, adjust 6 arg
5693         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5694         false.
5695         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5696         (gimple_fold_builtin_strlen): Pass true as last argument to
5697         get_range_strlen.
5698
5699 2018-02-20  Martin Sebor  <msebor@redhat.com>
5700
5701         PR middle-end/84095
5702         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5703         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
5704         (builtin_memref::builtin_memref): Factor out parts into
5705         set_base_and_offset and call it.
5706
5707 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
5708
5709         PR middle-end/84406
5710         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5711         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5712         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
5713         search at the associated MODE_INT.
5714
5715 2018-02-20  Jeff Law  <law@redhat.com>
5716
5717         PR middle-end/82123
5718         PR tree-optimization/81592
5719         PR middle-end/79257
5720         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5721         for range data rather than using global data.
5722         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5723         range data rather than using global data.
5724         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5725         pass it to children as needed.
5726         (struct directive::fmtresult): Similarly.
5727         (struct directive::set_width): Similarly.
5728         (struct directive::set_precision): Similarly.
5729         (format_integer, format_directive, parse_directive): Similarly.
5730         (format_none): Accept unnamed vr_values parameter.
5731         (format_percent, format_floating, format_character): Similarly.
5732         (format_string, format_plain): Similarly.
5733         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5734         the EVRP range analyzer for range data rather than using global data.
5735         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5736         gimple-ssa-evrp-analyze.h
5737         (class sprintf_dom_walker): Add after_dom_children member function.
5738         Add evrp_range_analyzer member.
5739         (sprintf_dom_walker::before_dom_children): Call into the EVRP
5740         range analyzer as needed.
5741         (sprintf_dom_walker::after_dom_children): New member function.
5742         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5743         if not optimizing.
5744         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5745         (evrp_range_analyzer::pop_to_marker): Likewise.
5746
5747 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
5748
5749         PR tree-optimization/84419
5750         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5751         with the required type if its current type is compatible but
5752         different.
5753
5754 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
5755
5756         PR middle-end/82004
5757         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5758         after vectorization.
5759
5760 2018-02-20  Martin Liska  <mliska@suse.cz>
5761
5762         PR driver/83193
5763         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5764         possible values if we don't have a hint.
5765
5766 2018-02-20  Martin Liska  <mliska@suse.cz>
5767
5768         PR c/84310
5769         PR target/79747
5770         * final.c (shorten_branches): Build align_tab array with one
5771         more element.
5772         * opts.c (finish_options): Add alignment option limit check.
5773         (MAX_CODE_ALIGN): Likewise.
5774         (MAX_CODE_ALIGN_VALUE): Likewise.
5775         * doc/invoke.texi: Document maximum allowed option value for
5776         all -falign-* options.
5777
5778 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
5779
5780         PR target/84146
5781         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5782         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5783         * var-tracking.c (emit_note_insn_var_location): Remove all references
5784         to NOTE_INSN_CALL_ARG_LOCATION.
5785         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5786         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5787         Use copy_rtx_if_shared.
5788         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5789         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5790         (dwarf2out_var_location): Remove handling of
5791         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5792         on call_insn.
5793         * final.c (final_scan_insn): Remove all references to
5794         NOTE_INSN_CALL_ARG_LOCATION.
5795         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
5796         before dumping final insns.
5797         * except.c (emit_note_eh_region_end): Remove all references to
5798         NOTE_INSN_CALL_ARG_LOCATION.
5799         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5800         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5801         * config/arc/arc.c (hwloop_optimize): Likewise.
5802         * config/arm/arm.c (create_fix_barrier): Likewise.
5803         * config/s390/s390.c (s390_chunkify_start): Likewise.
5804         * config/sh/sh.c (find_barrier): Likewise.
5805         * config/i386/i386.c (rest_of_insert_endbranch,
5806         ix86_seh_fixup_eh_fallthru): Likewise.
5807         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5808         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5809         * config/frv/frv.c (frv_function_prologue): Likewise.
5810         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
5811         reg note.
5812         (note_outside_basic_block_p): Remove all references to
5813         NOTE_INSN_CALL_ARG_LOCATION.
5814         * gengtype.c (adjust_field_rtx_def): Likewise.
5815         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5816         Likewise.
5817         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5818         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5819
5820         PR c++/84444
5821         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5822         is ADDR_EXPR.
5823
5824         PR tree-optimization/84452
5825         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5826         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5827         is NULL.
5828
5829 2018-02-19  Martin Liska  <mliska@suse.cz>
5830
5831         PR sanitizer/82183
5832         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5833
5834 2018-02-19  Martin Liska  <mliska@suse.cz>
5835             Richard Sandiford  <richard.sandiford@linaro.org>
5836
5837         PR tree-optimization/82491
5838         * gimple-fold.c (get_base_constructor): Make earlier bail out
5839         to prevent ubsan.
5840
5841 2018-02-19  Carl Love  <cel@us.ibm.com>
5842
5843         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5844         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5845         BU_P8V_OVERLOAD_1.
5846         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5847         P8V_BUILTIN_VEC_NEG.
5848
5849 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
5850
5851         * config/rl78/rl78.md (movdf): New define expand.
5852
5853 2018-02-19  Martin Liska  <mliska@suse.cz>
5854
5855         PR other/80589
5856         * doc/invoke.texi: Fix typo.
5857         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5858
5859 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
5860
5861         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5862         handle rs6000_single_float and rs6000_double_float specially for
5863         e500 family CPUs.
5864
5865 2018-02-16  Jeff Law  <law@redhat.com>
5866
5867         * config/rx/rx.c (add_pop_cfi_notes): New function.;
5868         (pop_regs): Use it.
5869
5870 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
5871
5872         PR ipa/84425
5873         * ipa-inline.c (inline_small_functions): Fix a typo.
5874
5875 2018-02-16  Nathan Sidwell  <nathan@acm.org>
5876
5877         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5878
5879 2018-02-16  Carl Love  <cel@us.ibm.com>
5880
5881         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5882         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5883         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5884         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5885         expansion to P8V_BUILTIN_VEC_FLOAT2.
5886
5887 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
5888
5889         PR rtl-optimization/70023
5890         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5891         src_regno into account.
5892
5893 2018-02-16  Carl Love  <cel@us.ibm.com>
5894
5895         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5896         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5897         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5898         * config/rs6000/rs6000.c: Remove case statements for
5899         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5900         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5901         and P9V_BUILTIN_VEC_VINSERT4B.
5902         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5903         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5904         * config/rs6000/vsx.md:
5905         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5906         vec_insert4b.
5907
5908 2018-02-16  Carl Love  <cel@us.ibm.com>
5909
5910         * config/rs6000/altivec.h: Add builtin names vec_extract4b
5911         vec_insert4b.
5912         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5913         definitions.
5914         * config/rs6000/rs6000-c.c: Add the definitions for
5915         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5916         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5917         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5918         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
5919         definition for insert4b and define insn *insert3b_internal.
5920         * doc/extend.texi: Add documentation for vec_extract4b.
5921
5922 2018-02-16  Nathan Sidwell  <nathan@acm.org>
5923
5924         * doc/extend.texi (Backwards Compatibility): Mention friend
5925         injection.  Note for-scope is deprecated.
5926         * doc/invoke.texi (-ffriend-injection): Deprecate.
5927
5928 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
5929
5930         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5931         that moved to I2, also allow destinations that are a paradoxical
5932         subreg (instead of a normal reg).
5933
5934 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
5935
5936         PR target/83831
5937         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5938         to QImode.
5939
5940 2018-02-16  Richard Biener  <rguenther@suse.de>
5941
5942         PR tree-optimization/84037
5943         PR tree-optimization/84016
5944         PR target/82862
5945         * config/i386/i386.c (ix86_builtin_vectorization_cost):
5946         Adjust vec_construct for the fact we need additional higher latency
5947         128bit inserts for AVX256 and AVX512 vector builds.
5948         (ix86_add_stmt_cost): Scale vector construction cost for
5949         elementwise loads.
5950
5951 2018-02-16  Richard Biener  <rguenther@suse.de>
5952
5953         PR tree-optimization/84417
5954         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5955         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5956         (non_rewritable_lvalue_p): Likewise, use poly-ints.
5957
5958 2018-02-16  Martin Liska  <mliska@suse.cz>
5959
5960         PR sanitizer/84307
5961         * internal-fn.def (ASAN_CHECK): Set proper flags.
5962         (ASAN_MARK): Likewise.
5963
5964 2018-02-16  Julia Koval  <julia.koval@intel.com>
5965
5966         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5967         from PTA_CANNONLAKE.
5968
5969 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
5970
5971         PR target/84272
5972         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5973         Use ++iter rather than iter++ for std::list iterators.
5974         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
5975         defer deleting them until all nodes in the forest are processed.  Do
5976         free even leaf nodes.  Change to_process into auto_vec.
5977
5978         PR bootstrap/84405
5979         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5980         * vec.h (vec_default_construct): Use memset instead of placement new
5981         if BROKEN_VALUE_INITIALIZATION is defined.
5982         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5983         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5984         is defined.
5985
5986         PR rtl-optimization/83723
5987         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5988         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
5989         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
5990         recursive calls.
5991         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5992         callers.
5993         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5994
5995 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
5996
5997         PR rtl-optimization/81443
5998         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5999         from inner REGs to paradoxical SUBREGs.
6000
6001 2018-02-16  Richard Biener  <rguenther@suse.de>
6002
6003         PR tree-optimization/84399
6004         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6005         For operands we can analyze at their definition make sure we can
6006         analyze them at each use as well.
6007
6008 2018-02-16  Richard Biener  <rguenther@suse.de>
6009
6010         PR tree-optimization/84190
6011         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6012         volatile accesses if the decl isn't volatile.
6013
6014 2018-02-15  Jason Merrill  <jason@redhat.com>
6015
6016         PR c++/84314 - ICE with templates and fastcall attribute.
6017         * attribs.c (build_type_attribute_qual_variant): Don't clobber
6018         TYPE_CANONICAL on an existing type.
6019
6020 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
6021
6022         PR tree-optimization/84383
6023         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6024         dstoff nor call operand_equal_p if dstbase is NULL.
6025
6026         PR tree-optimization/84334
6027         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6028         also a CONSTANT_CLASS_P, punt.
6029
6030 2018-02-14  Jim Wilson  <jimw@sifive.com>
6031
6032         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6033         first SMALL_OPERAND check.  New local min_second_step.  Move assert
6034         to where locals are set.  Add TARGET_RVC support.
6035         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6036
6037 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
6038
6039         * doc/invoke.texi: Correct -Wformat-overflow code sample.
6040
6041 2018-02-14  Martin Sebor  <msebor@redhat.com>
6042
6043         PR tree-optimization/83698
6044         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6045         arrays constrain the offset range to their bounds.
6046         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6047         (builtin_access::overlap): Avoid setting the size of overlap if it's
6048         already been set.
6049         (maybe_diag_overlap): Also consider arrays when deciding what values
6050         of offsets to include in diagnostics.
6051
6052 2018-02-14  Martin Sebor  <msebor@redhat.com>
6053
6054         PR c/84108
6055         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6056         that correspond to the kind of a declaration.
6057
6058 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
6059
6060         PR target/83984
6061         * config/pa/pa.md: Load address of PIC label using the linkage table
6062         if the label is nonlocal.
6063
6064 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6065
6066         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6067         warning message if user requests -maltivec=be.
6068         * doc/invoke.texi: Document deprecation of -maltivec=be.
6069
6070 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
6071
6072         PR target/84220
6073         * config/rs6000/rs6000-c.c: Update definitions for
6074         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6075         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6076
6077 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6078
6079         PR target/84239
6080         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6081         add _get_ssp intrinsics. Remove argument from
6082         __builtin_ia32_rdssp[d|q].
6083         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6084         * config/i386/i386-builtin.def: Remove argument from
6085         __builtin_ia32_rdssp[d|q].
6086         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6087         ix86_expand_special_args_builtin for _rdssp[d|q].
6088         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6089         Clear register before usage.
6090         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6091         Add documentation for new _get_ssp and _inc_ssp intrinsics.
6092
6093 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
6094
6095         PR tree-optimization/84357
6096         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6097         operand 1 of an ARRAY_REF too.
6098
6099 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
6100
6101         PR target/83831
6102         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6103         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6104         declarations.
6105         (set_of_reg): New struct.
6106         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6107         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6108         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6109         functions.
6110         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6111         Split into bitclr, bitset, bitinvert patterns if appropriate.
6112         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6113         use rx_fuse_in_memory_bitop.
6114         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6115         to named insn, correct maximum insn length.
6116
6117 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
6118
6119         PR target/79242
6120         * machmode.def: Define a complex mode for PARTIAL_INT.
6121         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6122         MODE_PARTIAL_INT.
6123         * doc/rtl.texi: Document CSPImode.
6124         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6125         handling.
6126         (msp430_hard_regno_nregs_with_padding): Likewise.
6127
6128 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
6129
6130         PR target/84279
6131         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6132
6133 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
6134
6135         PR rtl-optimization/84169
6136         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
6137         we generated a parallel as new i3 and we split that to new i2 and i3
6138         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
6139         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
6140         those to i2, not i1.  Partially rewrite this scan code.
6141
6142 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
6143
6144         PR c/82210
6145         * stor-layout.c (place_field): For variable length fields, adjust
6146         offset_align afterwards not just based on the field's alignment,
6147         but also on the size.
6148
6149         PR middle-end/84309
6150         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6151         of exps and logs in the use_exp2 case.
6152
6153 2018-02-13  Jeff Law  <law@redhat.com>
6154
6155         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6156         entry for "vector".
6157
6158         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6159         ARGS as unused.
6160
6161 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
6162
6163         PR debug/84342
6164         PR debug/84319
6165         * common.opt (gas-loc-support, gas-locview-support): New.
6166         (ginline-points, ginternal-reset-location-views): New.
6167         * doc/invoke.texi: Document them.  Use @itemx where intended.
6168         (gvariable-location-views): Adjust.
6169         * target.def (reset_location_view): New.
6170         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6171         (TARGET_RESET_LOCATION_VIEW): New.
6172         * doc/tm.texi: Rebuilt.
6173         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6174         (dwarf2out_default_as_locview_support): New.
6175         (output_asm_line_debug_info): Use option variables.
6176         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6177         (output_loc_list): Likewise.
6178         (add_high_low_attributes): Check option variables.
6179         Don't output entry view attribute in strict mode.
6180         (gen_inlined_subroutine_die): Check option variables.
6181         (dwarf2out_inline_entry): Likewise.
6182         (init_sections_and_labels): Likewise.
6183         (dwarf2out_early_finish): Likewise.
6184         (maybe_reset_location_view): New, from...
6185         (dwarf2out_var_location): ... here.  Call it.
6186         * debug.h (dwarf2out_default_as_loc_support): Declare.
6187         (dwarf2out_default_as_locview_support): Declare.
6188         * hooks.c (hook_int_rtx_insn_0): New.
6189         * hooks.h (hook_int_rtx_insn_0): Declare.
6190         * toplev.c (process_options): Take -gas-loc-support and
6191         -gas-locview-support from dwarf2out.  Enable
6192         -gvariable-location-views by default only with locview
6193         assembler support.  Enable -ginternal-reset-location-views by
6194         default only if the target defines the corresponding hook.
6195         Enable -ginline-points by default if location views are
6196         enabled; force it disabled if statement frontiers are
6197         disabled.
6198         * tree-inline.c (expand_call_inline): Check option variables.
6199         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6200
6201 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
6202
6203         PR tree-optimization/84321
6204         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6205         handling.  Also check whether the anti-range contains any values
6206         that satisfy the mask; switch to a VR_RANGE if not.
6207
6208 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
6209
6210         PR sanitizer/84340
6211         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6212
6213 2018-02-13  Martin Jambor  <mjambor@suse.cz>
6214
6215         PR c++/83990
6216         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6217         of call statements, also set location of a load to a temporary.
6218
6219 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
6220
6221         * config/rl78/rl78.c (add_vector_labels): New function.
6222         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6223         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6224         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6225         which checks that no arguments are passed.
6226         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6227         * doc/extend.texi: Documentation for the new attribute.
6228
6229 2018-02-13  Andreas Schwab  <schwab@suse.de>
6230
6231         * config/riscv/linux.h (CPP_SPEC): Define.
6232
6233 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
6234
6235         PR target/84335
6236         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6237         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6238         OPTION_MASK_ISA_AES as first argument to def_builtin_const
6239         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6240         instead of OPTION_MASK_ISA_PCLMUL as first argument to
6241         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6242         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6243         temporarily for AES and PCLMUL builtins.
6244
6245         PR tree-optimization/84339
6246         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6247         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6248         Formatting fixes.
6249
6250         PR middle-end/84309
6251         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6252         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6253         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6254         inline function.
6255         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6256         inline function.
6257         * omp-simd-clone.h: New file.
6258         * omp-simd-clone.c: Include omp-simd-clone.h.
6259         (expand_simd_clones): No longer static.
6260         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6261         cgraph.h and omp-simd-clone.h.
6262         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6263         (vect_recog_widen_shift_pattern): Formatting fix.
6264         (vect_pattern_recog_1): Don't check optab for calls.
6265
6266         PR target/84336
6267         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6268         operands[2] into a REG before using gen_lowpart on it.
6269
6270 2018-02-12  Jeff Law  <law@redhat.com>
6271
6272         PR target/83760
6273         * config/sh/sh.c (find_barrier): Consider a sibling call
6274         a barrier as well.
6275
6276         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6277         successfully back substituting a reg.
6278
6279 2018-02-12  Richard Biener  <rguenther@suse.de>
6280
6281         PR tree-optimization/84037
6282         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6283         parameter, move visited init to caller.
6284         (vect_slp_analyze_operations): Separate cost from validity
6285         check, initialize visited once for all instances.
6286         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6287         for all instances.
6288         * tree-vect-stmts.c (vect_model_simple_cost): Make early
6289         out an assert.
6290         (vect_model_promotion_demotion_cost): Likewise.
6291         (vectorizable_bswap): Guard cost modeling with !slp_node
6292         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6293         SLP stmts.
6294         (vectorizable_call): Likewise.
6295         (vectorizable_conversion): Likewise.
6296         (vectorizable_assignment): Likewise.
6297         (vectorizable_shift): Likewise.
6298         (vectorizable_operation): Likewise.
6299         (vectorizable_store): Likewise.
6300         (vectorizable_load): Likewise.
6301         (vectorizable_condition): Likewise.
6302         (vectorizable_comparison): Likewise.
6303
6304 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
6305
6306         PR sanitizer/84307
6307         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6308         (ASAN_MARK): Fix fnspec to account for return value, change pointer
6309         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6310
6311 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
6312
6313         PR middle-end/83665
6314         * params.def (inline-min-speedup): Increase from 8 to 15.
6315         (max-inline-insns-auto): Decrease from 40 to 30.
6316         * ipa-split.c (consider_split): Add some buffer for function to
6317         be considered inlining candidate.
6318         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6319         default values.
6320
6321 2018-02-12  Richard Biener  <rguenther@suse.de>
6322
6323         PR tree-optimization/84037
6324         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6325         matched stmts if we cannot swap the non-matched ones.
6326
6327 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
6328
6329         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6330         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6331         _mm_maskz_scalef_round_ss): New intrinsics.
6332         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6333         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6334         __builtin_ia32_scalefss_round): Remove.
6335         (__builtin_ia32_scalefsd_mask_round,
6336         __builtin_ia32_scalefss_mask_round): New intrinsics.
6337         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6338         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6339         ((match_operand:VF_128 2 "<round_nimm_predicate>"
6340         "<round_constraint>")): Changed to ...
6341         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6342         "<round_scalar_constraint>")): ... this.
6343         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6344         %0, %1, %2<round_op3>}"): Changed to ...
6345         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6346         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6347         %2<round_scalar_mask_op3>}"): ... this.
6348         * config/i386/subst.md (round_scalar_nimm_predicate): New.
6349
6350 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
6351
6352         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6353         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6354         (_mm_maskz_sqrt_round_ss): New intrinsics.
6355         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6356         (__builtin_ia32_sqrtsd_mask_round)
6357         (__builtin_ia32_sqrtss_mask_round): New builtins.
6358         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6359         (__builtin_ia32_sqrtss_round): Remove.
6360         (__builtin_ia32_sqrtsd_mask_round)
6361         (__builtin_ia32_sqrtss_mask_round): New builtins.
6362         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6363         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6364         ((match_operand:VF_128 1 "vector_operand"
6365         "xBm,<round_constraint>")): Changed to ...
6366         ((match_operand:VF_128 1 "vector_operand"
6367         "xBm,<round_scalar_constraint>")): ... this.
6368         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6369         %0, %2, %<iptr>1<round_op3>}): Changed to ...
6370         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6371         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6372         %<iptr>1<round_scalar_mask_op3>}): ... this.
6373         ((set_attr "prefix" "<round_prefix>")): Changed to ...
6374         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6375
6376 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
6377
6378         PR target/84266
6379         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6380         Cast vec_cmpeq result to correct type.
6381         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6382         Cast vec_cmpgt result to correct type.
6383
6384 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
6385
6386         * final.c (final_scan_insn_1): Renamed from...
6387         (final_scan_insn): ... this.  New wrapper, to recover
6388         seen from the outermost call in recursive ones.
6389         * config/sparc/sparc.c (output_return): Drop seen from call.
6390         (output_sibcall): Likewise.
6391         * config/visium/visium.c (output_branch): Likewise.
6392
6393 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
6394
6395         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6396         function label.
6397
6398 2018-02-10  Alan Modra  <amodra@gmail.com>
6399
6400         PR target/84300
6401         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6402         Specify LR as an input.
6403
6404 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
6405
6406         PR sanitizer/83987
6407         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6408         remove_member_access_dummy_vars): New functions.
6409         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6410         lower_omp_1, execute_lower_omp): Use them.
6411
6412         PR rtl-optimization/84308
6413         * shrink-wrap.c (spread_components): Release todo vector.
6414
6415 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
6416
6417         PR rtl-optimization/57193
6418         * ira-color.c (struct allocno_color_data): Add member
6419         conflict_allocno_hard_prefs.
6420         (update_conflict_allocno_hard_prefs): New.
6421         (bucket_allocno_compare_func): Add a preference based on
6422         conflict_allocno_hard_prefs.
6423         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6424         (color_allocnos): Remove a dead code.  Initiate
6425         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
6426
6427 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
6428
6429         PR target/84226
6430         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6431         constraint from =wa to wa.  Avoid a subreg on the output operand,
6432         instead use a pseudo and subreg it in a move.
6433         (p9_xxbrd_<mode>): Changed to ...
6434         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6435         (p9_xxbrd_v2df): New expander.
6436         (p9_xxbrw_<mode>): Changed to ...
6437         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6438         (p9_xxbrw_v4sf): New expander.
6439
6440 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
6441
6442         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6443
6444 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
6445
6446         PR target/83926
6447         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6448         multiply in 32-bit mode.
6449         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6450         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6451         mode.
6452
6453 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
6454
6455         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6456         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6457         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6458         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6459
6460 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
6461
6462         PR lto/84213
6463         * dwarf2out.c (is_trivial_indirect_ref): New function.
6464         (dwarf2out_late_global_decl): Do not generate a location
6465         attribute for variables that have a non-trivial DECL_VALUE_EXPR
6466         and that are not defined in the current unit.
6467
6468 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
6469
6470         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6471         instead of a libcall for UNORDERED.
6472
6473 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
6474
6475         PR target/82641
6476         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6477         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6478
6479 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
6480
6481         PR target/PR84295
6482         * config/s390/s390.c (s390_set_current_function): Invoke
6483         s390_indirect_branch_settings also if fndecl didn't change.
6484
6485 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
6486
6487         * config/rs6000/rs6000.md (blockage): Set length to zero.
6488
6489 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
6490
6491         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6492
6493 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
6494
6495         PR sanitizer/84285
6496         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6497         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6498         -static-lib*san.
6499
6500         PR debug/84252
6501         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6502         PARALLEL incoming that failed vt_get_decl_and_offset check.
6503
6504         PR middle-end/84237
6505         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6506         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6507         TREE_READONLY bit.
6508         (get_variable_section): For decls in named .bss* sections pass true as
6509         second argument to bss_initializer_p.
6510
6511 2018-02-09  Marek Polacek  <polacek@redhat.com>
6512             Jakub Jelinek  <jakub@redhat.com>
6513
6514         PR c++/83659
6515         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6516         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
6517         Sync some changes from cxx_fold_indirect_ref.
6518
6519 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
6520
6521         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6522         markers.
6523         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6524         (BLOCK_INLINE_ENTRY_LABEL): New.
6525         (dwarf2out_var_location): Disregard inline entry markers.
6526         (inline_entry_data): New struct.
6527         (inline_entry_data_hasher): New hashtable type.
6528         (inline_entry_data_hasher::hash): New.
6529         (inline_entry_data_hasher::equal): New.
6530         (inline_entry_data_table): New variable.
6531         (add_high_low_attributes): Add DW_AT_entry_pc and
6532         DW_AT_GNU_entry_view attributes if a pending entry is found
6533         in inline_entry_data_table.  Add old entry_pc attribute only
6534         if debug nonbinding markers are disabled.
6535         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6536         markers are enabled.
6537         (block_within_block_p, dwarf2out_inline_entry): New.
6538         (dwarf2out_finish): Check that no entries remained in
6539         inline_entry_data_table.
6540         * final.c (reemit_insn_block_notes): Handle inline entry notes.
6541         (final_scan_insn, notice_source_line): Likewise.
6542         (rest_of_clean_state): Skip inline entry markers.
6543         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6544         markers.
6545         * gimple.c (gimple_build_debug_inline_entry): New.
6546         * gimple.h (enum gimple_debug_subcode): Add
6547         GIMPLE_DEBUG_INLINE_ENTRY.
6548         (gimple_build_debug_inline_entry): Declare.
6549         (gimple_debug_inline_entry_p): New.
6550         (gimple_debug_nonbind_marker_p): Adjust.
6551         * insn-notes.def (INLINE_ENTRY): New.
6552         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6553         inline entry marker notes.
6554         (print_insn): Likewise.
6555         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6556         (INSN_DEBUG_MARKER_KIND): Likewise.
6557         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6558         * tree-inline.c (expand_call_inline): Build and insert
6559         debug_inline_entry stmt.
6560         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6561         inline entry blocks early, if nonbind markers are enabled.
6562         (dump_scope_block): Dump fragment info.
6563         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6564         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6565         (gimple_build_debug_inline_entry): New.
6566         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6567         Enable/disable inline entry points too.
6568         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6569         (DEBUG_INSN): Describe inline entry markers.
6570
6571         * common.opt (gvariable-location-views): New.
6572         (gvariable-location-views=incompat5): New.
6573         * config.in: Rebuilt.
6574         * configure: Rebuilt.
6575         * configure.ac: Test assembler for view support.
6576         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6577         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6578         * dwarf2out.c (var_loc_view): New typedef.
6579         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6580         (dwarf2out_locviews_in_attribute): New.
6581         (dwarf2out_locviews_in_loclist): New.
6582         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6583         (enum dw_line_info_opcode): Add LI_adv_address.
6584         (struct dw_line_info_table): Add view.
6585         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6586         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6587         (zero_view_p): New variable.
6588         (ZERO_VIEW_P): New macro.
6589         (output_asm_line_debug_info): New.
6590         (struct var_loc_node): Add view.
6591         (add_AT_view_list, AT_loc_list): New.
6592         (add_var_loc_to_decl): Add view param.  Test it against last.
6593         (new_loc_list): Add view params.  Record them.
6594         (AT_loc_list_ptr): Handle loc and view lists.
6595         (view_list_to_loc_list_val_node): New.
6596         (print_dw_val): Handle dw_val_class_view_list.
6597         (size_of_die): Likewise.
6598         (value_format): Likewise.
6599         (loc_list_has_views): New.
6600         (gen_llsym): Set vl_symbol too.
6601         (maybe_gen_llsym, skip_loc_list_entry): New.
6602         (dwarf2out_maybe_output_loclist_view_pair): New.
6603         (output_loc_list): Output view list or entries too.
6604         (output_view_list_offset): New.
6605         (output_die): Handle dw_val_class_view_list.
6606         (output_dwarf_version): New.
6607         (output_compilation_unit_header): Use it.
6608         (output_skeleton_debug_sections): Likewise.
6609         (output_rnglists, output_line_info): Likewise.
6610         (output_pubnames, output_aranges): Update version comments.
6611         (output_one_line_info_table): Output view numbers in asm comments.
6612         (dw_loc_list): Determine current endview, pass it to new_loc_list.
6613         Call maybe_gen_llsym.
6614         (loc_list_from_tree_1): Adjust.
6615         (add_AT_location_description): Create view list attribute if
6616         needed, check it's absent otherwise.
6617         (convert_cfa_to_fb_loc_list): Adjust.
6618         (maybe_emit_file): Call output_asm_line_debug_info for test.
6619         (dwarf2out_var_location): Reset views as needed.  Precompute
6620         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
6621         attribute.  Set view.
6622         (new_line_info_table): Reset next view.
6623         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6624         (dwarf2out_source_line): Likewise.  Output view resets and labels to
6625         the assembler, or select appropriate line info opcodes.
6626         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6627         (optimize_string_length): Catch it.  Adjust.
6628         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
6629         dw_val_class_view_list, and remove it if no longer needed.
6630         (hash_loc_list): Hash view numbers.
6631         (loc_list_hasher::equal): Compare them.
6632         (optimize_location_lists): Check whether a view list symbol is
6633         needed, and whether the locview attribute is present, and
6634         whether they match.  Remove the locview attribute if no longer
6635         needed.
6636         (index_location_lists): Call skip_loc_list_entry for test.
6637         (dwarf2out_finish): Call output_asm_line_debug_info for test.
6638         Use output_dwarf_version.
6639         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6640         (struct dw_val_node): Add val_view_list.
6641         * final.c (SEEN_NEXT_VIEW): New.
6642         (set_next_view_needed): New.
6643         (clear_next_view_needed): New.
6644         (maybe_output_next_view): New.
6645         (final_start_function): Rename to...
6646         (final_start_function_1): ... this.  Take pointer to FIRST,
6647         add SEEN parameter.  Emit param bindings in the initial view.
6648         (final_start_function): Reintroduce SEEN-less interface.
6649         (final): Rename to...
6650         (final_1): ... this.  Take SEEN parameter.  Output final pending
6651         next view at the end.
6652         (final): Reintroduce seen-less interface.
6653         (final_scan_insn): Output pending next view before switching
6654         sections or ending a block.  Mark the next view as needed when
6655         outputting variable locations.  Notify debug backend of section
6656         changes, and of location view changes.
6657         (rest_of_handle_final): Adjust.
6658         * toplev.c (process_options): Autodetect value for debug variable
6659         location views option.  Warn on incompat5 without -gdwarf-5.
6660         * doc/invoke.texi (gvariable-location-views): New.
6661         (gvariable-location-views=incompat5): New.
6662         (gno-variable-location-views): New.
6663
6664 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
6665
6666         PR tree-optimization/84136
6667         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6668         that the result of find_edge is non-NULL.
6669
6670 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
6671
6672         PR target/83008
6673         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6674         storing integer register in SImode.  Fix cost of 256 and 512
6675         byte aligned SSE register store.
6676
6677 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
6678
6679         * config/i386/i386.c (ix86_multiplication_cost): Fix
6680         multiplication cost for TARGET_AVX512DQ.
6681
6682 2018-02-08  Marek Polacek  <polacek@redhat.com>
6683
6684         PR tree-optimization/84238
6685         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6686         get_range_strlen.
6687
6688 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
6689
6690         PR tree-optimization/84265
6691         * tree-vect-stmts.c (vectorizable_store): Don't treat
6692         VMAT_CONTIGUOUS accesses as grouped.
6693         (vectorizable_load): Likewise.
6694
6695 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
6696
6697         PR tree-optimization/81635
6698         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6699         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6700         (test_round_for_mask): New functions.
6701         (wide_int_cc_tests): Call test_round_for_mask.
6702         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6703         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6704         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6705         range returned by get_range_info.
6706
6707 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
6708
6709         PR ipa/81360
6710         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6711         * symtab.c: Include builtins.h
6712         (symtab_node::output_to_lto_symbol_table_p): Move here
6713         from lto-streamer-out.c:output_symbol_p.
6714         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6715         (output_symbol_p): Move all logic to symtab.c
6716         (produce_symtab): Update.
6717
6718 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
6719
6720         * config/s390/s390-opts.h (enum indirect_branch): Define.
6721         * config/s390/s390-protos.h (s390_return_addr_from_memory)
6722         (s390_indirect_branch_via_thunk)
6723         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6724         (enum s390_indirect_branch_type): Define.
6725         * config/s390/s390.c (struct s390_frame_layout, struct
6726         machine_function): Remove.
6727         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6728         (indirect_branch_table_label_no, indirect_branch_table_name):
6729         Define variables.
6730         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6731         (enum s390_indirect_branch_option): Define.
6732         (s390_return_addr_from_memory): New function.
6733         (s390_handle_string_attribute): New function.
6734         (s390_attribute_table): Add new attribute handler.
6735         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6736         (s390_indirect_branch_via_thunk): New function.
6737         (s390_indirect_branch_via_inline_thunk): New function.
6738         (s390_function_ok_for_sibcall): When jumping via thunk disallow
6739         sibling call optimization for non z10 compiles.
6740         (s390_emit_call): Force indirect branch target to be a single
6741         register.  Add r1 clobber for non-z10 compiles.
6742         (s390_emit_epilogue): Emit return jump via return_use expander.
6743         (s390_reorg): Handle JUMP_INSNs as execute targets.
6744         (s390_option_override_internal): Perform validity checks for the
6745         new command line options.
6746         (s390_indirect_branch_attrvalue): New function.
6747         (s390_indirect_branch_settings): New function.
6748         (s390_set_current_function): Invoke s390_indirect_branch_settings.
6749         (s390_output_indirect_thunk_function):  New function.
6750         (s390_code_end): Implement target hook.
6751         (s390_case_values_threshold): Implement target hook.
6752         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6753         macros.
6754         * config/s390/s390.h (struct s390_frame_layout)
6755         (struct machine_function): Move here from s390.c.
6756         (TARGET_INDIRECT_BRANCH_NOBP_RET)
6757         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6758         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6759         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6760         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6761         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6762         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6763         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6764         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6765         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6766         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6767         (mnemonic attribute): Add values which aren't recognized
6768         automatically.
6769         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6770         pattern for branch conversion.  Fix mnemonic attribute.
6771         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6772         indirect branch via thunk if requested.
6773         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6774         ("*indirect_jump"): Disable for branch conversion using out of
6775         line thunks.
6776         ("indirect_jump_via_thunk<mode>_z10")
6777         ("indirect_jump_via_thunk<mode>")
6778         ("indirect_jump_via_inlinethunk<mode>_z10")
6779         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6780         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6781         ("casesi_jump_via_inlinethunk<mode>_z10")
6782         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6783         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6784         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6785         ("*indirect2_jump"): Disable for branch conversion.
6786         ("casesi_jump"): Turn into expander and expand patterns for branch
6787         conversion.
6788         ("return_use"): New expander.
6789         ("*return"): Emit return via thunk and rename it to ...
6790         ("*return<mode>"): ... this one.
6791         * config/s390/s390.opt: Add new options and and enum for the
6792         option values.
6793
6794 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
6795
6796         * lra-constraints.c (match_reload): Unconditionally use
6797         gen_lowpart_SUBREG, rather than selecting between that
6798         and equivalent gen_rtx_SUBREG code.
6799
6800 2018-02-08  Richard Biener  <rguenther@suse.de>
6801
6802         PR tree-optimization/84233
6803         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6804         changed flag instead of boguously re-using phi_inserted.
6805
6806 2018-02-08  Martin Jambor  <mjambor@suse.cz>
6807
6808         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6809         static local variables.
6810
6811 2018-02-08  Richard Biener  <rguenther@suse.de>
6812
6813         PR tree-optimization/84278
6814         * tree-vect-stmts.c (vectorizable_store): When looking for
6815         smaller vector types to perform grouped strided loads/stores
6816         make sure the mode is supported by the target.
6817         (vectorizable_load): Likewise.
6818
6819 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
6820
6821         * config/aarch64/aarch64.c (aarch64_components_for_bb):
6822         Increase LDP/STP opportunities by adding adjacent callee-saves.
6823
6824 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
6825
6826         PR rtl-optimization/84068
6827         PR rtl-optimization/83459
6828         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6829
6830 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
6831
6832         PR tree-optimization/84224
6833         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6834         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6835         non-zero arguments.
6836
6837 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
6838
6839         PR target/84113
6840         * config/rs6000/altivec.md (*restore_world): Remove LR use.
6841         * config/rs6000/predicates.md (restore_world_operation): Adjust op
6842         count, remove one USE.
6843
6844 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
6845
6846         * doc/install.texi (Configuration): Document the
6847         --with-long-double-format={ibm,ieee} PowerPC configuration
6848         options.
6849
6850         PR target/84154
6851         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6852         Convert from define_expand to be define_insn_and_split.  Rework
6853         float/double/_Float128 conversions to QI/HI/SImode to work with
6854         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
6855         conversions to QI/HImode types did a store and then a load to
6856         truncate the value.  For conversions to VSX registers, don't split
6857         the insn, instead emit the code directly.  Use the code iterator
6858         any_fix to combine signed and unsigned conversions.
6859         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6860         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6861         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6862         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6863         (fix_<mode>di2_hw): Likewise.
6864         (fixuns_<mode>di2_hw): Likewise.
6865         (fix_<mode>si2_hw): Likewise.
6866         (fixuns_<mode>si2_hw): Likewise.
6867         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6868         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6869         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6870         fix<uns>_trunc<SFDF:mode>si2_p8.
6871         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6872         used.
6873         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6874         (fix<uns>_<mode>_mem): Likewise.
6875         (fctiw<u>z_<mode>_mem): Likewise.
6876         (fix<uns>_<mode>_mem): Likewise.
6877         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6878         the register allocator from doing a direct move to the GPRs to do
6879         a store, and instead use the ISA 3.0 store byte/half-word from
6880         vector register instruction.  For IEEE 128-bit floating point,
6881         also optimize stores of 32-bit ints.
6882         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6883
6884 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
6885
6886         * genextract.c (push_pathstr_operand): New function to support
6887         [a-zA-Z].
6888         (walk_rtx): Call push_pathstr_operand.
6889         (print_path): Support [a-zA-Z].
6890
6891 2018-02-07  Richard Biener  <rguenther@suse.de>
6892
6893         PR tree-optimization/84037
6894         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6895         (cse_and_gimplify_to_preheader): Declare.
6896         (vect_get_place_in_interleaving_chain): Likewise.
6897         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6898         ivexpr_map.
6899         (_loop_vec_info::~_loop_vec_info): Delete it.
6900         (cse_and_gimplify_to_preheader): New function.
6901         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6902         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6903         (vectorizable_load): Likewise.  For grouped stores always base
6904         the IV on the first element.
6905         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6906         condition before gimplifying.
6907
6908 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
6909
6910         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6911         *DIV_EXPR and *MOD_EXPR.
6912
6913 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
6914
6915         PR target/84248
6916         * config/i386/i386.c (ix86_option_override_internal): Mask out
6917         the CF_SET bit when checking -fcf-protection.
6918
6919 2018-02-07  Tom de Vries  <tom@codesourcery.com>
6920
6921         PR libgomp/84217
6922         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6923         enough.
6924
6925 2018-02-07  Richard Biener  <rguenther@suse.de>
6926
6927         PR tree-optimization/84204
6928         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6929         this place.
6930
6931         PR tree-optimization/84205
6932         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6933         special-case isl_ast_op_zdiv_r.
6934
6935         PR tree-optimization/84223
6936         * graphite-scop-detection.c (gather_bbs::before_dom_children):
6937         Only add conditions from within the region.
6938         (gather_bbs::after_dom_children): Adjust.
6939
6940 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
6941
6942         PR target/84209
6943         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6944         * config/avr/avr.md: Only post-reload split REG-REG moves if
6945         either register is GENERAL_REG_P.
6946
6947 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
6948
6949         PR tree-optimization/84235
6950         * tree-ssa-scopedtables.c
6951         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6952         if the subtraction is performed in floating point type where NaNs are
6953         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6954         build 1.  Formatting fix.
6955
6956 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
6957
6958         PR target/84146
6959         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6960         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6961         and skip it regardless of bb boundaries.  Use CALL_P macro,
6962         don't test INSN_P (insn) together with CALL_P or JUMP_P check
6963         unnecessarily, formatting fix.
6964
6965 2018-02-06  Michael Collison  <michael.collison@arm.com>
6966
6967         * config/arm/thumb2.md:
6968         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6969         (*thumb_mov_notscc): Ditto.
6970
6971 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
6972
6973         PR target/84154
6974         * config/rs6000/rs6000.md (su code attribute): Use "u" for
6975         unsigned_fix, not "s".
6976
6977 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6978
6979         * configure.ac (gcc_fn_eh_frame_ro): New function.
6980         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6981         correct .eh_frame permissions.
6982         * configure: Regenerate.
6983
6984 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
6985
6986         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6987         irrelevant options.
6988
6989 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6990
6991         * config/rs6000/rs6000.c (rs6000_option_override_internal):
6992         Display warning message for -mno-speculate-indirect-jumps.
6993
6994 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
6995
6996         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6997         Undocumented.
6998         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6999
7000 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
7001
7002         PR tree-optimization/84225
7003         * tree-eh.c (find_trapping_overflow): Only call
7004         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7005
7006 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7007
7008         PR target/84145
7009         * config/i386/i386.c: Reimplement the check of possible options
7010         -mibt/-mshstk conbination. Change error messages.
7011         * doc/invoke.texi: Fix a typo: remove extra '='.
7012
7013 2018-02-06  Marek Polacek  <polacek@redhat.com>
7014
7015         PR tree-optimization/84228
7016         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7017
7018 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
7019
7020         PR target/82641
7021         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7022         emitted arch directives.
7023         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7024         __ARM_FEATURE_COPROC before changing architectures.
7025
7026 2018-02-06  Richard Biener  <rguenther@suse.de>
7027
7028         * config/i386/i386.c (print_reg): Fix typo.
7029         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7030
7031 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
7032
7033         * configure: Regenerate.
7034
7035 2018-02-05  Martin Sebor  <msebor@redhat.com>
7036
7037         PR tree-optimization/83369
7038         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7039         inlining context.
7040
7041 2018-02-05  Martin Liska  <mliska@suse.cz>
7042
7043         * doc/invoke.texi: Cherry-pick upstream r323995.
7044
7045 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
7046
7047         * ira.c (ira_init_register_move_cost): Adjust comment.
7048
7049 2018-02-05  Martin Liska  <mliska@suse.cz>
7050
7051         PR gcov-profile/84137
7052         * doc/gcov.texi: Fix typo in documentation.
7053
7054 2018-02-05  Martin Liska  <mliska@suse.cz>
7055
7056         PR gcov-profile/83879
7057         * doc/gcov.texi: Document necessity of --dynamic-list-data when
7058         using dlopen functionality.
7059
7060 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
7061
7062         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7063         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7064         _mm_maskz_range_ss, _mm_mask_range_round_ss,
7065         _mm_maskz_range_round_ss): New intrinsics.
7066         (__builtin_ia32_rangesd128_round)
7067         (__builtin_ia32_rangess128_round): Remove.
7068         (__builtin_ia32_rangesd128_mask_round,
7069         __builtin_ia32_rangess128_mask_round): New builtins.
7070         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7071         __builtin_ia32_rangess128_round): Remove.
7072         (__builtin_ia32_rangesd128_mask_round,
7073         __builtin_ia32_rangess128_mask_round): New builtins.
7074         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7075         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7076         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7077         "<round_saeonly_constraint>")): Changed to ...
7078         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7079         "<round_saeonly_scalar_constraint>")): ... this.
7080         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7081         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7082         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7083         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7084         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7085
7086 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
7087
7088         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7089         options.
7090         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7091         Remove all values except native, 8540 and 8548.
7092
7093 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
7094
7095         * config/i386/i386.c (ix86_output_function_return): Pass
7096         INVALID_REGNUM, instead of -1, as invalid register number to
7097         indirect_thunk_name and output_indirect_thunk.
7098
7099 2018-02-02  Julia Koval  <julia.koval@intel.com>
7100
7101         * config.gcc: Add -march=icelake.
7102         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7103         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7104         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7105         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7106         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7107         (processor_target_table): Add icelake.
7108         (ix86_option_override_internal): Handle new PTAs.
7109         (get_builtin_code_for_version): Handle icelake.
7110         (M_INTEL_COREI7_ICELAKE): New.
7111         (fold_builtin_cpu): Handle icelake.
7112         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7113         * doc/invoke.texi: Add -march=icelake.
7114
7115 2018-02-02  Julia Koval  <julia.koval@intel.com>
7116
7117         * config/i386/i386.c (ix86_option_override_internal): Change flags type
7118         to wide_int_bitmask.
7119         * wide-int-bitmask.h: New.
7120
7121 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7122
7123         PR target/84066
7124         * config/i386/i386.md: Replace Pmode with word_mode in
7125         builtin_setjmp_setup and builtin_longjmp to support x32.
7126
7127 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
7128
7129         PR target/56010
7130         PR target/83743
7131         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7132         #include "opts.h".
7133         (rs6000_supported_cpu_names): New static variable.
7134         (linux_cpu_translation_table): Likewise.
7135         (elf_platform) <cpu>: Define new static variable and use it.
7136         Translate kernel AT_PLATFORM name to canonical name if needed.
7137         Error if platform name is unknown.
7138
7139 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
7140
7141         PR target/84089
7142         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7143
7144 2018-02-01  Jeff Law  <law@redhat.com>
7145
7146         PR target/84128
7147         * config/i386/i386.c (release_scratch_register_on_entry): Add new
7148         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
7149         the scratch if RELEASE_VIA_POP is false.
7150         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7151         If we have to save a temporary register, decrement SIZE appropriately.
7152         Pass new arguments to release_scratch_register_on_entry.
7153         (ix86_adjust_stack_and_probe): Likewise.
7154         (ix86_emit_probe_stack_range): Pass new arguments to
7155         release_scratch_register_on_entry.
7156
7157 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
7158
7159         PR rtl-optimization/84157
7160         * combine.c (change_zero_ext): Use REG_P predicate in
7161         front of HARD_REGISTER_P predicate.
7162
7163 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
7164
7165         * config/avr/avr.c (avr_option_override): Move disabling of
7166         -fdelete-null-pointer-checks to...
7167         * common/config/avr/avr-common.c (avr_option_optimization_table):
7168         ...here.
7169
7170 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7171
7172         PR tree-optimization/81635
7173         * tree-data-ref.c (split_constant_offset_1): For types that
7174         wrap on overflow, try to use range info to prove that wrapping
7175         cannot occur.
7176
7177 2018-02-01  Renlin Li  <renlin.li@arm.com>
7178
7179         PR target/83370
7180         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7181         TAILCALL_ADDR_REGS.
7182         (aarch64_register_move_cost): Likewise.
7183         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7184         TAILCALL_ADDR_REGS.
7185         (REG_CLASS_NAMES): Likewise.
7186         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7187         TAILCALL_ADDR_REGS. Remove IP registers.
7188         * config/aarch64/aarch64.md (Ucs): Update register constraint.
7189
7190 2018-02-01  Richard Biener  <rguenther@suse.de>
7191
7192         * domwalk.h (dom_walker::dom_walker): Add additional constructor
7193         for specifying RPO order and allow NULL for that.
7194         * domwalk.c (dom_walker::dom_walker): Likewise.
7195         (dom_walker::walk): Handle NULL RPO order.
7196         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7197         in RPO order.
7198         (rewrite_update_dom_walker): Likewise.
7199         (mark_def_dom_walker): Likewise.
7200
7201 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7202
7203         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7204         (aarch64_maybe_expand_sve_subreg_move): Declare.
7205         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7206         * config/aarch64/predicates.md (aarch64_any_register_operand): New
7207         predicate.
7208         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7209         that are semantically a reverse operation.
7210         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7211         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7212         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7213         functions.
7214         (aarch64_can_change_mode_class): For big-endian, forbid changes
7215         between two SVE modes if they have different element sizes.
7216
7217 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7218
7219         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7220         the TImode handling for big-endian targets.
7221
7222 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7223
7224         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7225         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
7226         not just bytes.
7227         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7228         Remove BSWAP handing for big-endian targets and use the form of
7229         LD1RQ appropariate for the mode.
7230
7231 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7232
7233         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7234         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7235         duplicated element.
7236
7237 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7238
7239         PR tearget/83845
7240         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7241         check for operands that need to go through aarch64_sve_reload_be.
7242
7243 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
7244
7245         PR tree-optimization/81661
7246         PR tree-optimization/84117
7247         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7248         * tree-eh.c: Include gimplify.h.
7249         (find_trapping_overflow, replace_trapping_overflow,
7250         rewrite_to_non_trapping_overflow): New functions.
7251         * tree-vect-loop.c: Include tree-eh.h.
7252         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7253         * tree-data-ref.c: Include tree-eh.h.
7254         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7255
7256 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
7257
7258         PR rtl-optimization/84123
7259         * combine.c (change_zero_ext): Check if hard register satisfies
7260         can_change_dest_mode before calling gen_lowpart_SUBREG.
7261
7262 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
7263
7264         PR target/82444
7265         * ira.c (ira_init_register_move_cost): Remove assert.
7266
7267 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
7268
7269         PR rtl-optimization/84071
7270         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7271         * doc/tm.texi: Regenerate.
7272
7273 2018-01-31  Richard Biener  <rguenther@suse.de>
7274
7275         PR tree-optimization/84132
7276         * tree-data-ref.c (analyze_miv_subscript): Properly
7277         check whether evolution_function_is_affine_multivariate_p
7278         before calling gcd_of_steps_may_divide_p.
7279
7280 2018-01-31  Julia Koval  <julia.koval@intel.com>
7281
7282         PR target/83618
7283         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7284         * config/i386/i386.md (rdpid_rex64) New.
7285         (rdpid): Make 32bit only.
7286
7287 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
7288
7289         PR lto/84105
7290         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7291         an IDENTIFIER_NODE for FUNCTION_TYPE's.
7292
7293 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
7294
7295         Revert
7296         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
7297
7298         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7299
7300 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
7301
7302         PR rtl-optimization/84071
7303         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7304         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7305
7306 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
7307
7308         * config/arc/arc.c (arc_handle_aux_attribute): New function.
7309         (arc_attribute_table): Add 'aux' attribute.
7310         (arc_in_small_data_p): Consider aux like variables.
7311         (arc_is_aux_reg_p): New function.
7312         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7313         (arc_get_aux_arg): New function.
7314         (prepare_move_operands): Handle aux-register access.
7315         (arc_handle_aux_attribute): New function.
7316         * doc/extend.texi (ARC Variable attributes): Add subsection.
7317
7318 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
7319
7320         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7321         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7322         (arc_attribute_table): Add 'uncached' attribute.
7323         (arc_print_operand): Print '.di' flag for uncached memory
7324         accesses.
7325         (arc_in_small_data_p): Do not consider for small data the uncached
7326         types.
7327         (arc_is_uncached_mem_p): New function.
7328         * config/arc/predicates.md (compact_store_memory_operand): Check
7329         for uncached memory accesses.
7330         (nonvol_nonimm_operand): Likewise.
7331         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7332
7333 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
7334
7335         PR c/84100
7336         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7337         falign-loops=): Add Optimization flag.
7338
7339 2018-01-30  Jeff Law  <law@redhat.com>
7340
7341         PR target/84064
7342         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7343         INT_REGISTERS_SAVED.  Check it prior to calling
7344         get_scratch_register_on_entry.
7345         (ix86_adjust_stack_and_probe): Similarly.
7346         (ix86_emit_probe_stack_range): Similarly.
7347         (ix86_expand_prologue): Corresponding changes.
7348
7349 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7350
7351         PR target/40411
7352         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7353         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7354
7355 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
7356
7357         PR target/84112
7358         * lra-constraints.c (curr_insn_transform): Process AND in the
7359         address.
7360
7361 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
7362
7363         PR rtl-optimization/83986
7364         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7365         dependence against last_pending_memory_flush in addition to
7366         pending_jump_insns.
7367
7368 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
7369
7370         PR tree-optimization/81611
7371         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7372         copies.
7373
7374 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
7375
7376         PR target/83758
7377         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7378         a reg rtx.
7379
7380 2018-01-30  Richard Biener  <rguenther@suse.de>
7381             Jakub Jelinek  <jakub@redhat.com>
7382
7383         PR tree-optimization/84111
7384         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7385         inner loops added during recursion, as they don't have up-to-date
7386         SSA form.
7387
7388 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
7389
7390         PR ipa/81360
7391         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7392         (can_inline_edge_by_limits_p): ... here.
7393         (can_early_inline_edge_p, check_callers,
7394         update_caller_keys, update_callee_keys, recursive_inlining,
7395         add_new_edges_to_heap, speculation_useful_p,
7396         inline_small_functions,
7397         inline_small_functions, flatten_function,
7398         inline_to_all_callers_1): Update.
7399
7400 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
7401
7402         * profile-count.c (profile_count::combine_with_ipa_count): Handle
7403         zeros correctly.
7404
7405 2018-01-30  Richard Biener  <rguenther@suse.de>
7406
7407         PR tree-optimization/83008
7408         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7409         invariant and constant vector uses in stmts when they need
7410         more than one stmt.
7411
7412 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7413
7414         PR bootstrap/84017
7415         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7416         * configure: Regenerate.
7417
7418 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
7419
7420         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7421         pattern.
7422         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7423         Use gen_rtx_REG rather than gen_lowpart.
7424
7425 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
7426
7427         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7428         rather than 0 when creating partial subregs.
7429
7430 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
7431
7432         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7433         of usage.
7434
7435 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
7436
7437         PR target/81550
7438         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7439         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7440         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7441         flags.  This restores the settings used before the 2017-07-24.
7442         Turning off pre increment/decrement/modify allows IVOPTS to
7443         optimize DF/SF loops where the index is an int.
7444
7445 2018-01-29  Richard Biener  <rguenther@suse.de>
7446             Kelvin Nilsen  <kelvin@gcc.gnu.org>
7447
7448         PR bootstrap/80867
7449         * tree-vect-stmts.c (vectorizable_call): Don't call
7450         targetm.vectorize_builtin_md_vectorized_function if callee is
7451         NULL.
7452
7453 2018-01-22  Carl Love  <cel@us.ibm.com>
7454
7455         * doc/extend.tex: Fix typo in second arg in
7456         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7457
7458 2018-01-29  Richard Biener  <rguenther@suse.de>
7459
7460         PR tree-optimization/84086
7461         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7462         (flush_ssaname_freelist): When SSA names were released reset
7463         the SCEV hash table.
7464
7465 2018-01-29  Richard Biener  <rguenther@suse.de>
7466
7467         PR tree-optimization/84057
7468         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7469         removed paths when removing edges.
7470
7471 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
7472
7473         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7474         -mfunction-return=@var{choice}.
7475
7476 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7477
7478         PR diagnostic/84034
7479         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7480         Handle CR like TAB.
7481         (layout::print_source_line): Likewise.
7482         (test_get_line_width_without_trailing_whitespace): Add test cases.
7483
7484 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
7485
7486         PR middle-end/84040
7487         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7488         debug insns.
7489
7490 2018-01-26  Jim Wilson  <jimw@sifive.com>
7491
7492         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7493
7494         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7495         specified.
7496
7497 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7498
7499         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7500         and CMP + SUB-immediate -> SUBS.
7501
7502 2018-01-26  Martin Sebor  <msebor@redhat.com>
7503
7504         PR tree-optimization/83896
7505         * tree-ssa-strlen.c (get_string_len): Rename...
7506         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
7507         Avoid assuming length is constant.
7508         (handle_char_store): Use HOST_WIDE_INT for string length.
7509
7510 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
7511
7512         PR target/81763
7513         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7514         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7515
7516 2018-01-26  Richard Biener  <rguenther@suse.de>
7517
7518         PR rtl-optimization/84003
7519         * dse.c (record_store): Only record redundant stores when
7520         the earlier store aliases at least all accesses the later one does.
7521
7522 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
7523
7524         PR rtl-optimization/83985
7525         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7526         REG_CFA_RESTORE insns.
7527         (delete_unmarked_insns): Don't ignore separate shrink wrapping
7528         REG_CFA_RESTORE insns here.
7529
7530         PR c/83989
7531         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7532         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7533
7534 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
7535
7536         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7537         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7538         (arc_init): Likewise.
7539         (arc_override_options): Likewise.
7540         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7541         value.
7542         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7543         support.
7544         * config/arc/arc.h (TARGET_DBNZ): Define.
7545         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7546         properly set the tune attribute.
7547         (dbnz): Use TARGET_DBNZ guard.
7548         * config/arc/arc.opt (mtune): Add core3 option.
7549
7550 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
7551
7552         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7553         recognize new pic like addresses.
7554         (arc_delegitimize_address): Clean up.
7555
7556 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
7557
7558         * config/arc/arc-arches.def: Option mrf16 valid for all
7559         architectures.
7560         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7561         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7562         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7563         * config/arc/arc-tables.opt: Regenerate.
7564         * config/arc/arc.c (arc_conditional_register_usage): Handle
7565         reduced register file case.
7566         (arc_file_start): Set must have build attributes.
7567         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7568         mrf16 option value.
7569         * config/arc/arc.opt (mrf16): Add new option.
7570         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7571         * config/arc/genmultilib.awk: Handle new mrf16 option.
7572         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7573         * config/arc/t-multilib: Regenerate.
7574         * doc/invoke.texi (ARC Options): Document mrf16 option.
7575
7576 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
7577
7578         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7579         * config/arc/arc.c (arc_handle_secure_attribute): New function.
7580         (arc_attribute_table): Add 'secure_call' attribute.
7581         (arc_print_operand): Print secure call operand.
7582         (arc_function_ok_for_sibcall): Don't optimize tail calls when
7583         secure.
7584         (arc_is_secure_call_p): New function.  * config/arc/arc.md
7585         (call_i): Add support for sjli instruction.
7586         (call_value_i): Likewise.
7587         * config/arc/constraints.md (Csc): New constraint.
7588
7589 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
7590             John Eric Martin  <John.Martin@emmicro-us.com>
7591
7592         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7593         * config/arc/arc.c (_arc_jli_section): New struct.
7594         (arc_jli_section): New type.
7595         (rc_jli_sections): New static variable.
7596         (arc_handle_jli_attribute): New function.
7597         (arc_attribute_table): Add jli_always and jli_fixed attribute.
7598         (arc_file_end): New function.
7599         (TARGET_ASM_FILE_END): Define.
7600         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7601         (arc_add_jli_section): New function.
7602         (jli_call_scan): Likewise.
7603         (arc_reorg): Call jli_call_scan.
7604         (arc_output_addsi): Remove 'S' from printing asm operand.
7605         (arc_is_jli_call_p): New function.
7606         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7607         operand.
7608         (movhi_insn): Likewise.
7609         (movsi_insn): Likewise.
7610         (movsi_set_cc_insn): Likewise.
7611         (loadqi_update): Likewise.
7612         (load_zeroextendqisi_update): Likewise.
7613         (load_signextendqisi_update): Likewise.
7614         (loadhi_update): Likewise.
7615         (load_zeroextendhisi_update): Likewise.
7616         (load_signextendhisi_update): Likewise.
7617         (loadsi_update): Likewise.
7618         (loadsf_update): Likewise.
7619         (movsicc_insn): Likewise.
7620         (bset_insn): Likewise.
7621         (bxor_insn): Likewise.
7622         (bclr_insn): Likewise.
7623         (bmsk_insn): Likewise.
7624         (bicsi3_insn): Likewise.
7625         (cmpsi_cc_c_insn): Likewise.
7626         (movsi_ne): Likewise.
7627         (movsi_cond_exec): Likewise.
7628         (clrsbsi2): Likewise.
7629         (norm_f): Likewise.
7630         (normw): Likewise.
7631         (swap): Likewise.
7632         (divaw): Likewise.
7633         (flag): Likewise.
7634         (sr): Likewise.
7635         (kflag): Likewise.
7636         (ffs): Likewise.
7637         (ffs_f): Likewise.
7638         (fls): Likewise.
7639         (call_i): Remove 'S' asm letter, add jli instruction.
7640         (call_value_i): Likewise.
7641         * config/arc/arc.op (mjli-always): New option.
7642         * config/arc/constraints.md (Cji): New constraint.
7643         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7644         operand.
7645         (subsf3_fpx): Likewise.
7646         (mulsf3_fpx): Likewise.
7647         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7648         asm operand.
7649         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7650         function attrbutes.
7651         * doc/invoke.texi (ARC): Document mjli-always option.
7652
7653 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
7654
7655         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7656         avoid addition with 0 and use incw and decw where possible.
7657
7658 2018-01-26  Richard Biener  <rguenther@suse.de>
7659
7660         PR tree-optimization/81082
7661         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7662         association if it requires casting to unsigned.
7663         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7664         from fold_plusminus_mult_expr to catch important cases late when
7665         range info is available.
7666
7667 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7668
7669         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7670         * configure.ac (hidden_linkonce): New test.
7671         * configure: Regenerate.
7672         * config.in: Regenerate.
7673
7674 2018-01-26  Julia Koval  <julia.koval@intel.com>
7675
7676         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7677         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7678         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7679         _mm_mask_bitshuffle_epi64_mask): Fix type.
7680         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7681         USI_FTYPE_V4DI_V4DI_USI): Remove.
7682         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7683         __builtin_ia32_vpshufbitqmb256_mask,
7684         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7685         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7686         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7687
7688 2018-01-26  Alan Modra  <amodra@gmail.com>
7689
7690         PR target/84033
7691         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7692         UNSPEC_VBPERMQ.  Sort other unspecs.
7693
7694 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
7695
7696         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7697
7698 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
7699
7700         PR middle-end/83055
7701         * predict.c (drop_profile): Do not push/pop cfun; update also
7702         node->count.
7703         (handle_missing_profiles): Fix logic looking for zero profiles.
7704
7705 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
7706
7707         PR middle-end/83977
7708         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7709         on functions with #pragma omp declare simd or functions with simd
7710         attribute.
7711         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7712         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7713         Remove trailing \n from warning_at calls.
7714
7715 2018-01-25  Tom de Vries  <tom@codesourcery.com>
7716
7717         PR target/84028
7718         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7719         for neutered workers.
7720
7721 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
7722
7723         PR target/68467
7724         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7725         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7726
7727 2018-01-24  Jeff Law  <law@redhat.com>
7728
7729         PR target/83994
7730         * i386.c (get_probe_interval): Move to earlier point.
7731         (ix86_compute_frame_layout): If -fstack-clash-protection and
7732         the frame is larger than the probe interval, then use pushes
7733         to save registers rather than reg->mem moves.
7734         (ix86_expand_prologue): Remove conditional for int_registers_saved
7735         assertion.
7736
7737 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
7738
7739         PR target/84014
7740         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7741         min/max for never referenced object.
7742
7743 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
7744
7745         PR middle-end/83977
7746         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7747         here.
7748         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7749         attributes from DECL_ATTRIBUTES (decl) without affecting
7750         DECL_ATTRIBUTES (current_function_decl).
7751         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7752         functions with non-NULL DECL_ABSTRACT_ORIGIN.
7753
7754 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
7755
7756         PR tree-optimization/83979
7757         * fold-const.c (fold_comparison): Use constant_boolean_node
7758         instead of boolean_{true,false}_node.
7759
7760 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
7761
7762         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7763         with zero counts.
7764
7765 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7766
7767         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7768         Simplify the clause that sets the length attribute.
7769         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7770         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7771         clause that sets the length attribute.
7772         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7773
7774 2018-01-24  Tom de Vries  <tom@codesourcery.com>
7775
7776         PR target/83589
7777         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7778         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7779         Add strict parameter.
7780         (prevent_branch_around_nothing): Insert dummy insn between branch to
7781         label and label with no ptx insn inbetween.
7782         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7783
7784 2018-01-24  Tom de Vries  <tom@codesourcery.com>
7785
7786         PR target/81352
7787         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7788         for neutered threads in warp.
7789         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7790
7791 2018-01-24  Richard Biener  <rguenther@suse.de>
7792
7793         PR tree-optimization/83176
7794         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7795         operands.
7796
7797 2018-01-24  Richard Biener  <rguenther@suse.de>
7798
7799         PR tree-optimization/82819
7800         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7801         code generating pluses that are no-ops in the target precision.
7802
7803 2018-01-24  Richard Biener  <rguenther@suse.de>
7804
7805         PR middle-end/84000
7806         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7807
7808 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
7809
7810         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7811         to merge probabilities.
7812         * predict.c (probably_never_executed): Also mark as cold functions
7813         with global 0 profile and guessed local profile.
7814         * profile-count.c (profile_probability::combine_with_count): New
7815         member function.
7816         * profile-count.h (profile_probability::operator*,
7817         profile_probability::operator*=, profile_probability::operator/,
7818         profile_probability::operator/=): Reduce precision to adjusted
7819         and set value to guessed on contradictory divisions.
7820         (profile_probability::combine_with_freq): Remove.
7821         (profile_probability::combine_wiht_count): Declare.
7822         (profile_count::force_nonzero):: Set to adjusted.
7823         (profile_count::probability_in):: Set quality to adjusted.
7824         * tree-ssa-tail-merge.c (replace_block_by): Use
7825         combine_with_count.
7826
7827 2018-01-23  Andrew Waterman  <andrew@sifive.com>
7828             Jim Wilson  <jimw@sifive.com>
7829
7830         * config/riscv/riscv.c (riscv_stack_boundary): New.
7831         (riscv_option_override): Set riscv_stack_boundary.  Handle
7832         riscv_preferred_stack_boundary_arg.
7833         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7834         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7835         (STACK_BOUNDARY): Set to riscv_stack_boundary.
7836         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7837         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7838         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7839
7840 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
7841
7842         PR target/83905
7843         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7844         of struct ix86_frame.
7845         (ix86_expand_epilogue): Likewise.  Add a local variable for
7846         the reg_save_offset field in struct ix86_frame.
7847
7848 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
7849
7850         PR tree-optimization/82604
7851         * tree-loop-distribution.c (enum partition_kind): New enum item
7852         PKIND_PARTIAL_MEMSET.
7853         (partition_builtin_p): Support above new enum item.
7854         (generate_code_for_partition): Ditto.
7855         (compute_access_range): Differentiate cases that equality can be
7856         proven at all loops, the innermost loops or no loops.
7857         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7858         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
7859         (finalize_partitions, distribute_loop): Don't fuse partition of
7860         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7861         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7862         parloop is enabled.
7863
7864 2018-01-23  Martin Liska  <mliska@suse.cz>
7865
7866         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7867         order to ignore the predictor.
7868         (PRED_POLYMORPHIC_CALL): Likewise.
7869         (PRED_RECURSIVE_CALL): Likewise.
7870
7871 2018-01-23  Martin Liska  <mliska@suse.cz>
7872
7873         * tree-profile.c (tree_profiling): Print function header to
7874         aware reader which function we are working on.
7875         * value-prof.c (gimple_find_values_to_profile): Do not print
7876         not interesting value histograms.
7877
7878 2018-01-23  Martin Liska  <mliska@suse.cz>
7879
7880         * profile-count.h (enum profile_quality): Add
7881         profile_uninitialized as the first value. Do not number values
7882         as they are zero based.
7883         (profile_count::verify): Update sanity check.
7884         (profile_probability::verify): Likewise.
7885
7886 2018-01-23  Nathan Sidwell  <nathan@acm.org>
7887
7888         * doc/invoke.texi (ffor-scope): Deprecate.
7889
7890 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
7891
7892         PR tree-optimization/83510
7893         * domwalk.c (set_all_edges_as_executable): New function.
7894         (dom_walker::dom_walker): Convert bool param
7895         "skip_unreachable_blocks" to enum reachability.  Move setup of
7896         edge flags to set_all_edges_as_executable and only do it when
7897         reachability is REACHABLE_BLOCKS.
7898         * domwalk.h (enum dom_walker::reachability): New enum.
7899         (dom_walker::dom_walker): Convert bool param
7900         "skip_unreachable_blocks" to enum reachability.
7901         (set_all_edges_as_executable): New decl.
7902         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
7903         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7904         "reachability".
7905         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7906         but converting true to REACHABLE_BLOCKS.
7907         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7908         * tree-vrp.c
7909         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7910         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7911         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7912         REACHABLE_BLOCKS.
7913         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7914         if check_all_array_refs will be called.
7915
7916 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
7917
7918         * tree.c (selftest::test_location_wrappers): Add more test
7919         coverage.
7920
7921 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
7922
7923         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7924         (selftest::test_bit_in_range): Likewise.
7925
7926 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
7927
7928         PR testsuite/83888
7929         * doc/sourcebuild.texi (vect_float): Say that the selector
7930         only describes the situation when -funsafe-math-optimizations is on.
7931         (vect_float_strict): Document.
7932
7933 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
7934
7935         PR tree-optimization/83965
7936         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7937         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7938         instead of checking only for a reduction.
7939         (vect_recog_widen_sum_pattern): Likewise.
7940
7941 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
7942
7943         * predict.c (probably_never_executed): Only use precise profile info.
7944         (compute_function_frequency): Skip after inlining hack since we now
7945         have quality checking.
7946
7947 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
7948
7949         * profile-count.h (profile_probability::very_unlikely,
7950         profile_probability::unlikely, profile_probability::even): Set
7951         precision to guessed.
7952
7953 2018-01-23  Richard Biener  <rguenther@suse.de>
7954
7955         PR tree-optimization/83963
7956         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7957         Properly terminate dominator walk when crossing the exit edge not
7958         when visiting its source block.
7959
7960 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
7961
7962         PR c++/83918
7963         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7964         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7965
7966 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
7967
7968         PR tree-optimization/83957
7969         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
7970         semicolon after for body surrounded by braces.
7971
7972         PR tree-optimization/83081
7973         * profile-count.h (profile_probability::split): New method.
7974         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7975         Use profile_probability::split.
7976         (do_compare_rtx_and_jump): Fix adjustment of probabilities
7977         when splitting a single conditional jump into 2.
7978
7979 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
7980
7981         PR tree-optimization/69452
7982         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7983         decl.
7984
7985 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
7986
7987         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7988         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7989         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7990
7991 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
7992
7993         * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7994         * config/rl78/rl78.md (movdi): New define_expand.
7995         * config/rl78/rl78.c (rl78_split_movdi): New function.
7996
7997 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
7998
7999         PR target/83862
8000         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8001         no longer used.
8002         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8003         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8004         128-bit to produce an UNSPEC move to get the double word with the
8005         signbit and then a shift directly to do signbit.
8006         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8007         implementation with a new version that just does either a direct
8008         move or a regular move.  Move memory interface to separate insns.
8009         Move insns so they are next to the expander.
8010         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8011         with signbit move.  Split big and little endian case.
8012         (signbit<mode>2_dm_mem_le): Likewise.
8013         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8014         (signbit<mode>2_dm2): Likewise.
8015
8016 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8017
8018         * config/rl78/rl78.md (anddi3): New define_expand.
8019
8020 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8021
8022         * config/rl78/rl78.md (umindi3): New define_expand.
8023
8024 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8025
8026         * config/rl78/rl78.md (smindi3): New define_expand.
8027
8028 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8029
8030         * config/rl78/rl78.md (smaxdi3): New define_expand.
8031
8032 2018-01-22  Carl Love  <cel@us.ibm.com>
8033
8034         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8035         LVX_V1TI): Add macro expansion.
8036         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8037         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8038         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8039         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8040         Change check to determine if the instruction is a byte reversing
8041         entry.  Fix typo in comment.
8042         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8043         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8044         Add def_builtin calls for new builtins.
8045         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8046         Add define_insn expansion.
8047
8048 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8049
8050         * config/rl78/rl78.md (umaxdi3): New define_expand.
8051
8052 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8053
8054         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8055         for non-QImode registers.
8056
8057 2018-01-22  Richard Biener  <rguenther@suse.de>
8058
8059         PR tree-optimization/83963
8060         * graphite-scop-detection.c (scop_detection::get_sese): Delay
8061         including the loop exit block.
8062         (scop_detection::merge_sese): Likewise.
8063         (scop_detection::add_scop): Do it here instead.
8064
8065 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8066
8067         * doc/sourcebuild.texi (arm_softfloat): Document.
8068
8069 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
8070
8071         PR gcc/77734
8072         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8073         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8074         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8075
8076 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8077             David Edelsohn  <dje.gcc@gmail.com>
8078
8079         PR target/83946
8080         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8081         Change "crset eq" to "crset 2".
8082         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8083         (*call_indirect_aix<mode>_nospec): Likewise.
8084         (*call_value_indirect_aix<mode>_nospec): Likewise.
8085         (*call_indirect_elfv2<mode>_nospec): Likewise.
8086         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8087         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8088         change assembly output from . to $.
8089         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8090         (indirect_jump<mode>_nospec): Change assembly output from . to $.
8091         (*tablejump<mode>_internal1_nospec): Likewise.
8092
8093 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
8094
8095         PR target/80870
8096         * config/sh/sh_optimize_sett_clrt.cc:
8097         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8098
8099 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
8100
8101         PR tree-optimization/83940
8102         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8103         offset_dt to vect_constant_def rather than vect_unknown_def_type.
8104         (vect_check_load_store_mask): Add a mask_dt_out parameter and
8105         use it to pass back the definition type.
8106         (vect_check_store_rhs): Likewise rhs_dt_out.
8107         (vect_build_gather_load_calls): Add a mask_dt argument and use
8108         it instead of a call to vect_is_simple_use.
8109         (vectorizable_store): Update calls to vect_check_load_store_mask
8110         and vect_check_store_rhs.  Use the dt returned by the latter instead
8111         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
8112         instead of calls to vect_is_simple_use.  Pass the scalar rather
8113         than the vector operand to vect_is_simple_use when handling
8114         second and subsequent copies of an rhs value.
8115         (vectorizable_load): Update calls to vect_check_load_store_mask
8116         and vect_build_gather_load_calls.  Use the cached mask_dt and
8117         gs_info.offset_dt instead of calls to vect_is_simple_use.
8118
8119 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
8120
8121         PR middle-end/83945
8122         * tree-emutls.c: Include gimplify.h.
8123         (lower_emutls_2): New function.
8124         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8125         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8126         it before further processing.
8127
8128         PR target/83930
8129         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8130         UINTVAL (trueop1) instead of INTVAL (op1).
8131
8132 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
8133
8134         PR debug/81570
8135         PR debug/83728
8136         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8137         INCOMING_FRAME_SP_OFFSET if not defined.
8138         (scan_trace): Add ENTRY argument.  If true and
8139         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8140         emit a note to adjust the CFA offset.
8141         (create_cfi_notes): Adjust scan_trace callers.
8142         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8143         INCOMING_FRAME_SP_OFFSET in the CIE.
8144         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8145         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8146         Likewise.
8147         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8148         * doc/tm.texi: Regenerated.
8149
8150 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8151
8152         PR rtl-optimization/83147
8153         * lra-constraints.c (remove_inheritance_pseudos): Use
8154         lra_substitute_pseudo_within_insn.
8155
8156 2018-01-19  Tom de Vries  <tom@codesourcery.com>
8157             Cesar Philippidis  <cesar@codesourcery.com>
8158
8159         PR target/83920
8160         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8161
8162 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
8163
8164         PR target/83790
8165         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8166         spaces for function labels.
8167
8168 2018-01-19  Martin Liska  <mliska@suse.cz>
8169
8170         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8171         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8172         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8173         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8174         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8175         (PRED_CONST_RETURN): Change from 69 to 65.
8176         (PRED_NULL_RETURN): Change from 91 to 71.
8177         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8178         (PRED_LOOP_GUARD): Change from 66 to 73.
8179
8180 2018-01-19  Martin Liska  <mliska@suse.cz>
8181
8182         * predict.c (predict_insn_def): Add new assert.
8183         (struct branch_predictor): Change type to signed integer.
8184         (test_prediction_value_range): Amend test to cover
8185         PROB_UNINITIALIZED.
8186         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8187         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8188         (PRED_LOOP_ITERATIONS_MAX): Likewise.
8189         (PRED_LOOP_IV_COMPARE): Likewise.
8190         * predict.h (PROB_UNINITIALIZED): Define new constant.
8191
8192 2018-01-19  Martin Liska  <mliska@suse.cz>
8193
8194         * predict.c (dump_prediction): Add new format for
8195         analyze_brprob.py script which is enabled with -details
8196         suboption.
8197         * profile-count.h (precise_p): New function.
8198
8199 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
8200
8201         PR tree-optimization/83922
8202         * tree-vect-loop.c (vect_verify_full_masking): Return false if
8203         there are no statements that need masking.
8204         (vect_active_double_reduction_p): New function.
8205         (vect_analyze_loop_operations): Use it when handling phis that
8206         are not in the loop header.
8207
8208 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
8209
8210         PR tree-optimization/83914
8211         * tree-vect-loop.c (vectorizable_induction): Don't convert
8212         init_expr or apply the peeling adjustment for inductions
8213         that are nested within the vectorized loop.
8214
8215 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8216
8217         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8218         instead of NEG.
8219
8220 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
8221
8222         PR sanitizer/81715
8223         PR testsuite/83882
8224         * function.h (gimplify_parameters): Add gimple_seq * argument.
8225         * function.c: Include gimple.h and options.h.
8226         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8227         for the added local temporaries if needed.
8228         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8229         if there are any parameter cleanups, wrap whole body into a
8230         try/finally with the cleanups.
8231
8232 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
8233
8234         PR target/82964
8235         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8236         Use GET_MODE_CLASS for scalar floating point.
8237
8238 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
8239
8240         PR ipa/82256
8241         patch by PaX Team
8242         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8243         Fix call of call_cgraph_insertion_hooks.
8244
8245 2018-01-18  Martin Sebor  <msebor@redhat.com>
8246
8247         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8248
8249 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
8250
8251         PR ipa/83619
8252         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8253         frequencies.
8254
8255 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
8256
8257         PR other/70268
8258         * common.opt: (-ffile-prefix-map): New option.
8259         * opts.c (common_handle_option): Defer it.
8260         * opts-global.c (handle_common_deferred_options): Handle it.
8261         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8262         * file-prefix-map.h: New file.
8263         (remap_debug_filename, add_debug_prefix_map): ...here.
8264         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8265         * final.c (debug_prefix_map, add_debug_prefix_map
8266         remap_debug_filename): Move to...
8267         * file-prefix-map.c: New file.
8268         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8269         generalize, get rid of alloca(), use strrchr() instead of strchr().
8270         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8271         Implement in terms of add_prefix_map().
8272         (remap_macro_filename, remap_debug_filename): Implement in term of
8273         remap_filename().
8274         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8275         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8276         * dbxout.c: Include file-prefix-map.h.
8277         * varasm.c: Likewise.
8278         * vmsdbgout.c: Likewise.
8279         * xcoffout.c: Likewise.
8280         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8281         * doc/cppopts.texi (-fmacro-prefix-map): Document.
8282         * doc/invoke.texi (-ffile-prefix-map): Document.
8283         (-fdebug-prefix-map): Update description.
8284
8285 2018-01-18  Martin Liska  <mliska@suse.cz>
8286
8287         * config/i386/i386.c (indirect_thunk_name): Document that also
8288         lfence is emitted.
8289         (output_indirect_thunk): Document why both instructions
8290         (pause and lfence) are generated.
8291
8292 2018-01-18  Richard Biener  <rguenther@suse.de>
8293
8294         PR tree-optimization/83887
8295         * graphite-scop-detection.c
8296         (scop_detection::get_nearest_dom_with_single_entry): Remove.
8297         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8298         (scop_detection::merge_sese): Re-implement with a flood-fill
8299         algorithm that properly finds a SESE region if it exists.
8300
8301 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
8302
8303         PR c/61240
8304         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8305         pointer_diff optimizations use view_convert instead of convert.
8306
8307 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8308
8309         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8310         Generate different code for -mno-speculate-indirect-jumps.
8311         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8312         (*call_indirect_aix<mode>): Disable for
8313         -mno-speculate-indirect-jumps.
8314         (*call_indirect_aix<mode>_nospec): New define_insn.
8315         (*call_value_indirect_aix<mode>): Disable for
8316         -mno-speculate-indirect-jumps.
8317         (*call_value_indirect_aix<mode>_nospec): New define_insn.
8318         (*sibcall_nonlocal_sysv<mode>): Generate different code for
8319         -mno-speculate-indirect-jumps.
8320         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8321
8322 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
8323
8324         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8325         long double type, set the flags for noting the default long double
8326         type, even if we don't pass or return a long double type.
8327
8328 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
8329
8330         PR ipa/83051
8331         * ipa-inline.c (flatten_function): Do not overwrite final inlining
8332         failure.
8333
8334 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
8335
8336         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8337         support for merge[hl].
8338         (fold_mergehl_helper): New helper function.
8339         (tree-vector-builder.h): New #include for tree_vector_builder usage.
8340         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8341         (altivec_vmrglw_direct): Add xxmrglw insn.
8342
8343 2018-01-17  Andrew Waterman  <andrew@sifive.com>
8344
8345         * config/riscv/riscv.c (riscv_conditional_register_usage): If
8346         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8347
8348 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
8349
8350         PR lto/83121
8351         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8352         call the lto_location_cache before reading the
8353         DECL_SOURCE_LOCATION of the types.
8354
8355 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
8356             Richard Sandiford  <richard.sandiford@linaro.org>
8357
8358         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8359         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8360         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8361         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8362         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8363         Add declaration.
8364         * config/aarch64/constraints.md (aarch64_movti_operand):
8365         Limit immediates.
8366         * config/aarch64/predicates.md (Uti): Add new constraint.
8367
8368 2018-01-17 Carl Love  <cel@us.ibm.com>
8369
8370         * config/rs6000/vsx.md (define_expand xl_len_r,
8371         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8372         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8373         lxvll.
8374         (define_expand, define_insn): Move the shift left from  the
8375         define_insn to the define_expand for lxvl and stxvl instructions.
8376         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8377         and XL_LEN_R definitions to PURE.
8378
8379 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
8380
8381         * config/i386/i386.c (indirect_thunk_name): Declare regno
8382         as unsigned int.  Compare regno with INVALID_REGNUM.
8383         (output_indirect_thunk): Ditto.
8384         (output_indirect_thunk_function): Ditto.
8385         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
8386         in the call to output_indirect_thunk_function.
8387
8388 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
8389
8390         PR middle-end/83884
8391         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8392         rather than the size of inner_type to determine the stack slot size
8393         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8394
8395 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
8396
8397         PR target/83546
8398         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8399         to PTA_SILVERMONT.
8400
8401 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
8402
8403         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8404         endian Linux systems to optionally enable multilibs for selecting
8405         the long double type if the user configured an explicit type.
8406         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8407         have no long double multilibs if not defined.
8408         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8409         warn if the user used -mabi={ieee,ibm}longdouble and we built
8410         multilibs for long double.
8411         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8412         appropriate multilib option.
8413         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8414         multilib options.
8415         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8416         for building long double multilibs.
8417         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8418
8419 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
8420
8421         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8422         copies.
8423
8424         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8425         64 bits.
8426         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8427         128 bits.
8428
8429         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8430         variables.
8431
8432         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8433         return value.
8434
8435 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
8436
8437         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8438         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8439
8440 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8441
8442         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8443         different rtl trees depending on TARGET_64BIT.
8444         (rs6000_gen_lvx): Likewise.
8445
8446 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
8447
8448         * config/visium/visium.md (nop): Tweak comment.
8449         (hazard_nop): Likewise.
8450
8451 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8452
8453         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8454         -mspeculate-indirect-jumps.
8455         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8456         for -mno-speculate-indirect-jumps.
8457         (*call_indirect_elfv2<mode>_nospec): New define_insn.
8458         (*call_value_indirect_elfv2<mode>): Disable for
8459         -mno-speculate-indirect-jumps.
8460         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8461         (indirect_jump): Emit different RTL for
8462         -mno-speculate-indirect-jumps.
8463         (*indirect_jump<mode>): Disable for
8464         -mno-speculate-indirect-jumps.
8465         (*indirect_jump<mode>_nospec): New define_insn.
8466         (tablejump): Emit different RTL for
8467         -mno-speculate-indirect-jumps.
8468         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8469         (tablejumpsi_nospec): New define_expand.
8470         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8471         (tablejumpdi_nospec): New define_expand.
8472         (*tablejump<mode>_internal1): Disable for
8473         -mno-speculate-indirect-jumps.
8474         (*tablejump<mode>_internal1_nospec): New define_insn.
8475         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8476         option.
8477
8478 2018-01-16  Artyom Skrobov tyomitch@gmail.com
8479
8480         * caller-save.c (insert_save): Drop unnecessary parameter.  All
8481         callers updated.
8482
8483 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
8484             Richard Biener  <rguenth@suse.de>
8485
8486         PR libgomp/83590
8487         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8488         return early, inline manually is_gimple_sizepos.  Make sure if we
8489         call gimplify_expr we don't end up with a gimple constant.
8490         * tree.c (variably_modified_type_p): Don't return true for
8491         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
8492         * gimplify.h (is_gimple_sizepos): Remove.
8493
8494 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
8495
8496         PR tree-optimization/83857
8497         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8498         vectorizable_live_operation for pure SLP statements.
8499         (vectorizable_live_operation): Handle PHIs.
8500
8501 2018-01-16  Richard Biener  <rguenther@suse.de>
8502
8503         PR tree-optimization/83867
8504         * tree-vect-stmts.c (vect_transform_stmt): Precompute
8505         nested_in_vect_loop_p since the scalar stmt may get invalidated.
8506
8507 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
8508
8509         PR c/83844
8510         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8511         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8512         If off is not INTEGER_CST, issue a may not be aligned warning
8513         rather than isn't aligned.  Use isn%'t rather than isn't.
8514         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8515         into MULT_EXPR.
8516         <case MULT_EXPR>: Improve the case when bottom and one of the
8517         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8518         operand, in that case check if the other operand is multiple of
8519         bottom divided by the INTEGER_CST operand.
8520
8521 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
8522
8523         PR target/83858
8524         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8525         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8526         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8527         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8528         * config/pa/pa.c (pa_function_arg_advance): Likewise.
8529         (pa_function_arg, pa_arg_partial_bytes): Likewise.
8530         (pa_function_arg_size): New function.
8531
8532 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
8533
8534         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8535         in a separate statement.
8536
8537 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
8538
8539         PR tree-optimization/83847
8540         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8541         group gathers and scatters.
8542
8543 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
8544
8545         PR rtl-optimization/86620
8546         * params.def (max-sched-ready-insns): Bump minimum value to 1.
8547
8548         PR rtl-optimization/83213
8549         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8550         to last if both are JUMP_INSNs.
8551
8552         PR tree-optimization/83843
8553         * gimple-ssa-store-merging.c
8554         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8555         store_immediate_info for bswap/nop orig_stores.
8556
8557 2018-01-15  Andrew Waterman  <andrew@sifive.com>
8558
8559         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8560         !TARGET_MUL.
8561         <UDIV>: Increase cost if !TARGET_DIV.
8562
8563 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
8564
8565         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8566         (define_attr "cr_logical_3op"): New.
8567         (cceq_ior_compare): Adjust.
8568         (cceq_ior_compare_complement): Adjust.
8569         (*cceq_rev_compare): Adjust.
8570         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8571         (is_cracked_insn): Adjust.
8572         (insn_must_be_first_in_group): Adjust.
8573         * config/rs6000/40x.md: Adjust.
8574         * config/rs6000/440.md: Adjust.
8575         * config/rs6000/476.md: Adjust.
8576         * config/rs6000/601.md: Adjust.
8577         * config/rs6000/603.md: Adjust.
8578         * config/rs6000/6xx.md: Adjust.
8579         * config/rs6000/7450.md: Adjust.
8580         * config/rs6000/7xx.md: Adjust.
8581         * config/rs6000/8540.md: Adjust.
8582         * config/rs6000/cell.md: Adjust.
8583         * config/rs6000/e300c2c3.md: Adjust.
8584         * config/rs6000/e500mc.md: Adjust.
8585         * config/rs6000/e500mc64.md: Adjust.
8586         * config/rs6000/e5500.md: Adjust.
8587         * config/rs6000/e6500.md: Adjust.
8588         * config/rs6000/mpc.md: Adjust.
8589         * config/rs6000/power4.md: Adjust.
8590         * config/rs6000/power5.md: Adjust.
8591         * config/rs6000/power6.md: Adjust.
8592         * config/rs6000/power7.md: Adjust.
8593         * config/rs6000/power8.md: Adjust.
8594         * config/rs6000/power9.md: Adjust.
8595         * config/rs6000/rs64.md: Adjust.
8596         * config/rs6000/titan.md: Adjust.
8597
8598 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
8599
8600         * config/i386/predicates.md (indirect_branch_operand): Rewrite
8601         ix86_indirect_branch_register logic.
8602
8603 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
8604
8605         * config/i386/constraints.md (Bs): Update
8606         ix86_indirect_branch_register check.  Don't check
8607         ix86_indirect_branch_register with GOT_memory_operand.
8608         (Bw): Likewise.
8609         * config/i386/predicates.md (GOT_memory_operand): Don't check
8610         ix86_indirect_branch_register here.
8611         (GOT32_symbol_operand): Likewise.
8612
8613 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
8614
8615         * config/i386/predicates.md (constant_call_address_operand):
8616         Rewrite ix86_indirect_branch_register logic.
8617         (sibcall_insn_operand): Likewise.
8618
8619 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
8620
8621         * config/i386/constraints.md (Bs): Replace
8622         ix86_indirect_branch_thunk_register with
8623         ix86_indirect_branch_register.
8624         (Bw): Likewise.
8625         * config/i386/i386.md (indirect_jump): Likewise.
8626         (tablejump): Likewise.
8627         (*sibcall_memory): Likewise.
8628         (*sibcall_value_memory): Likewise.
8629         Peepholes of indirect call and jump via memory: Likewise.
8630         * config/i386/i386.opt: Likewise.
8631         * config/i386/predicates.md (indirect_branch_operand): Likewise.
8632         (GOT_memory_operand): Likewise.
8633         (call_insn_operand): Likewise.
8634         (sibcall_insn_operand): Likewise.
8635         (GOT32_symbol_operand): Likewise.
8636
8637 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
8638
8639         PR middle-end/83837
8640         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8641         type rather than type addr's type points to.
8642         (expand_omp_atomic_mutex): Likewise.
8643         (expand_omp_atomic): Likewise.
8644
8645 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
8646
8647         PR target/83839
8648         * config/i386/i386.c (output_indirect_thunk_function): Use
8649         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8650         for  __x86_return_thunk.
8651
8652 2018-01-15  Richard Biener  <rguenther@suse.de>
8653
8654         PR middle-end/83850
8655         * expmed.c (extract_bit_field_1): Fix typo.
8656
8657 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8658
8659         PR target/83687
8660         * config/arm/iterators.md (VF): New mode iterator.
8661         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8662         Remove integer-related logic from pattern.
8663         (neon_vabd<mode>_3): Likewise.
8664
8665 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
8666
8667         PR middle-end/82694
8668         * common.opt (fstrict-overflow): No longer an alias.
8669         (fwrapv-pointer): New option.
8670         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8671         also for pointer types based on flag_wrapv_pointer.
8672         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8673         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8674         opts->x_flag_wrapv got set.
8675         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8676         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8677         POINTER_TYPE_OVERFLOW_UNDEFINED.
8678         * match.pd: Likewise in address comparison pattern.
8679         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8680
8681 2018-01-15  Richard Biener  <rguenther@suse.de>
8682
8683         PR lto/83804
8684         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8685         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
8686         Reset type names to their identifier if their TYPE_DECL doesn't
8687         have linkage (and thus is used for ODR and devirt).
8688         (save_debug_info_for_decl): Remove.
8689         (save_debug_info_for_type): Likewise.
8690         (add_tree_to_fld_list): Adjust.
8691         * tree-pretty-print.c (dump_generic_node): Make dumping of
8692         type names more robust.
8693
8694 2018-01-15  Richard Biener  <rguenther@suse.de>
8695
8696         * BASE-VER: Bump to 8.0.1.
8697
8698 2018-01-14  Martin Sebor  <msebor@redhat.com>
8699
8700         PR other/83508
8701         * builtins.c (check_access): Avoid warning when the no-warning bit
8702         is set.
8703
8704 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
8705
8706         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8707         * ira-color (allocno_hard_regs_compare): Likewise.
8708
8709 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
8710
8711         PR target/83013
8712         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8713         Use .pushsection/.popsection.
8714
8715 2018-01-14  Martin Sebor  <msebor@redhat.com>
8716
8717         PR c++/81327
8718         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8719
8720 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
8721
8722         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8723         entry from extra_headers.
8724         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8725         extra_headers, make the list bitwise identical to the i?86-*-* one.
8726
8727 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
8728
8729         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8730         -mcmodel=large with -mindirect-branch=thunk,
8731         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8732         -mfunction-return=thunk-extern.
8733         * doc/invoke.texi: Document -mcmodel=large is incompatible with
8734         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8735         -mfunction-return=thunk and -mfunction-return=thunk-extern.
8736
8737 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
8738
8739         * config/i386/i386.c (print_reg): Print the name of the full
8740         integer register without '%'.
8741         (ix86_print_operand): Handle 'V'.
8742          * doc/extend.texi: Document 'V' modifier.
8743
8744 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
8745
8746         * config/i386/constraints.md (Bs): Disallow memory operand for
8747         -mindirect-branch-register.
8748         (Bw): Likewise.
8749         * config/i386/predicates.md (indirect_branch_operand): Likewise.
8750         (GOT_memory_operand): Likewise.
8751         (call_insn_operand): Likewise.
8752         (sibcall_insn_operand): Likewise.
8753         (GOT32_symbol_operand): Likewise.
8754         * config/i386/i386.md (indirect_jump): Call convert_memory_address
8755         for -mindirect-branch-register.
8756         (tablejump): Likewise.
8757         (*sibcall_memory): Likewise.
8758         (*sibcall_value_memory): Likewise.
8759         Disallow peepholes of indirect call and jump via memory for
8760         -mindirect-branch-register.
8761         (*call_pop): Replace m with Bw.
8762         (*call_value_pop): Likewise.
8763         (*sibcall_pop_memory): Replace m with Bs.
8764         * config/i386/i386.opt (mindirect-branch-register): New option.
8765         * doc/invoke.texi: Document -mindirect-branch-register option.
8766
8767 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
8768
8769         * config/i386/i386-protos.h (ix86_output_function_return): New.
8770         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8771         set function_return_type.
8772         (indirect_thunk_name): Add ret_p to indicate thunk for function
8773         return.
8774         (output_indirect_thunk_function): Pass false to
8775         indirect_thunk_name.
8776         (ix86_output_indirect_branch_via_reg): Likewise.
8777         (ix86_output_indirect_branch_via_push): Likewise.
8778         (output_indirect_thunk_function): Create alias for function
8779         return thunk if regno < 0.
8780         (ix86_output_function_return): New function.
8781         (ix86_handle_fndecl_attribute): Handle function_return.
8782         (ix86_attribute_table): Add function_return.
8783         * config/i386/i386.h (machine_function): Add
8784         function_return_type.
8785         * config/i386/i386.md (simple_return_internal): Use
8786         ix86_output_function_return.
8787         (simple_return_internal_long): Likewise.
8788         * config/i386/i386.opt (mfunction-return=): New option.
8789         (indirect_branch): Mention -mfunction-return=.
8790         * doc/extend.texi: Document function_return function attribute.
8791         * doc/invoke.texi: Document -mfunction-return= option.
8792
8793 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
8794
8795         * config/i386/i386-opts.h (indirect_branch): New.
8796         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8797         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8798         with local indirect jump when converting indirect call and jump.
8799         (ix86_set_indirect_branch_type): New.
8800         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8801         (indirectlabelno): New.
8802         (indirect_thunk_needed): Likewise.
8803         (indirect_thunk_bnd_needed): Likewise.
8804         (indirect_thunks_used): Likewise.
8805         (indirect_thunks_bnd_used): Likewise.
8806         (INDIRECT_LABEL): Likewise.
8807         (indirect_thunk_name): Likewise.
8808         (output_indirect_thunk): Likewise.
8809         (output_indirect_thunk_function): Likewise.
8810         (ix86_output_indirect_branch_via_reg): Likewise.
8811         (ix86_output_indirect_branch_via_push): Likewise.
8812         (ix86_output_indirect_branch): Likewise.
8813         (ix86_output_indirect_jmp): Likewise.
8814         (ix86_code_end): Call output_indirect_thunk_function if needed.
8815         (ix86_output_call_insn): Call ix86_output_indirect_branch if
8816         needed.
8817         (ix86_handle_fndecl_attribute): Handle indirect_branch.
8818         (ix86_attribute_table): Add indirect_branch.
8819         * config/i386/i386.h (machine_function): Add indirect_branch_type
8820         and has_local_indirect_jump.
8821         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8822         to true.
8823         (tablejump): Likewise.
8824         (*indirect_jump): Use ix86_output_indirect_jmp.
8825         (*tablejump_1): Likewise.
8826         (simple_return_indirect_internal): Likewise.
8827         * config/i386/i386.opt (mindirect-branch=): New option.
8828         (indirect_branch): New.
8829         (keep): Likewise.
8830         (thunk): Likewise.
8831         (thunk-inline): Likewise.
8832         (thunk-extern): Likewise.
8833         * doc/extend.texi: Document indirect_branch function attribute.
8834         * doc/invoke.texi: Document -mindirect-branch= option.
8835
8836 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
8837
8838         PR ipa/83051
8839         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8840
8841 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
8842
8843         * ipa-inline.c (want_inline_small_function_p): Return false if
8844         inlining has already failed with CIF_FINAL_ERROR.
8845         (update_caller_keys): Call want_inline_small_function_p before
8846         can_inline_edge_p.
8847         (update_callee_keys): Likewise.
8848
8849 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8850
8851         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8852         New function.
8853         (rs6000_quadword_masked_address_p): Likewise.
8854         (quad_aligned_load_p): Likewise.
8855         (quad_aligned_store_p): Likewise.
8856         (const_load_sequence_p): Add comment to describe the outer-most loop.
8857         (mimic_memory_attributes_and_flags): New function.
8858         (rs6000_gen_stvx): Likewise.
8859         (replace_swapped_aligned_store): Likewise.
8860         (rs6000_gen_lvx): Likewise.
8861         (replace_swapped_aligned_load): Likewise.
8862         (replace_swapped_load_constant): Capitalize argument name in
8863         comment describing this function.
8864         (rs6000_analyze_swaps): Add a third pass to search for vector loads
8865         and stores that access quad-word aligned addresses and replace
8866         with stvx or lvx instructions when appropriate.
8867         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8868         New function prototype.
8869         (rs6000_quadword_masked_address_p): Likewise.
8870         (rs6000_gen_lvx): Likewise.
8871         (rs6000_gen_stvx): Likewise.
8872         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8873         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8874         when memory address is aligned.
8875         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8876         this split to select lvx instruction when memory address is aligned.
8877         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8878         instruction when memory address is aligned.
8879         (*vsx_le_perm_load_v16qi): Likewise.
8880         (four unnamed splitters): Modify to select the stvx instruction
8881         when memory is aligned.
8882
8883 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
8884
8885         * predict.c (determine_unlikely_bbs): Handle correctly BBs
8886         which appears in the queue multiple times.
8887
8888 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
8889             Alan Hayward  <alan.hayward@arm.com>
8890             David Sherwood  <david.sherwood@arm.com>
8891
8892         * tree-vectorizer.h (vec_lower_bound): New structure.
8893         (_loop_vec_info): Add check_nonzero and lower_bounds.
8894         (LOOP_VINFO_CHECK_NONZERO): New macro.
8895         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8896         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8897         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8898         fields.  Make seg_len the distance travelled, not including the
8899         access size.
8900         (dr_direction_indicator): Declare.
8901         (dr_zero_step_indicator): Likewise.
8902         (dr_known_forward_stride_p): Likewise.
8903         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8904         tree-ssanames.h.
8905         (runtime_alias_check_p): Allow runtime alias checks with
8906         variable strides.
8907         (operator ==): Compare access_size and align.
8908         (prune_runtime_alias_test_list): Rework for new distinction between
8909         the access_size and seg_len.
8910         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
8911         segment lengths.
8912         (get_segment_min_max): New function.
8913         (create_intersect_range_checks): Use it.
8914         (dr_step_indicator): New function.
8915         (dr_direction_indicator): Likewise.
8916         (dr_zero_step_indicator): Likewise.
8917         (dr_known_forward_stride_p): Likewise.
8918         * tree-loop-distribution.c (data_ref_segment_size): Return
8919         DR_STEP * (niters - 1).
8920         (compute_alias_check_pairs): Update call to the dr_with_seg_len
8921         constructor.
8922         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8923         (vect_preserves_scalar_order_p): New function, split out from...
8924         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
8925         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8926         (vect_vfa_access_size): New function.
8927         (vect_vfa_align): Likewise.
8928         (vect_compile_time_alias): Take access_size_a and access_b arguments.
8929         (dump_lower_bound): New function.
8930         (vect_check_lower_bound): Likewise.
8931         (vect_small_gap_p): Likewise.
8932         (vectorizable_with_step_bound_p): Likewise.
8933         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8934         depencies if the vectorization factor is 1.  Convert the checks
8935         for nonzero steps into checks on the bounds of DR_STEP.  Try using
8936         a bunds check for variable steps if the minimum required step is
8937         relatively small. Update calls to the dr_with_seg_len
8938         constructor and to vect_compile_time_alias.
8939         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8940         function.
8941         (vect_loop_versioning): Call it.
8942         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8943         when retrying.
8944         (vect_estimate_min_profitable_iters): Account for any bounds checks.
8945
8946 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
8947             Alan Hayward  <alan.hayward@arm.com>
8948             David Sherwood  <david.sherwood@arm.com>
8949
8950         * doc/sourcebuild.texi (vect_scatter_store): Document.
8951         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8952         optabs.
8953         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8954         Document.
8955         * genopinit.c (main): Add supports_vec_scatter_store and
8956         supports_vec_scatter_store_cached to target_optabs.
8957         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8958         IFN_MASK_SCATTER_STORE.
8959         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8960         functions.
8961         * internal-fn.h (internal_store_fn_p): Declare.
8962         (internal_fn_stored_value_index): Likewise.
8963         * internal-fn.c (scatter_store_direct): New macro.
8964         (expand_scatter_store_optab_fn): New function.
8965         (direct_scatter_store_optab_supported_p): New macro.
8966         (internal_store_fn_p): New function.
8967         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8968         IFN_MASK_SCATTER_STORE.
8969         (internal_fn_mask_index): Likewise.
8970         (internal_fn_stored_value_index): New function.
8971         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8972         for scatter stores.
8973         * optabs-query.h (supports_vec_scatter_store_p): Declare.
8974         * optabs-query.c (supports_vec_scatter_store_p): New function.
8975         * tree-vectorizer.h (vect_get_store_rhs): Declare.
8976         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8977         true for scatter stores.
8978         (vect_gather_scatter_fn_p): Handle scatter stores too.
8979         (vect_check_gather_scatter): Consider using scatter stores if
8980         supports_vec_scatter_store_p.
8981         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8982         scatter stores too.
8983         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8984         internal_fn_stored_value_index.
8985         (check_load_store_masking): Handle scatter stores too.
8986         (vect_get_store_rhs): Make public.
8987         (vectorizable_call): Use internal_store_fn_p.
8988         (vectorizable_store): Handle scatter store internal functions.
8989         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8990         when deciding whether the end of the group has been reached.
8991         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8992         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8993         (mask_scatter_store<mode>): New insns.
8994
8995 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
8996             Alan Hayward  <alan.hayward@arm.com>
8997             David Sherwood  <david.sherwood@arm.com>
8998
8999         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9000         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9001         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9002         function.
9003         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9004         Use vect_truncate_gather_scatter_offset if we can't treat the
9005         operation as a normal gather load or scatter store.
9006         (get_group_load_store_type): Take the gather_scatter_info
9007         as argument.  Try using a gather load or scatter store for
9008         single-element groups.
9009         (get_load_store_type): Update calls to get_group_load_store_type
9010         and vect_use_strided_gather_scatters_p.
9011
9012 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9013             Alan Hayward  <alan.hayward@arm.com>
9014             David Sherwood  <david.sherwood@arm.com>
9015
9016         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9017         optional tree argument.
9018         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9019         null target hooks.
9020         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9021         but continue to use the current value as a fallback.
9022         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9023         to compare the updates.
9024         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9025         (get_load_store_type): Use it when handling a strided access.
9026         (vect_get_strided_load_store_ops): New function.
9027         (vect_get_data_ptr_increment): Likewise.
9028         (vectorizable_load): Handle strided gather loads.  Always pass
9029         a step to vect_create_data_ref_ptr and bump_vector_ptr.
9030
9031 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9032             Alan Hayward  <alan.hayward@arm.com>
9033             David Sherwood  <david.sherwood@arm.com>
9034
9035         * doc/md.texi (gather_load@var{m}): Document.
9036         (mask_gather_load@var{m}): Likewise.
9037         * genopinit.c (main): Add supports_vec_gather_load and
9038         supports_vec_gather_load_cached to target_optabs.
9039         * optabs-tree.c (init_tree_optimization_optabs): Use
9040         ggc_cleared_alloc to allocate target_optabs.
9041         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9042         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9043         functions.
9044         * internal-fn.h (internal_load_fn_p): Declare.
9045         (internal_gather_scatter_fn_p): Likewise.
9046         (internal_fn_mask_index): Likewise.
9047         (internal_gather_scatter_fn_supported_p): Likewise.
9048         * internal-fn.c (gather_load_direct): New macro.
9049         (expand_gather_load_optab_fn): New function.
9050         (direct_gather_load_optab_supported_p): New macro.
9051         (direct_internal_fn_optab): New function.
9052         (internal_load_fn_p): Likewise.
9053         (internal_gather_scatter_fn_p): Likewise.
9054         (internal_fn_mask_index): Likewise.
9055         (internal_gather_scatter_fn_supported_p): Likewise.
9056         * optabs-query.c (supports_at_least_one_mode_p): New function.
9057         (supports_vec_gather_load_p): Likewise.
9058         * optabs-query.h (supports_vec_gather_load_p): Declare.
9059         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9060         and memory_type field.
9061         (NUM_PATTERNS): Bump to 15.
9062         * tree-vect-data-refs.c: Include internal-fn.h.
9063         (vect_gather_scatter_fn_p): New function.
9064         (vect_describe_gather_scatter_call): Likewise.
9065         (vect_check_gather_scatter): Try using internal functions for
9066         gather loads.  Recognize existing calls to a gather load function.
9067         (vect_analyze_data_refs): Consider using gather loads if
9068         supports_vec_gather_load_p.
9069         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9070         (vect_get_gather_scatter_offset_type): Likewise.
9071         (vect_convert_mask_for_vectype): Likewise.
9072         (vect_add_conversion_to_patterm): Likewise.
9073         (vect_try_gather_scatter_pattern): Likewise.
9074         (vect_recog_gather_scatter_pattern): New pattern recognizer.
9075         (vect_vect_recog_func_ptrs): Add it.
9076         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9077         internal_fn_mask_index and internal_gather_scatter_fn_p.
9078         (check_load_store_masking): Take the gather_scatter_info as an
9079         argument and handle gather loads.
9080         (vect_get_gather_scatter_ops): New function.
9081         (vectorizable_call): Check internal_load_fn_p.
9082         (vectorizable_load): Likewise.  Handle gather load internal
9083         functions.
9084         (vectorizable_store): Update call to check_load_store_masking.
9085         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9086         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9087         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9088         (aarch64_gather_scale_operand_d): New predicates.
9089         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9090         (mask_gather_load<mode>): New insns.
9091
9092 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9093             Alan Hayward  <alan.hayward@arm.com>
9094             David Sherwood  <david.sherwood@arm.com>
9095
9096         * optabs.def (fold_left_plus_optab): New optab.
9097         * doc/md.texi (fold_left_plus_@var{m}): Document.
9098         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9099         * internal-fn.c (fold_left_direct): Define.
9100         (expand_fold_left_optab_fn): Likewise.
9101         (direct_fold_left_optab_supported_p): Likewise.
9102         * fold-const-call.c (fold_const_fold_left): New function.
9103         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9104         * tree-parloops.c (valid_reduction_p): New function.
9105         (gather_scalar_reductions): Use it.
9106         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9107         (vect_finish_replace_stmt): Declare.
9108         * tree-vect-loop.c (fold_left_reduction_fn): New function.
9109         (needs_fold_left_reduction_p): New function, split out from...
9110         (vect_is_simple_reduction): ...here.  Accept reductions that
9111         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9112         (vect_force_simple_reduction): Also store the reduction type in
9113         the assignment's STMT_VINFO_REDUC_TYPE.
9114         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9115         (merge_with_identity): New function.
9116         (vect_expand_fold_left): Likewise.
9117         (vectorize_fold_left_reduction): Likewise.
9118         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
9119         scalar phi in place for it.  Check for target support and reject
9120         cases that would reassociate the operation.  Defer the transform
9121         phase to vectorize_fold_left_reduction.
9122         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9123         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9124         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9125
9126 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9127
9128         * tree-if-conv.c (predicate_mem_writes): Remove redundant
9129         call to ifc_temp_var.
9130
9131 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9132             Alan Hayward  <alan.hayward@arm.com>
9133             David Sherwood  <david.sherwood@arm.com>
9134
9135         * target.def (legitimize_address_displacement): Take the original
9136         offset as a poly_int.
9137         * targhooks.h (default_legitimize_address_displacement): Update
9138         accordingly.
9139         * targhooks.c (default_legitimize_address_displacement): Likewise.
9140         * doc/tm.texi: Regenerate.
9141         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9142         as an argument, moving assert of ad->disp == ad->disp_term to...
9143         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
9144         Try calling targetm.legitimize_address_displacement before expanding
9145         the address rather than afterwards, and adjust for the new interface.
9146         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9147         Match the new hook interface.  Handle SVE addresses.
9148         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9149         new hook interface.
9150
9151 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9152
9153         * Makefile.in (OBJS): Add early-remat.o.
9154         * target.def (select_early_remat_modes): New hook.
9155         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9156         * doc/tm.texi: Regenerate.
9157         * targhooks.h (default_select_early_remat_modes): Declare.
9158         * targhooks.c (default_select_early_remat_modes): New function.
9159         * timevar.def (TV_EARLY_REMAT): New timevar.
9160         * passes.def (pass_early_remat): New pass.
9161         * tree-pass.h (make_pass_early_remat): Declare.
9162         * early-remat.c: New file.
9163         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9164         function.
9165         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9166
9167 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9168             Alan Hayward  <alan.hayward@arm.com>
9169             David Sherwood  <david.sherwood@arm.com>
9170
9171         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9172         vfm1 with a bound_epilog parameter.
9173         (vect_do_peeling): Update calls accordingly, and move the prologue
9174         call earlier in the function.  Treat the base bound_epilog as 0 for
9175         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
9176         this base when peeling for gaps.
9177         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9178         with fully-masked loops.
9179         (vect_estimate_min_profitable_iters): Handle the single peeled
9180         iteration in that case.
9181
9182 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9183             Alan Hayward  <alan.hayward@arm.com>
9184             David Sherwood  <david.sherwood@arm.com>
9185
9186         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9187         single-element interleaving even if the size is not a power of 2.
9188         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9189         accesses for single-element interleaving if the group size is
9190         not a power of 2.
9191
9192 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9193             Alan Hayward  <alan.hayward@arm.com>
9194             David Sherwood  <david.sherwood@arm.com>
9195
9196         * doc/md.texi (fold_extract_last_@var{m}): Document.
9197         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9198         * optabs.def (fold_extract_last_optab): New optab.
9199         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9200         * internal-fn.c (fold_extract_direct): New macro.
9201         (expand_fold_extract_optab_fn): Likewise.
9202         (direct_fold_extract_optab_supported_p): Likewise.
9203         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9204         * tree-vect-loop.c (vect_model_reduction_cost): Handle
9205         EXTRACT_LAST_REDUCTION.
9206         (get_initial_def_for_reduction): Do not create an initial vector
9207         for EXTRACT_LAST_REDUCTION reductions.
9208         (vectorizable_reduction): Leave the scalar phi in place for
9209         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
9210         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
9211         epilogue code for EXTRACT_LAST_REDUCTION and defer the
9212         transform phase to vectorizable_condition.
9213         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9214         split out from...
9215         (vect_finish_stmt_generation): ...here.
9216         (vect_finish_replace_stmt): New function.
9217         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9218         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9219         pattern.
9220         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9221
9222 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9223             Alan Hayward  <alan.hayward@arm.com>
9224             David Sherwood  <david.sherwood@arm.com>
9225
9226         * doc/md.texi (extract_last_@var{m}): Document.
9227         * optabs.def (extract_last_optab): New optab.
9228         * internal-fn.def (EXTRACT_LAST): New internal function.
9229         * internal-fn.c (cond_unary_direct): New macro.
9230         (expand_cond_unary_optab_fn): Likewise.
9231         (direct_cond_unary_optab_supported_p): Likewise.
9232         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9233         loops using EXTRACT_LAST.
9234         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9235         (extract_last_<mode>): ...this optab.
9236         (vec_extract<mode><Vel>): Update accordingly.
9237
9238 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9239             Alan Hayward  <alan.hayward@arm.com>
9240             David Sherwood  <david.sherwood@arm.com>
9241
9242         * target.def (empty_mask_is_expensive): New hook.
9243         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9244         * doc/tm.texi: Regenerate.
9245         * targhooks.h (default_empty_mask_is_expensive): Declare.
9246         * targhooks.c (default_empty_mask_is_expensive): New function.
9247         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9248         if the target says that empty masks are expensive.
9249         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9250         New function.
9251         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9252
9253 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9254             Alan Hayward  <alan.hayward@arm.com>
9255             David Sherwood  <david.sherwood@arm.com>
9256
9257         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9258         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9259         (vect_use_loop_mask_for_alignment_p): New function.
9260         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9261         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9262         niters_skip argument.  Make sure that the first niters_skip elements
9263         of the first iteration are inactive.
9264         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9265         Update call to vect_set_loop_masks_directly.
9266         (get_misalign_in_elems): New function, split out from...
9267         (vect_gen_prolog_loop_niters): ...here.
9268         (vect_update_init_of_dr): Take a code argument that specifies whether
9269         the adjustment should be added or subtracted.
9270         (vect_update_init_of_drs): Likewise.
9271         (vect_prepare_for_masked_peels): New function.
9272         (vect_do_peeling): Skip prologue peeling if we're using a mask
9273         instead.  Update call to vect_update_inits_of_drs.
9274         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9275         mask_skip_niters.
9276         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9277         alignment.  Do not include the number of peeled iterations in
9278         the minimum threshold in that case.
9279         (vectorizable_induction): Adjust the start value down by
9280         LOOP_VINFO_MASK_SKIP_NITERS iterations.
9281         (vect_transform_loop): Call vect_prepare_for_masked_peels.
9282         Take the number of skipped iterations into account when calculating
9283         the loop bounds.
9284         * tree-vect-stmts.c (vect_gen_while_not): New function.
9285
9286 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9287             Alan Hayward  <alan.hayward@arm.com>
9288             David Sherwood  <david.sherwood@arm.com>
9289
9290         * doc/sourcebuild.texi (vect_fully_masked): Document.
9291         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9292         default value to 0.
9293         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9294         split out from...
9295         (vect_analyze_loop_2): ...here. Don't check the vectorization
9296         factor against the number of loop iterations if the loop is
9297         fully-masked.
9298
9299 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9300             Alan Hayward  <alan.hayward@arm.com>
9301             David Sherwood  <david.sherwood@arm.com>
9302
9303         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9304         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9305         (dump_groups): Update accordingly.
9306         (iv_use::mem_type): New member variable.
9307         (address_p): New function.
9308         (record_use): Add a mem_type argument and initialize the new
9309         mem_type field.
9310         (record_group_use): Add a mem_type argument.  Use address_p.
9311         Remove obsolete null checks of base_object.  Update call to record_use.
9312         (find_interesting_uses_op): Update call to record_group_use.
9313         (find_interesting_uses_cond): Likewise.
9314         (find_interesting_uses_address): Likewise.
9315         (get_mem_type_for_internal_fn): New function.
9316         (find_address_like_use): Likewise.
9317         (find_interesting_uses_stmt): Try find_address_like_use before
9318         calling find_interesting_uses_op.
9319         (addr_offset_valid_p): Use the iv mem_type field as the type
9320         of the addressed memory.
9321         (add_autoinc_candidates): Likewise.
9322         (get_address_cost): Likewise.
9323         (split_small_address_groups_p): Use address_p.
9324         (split_address_groups): Likewise.
9325         (add_iv_candidate_for_use): Likewise.
9326         (autoinc_possible_for_pair): Likewise.
9327         (rewrite_groups): Likewise.
9328         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9329         (determine_group_iv_cost): Update after split of USE_ADDRESS.
9330         (get_alias_ptr_type_for_ptr_address): New function.
9331         (rewrite_use_address): Rewrite address uses in calls that were
9332         identified by find_address_like_use.
9333
9334 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9335             Alan Hayward  <alan.hayward@arm.com>
9336             David Sherwood  <david.sherwood@arm.com>
9337
9338         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9339         TARGET_MEM_REFs.
9340         * gimple-expr.h (is_gimple_addressable: Likewise.
9341         * gimple-expr.c (is_gimple_address): Likewise.
9342         * internal-fn.c (expand_call_mem_ref): New function.
9343         (expand_mask_load_optab_fn): Use it.
9344         (expand_mask_store_optab_fn): Likewise.
9345
9346 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9347             Alan Hayward  <alan.hayward@arm.com>
9348             David Sherwood  <david.sherwood@arm.com>
9349
9350         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9351         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9352         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9353         (cond_umax@var{mode}): Document.
9354         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9355         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9356         (cond_umin_optab, cond_umax_optab): New optabs.
9357         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9358         (COND_IOR, COND_XOR): New internal functions.
9359         * internal-fn.h (get_conditional_internal_fn): Declare.
9360         * internal-fn.c (cond_binary_direct): New macro.
9361         (expand_cond_binary_optab_fn): Likewise.
9362         (direct_cond_binary_optab_supported_p): Likewise.
9363         (get_conditional_internal_fn): New function.
9364         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9365         Cope with reduction statements that are vectorized as calls rather
9366         than assignments.
9367         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9368         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9369         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9370         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9371         (UNSPEC_COND_EOR): New unspecs.
9372         (optab): Add mappings for them.
9373         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9374         (sve_int_op, sve_fp_op): New int attributes.
9375
9376 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9377             Alan Hayward  <alan.hayward@arm.com>
9378             David Sherwood  <david.sherwood@arm.com>
9379
9380         * optabs.def (while_ult_optab): New optab.
9381         * doc/md.texi (while_ult@var{m}@var{n}): Document.
9382         * internal-fn.def (WHILE_ULT): New internal function.
9383         * internal-fn.h (direct_internal_fn_supported_p): New override
9384         that takes two types as argument.
9385         * internal-fn.c (while_direct): New macro.
9386         (expand_while_optab_fn): New function.
9387         (convert_optab_supported_p): Likewise.
9388         (direct_while_optab_supported_p): New macro.
9389         * wide-int.h (wi::udiv_ceil): New function.
9390         * tree-vectorizer.h (rgroup_masks): New structure.
9391         (vec_loop_masks): New typedef.
9392         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9393         and fully_masked_p.
9394         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9395         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9396         (vect_max_vf): New function.
9397         (slpeel_make_loop_iterate_ntimes): Delete.
9398         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9399         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9400         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9401         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9402         internal-fn.h, stor-layout.h and optabs-query.h.
9403         (vect_set_loop_mask): New function.
9404         (add_preheader_seq): Likewise.
9405         (add_header_seq): Likewise.
9406         (interleave_supported_p): Likewise.
9407         (vect_maybe_permute_loop_masks): Likewise.
9408         (vect_set_loop_masks_directly): Likewise.
9409         (vect_set_loop_condition_masked): Likewise.
9410         (vect_set_loop_condition_unmasked): New function, split out from
9411         slpeel_make_loop_iterate_ntimes.
9412         (slpeel_make_loop_iterate_ntimes): Rename to..
9413         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
9414         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9415         (vect_do_peeling): Update call accordingly.
9416         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9417         loops.
9418         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9419         mask_compare_type, can_fully_mask_p and fully_masked_p.
9420         (release_vec_loop_masks): New function.
9421         (_loop_vec_info): Use it to free the loop masks.
9422         (can_produce_all_loop_masks_p): New function.
9423         (vect_get_max_nscalars_per_iter): Likewise.
9424         (vect_verify_full_masking): Likewise.
9425         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9426         retries, and free the mask rgroups before retrying.  Check loop-wide
9427         reasons for disallowing fully-masked loops.  Make the final decision
9428         about whether use a fully-masked loop or not.
9429         (vect_estimate_min_profitable_iters): Do not assume that peeling
9430         for the number of iterations will be needed for fully-masked loops.
9431         (vectorizable_reduction): Disable fully-masked loops.
9432         (vectorizable_live_operation): Likewise.
9433         (vect_halve_mask_nunits): New function.
9434         (vect_double_mask_nunits): Likewise.
9435         (vect_record_loop_mask): Likewise.
9436         (vect_get_loop_mask): Likewise.
9437         (vect_transform_loop): Handle the case in which the final loop
9438         iteration might handle a partial vector.  Call vect_set_loop_condition
9439         instead of slpeel_make_loop_iterate_ntimes.
9440         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9441         (check_load_store_masking): New function.
9442         (prepare_load_store_mask): Likewise.
9443         (vectorizable_store): Handle fully-masked loops.
9444         (vectorizable_load): Likewise.
9445         (supportable_widening_operation): Use vect_halve_mask_nunits for
9446         booleans.
9447         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9448         (vect_gen_while): New function.
9449         * config/aarch64/aarch64.md (umax<mode>3): New expander.
9450         (aarch64_uqdec<mode>): New insn.
9451
9452 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9453             Alan Hayward  <alan.hayward@arm.com>
9454             David Sherwood  <david.sherwood@arm.com>
9455
9456         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9457         (reduc_xor_scal_optab): New optabs.
9458         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9459         (reduc_xor_scal_@var{m}): Document.
9460         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9461         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9462         internal functions.
9463         * fold-const-call.c (fold_const_call): Handle them.
9464         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9465         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9466         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9467         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9468         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9469         (UNSPEC_XORV): New unspecs.
9470         (optab): Add entries for them.
9471         (BITWISEV): New int iterator.
9472         (bit_reduc_op): New int attributes.
9473
9474 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9475             Alan Hayward  <alan.hayward@arm.com>
9476             David Sherwood  <david.sherwood@arm.com>
9477
9478         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9479         * internal-fn.def (VEC_SHL_INSERT): New internal function.
9480         * optabs.def (vec_shl_insert_optab): New optab.
9481         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9482         (duplicate_and_interleave): Likewise.
9483         * tree-vect-loop.c: Include internal-fn.h.
9484         (neutral_op_for_slp_reduction): New function, split out from
9485         get_initial_defs_for_reduction.
9486         (get_initial_def_for_reduction): Handle option 2 for variable-length
9487         vectors by loading the neutral value into a vector and then shifting
9488         the initial value into element 0.
9489         (get_initial_defs_for_reduction): Replace the code argument with
9490         the neutral value calculated by neutral_op_for_slp_reduction.
9491         Use gimple_build_vector for constant-length vectors.
9492         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9493         but the first group_size elements have a neutral value.
9494         Use duplicate_and_interleave otherwise.
9495         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9496         Update call to get_initial_defs_for_reduction.  Handle SLP
9497         reductions for variable-length vectors by creating one vector
9498         result for each scalar result, with the elements associated
9499         with other scalar results stubbed out with the neutral value.
9500         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9501         Require IFN_VEC_SHL_INSERT for double reductions on
9502         variable-length vectors, or SLP reductions that have
9503         a neutral value.  Require can_duplicate_and_interleave_p
9504         support for variable-length unchained SLP reductions if there
9505         is no neutral value, such as for MIN/MAX reductions.  Also require
9506         the number of vector elements to be a multiple of the number of
9507         SLP statements when doing variable-length unchained SLP reductions.
9508         Update call to vect_create_epilog_for_reduction.
9509         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9510         and remove initial values.
9511         (duplicate_and_interleave): Make public.
9512         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9513         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9514
9515 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9516             Alan Hayward  <alan.hayward@arm.com>
9517             David Sherwood  <david.sherwood@arm.com>
9518
9519         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9520         (can_duplicate_and_interleave_p): New function.
9521         (vect_get_and_check_slp_defs): Take the vector of statements
9522         rather than just the current one.  Remove excess parentheses.
9523         Restriction rejectinon of vect_constant_def and vect_external_def
9524         for variable-length vectors to boolean types, or types for which
9525         can_duplicate_and_interleave_p is false.
9526         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9527         (duplicate_and_interleave): New function.
9528         (vect_get_constant_vectors): Use gimple_build_vector for
9529         constant-length vectors and suitable variable-length constant
9530         vectors.  Use duplicate_and_interleave for other variable-length
9531         vectors.  Don't defer the update when inserting new statements.
9532
9533 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9534             Alan Hayward  <alan.hayward@arm.com>
9535             David Sherwood  <david.sherwood@arm.com>
9536
9537         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9538         min_profitable_iters doesn't go negative.
9539
9540 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9541             Alan Hayward  <alan.hayward@arm.com>
9542             David Sherwood  <david.sherwood@arm.com>
9543
9544         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9545         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9546         * optabs.def (vec_mask_load_lanes_optab): New optab.
9547         (vec_mask_store_lanes_optab): Likewise.
9548         * internal-fn.def (MASK_LOAD_LANES): New internal function.
9549         (MASK_STORE_LANES): Likewise.
9550         * internal-fn.c (mask_load_lanes_direct): New macro.
9551         (mask_store_lanes_direct): Likewise.
9552         (expand_mask_load_optab_fn): Handle masked operations.
9553         (expand_mask_load_lanes_optab_fn): New macro.
9554         (expand_mask_store_optab_fn): Handle masked operations.
9555         (expand_mask_store_lanes_optab_fn): New macro.
9556         (direct_mask_load_lanes_optab_supported_p): Likewise.
9557         (direct_mask_store_lanes_optab_supported_p): Likewise.
9558         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9559         parameter.
9560         (vect_load_lanes_supported): Likewise.
9561         * tree-vect-data-refs.c (strip_conversion): New function.
9562         (can_group_stmts_p): Likewise.
9563         (vect_analyze_data_ref_accesses): Use it instead of checking
9564         for a pair of assignments.
9565         (vect_store_lanes_supported): Take a masked_p parameter.
9566         (vect_load_lanes_supported): Likewise.
9567         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9568         vect_store_lanes_supported and vect_load_lanes_supported.
9569         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9570         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9571         parameter.  Don't allow gaps for masked accesses.
9572         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
9573         and vect_load_lanes_supported.
9574         (get_load_store_type): Take a masked_p parameter and update
9575         call to get_group_load_store_type.
9576         (vectorizable_store): Update call to get_load_store_type.
9577         Handle IFN_MASK_STORE_LANES.
9578         (vectorizable_load): Update call to get_load_store_type.
9579         Handle IFN_MASK_LOAD_LANES.
9580
9581 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9582             Alan Hayward  <alan.hayward@arm.com>
9583             David Sherwood  <david.sherwood@arm.com>
9584
9585         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9586         modes for SVE.
9587         * config/aarch64/aarch64-protos.h
9588         (aarch64_sve_struct_memory_operand_p): Declare.
9589         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9590         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9591         (VPRED, vpred): Handle SVE structure modes.
9592         * config/aarch64/constraints.md (Utx): New constraint.
9593         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9594         (aarch64_sve_struct_nonimmediate_operand): New predicates.
9595         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9596         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9597         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9598         structure modes.  Split into pieces after RA.
9599         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9600         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9601         New patterns.
9602         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9603         SVE structure modes.
9604         (aarch64_classify_address): Likewise.
9605         (sizetochar): Move earlier in file.
9606         (aarch64_print_operand): Handle SVE register lists.
9607         (aarch64_array_mode): New function.
9608         (aarch64_sve_struct_memory_operand_p): Likewise.
9609         (TARGET_ARRAY_MODE): Redefine.
9610
9611 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9612             Alan Hayward  <alan.hayward@arm.com>
9613             David Sherwood  <david.sherwood@arm.com>
9614
9615         * target.def (array_mode): New target hook.
9616         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9617         * doc/tm.texi: Regenerate.
9618         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9619         * hooks.c (hook_optmode_mode_uhwi_none): New function.
9620         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9621         targetm.array_mode.
9622         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
9623         type sizes.
9624
9625 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9626             Alan Hayward  <alan.hayward@arm.com>
9627             David Sherwood  <david.sherwood@arm.com>
9628
9629         * fold-const.c (fold_binary_loc): Check the argument types
9630         rather than the result type when testing for a vector operation.
9631
9632 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9633
9634         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9635         * doc/tm.texi: Regenerate.
9636
9637 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9638             Alan Hayward  <alan.hayward@arm.com>
9639             David Sherwood  <david.sherwood@arm.com>
9640
9641         * doc/invoke.texi (-msve-vector-bits=): Document new option.
9642         (sve): Document new AArch64 extension.
9643         * doc/md.texi (w): Extend the description of the AArch64
9644         constraint to include SVE vectors.
9645         (Upl, Upa): Document new AArch64 predicate constraints.
9646         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9647         enum.
9648         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9649         (msve-vector-bits=): New option.
9650         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9651         SVE when these are disabled.
9652         (sve): New extension.
9653         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9654         modes.  Adjust their number of units based on aarch64_sve_vg.
9655         (MAX_BITSIZE_MODE_ANY_MODE): Define.
9656         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9657         aarch64_addr_query_type.
9658         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9659         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9660         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9661         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9662         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9663         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9664         (aarch64_simd_imm_zero_p): Delete.
9665         (aarch64_check_zero_based_sve_index_immediate): Declare.
9666         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9667         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9668         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9669         (aarch64_sve_float_mul_immediate_p): Likewise.
9670         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9671         rather than an rtx.
9672         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9673         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9674         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9675         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9676         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9677         (aarch64_regmode_natural_size): Likewise.
9678         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9679         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9680         left one place.
9681         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9682         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9683         for VG and the SVE predicate registers.
9684         (V_ALIASES): Add a "z"-prefixed alias.
9685         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9686         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9687         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9688         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9689         (REG_CLASS_NAMES): Add entries for them.
9690         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
9691         and the predicate registers.
9692         (aarch64_sve_vg): Declare.
9693         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9694         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9695         (REGMODE_NATURAL_SIZE): Define.
9696         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9697         SVE macros.
9698         * config/aarch64/aarch64.c: Include cfgrtl.h.
9699         (simd_immediate_info): Add a constructor for series vectors,
9700         and an associated step field.
9701         (aarch64_sve_vg): New variable.
9702         (aarch64_dbx_register_number): Handle VG and the predicate registers.
9703         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9704         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9705         (VEC_ANY_DATA, VEC_STRUCT): New constants.
9706         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9707         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9708         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9709         (aarch64_get_mask_mode): New functions.
9710         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9711         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9712         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
9713         predicate modes and predicate registers.  Explicitly restrict
9714         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
9715         to store a vector mode if it is recognized by
9716         aarch64_classify_vector_mode.
9717         (aarch64_regmode_natural_size): New function.
9718         (aarch64_hard_regno_caller_save_mode): Return the original mode
9719         for predicates.
9720         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9721         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9722         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9723         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9724         functions.
9725         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
9726         does not overlap dest if the function is frame-related.  Handle
9727         SVE constants.
9728         (aarch64_split_add_offset): New function.
9729         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9730         them aarch64_add_offset.
9731         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9732         and update call to aarch64_sub_sp.
9733         (aarch64_add_cfa_expression): New function.
9734         (aarch64_expand_prologue): Pass extra temporary registers to the
9735         functions above.  Handle the case in which we need to emit new
9736         DW_CFA_expressions for registers that were originally saved
9737         relative to the stack pointer, but now have to be expressed
9738         relative to the frame pointer.
9739         (aarch64_output_mi_thunk): Pass extra temporary registers to the
9740         functions above.
9741         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
9742         IP0 and IP1 values for SVE frames.
9743         (aarch64_expand_vec_series): New function.
9744         (aarch64_expand_sve_widened_duplicate): Likewise.
9745         (aarch64_expand_sve_const_vector): Likewise.
9746         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9747         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
9748         into the register, rather than emitting a SET directly.
9749         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9750         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9751         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9752         (offset_9bit_signed_scaled_p): New functions.
9753         (aarch64_replicate_bitmask_imm): New function.
9754         (aarch64_bitmask_imm): Use it.
9755         (aarch64_cannot_force_const_mem): Reject expressions involving
9756         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
9757         (aarch64_classify_index): Handle SVE indices, by requiring
9758         a plain register index with a scale that matches the element size.
9759         (aarch64_classify_address): Handle SVE addresses.  Assert that
9760         the mode of the address is VOIDmode or an integer mode.
9761         Update call to aarch64_classify_symbol.
9762         (aarch64_classify_symbolic_expression): Update call to
9763         aarch64_classify_symbol.
9764         (aarch64_const_vec_all_in_range_p): New function.
9765         (aarch64_print_vector_float_operand): Likewise.
9766         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
9767         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
9768         and the FP immediates 1.0 and 0.5.
9769         (aarch64_print_address_internal): Handle SVE addresses.
9770         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9771         (aarch64_regno_regclass): Handle predicate registers.
9772         (aarch64_secondary_reload): Handle big-endian reloads of SVE
9773         data modes.
9774         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9775         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9776         (aarch64_convert_sve_vector_bits): New function.
9777         (aarch64_override_options): Use it to handle -msve-vector-bits=.
9778         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9779         rather than an rtx.
9780         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9781         Handle SVE vector and predicate modes.  Accept VL-based constants
9782         that need only one temporary register, and VL offsets that require
9783         no temporary registers.
9784         (aarch64_conditional_register_usage): Mark the predicate registers
9785         as fixed if SVE isn't available.
9786         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9787         Return true for SVE vector and predicate modes.
9788         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9789         rather than an unsigned int.  Handle SVE modes.
9790         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
9791         SVE modes.
9792         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9793         if SVE is enabled.
9794         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9795         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9796         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9797         (aarch64_sve_float_mul_immediate_p): New functions.
9798         (aarch64_sve_valid_immediate): New function.
9799         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9800         Explicitly reject structure modes.  Check for INDEX constants.
9801         Handle PTRUE and PFALSE constants.
9802         (aarch64_check_zero_based_sve_index_immediate): New function.
9803         (aarch64_simd_imm_zero_p): Delete.
9804         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9805         vector modes.  Accept constants in the range of CNT[BHWD].
9806         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9807         ask for an Advanced SIMD mode.
9808         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9809         (aarch64_simd_vector_alignment): Handle SVE predicates.
9810         (aarch64_vectorize_preferred_vector_alignment): New function.
9811         (aarch64_simd_vector_alignment_reachable): Use it instead of
9812         the vector size.
9813         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9814         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9815         functions.
9816         (MAX_VECT_LEN): Delete.
9817         (expand_vec_perm_d): Add a vec_flags field.
9818         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9819         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9820         (aarch64_evpc_ext): Don't apply a big-endian lane correction
9821         for SVE modes.
9822         (aarch64_evpc_rev): Rename to...
9823         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
9824         (aarch64_evpc_rev_global): New function.
9825         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9826         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9827         MAX_VECT_LEN.
9828         (aarch64_evpc_sve_tbl): New function.
9829         (aarch64_expand_vec_perm_const_1): Update after rename of
9830         aarch64_evpc_rev.  Handle SVE permutes too, trying
9831         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9832         than aarch64_evpc_tbl.
9833         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9834         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9835         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9836         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9837         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9838         (aarch64_expand_sve_vcond): New functions.
9839         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9840         of aarch64_vector_mode_p.
9841         (aarch64_dwarf_poly_indeterminate_value): New function.
9842         (aarch64_compute_pressure_classes): Likewise.
9843         (aarch64_can_change_mode_class): Likewise.
9844         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9845         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9846         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9847         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9848         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9849         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9850         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9851         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9852         constraints.
9853         (Dn, Dl, Dr): Accept const as well as const_vector.
9854         (Dz): Likewise.  Compare against CONST0_RTX.
9855         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9856         of "vector" where appropriate.
9857         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9858         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9859         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9860         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9861         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9862         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9863         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9864         (v_int_equiv): Extend to SVE modes.
9865         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9866         mode attributes.
9867         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9868         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9869         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9870         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9871         (SVE_COND_FP_CMP): New int iterators.
9872         (perm_hilo): Handle the new unpack unspecs.
9873         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9874         attributes.
9875         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9876         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9877         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9878         (aarch64_equality_operator, aarch64_constant_vector_operand)
9879         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9880         (aarch64_sve_nonimmediate_operand): Likewise.
9881         (aarch64_sve_general_operand): Likewise.
9882         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9883         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9884         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9885         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9886         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9887         (aarch64_sve_float_arith_immediate): Likewise.
9888         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9889         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9890         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9891         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9892         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9893         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9894         (aarch64_sve_float_arith_operand): Likewise.
9895         (aarch64_sve_float_arith_with_sub_operand): Likewise.
9896         (aarch64_sve_float_mul_operand): Likewise.
9897         (aarch64_sve_vec_perm_operand): Likewise.
9898         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9899         (aarch64_mov_operand): Accept const_poly_int and const_vector.
9900         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9901         as well as const_vector.
9902         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9903         in file.  Use CONST0_RTX and CONSTM1_RTX.
9904         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
9905         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9906         Use aarch64_simd_imm_zero.
9907         * config/aarch64/aarch64-sve.md: New file.
9908         * config/aarch64/aarch64.md: Include it.
9909         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9910         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9911         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9912         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9913         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9914         (sve): New attribute.
9915         (enabled): Disable instructions with the sve attribute unless
9916         TARGET_SVE.
9917         (movqi, movhi): Pass CONST_POLY_INT operaneds through
9918         aarch64_expand_mov_immediate.
9919         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9920         CNT[BHSD] immediates.
9921         (movti): Split CONST_POLY_INT moves into two halves.
9922         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9923         Split additions that need a temporary here if the destination
9924         is the stack pointer.
9925         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9926         (*add<mode>3_poly_1): New instruction.
9927         (set_clobber_cc): New expander.
9928
9929 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9930
9931         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9932         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
9933         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9934         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9935         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
9936         Change innermode from fixed_mode_size to machine_mode.
9937         (simplify_subreg): Update call accordingly.  Handle a constant-sized
9938         subreg of a variable-length CONST_VECTOR.
9939
9940 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9941             Alan Hayward  <alan.hayward@arm.com>
9942             David Sherwood  <david.sherwood@arm.com>
9943
9944         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9945         (add_offset_to_base): New function, split out from...
9946         (create_mem_ref): ...here.  When handling a scale other than 1,
9947         check first whether the address is valid without the offset.
9948         Add it into the base if so, leaving the index and scale as-is.
9949
9950 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
9951
9952         PR c++/83778
9953         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9954         fold_for_warn before checking if arg2 is INTEGER_CST.
9955
9956 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
9957
9958         * config/rs6000/predicates.md (load_multiple_operation): Delete.
9959         (store_multiple_operation): Delete.
9960         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9961         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9962         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9963         guarded by TARGET_STRING.
9964         (rs6000_output_load_multiple): Delete.
9965         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9966         OPTION_MASK_STRING / TARGET_STRING handling.
9967         (print_operand) <'N', 'O'>: Add comment that these are unused now.
9968         (const rs6000_opt_masks) <"string">: Change mask to 0.
9969         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9970         (MASK_STRING): Delete.
9971         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9972         parts.  Simplify.
9973         (load_multiple): Delete.
9974         (*ldmsi8): Delete.
9975         (*ldmsi7): Delete.
9976         (*ldmsi6): Delete.
9977         (*ldmsi5): Delete.
9978         (*ldmsi4): Delete.
9979         (*ldmsi3): Delete.
9980         (store_multiple): Delete.
9981         (*stmsi8): Delete.
9982         (*stmsi7): Delete.
9983         (*stmsi6): Delete.
9984         (*stmsi5): Delete.
9985         (*stmsi4): Delete.
9986         (*stmsi3): Delete.
9987         (movmemsi_8reg): Delete.
9988         (corresponding unnamed define_insn): Delete.
9989         (movmemsi_6reg): Delete.
9990         (corresponding unnamed define_insn): Delete.
9991         (movmemsi_4reg): Delete.
9992         (corresponding unnamed define_insn): Delete.
9993         (movmemsi_2reg): Delete.
9994         (corresponding unnamed define_insn): Delete.
9995         (movmemsi_1reg): Delete.
9996         (corresponding unnamed define_insn): Delete.
9997         * config/rs6000/rs6000.opt (mno-string): New.
9998         (mstring): Replace by deprecation warning stub.
9999         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10000
10001 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
10002
10003         * regrename.c (regrename_do_replace): If replacing the same
10004         reg multiple times, try to reuse last created gen_raw_REG.
10005
10006         PR debug/81155
10007         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10008         main to workaround a bug in GDB.
10009
10010 2018-01-12  Tom de Vries  <tom@codesourcery.com>
10011
10012         PR target/83737
10013         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10014
10015 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
10016
10017         PR rtl-optimization/80481
10018         * ira-color.c (get_cap_member): New function.
10019         (allocnos_conflict_by_live_ranges_p): Use it.
10020         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10021         (setup_slot_coalesced_allocno_live_ranges): Ditto.
10022
10023 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
10024
10025         PR target/83628
10026         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10027         (*saddl_se_1): Ditto.
10028         (*ssubsi_1): Ditto.
10029         (*ssubl_se_1): Ditto.
10030
10031 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
10032
10033         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10034         rather than wi::to_widest for DR_INITs.
10035         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10036         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10037         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10038         INTEGER_CSTs.
10039         (vect_analyze_group_access_1): Note that here.
10040
10041 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
10042
10043         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10044         polynomial type sizes.
10045
10046 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
10047
10048         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10049         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10050         (gimple_add_tmp_var): Likewise.
10051
10052 2018-01-12  Martin Liska  <mliska@suse.cz>
10053
10054         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10055         (gimple_alloc_sizes): Likewise.
10056         (dump_gimple_statistics): Use PRIu64 in printf format.
10057         * gimple.h: Change uint64_t to int.
10058
10059 2018-01-12  Martin Liska  <mliska@suse.cz>
10060
10061         * tree-core.h: Use uint64_t instead of int.
10062         * tree.c (tree_node_counts): Likewise.
10063         (tree_node_sizes): Likewise.
10064         (dump_tree_statistics): Use PRIu64 in printf format.
10065
10066 2018-01-12  Martin Liska  <mliska@suse.cz>
10067
10068         * Makefile.in: As qsort_chk is implemented in vec.c, add
10069         vec.o to linkage of gencfn-macros.
10070         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10071         passing the info to record_node_allocation_statistics.
10072         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10073         and pass the info.
10074         * ggc-common.c (struct ggc_usage): Add operator== and use
10075         it in operator< and compare function.
10076         * mem-stats.h (struct mem_usage): Likewise.
10077         * vec.c (struct vec_usage): Remove operator< and compare
10078         function. Can be simply inherited.
10079
10080 2018-01-12  Martin Jambor  <mjambor@suse.cz>
10081
10082         PR target/81616
10083         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10084         * tree-ssa-math-opts.c: Include domwalk.h.
10085         (convert_mult_to_fma_1): New function.
10086         (fma_transformation_info): New type.
10087         (fma_deferring_state): Likewise.
10088         (cancel_fma_deferring): New function.
10089         (result_of_phi): Likewise.
10090         (last_fma_candidate_feeds_initial_phi): Likewise.
10091         (convert_mult_to_fma): Added deferring logic, split actual
10092         transformation to convert_mult_to_fma_1.
10093         (math_opts_dom_walker): New type.
10094         (math_opts_dom_walker::after_dom_children): New method, body moved
10095         here from pass_optimize_widening_mul::execute, added deferring logic
10096         bits.
10097         (pass_optimize_widening_mul::execute): Moved most of code to
10098         math_opts_dom_walker::after_dom_children.
10099         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10100         * config/i386/i386.c (ix86_option_override_internal): Added
10101         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10102
10103 2018-01-12  Richard Biener  <rguenther@suse.de>
10104
10105         PR debug/83157
10106         * dwarf2out.c (gen_variable_die): Do not reset old_die for
10107         inline instance vars.
10108
10109 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
10110
10111         PR target/81819
10112         * config/rx/rx.c (rx_is_restricted_memory_address):
10113         Handle SUBREG case.
10114
10115 2018-01-12  Richard Biener  <rguenther@suse.de>
10116
10117         PR tree-optimization/80846
10118         * target.def (split_reduction): New target hook.
10119         * targhooks.c (default_split_reduction): New function.
10120         * targhooks.h (default_split_reduction): Declare.
10121         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10122         target requests first reduce vectors by combining low and high
10123         parts.
10124         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10125         (get_vectype_for_scalar_type_and_size): Export.
10126         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10127         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10128         * doc/tm.texi: Regenerate.
10129         * config/i386/i386.c (ix86_split_reduction): Implement
10130         TARGET_VECTORIZE_SPLIT_REDUCTION.
10131
10132 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
10133
10134         PR target/83368
10135         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10136         in PIC mode except for TARGET_VXWORKS_RTP.
10137         * config/sparc/sparc.c: Include cfgrtl.h.
10138         (TARGET_INIT_PIC_REG): Define.
10139         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10140         (sparc_pic_register_p): New predicate.
10141         (sparc_legitimate_address_p): Use it.
10142         (sparc_legitimize_pic_address): Likewise.
10143         (sparc_delegitimize_address): Likewise.
10144         (sparc_mode_dependent_address_p): Likewise.
10145         (gen_load_pcrel_sym): Remove 4th parameter.
10146         (load_got_register): Adjust call to above.  Remove obsolete stuff.
10147         (sparc_expand_prologue): Do not call load_got_register here.
10148         (sparc_flat_expand_prologue): Likewise.
10149         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10150         (sparc_use_pseudo_pic_reg): New function.
10151         (sparc_init_pic_reg): Likewise.
10152         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10153         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10154
10155 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
10156
10157         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10158         Add item for branch_cost.
10159
10160 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
10161
10162         PR rtl-optimization/83565
10163         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10164         not extend the result to a larger mode for rotate operations.
10165         (num_sign_bit_copies1): Likewise.
10166
10167 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10168
10169         PR target/40411
10170         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10171         -symbolic.
10172         Use values-Xc.o for -pedantic.
10173         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10174
10175 2018-01-12  Martin Liska  <mliska@suse.cz>
10176
10177         PR ipa/83054
10178         * ipa-devirt.c (final_warning_record::grow_type_warnings):
10179         New function.
10180         (possible_polymorphic_call_targets): Use it.
10181         (ipa_devirt): Likewise.
10182
10183 2018-01-12  Martin Liska  <mliska@suse.cz>
10184
10185         * profile-count.h (enum profile_quality): Use 0 as invalid
10186         enum value of profile_quality.
10187
10188 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10189
10190         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10191         -mext-string options.
10192
10193 2018-01-12  Richard Biener  <rguenther@suse.de>
10194
10195         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10196         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10197         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10198         Likewise.
10199         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10200
10201 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
10202
10203         * configure.ac (--with-long-double-format): Add support for the
10204         configuration option to change the default long double format on
10205         PowerPC systems.
10206         * config.gcc (powerpc*-linux*-*): Likewise.
10207         * configure: Regenerate.
10208         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10209         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10210         used without modification.
10211
10212 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10213
10214         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10215         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10216         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10217         MISC_BUILTIN_SPEC_BARRIER.
10218         (rs6000_init_builtins): Likewise.
10219         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10220         enum value.
10221         (speculation_barrier): New define_insn.
10222         * doc/extend.texi: Document __builtin_speculation_barrier.
10223
10224 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
10225
10226         PR target/83203
10227         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10228         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10229         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10230         iterators.
10231         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
10232         integral modes instead of "ss" and "sd".
10233         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10234         vectors with 32-bit and 64-bit elements.
10235         (vecdupssescalarmodesuffix): New mode attribute.
10236         (vec_dup<mode>): Use it.
10237
10238 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
10239
10240         PR target/83330
10241         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10242         frame if argument is passed on stack.
10243
10244 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
10245
10246         PR target/82682
10247         * ree.c (combine_reaching_defs): Optimize also
10248         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10249         reg2=any_extend(exp); reg1=reg2;, formatting fix.
10250
10251 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
10252
10253         PR middle-end/83189
10254         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10255
10256 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
10257
10258         PR middle-end/83718
10259         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10260         after they are computed.
10261
10262 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
10263
10264         PR tree-optimization/83695
10265         * gimple-loop-linterchange.cc
10266         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10267         reset cached scev information after interchange.
10268         (pass_linterchange::execute): Remove call to scev_reset_htab.
10269
10270 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10271
10272         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10273         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10274         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10275         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10276         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10277         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10278         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10279         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10280         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10281         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10282         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10283         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10284         (V_lane_reg): Likewise.
10285         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10286         New define_expand.
10287         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10288         (vfmal_lane_low<mode>_intrinsic,
10289         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10290         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10291         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10292         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10293         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10294         vfmsl_lane_high<mode>_intrinsic): New define_insns.
10295
10296 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10297
10298         * config/arm/arm-cpus.in (fp16fml): New feature.
10299         (ALL_SIMD): Add fp16fml.
10300         (armv8.2-a): Add fp16fml as an option.
10301         (armv8.3-a): Likewise.
10302         (armv8.4-a): Add fp16fml as part of fp16.
10303         * config/arm/arm.h (TARGET_FP16FML): Define.
10304         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10305         when appropriate.
10306         * config/arm/arm-modes.def (V2HF): Define.
10307         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10308         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10309         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10310         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10311         vfmsl_low, vfmsl_high): New set of builtins.
10312         * config/arm/iterators.md (PLUSMINUS): New code iterator.
10313         (vfml_op): New code attribute.
10314         (VFMLHALVES): New int iterator.
10315         (VFML, VFMLSEL): New mode attributes.
10316         (V_reg): Define mapping for V2HF.
10317         (V_hi, V_lo): New mode attributes.
10318         (VF_constraint): Likewise.
10319         (vfml_half, vfml_half_selector): New int attributes.
10320         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10321         define_expand.
10322         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10323         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10324         New define_insn.
10325         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10326         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10327         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10328         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
10329         documentation.
10330         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10331         Document new effective target and option set.
10332
10333 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10334
10335         * config/arm/arm-cpus.in (armv8_4): New feature.
10336         (ARMv8_4a): New fgroup.
10337         (armv8.4-a): New arch.
10338         * config/arm/arm-tables.opt: Regenerate.
10339         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10340         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10341         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10342         Add matching rules for -march=armv8.4-a and extensions.
10343         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10344
10345 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
10346
10347         PR target/81821
10348         * config/rx/rx.md (BW): New mode attribute.
10349         (sync_lock_test_and_setsi): Add mode suffix to insn output.
10350
10351 2018-01-11  Richard Biener  <rguenther@suse.de>
10352
10353         PR tree-optimization/83435
10354         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10355         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10356         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10357
10358 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
10359             Alan Hayward  <alan.hayward@arm.com>
10360             David Sherwood  <david.sherwood@arm.com>
10361
10362         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10363         field.
10364         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
10365         (aarch64_print_address_internal): Use it to check for a zero offset.
10366
10367 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
10368             Alan Hayward  <alan.hayward@arm.com>
10369             David Sherwood  <david.sherwood@arm.com>
10370
10371         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10372         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10373         Return a poly_int64 rather than a HOST_WIDE_INT.
10374         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10375         rather than a HOST_WIDE_INT.
10376         * config/aarch64/aarch64.h (aarch64_frame): Protect with
10377         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
10378         hard_fp_offset, frame_size, initial_adjust, callee_offset and
10379         final_offset from HOST_WIDE_INT to poly_int64.
10380         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10381         to_constant when getting the number of units in an Advanced SIMD
10382         mode.
10383         (aarch64_builtin_vectorized_function): Check for a constant number
10384         of units.
10385         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10386         GET_MODE_SIZE.
10387         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10388         attribute instead of GET_MODE_NUNITS.
10389         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10390         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10391         GET_MODE_SIZE for fixed-size registers.
10392         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10393         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10394         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10395         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10396         (aarch64_print_operand, aarch64_print_address_internal)
10397         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10398         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10399         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10400         Handle polynomial GET_MODE_SIZE.
10401         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
10402         wider than SImode without modification.
10403         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10404         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10405         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10406         passing and returning SVE modes.
10407         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10408         rather than GEN_INT.
10409         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10410         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10411         (aarch64_allocate_and_probe_stack_space): Likewise.
10412         (aarch64_layout_frame): Cope with polynomial offsets.
10413         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10414         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
10415         polynomial offsets.
10416         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10417         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10418         poly_int64 rather than a HOST_WIDE_INT.
10419         (aarch64_get_separate_components, aarch64_process_components)
10420         (aarch64_expand_prologue, aarch64_expand_epilogue)
10421         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10422         (aarch64_anchor_offset): New function, split out from...
10423         (aarch64_legitimize_address): ...here.
10424         (aarch64_builtin_vectorization_cost): Handle polynomial
10425         TYPE_VECTOR_SUBPARTS.
10426         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10427         GET_MODE_NUNITS.
10428         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10429         number of elements from the PARALLEL rather than the mode.
10430         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10431         rather than GET_MODE_BITSIZE.
10432         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10433         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10434         (aarch64_expand_vec_perm_const_1): Handle polynomial
10435         d->perm.length () and d->perm elements.
10436         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
10437         Apply to_constant to d->perm elements.
10438         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10439         polynomial CONST_VECTOR_NUNITS.
10440         (aarch64_move_pointer): Take amount as a poly_int64 rather
10441         than an int.
10442         (aarch64_progress_pointer): Avoid temporary variable.
10443         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10444         the mode attribute instead of GET_MODE.
10445
10446 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
10447             Alan Hayward  <alan.hayward@arm.com>
10448             David Sherwood  <david.sherwood@arm.com>
10449
10450         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10451         x exists before using it.
10452         (aarch64_add_constant_internal): Rename to...
10453         (aarch64_add_offset_1): ...this.  Replace regnum with separate
10454         src and dest rtxes.  Handle the case in which they're different,
10455         including when the offset is zero.  Replace scratchreg with an rtx.
10456         Use 2 additions if there is no spare register into which we can
10457         move a 16-bit constant.
10458         (aarch64_add_constant): Delete.
10459         (aarch64_add_offset): Replace reg with separate src and dest
10460         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
10461         Use aarch64_add_offset_1.
10462         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10463         an rtx rather than an int.  Take the delta as a poly_int64
10464         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
10465         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10466         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10467         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10468         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10469         and aarch64_add_sp.
10470         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10471         aarch64_add_constant.
10472
10473 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
10474
10475         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10476         Use scalar_float_mode.
10477
10478 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
10479
10480         * config/aarch64/aarch64-simd.md
10481         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10482         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10483         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10484         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10485         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10486         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10487         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10488         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10489         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10490         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10491
10492 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10493
10494         PR target/83514
10495         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10496         targ_options->x_arm_arch_string is non NULL.
10497
10498 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
10499
10500         * config/aarch64/aarch64.h
10501         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
10502
10503 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
10504
10505         PR target/82096
10506         * expmed.c (emit_store_flag_force): Swap if const op0
10507         and change VOIDmode to mode of op0.
10508
10509 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
10510
10511         PR rtl-optimization/83761
10512         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10513         than bytes to mode_for_size.
10514
10515 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
10516
10517         PR middle-end/83189
10518         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10519         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10520         profile.
10521
10522 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
10523
10524         PR middle-end/83575
10525         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10526         when in layout mode.
10527         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10528         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10529         partition fixup.
10530
10531 2018-01-10  Michael Collison  <michael.collison@arm.com>
10532
10533         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10534         * config/aarch64/aarch64-option-extension.def: Add
10535         AARCH64_OPT_EXTENSION of 'fp16fml'.
10536         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10537         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10538         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10539         * config/aarch64/constraints.md (Ui7): New constraint.
10540         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10541         (VFMLA_SEL_W): Ditto.
10542         (f16quad): Ditto.
10543         (f16mac1): Ditto.
10544         (VFMLA16_LOW): New int iterator.
10545         (VFMLA16_HIGH): Ditto.
10546         (UNSPEC_FMLAL): New unspec.
10547         (UNSPEC_FMLSL): Ditto.
10548         (UNSPEC_FMLAL2): Ditto.
10549         (UNSPEC_FMLSL2): Ditto.
10550         (f16mac): New code attribute.
10551         * config/aarch64/aarch64-simd-builtins.def
10552         (aarch64_fmlal_lowv2sf): Ditto.
10553         (aarch64_fmlsl_lowv2sf): Ditto.
10554         (aarch64_fmlalq_lowv4sf): Ditto.
10555         (aarch64_fmlslq_lowv4sf): Ditto.
10556         (aarch64_fmlal_highv2sf): Ditto.
10557         (aarch64_fmlsl_highv2sf): Ditto.
10558         (aarch64_fmlalq_highv4sf): Ditto.
10559         (aarch64_fmlslq_highv4sf): Ditto.
10560         (aarch64_fmlal_lane_lowv2sf): Ditto.
10561         (aarch64_fmlsl_lane_lowv2sf): Ditto.
10562         (aarch64_fmlal_laneq_lowv2sf): Ditto.
10563         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10564         (aarch64_fmlalq_lane_lowv4sf): Ditto.
10565         (aarch64_fmlsl_lane_lowv4sf): Ditto.
10566         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10567         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10568         (aarch64_fmlal_lane_highv2sf): Ditto.
10569         (aarch64_fmlsl_lane_highv2sf): Ditto.
10570         (aarch64_fmlal_laneq_highv2sf): Ditto.
10571         (aarch64_fmlsl_laneq_highv2sf): Ditto.
10572         (aarch64_fmlalq_lane_highv4sf): Ditto.
10573         (aarch64_fmlsl_lane_highv4sf): Ditto.
10574         (aarch64_fmlalq_laneq_highv4sf): Ditto.
10575         (aarch64_fmlsl_laneq_highv4sf): Ditto.
10576         * config/aarch64/aarch64-simd.md:
10577         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10578         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10579         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10580         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10581         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10582         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10583         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10584         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10585         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10586         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10587         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10588         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10589         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10590         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10591         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10592         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10593         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10594         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10595         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10596         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10597         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10598         (vfmlsl_low_u32): Ditto.
10599         (vfmlalq_low_u32): Ditto.
10600         (vfmlslq_low_u32): Ditto.
10601         (vfmlal_high_u32): Ditto.
10602         (vfmlsl_high_u32): Ditto.
10603         (vfmlalq_high_u32): Ditto.
10604         (vfmlslq_high_u32): Ditto.
10605         (vfmlal_lane_low_u32): Ditto.
10606         (vfmlsl_lane_low_u32): Ditto.
10607         (vfmlal_laneq_low_u32): Ditto.
10608         (vfmlsl_laneq_low_u32): Ditto.
10609         (vfmlalq_lane_low_u32): Ditto.
10610         (vfmlslq_lane_low_u32): Ditto.
10611         (vfmlalq_laneq_low_u32): Ditto.
10612         (vfmlslq_laneq_low_u32): Ditto.
10613         (vfmlal_lane_high_u32): Ditto.
10614         (vfmlsl_lane_high_u32): Ditto.
10615         (vfmlal_laneq_high_u32): Ditto.
10616         (vfmlsl_laneq_high_u32): Ditto.
10617         (vfmlalq_lane_high_u32): Ditto.
10618         (vfmlslq_lane_high_u32): Ditto.
10619         (vfmlalq_laneq_high_u32): Ditto.
10620         (vfmlslq_laneq_high_u32): Ditto.
10621         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10622         (AARCH64_FL_FOR_ARCH8_4): New.
10623         (AARCH64_ISA_F16FML): New ISA flag.
10624         (TARGET_F16FML): New feature flag for fp16fml.
10625         (doc/invoke.texi): Document new fp16fml option.
10626
10627 2018-01-10  Michael Collison  <michael.collison@arm.com>
10628
10629         * config/aarch64/aarch64-builtins.c:
10630         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10631         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10632         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10633         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10634         (AARCH64_ISA_SHA3): New ISA flag.
10635         (TARGET_SHA3): New feature flag for sha3.
10636         * config/aarch64/iterators.md (sha512_op): New int attribute.
10637         (CRYPTO_SHA512): New int iterator.
10638         (UNSPEC_SHA512H): New unspec.
10639         (UNSPEC_SHA512H2): Ditto.
10640         (UNSPEC_SHA512SU0): Ditto.
10641         (UNSPEC_SHA512SU1): Ditto.
10642         * config/aarch64/aarch64-simd-builtins.def
10643         (aarch64_crypto_sha512hqv2di): New builtin.
10644         (aarch64_crypto_sha512h2qv2di): Ditto.
10645         (aarch64_crypto_sha512su0qv2di): Ditto.
10646         (aarch64_crypto_sha512su1qv2di): Ditto.
10647         (aarch64_eor3qv8hi): Ditto.
10648         (aarch64_rax1qv2di): Ditto.
10649         (aarch64_xarqv2di): Ditto.
10650         (aarch64_bcaxqv8hi): Ditto.
10651         * config/aarch64/aarch64-simd.md:
10652         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10653         (aarch64_crypto_sha512su0qv2di): Ditto.
10654         (aarch64_crypto_sha512su1qv2di): Ditto.
10655         (aarch64_eor3qv8hi): Ditto.
10656         (aarch64_rax1qv2di): Ditto.
10657         (aarch64_xarqv2di): Ditto.
10658         (aarch64_bcaxqv8hi): Ditto.
10659         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10660         (vsha512h2q_u64): Ditto.
10661         (vsha512su0q_u64): Ditto.
10662         (vsha512su1q_u64): Ditto.
10663         (veor3q_u16): Ditto.
10664         (vrax1q_u64): Ditto.
10665         (vxarq_u64): Ditto.
10666         (vbcaxq_u16): Ditto.
10667         * config/arm/types.md (crypto_sha512): New type attribute.
10668         (crypto_sha3): Ditto.
10669         (doc/invoke.texi): Document new sha3 option.
10670
10671 2018-01-10  Michael Collison  <michael.collison@arm.com>
10672
10673         * config/aarch64/aarch64-builtins.c:
10674         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10675         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10676         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10677         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10678         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10679         (AARCH64_ISA_SM4): New ISA flag.
10680         (TARGET_SM4): New feature flag for sm4.
10681         * config/aarch64/aarch64-simd-builtins.def
10682         (aarch64_sm3ss1qv4si): Ditto.
10683         (aarch64_sm3tt1aq4si): Ditto.
10684         (aarch64_sm3tt1bq4si): Ditto.
10685         (aarch64_sm3tt2aq4si): Ditto.
10686         (aarch64_sm3tt2bq4si): Ditto.
10687         (aarch64_sm3partw1qv4si): Ditto.
10688         (aarch64_sm3partw2qv4si): Ditto.
10689         (aarch64_sm4eqv4si): Ditto.
10690         (aarch64_sm4ekeyqv4si): Ditto.
10691         * config/aarch64/aarch64-simd.md:
10692         (aarch64_sm3ss1qv4si): Ditto.
10693         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10694         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10695         (aarch64_sm4eqv4si): Ditto.
10696         (aarch64_sm4ekeyqv4si): Ditto.
10697         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10698         (sm3part_op): Ditto.
10699         (CRYPTO_SM3TT): Ditto.
10700         (CRYPTO_SM3PART): Ditto.
10701         (UNSPEC_SM3SS1): New unspec.
10702         (UNSPEC_SM3TT1A): Ditto.
10703         (UNSPEC_SM3TT1B): Ditto.
10704         (UNSPEC_SM3TT2A): Ditto.
10705         (UNSPEC_SM3TT2B): Ditto.
10706         (UNSPEC_SM3PARTW1): Ditto.
10707         (UNSPEC_SM3PARTW2): Ditto.
10708         (UNSPEC_SM4E): Ditto.
10709         (UNSPEC_SM4EKEY): Ditto.
10710         * config/aarch64/constraints.md (Ui2): New constraint.
10711         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10712         * config/arm/types.md (crypto_sm3): New type attribute.
10713         (crypto_sm4): Ditto.
10714         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10715         (vsm3tt1aq_u32): Ditto.
10716         (vsm3tt1bq_u32): Ditto.
10717         (vsm3tt2aq_u32): Ditto.
10718         (vsm3tt2bq_u32): Ditto.
10719         (vsm3partw1q_u32): Ditto.
10720         (vsm3partw2q_u32): Ditto.
10721         (vsm4eq_u32): Ditto.
10722         (vsm4ekeyq_u32): Ditto.
10723         (doc/invoke.texi): Document new sm4 option.
10724
10725 2018-01-10  Michael Collison  <michael.collison@arm.com>
10726
10727         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10728         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10729         (AARCH64_FL_FOR_ARCH8_4): New.
10730         (AARCH64_FL_V8_4): New flag.
10731         (doc/invoke.texi): Document new armv8.4-a option.
10732
10733 2018-01-10  Michael Collison  <michael.collison@arm.com>
10734
10735         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10736         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10737         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10738         * config/aarch64/aarch64-option-extension.def: Add
10739         AARCH64_OPT_EXTENSION of 'sha2'.
10740         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10741         (crypto): Disable sha2 and aes if crypto disabled.
10742         (crypto): Enable aes and sha2 if enabled.
10743         (simd): Disable sha2 and aes if simd disabled.
10744         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10745         New flags.
10746         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10747         (TARGET_SHA2): New feature flag for sha2.
10748         (TARGET_AES): New feature flag for aes.
10749         * config/aarch64/aarch64-simd.md:
10750         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10751         conditional on TARGET_AES.
10752         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10753         (aarch64_crypto_sha1hsi): Make pattern conditional
10754         on TARGET_SHA2.
10755         (aarch64_crypto_sha1hv4si): Ditto.
10756         (aarch64_be_crypto_sha1hv4si): Ditto.
10757         (aarch64_crypto_sha1su1v4si): Ditto.
10758         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10759         (aarch64_crypto_sha1su0v4si): Ditto.
10760         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10761         (aarch64_crypto_sha256su0v4si): Ditto.
10762         (aarch64_crypto_sha256su1v4si): Ditto.
10763         (doc/invoke.texi): Document new aes and sha2 options.
10764
10765 2018-01-10  Martin Sebor  <msebor@redhat.com>
10766
10767         PR tree-optimization/83781
10768         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10769         as string arrays.
10770
10771 2018-01-11  Martin Sebor  <msebor@gmail.com>
10772             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10773
10774         PR tree-optimization/83501
10775         PR tree-optimization/81703
10776
10777         * tree-ssa-strlen.c (get_string_cst): Rename...
10778         (get_string_len): ...to this.  Handle global constants.
10779         (handle_char_store): Adjust.
10780
10781 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
10782             Jim Wilson  <jimw@sifive.com>
10783
10784         * config/riscv/riscv-protos.h (riscv_output_return): New.
10785         * config/riscv/riscv.c (struct machine_function): New naked_p field.
10786         (riscv_attribute_table, riscv_output_return),
10787         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10788         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10789         (riscv_compute_frame_info): Only compute frame->mask if not a naked
10790         function.
10791         (riscv_expand_prologue): Add early return for naked function.
10792         (riscv_expand_epilogue): Likewise.
10793         (riscv_function_ok_for_sibcall): Return false for naked function.
10794         (riscv_set_current_function): New.
10795         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10796         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10797         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10798         * doc/extend.texi (RISC-V Function Attributes): New.
10799
10800 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
10801
10802         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10803         check for 128-bit long double before checking TCmode.
10804         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10805         128-bit long doubles before checking TFmode or TCmode.
10806         (FLOAT128_IBM_P): Likewise.
10807
10808 2018-01-10  Martin Sebor  <msebor@redhat.com>
10809
10810         PR tree-optimization/83671
10811         * builtins.c (c_strlen): Unconditionally return zero for the empty
10812         string.
10813         Use -Warray-bounds for warnings.
10814         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10815         for non-constant array indices with COMPONENT_REF, arrays of
10816         arrays, and pointers to arrays.
10817         (gimple_fold_builtin_strlen): Determine and set length range for
10818         non-constant character arrays.
10819
10820 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
10821
10822         PR middle-end/81897
10823         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10824         empty blocks.
10825
10826 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
10827
10828         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10829
10830 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
10831
10832         PR target/83399
10833         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10834         VECTOR_MEM_ALTIVEC_OR_VSX_P.
10835         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10836         indexed_or_indirect_operand predicate.
10837         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10838         (*vsx_le_perm_load_v8hi): Likewise.
10839         (*vsx_le_perm_load_v16qi): Likewise.
10840         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10841         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10842         (*vsx_le_perm_store_v8hi): Likewise.
10843         (*vsx_le_perm_store_v16qi): Likewise.
10844         (eight unnamed splitters): Likewise.
10845
10846 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
10847
10848         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10849         * config/rs6000/emmintrin.h: Likewise.
10850         * config/rs6000/mmintrin.h: Likewise.
10851         * config/rs6000/xmmintrin.h: Likewise.
10852
10853 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
10854
10855         PR c++/43486
10856         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10857         "public_flag".
10858         * tree.c (tree_nop_conversion): Return true for location wrapper
10859         nodes.
10860         (maybe_wrap_with_location): New function.
10861         (selftest::check_strip_nops): New function.
10862         (selftest::test_location_wrappers): New function.
10863         (selftest::tree_c_tests): Call it.
10864         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10865         (maybe_wrap_with_location): New decl.
10866         (EXPR_LOCATION_WRAPPER_P): New macro.
10867         (location_wrapper_p): New inline function.
10868         (tree_strip_any_location_wrapper): New inline function.
10869
10870 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
10871
10872         PR target/83735
10873         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10874         stack_realign_offset for the largest alignment of stack slot
10875         actually used.
10876         (ix86_find_max_used_stack_alignment): New function.
10877         (ix86_finalize_stack_frame_flags): Use it.  Set
10878         max_used_stack_alignment if we don't realign stack.
10879         * config/i386/i386.h (machine_function): Add
10880         max_used_stack_alignment.
10881
10882 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
10883
10884         * config/arm/arm.opt (-mbranch-cost): New option.
10885         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10886         account.
10887
10888 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
10889
10890         PR target/83629
10891         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10892         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10893
10894 2018-01-10  Richard Biener  <rguenther@suse.de>
10895
10896         PR debug/83765
10897         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10898         early out so it also covers the case where we have a non-NULL
10899         origin.
10900
10901 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
10902
10903         PR tree-optimization/83753
10904         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10905         for non-strided grouped accesses if the number of elements is 1.
10906
10907 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
10908
10909         PR target/81616
10910         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10911         * i386.h (TARGET_USE_GATHER): Define.
10912         * x86-tune.def (X86_TUNE_USE_GATHER): New.
10913
10914 2018-01-10  Martin Liska  <mliska@suse.cz>
10915
10916         PR bootstrap/82831
10917         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10918         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10919         partitioning.
10920         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10921         CLEANUP_NO_PARTITIONING is not set.
10922
10923 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
10924
10925         * doc/rtl.texi: Remove documentation of (const ...) wrappers
10926         for vectors, as a partial revert of r254296.
10927         * rtl.h (const_vec_p): Delete.
10928         (const_vec_duplicate_p): Don't test for vector CONSTs.
10929         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10930         * expmed.c (make_tree): Likewise.
10931
10932         Revert:
10933         * common.md (E, F): Use CONSTANT_P instead of checking for
10934         CONST_VECTOR.
10935         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10936         checking for CONST_VECTOR.
10937
10938 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
10939
10940         PR middle-end/83575
10941         * predict.c (force_edge_cold): Handle in more sane way edges
10942         with no prediction.
10943
10944 2018-01-09  Carl Love  <cel@us.ibm.com>
10945
10946         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10947         V4SI, V4SF types.
10948         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10949         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10950         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10951         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
10952         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10953         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
10954         * config/rs6000/rs6000-protos.h: Add extern defition for
10955         rs6000_generate_float2_double_code.
10956         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10957         function.
10958         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10959         (float2_v2df): Add define_expand.
10960
10961 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
10962
10963         PR target/83628
10964         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10965         op_mode in the force_to_mode call.
10966
10967 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
10968
10969         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10970         instead of checking each element individually.
10971         (aarch64_evpc_uzp): Likewise.
10972         (aarch64_evpc_zip): Likewise.
10973         (aarch64_evpc_ext): Likewise.
10974         (aarch64_evpc_rev): Likewise.
10975         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10976         instead of checking each element individually.  Return true without
10977         generating rtl if
10978         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10979         whether all selected elements come from the same input, instead of
10980         checking each element individually.  Remove calls to gen_rtx_REG,
10981         start_sequence and end_sequence and instead assert that no rtl is
10982         generated.
10983
10984 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
10985
10986         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10987         order of HIGH and CONST checks.
10988
10989 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
10990
10991         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10992         if the destination isn't an SSA_NAME.
10993
10994 2018-01-09  Richard Biener  <rguenther@suse.de>
10995
10996         PR tree-optimization/83668
10997         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10998         move prologue...
10999         (canonicalize_loop_form): ... here, renamed from ...
11000         (canonicalize_loop_closed_ssa_form): ... this and amended to
11001         swap successor edges for loop exit blocks to make us use
11002         the RPO order we need for initial schedule generation.
11003
11004 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
11005
11006         PR tree-optimization/64811
11007         * match.pd: When optimizing comparisons with Inf, avoid
11008         introducing or losing exceptions from comparisons with NaN.
11009
11010 2018-01-09  Martin Liska  <mliska@suse.cz>
11011
11012         PR sanitizer/82517
11013         * asan.c (shadow_mem_size): Add gcc_assert.
11014
11015 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
11016
11017         Don't save registers in main().
11018
11019         PR target/83738
11020         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11021         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11022         * config/avr/avr.c (avr_set_current_function): Don't error if
11023         naked, OS_task or OS_main are specified at the same time.
11024         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11025         OS_main.
11026         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11027         attribute.
11028         * common/config/avr/avr-common.c (avr_option_optimization_table):
11029         Switch on -mmain-is-OS_task for optimizing compilations.
11030
11031 2018-01-09  Richard Biener  <rguenther@suse.de>
11032
11033         PR tree-optimization/83572
11034         * graphite.c: Include cfganal.h.
11035         (graphite_transform_loops): Connect infinite loops to exit
11036         and remove fake edges at the end.
11037
11038 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
11039
11040         * ipa-inline.c (edge_badness): Revert accidental checkin.
11041
11042 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
11043
11044         PR ipa/80763
11045         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11046         symbols; not inline clones.
11047
11048 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
11049
11050         PR target/83507
11051         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11052         hard registers.  Formatting fixes.
11053
11054         PR preprocessor/83722
11055         * gcc.c (try_generate_repro): Pass
11056         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11057         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11058         do_report_bug.
11059
11060 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
11061             Kito Cheng  <kito.cheng@gmail.com>
11062
11063         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11064         (riscv_leaf_function_p): Delete.
11065         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11066
11067 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
11068
11069         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11070         function.
11071         (do_ifelse): New function.
11072         (do_isel): New function.
11073         (do_sub3): New function.
11074         (do_add3): New function.
11075         (do_load_mask_compare): New function.
11076         (do_overlap_load_compare): New function.
11077         (expand_compare_loop): New function.
11078         (expand_block_compare): Call expand_compare_loop() when appropriate.
11079         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11080         option description.
11081         (-mblock-compare-inline-loop-limit): New option.
11082
11083 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11084
11085         PR target/83677
11086         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11087         Reverse order of second and third operands in first alternative.
11088         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11089         of first and second elements in UNSPEC_VPERMR vector.
11090         (altivec_expand_vec_perm_le): Likewise.
11091
11092 2018-01-08  Jeff Law  <law@redhat.com>
11093
11094         PR rtl-optimizatin/81308
11095         * tree-switch-conversion.c (cfg_altered): New file scoped static.
11096         (process_switch): If group_case_labels makes a change, then set
11097         cfg_altered.
11098         (pass_convert_switch::execute): If a switch is converted, then
11099         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
11100
11101         PR rtl-optimization/81308
11102         * recog.c (split_all_insns): Conditionally cleanup the CFG after
11103         splitting insns.
11104
11105 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
11106
11107         PR target/83663 - Revert r255946
11108         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11109         generation for cases where splatting a value is not useful.
11110         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11111         across a vec_duplicate and a paradoxical subreg forming a vector
11112         mode to a vec_concat.
11113
11114 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11115
11116         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11117         -march=armv8.3-a variants.
11118         * config/arm/t-multilib: Likewise.
11119         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
11120
11121 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
11122
11123         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11124         to generate rtl.
11125         (cceq_ior_compare_complement): Give it a name so I can use it, and
11126         change boolean_or_operator predicate to boolean_operator so it can
11127         be used to generate a crand.
11128         (eqne): New code iterator.
11129         (bd/bd_neg): New code_attrs.
11130         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11131         a single define_insn.
11132         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11133         decrement (bdnzt/bdnzf/bdzt/bdzf).
11134         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11135         with the new names of the branch decrement patterns, and added the
11136         names of the branch decrement conditional patterns.
11137
11138 2018-01-08  Richard Biener  <rguenther@suse.de>
11139
11140         PR tree-optimization/83563
11141         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11142         cache.
11143
11144 2018-01-08  Richard Biener  <rguenther@suse.de>
11145
11146         PR middle-end/83713
11147         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11148
11149 2018-01-08  Richard Biener  <rguenther@suse.de>
11150
11151         PR tree-optimization/83685
11152         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11153         references to abnormals.
11154
11155 2018-01-08  Richard Biener  <rguenther@suse.de>
11156
11157         PR lto/83719
11158         * dwarf2out.c (output_indirect_strings): Handle empty
11159         skeleton_debug_str_hash.
11160         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11161
11162 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
11163
11164         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11165         (emit_store_direct): Likewise.
11166         (arc_trampoline_adjust_address): Likewise.
11167         (arc_asm_trampoline_template): New function.
11168         (arc_initialize_trampoline): Use asm_trampoline_template.
11169         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11170         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11171         * config/arc/arc.md (flush_icache): Delete pattern.
11172
11173 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
11174
11175         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11176         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11177         munaligned-access.
11178
11179 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11180
11181         PR target/83681
11182         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11183         by not USED_FOR_TARGET.
11184         (make_pass_resolve_sw_modes): Likewise.
11185
11186 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11187
11188         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11189         USED_FOR_TARGET.
11190
11191 2018-01-08  Richard Biener  <rguenther@suse.de>
11192
11193         PR middle-end/83580
11194         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11195
11196 2018-01-08  Richard Biener  <rguenther@suse.de>
11197
11198         PR middle-end/83517
11199         * match.pd ((t * 2) / 2) -> t): Add missing :c.
11200
11201 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
11202
11203         PR middle-end/81897
11204         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11205         basic blocks with a small number of successors.
11206         (convert_control_dep_chain_into_preds): Improve handling of
11207         forwarder blocks.
11208         (dump_predicates): Split apart into...
11209         (dump_pred_chain): ...here...
11210         (dump_pred_info): ...and here.
11211         (can_one_predicate_be_invalidated_p): Add debugging printfs.
11212         (can_chain_union_be_invalidated_p): Improve check for invalidation
11213         of paths.
11214         (uninit_uses_cannot_happen): Avoid unnecessary if
11215         convert_control_dep_chain_into_preds yielded nothing.
11216
11217 2018-01-06  Martin Sebor  <msebor@redhat.com>
11218
11219         PR tree-optimization/83640
11220         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11221         subtracting negative offset from size.
11222         (builtin_access::overlap): Adjust offset bounds of the access to fall
11223         within the size of the object if possible.
11224
11225 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
11226
11227         PR rtl-optimization/83699
11228         * expmed.c (extract_bit_field_1): Restrict the vector usage of
11229         extract_bit_field_as_subreg to cases in which the extracted
11230         value is also a vector.
11231
11232         * lra-constraints.c (process_alt_operands): Test for the equivalence
11233         substitutions when detecting a possible reload cycle.
11234
11235 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
11236
11237         PR debug/83480
11238         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11239         by default if flag_selective_schedling{,2}.  Formatting fixes.
11240
11241         PR rtl-optimization/83682
11242         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11243         if it has non-VECTOR_MODE element mode.
11244         (vec_duplicate_p): Likewise.
11245
11246         PR middle-end/83694
11247         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11248         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11249
11250 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
11251
11252         PR target/83604
11253         * config/i386/i386-builtin.def
11254         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11255         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11256         Require also OPTION_MASK_ISA_AVX512F in addition to
11257         OPTION_MASK_ISA_GFNI.
11258         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11259         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11260         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11261         to OPTION_MASK_ISA_GFNI.
11262         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11263         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11264         OPTION_MASK_ISA_AVX512BW.
11265         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11266         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11267         addition to OPTION_MASK_ISA_GFNI.
11268         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11269         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11270         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11271         to OPTION_MASK_ISA_GFNI.
11272         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11273         a requirement for all ISAs rather than any of them with a few
11274         exceptions.
11275         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11276         processing.
11277         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11278         bitmasks to be enabled with 3 exceptions, instead of requiring any
11279         enabled ISA with lots of exceptions.
11280         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11281         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11282         Change avx512bw in isa attribute to avx512f.
11283         * config/i386/sgxintrin.h: Add license boilerplate.
11284         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
11285         to __AVX512F__ and __AVX512VL to __AVX512VL__.
11286         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11287         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11288         defined.
11289         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11290         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11291         temporarily sse2 rather than sse if not enabled already.
11292
11293         PR target/83604
11294         * config/i386/sse.md (VI248_VLBW): Rename to ...
11295         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
11296         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11297         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11298         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11299         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11300         mode iterator instead of VI248_VLBW.
11301
11302 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
11303
11304         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11305         (record_modified): Skip clobbers; add debug output.
11306         (param_change_prob): Use sreal frequencies.
11307
11308 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
11309
11310         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11311         punt for user-aligned variables.
11312
11313 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
11314
11315         * tree-chrec.c (chrec_contains_symbols): Return true for
11316         POLY_INT_CST.
11317
11318 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
11319
11320         PR target/82439
11321         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11322         of (x|y) == x for BICS pattern.
11323
11324 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
11325
11326         PR tree-optimization/83605
11327         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11328         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11329         can throw.
11330
11331 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11332
11333         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11334         * config/epiphany/rtems.h: New file.
11335
11336 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
11337             Uros Bizjak  <ubizjak@gmail.com>
11338
11339         PR target/83554
11340         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11341         QIreg_operand instead of register_operand predicate.
11342         * config/i386/i386.c (ix86_rop_should_change_byte_p,
11343         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11344         comments instead of -fmitigate[-_]rop.
11345
11346 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11347
11348         PR bootstrap/81926
11349         * cgraphunit.c (symbol_table::compile): Switch to text_section
11350         before calling assembly_start debug hook.
11351         * run-rtl-passes.c (run_rtl_passes): Likewise.
11352         Include output.h.
11353
11354 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11355
11356         * tree-vrp.c (extract_range_from_binary_expr_1): Check
11357         range_int_cst_p rather than !symbolic_range_p before calling
11358         extract_range_from_multiplicative_op_1.
11359
11360 2018-01-04  Jeff Law  <law@redhat.com>
11361
11362         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11363         redundant test in assertion.
11364
11365 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11366
11367         * doc/rtl.texi: Document machine_mode wrapper classes.
11368
11369 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11370
11371         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11372         using tree_to_uhwi.
11373
11374 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11375
11376         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11377         the VEC_PERM_EXPR fold to fail.
11378
11379 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
11380
11381         PR debug/83585
11382         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11383         to switched_sections.
11384
11385 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11386
11387         PR target/83680
11388         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11389         test for d.testing.
11390
11391 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
11392
11393         PR target/83387
11394         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11395         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11396
11397 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
11398
11399         PR debug/83666
11400         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11401         is BLKmode and bitpos not zero or mode change is needed.
11402
11403 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11404
11405         PR target/83675
11406         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11407         TARGET_VIS2.
11408
11409 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
11410
11411         PR target/83628
11412         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11413         instead of MULT rtx.  Update all corresponding splitters.
11414         (*saddl_se): Ditto.
11415         (*ssub<modesuffix>): Ditto.
11416         (*ssubl_se): Ditto.
11417         (*cmp_sadd_di): Update split patterns.
11418         (*cmp_sadd_si): Ditto.
11419         (*cmp_sadd_sidi): Ditto.
11420         (*cmp_ssub_di): Ditto.
11421         (*cmp_ssub_si): Ditto.
11422         (*cmp_ssub_sidi): Ditto.
11423         * config/alpha/predicates.md (const23_operand): New predicate.
11424         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11425         Look for ASHIFT, not MULT inner operand.
11426         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11427
11428 2018-01-04  Martin Liska  <mliska@suse.cz>
11429
11430         PR gcov-profile/83669
11431         * gcov.c (output_intermediate_file): Add version to intermediate
11432         gcov file.
11433         * doc/gcov.texi: Document new field 'version' in intermediate
11434         file format. Fix location of '-k' option of gcov command.
11435
11436 2018-01-04  Martin Liska  <mliska@suse.cz>
11437
11438         PR ipa/82352
11439         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11440
11441 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
11442
11443         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11444
11445 2018-01-03  Martin Sebor  <msebor@redhat.com>
11446
11447         PR tree-optimization/83655
11448         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11449         checking calls with invalid arguments.
11450
11451 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11452
11453         * tree-vect-stmts.c (vect_get_store_rhs): New function.
11454         (vectorizable_mask_load_store): Delete.
11455         (vectorizable_call): Return false for masked loads and stores.
11456         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
11457         instead of gimple_assign_rhs1.
11458         (vectorizable_load): Handle IFN_MASK_LOAD.
11459         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11460
11461 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11462
11463         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11464         split out from..,
11465         (vectorizable_mask_load_store): ...here.
11466         (vectorizable_load): ...and here.
11467
11468 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11469
11470         * tree-vect-stmts.c (vect_build_all_ones_mask)
11471         (vect_build_zero_merge_argument): New functions, split out from...
11472         (vectorizable_load): ...here.
11473
11474 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11475
11476         * tree-vect-stmts.c (vect_check_store_rhs): New function,
11477         split out from...
11478         (vectorizable_mask_load_store): ...here.
11479         (vectorizable_store): ...and here.
11480
11481 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11482
11483         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11484         split out from...
11485         (vectorizable_mask_load_store): ...here.
11486
11487 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11488
11489         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11490         (vect_model_store_cost): Take a vec_load_store_type instead of a
11491         vect_def_type.
11492         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11493         (vect_model_store_cost): Take a vec_load_store_type instead of a
11494         vect_def_type.
11495         (vectorizable_mask_load_store): Update accordingly.
11496         (vectorizable_store): Likewise.
11497         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11498
11499 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11500
11501         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11502         IFN_MASK_LOAD calls here rather than...
11503         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11504
11505 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11506             Alan Hayward  <alan.hayward@arm.com>
11507             David Sherwood  <david.sherwood@arm.com>
11508
11509         * expmed.c (extract_bit_field_1): For vector extracts,
11510         fall back to extract_bit_field_as_subreg if vec_extract
11511         isn't available.
11512
11513 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11514             Alan Hayward  <alan.hayward@arm.com>
11515             David Sherwood  <david.sherwood@arm.com>
11516
11517         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11518         they are variable or constant sized.
11519         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11520         slots for constant-sized data.
11521
11522 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11523             Alan Hayward  <alan.hayward@arm.com>
11524             David Sherwood  <david.sherwood@arm.com>
11525
11526         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11527         handling COND_EXPRs with boolean comparisons, try to find a better
11528         basis for the mask type than the boolean itself.
11529
11530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11531
11532         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11533         is calculated and how it can be overridden.
11534         * genmodes.c (max_bitsize_mode_any_mode): New variable.
11535         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11536         if defined.
11537         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11538         if nonzero.
11539
11540 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11541             Alan Hayward  <alan.hayward@arm.com>
11542             David Sherwood  <david.sherwood@arm.com>
11543
11544         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11545         Remove the mode argument.
11546         (aarch64_simd_valid_immediate): Remove the mode and inverse
11547         arguments.
11548         * config/aarch64/iterators.md (bitsize): New iterator.
11549         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11550         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11551         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11552         aarch64_simd_valid_immediate.
11553         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11554         (aarch64_reg_or_bic_imm): Likewise.
11555         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11556         with an insn_type enum and msl with a modifier_type enum.
11557         Replace element_width with a scalar_mode.  Change the shift
11558         to unsigned int.  Add constructors for scalar_float_mode and
11559         scalar_int_mode elements.
11560         (aarch64_vect_float_const_representable_p): Delete.
11561         (aarch64_can_const_movi_rtx_p)
11562         (aarch64_simd_scalar_immediate_valid_for_move)
11563         (aarch64_simd_make_constant): Update call to
11564         aarch64_simd_valid_immediate.
11565         (aarch64_advsimd_valid_immediate_hs): New function.
11566         (aarch64_advsimd_valid_immediate): Likewise.
11567         (aarch64_simd_valid_immediate): Remove mode and inverse
11568         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
11569         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11570         and aarch64_float_const_representable_p on the result.
11571         (aarch64_output_simd_mov_immediate): Remove mode argument.
11572         Update call to aarch64_simd_valid_immediate and use of
11573         simd_immediate_info.
11574         (aarch64_output_scalar_simd_mov_immediate): Update call
11575         accordingly.
11576
11577 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11578             Alan Hayward  <alan.hayward@arm.com>
11579             David Sherwood  <david.sherwood@arm.com>
11580
11581         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11582         (mode_nunits): Likewise CONST_MODE_NUNITS.
11583         * machmode.def (ADJUST_NUNITS): Document.
11584         * genmodes.c (mode_data::need_nunits_adj): New field.
11585         (blank_mode): Update accordingly.
11586         (adj_nunits): New variable.
11587         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11588         parameter.
11589         (emit_mode_size_inline): Set need_bytesize_adj for all modes
11590         listed in adj_nunits.
11591         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11592         listed in adj_nunits.  Don't emit case statements for such modes.
11593         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11594         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
11595         nothing if adj_nunits is nonnull.
11596         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11597         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11598         (emit_mode_fbit): Update use of print_maybe_const_decl.
11599         (emit_move_size): Likewise.  Treat the array as non-const
11600         if adj_nunits.
11601         (emit_mode_adjustments): Handle adj_nunits.
11602
11603 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11604
11605         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11606         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11607         (VECTOR_MODES): Use it.
11608         (make_vector_modes): Take the prefix as an argument.
11609
11610 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11611             Alan Hayward  <alan.hayward@arm.com>
11612             David Sherwood  <david.sherwood@arm.com>
11613
11614         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11615         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11616         for MODE_VECTOR_BOOL.
11617         * machmode.def (VECTOR_BOOL_MODE): Document.
11618         * genmodes.c (VECTOR_BOOL_MODE): New macro.
11619         (make_vector_bool_mode): New function.
11620         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11621         MODE_VECTOR_BOOL.
11622         * lto-streamer-in.c (lto_input_mode_table): Likewise.
11623         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11624         Likewise.
11625         * stor-layout.c (int_mode_for_mode): Likewise.
11626         * tree.c (build_vector_type_for_mode): Likewise.
11627         * varasm.c (output_constant_pool_2): Likewise.
11628         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11629         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
11630         for MODE_VECTOR_BOOL.
11631         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11632         of mode class checks.
11633         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11634         instead of a list of mode class checks.
11635         (expand_vector_scalar_condition): Likewise.
11636         (type_for_widest_vector_mode): Handle BImode as an inner mode.
11637
11638 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11639             Alan Hayward  <alan.hayward@arm.com>
11640             David Sherwood  <david.sherwood@arm.com>
11641
11642         * machmode.h (mode_size): Change from unsigned short to
11643         poly_uint16_pod.
11644         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11645         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11646         or if measurement_type is not polynomial.
11647         (fixed_size_mode::includes_p): Check for constant-sized modes.
11648         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11649         return a poly_uint16 rather than an unsigned short.
11650         (emit_mode_size): Change the type of mode_size from unsigned short
11651         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
11652         (emit_mode_adjustments): Cope with polynomial vector sizes.
11653         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11654         for GET_MODE_SIZE.
11655         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11656         for GET_MODE_SIZE.
11657         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11658         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11659         * caller-save.c (setup_save_areas): Likewise.
11660         (replace_reg_with_saved_mem): Likewise.
11661         * calls.c (emit_library_call_value_1): Likewise.
11662         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11663         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11664         (gen_lowpart_for_combine): Likewise.
11665         * convert.c (convert_to_integer_1): Likewise.
11666         * cse.c (equiv_constant, cse_insn): Likewise.
11667         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11668         (cselib_subst_to_values): Likewise.
11669         * dce.c (word_dce_process_block): Likewise.
11670         * df-problems.c (df_word_lr_mark_ref): Likewise.
11671         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11672         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11673         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11674         (rtl_for_decl_location): Likewise.
11675         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11676         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11677         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11678         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11679         (expand_expr_real_1): Likewise.
11680         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11681         (pad_below): Likewise.
11682         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11683         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11684         * ira.c (get_subreg_tracking_sizes): Likewise.
11685         * ira-build.c (ira_create_allocno_objects): Likewise.
11686         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11687         (ira_sort_regnos_for_alter_reg): Likewise.
11688         * ira-costs.c (record_operand_costs): Likewise.
11689         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11690         (resolve_simple_move): Likewise.
11691         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11692         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11693         (lra_constraints): Likewise.
11694         (CONST_POOL_OK_P): Reject variable-sized modes.
11695         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11696         (add_pseudo_to_slot, lra_spill): Likewise.
11697         * omp-low.c (omp_clause_aligned_alignment): Likewise.
11698         * optabs-query.c (get_best_extraction_insn): Likewise.
11699         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11700         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11701         (expand_mult_highpart, valid_multiword_target_p): Likewise.
11702         * recog.c (offsettable_address_addr_space_p): Likewise.
11703         * regcprop.c (maybe_mode_change): Likewise.
11704         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11705         * regrename.c (build_def_use): Likewise.
11706         * regstat.c (dump_reg_info): Likewise.
11707         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11708         (find_reloads, find_reloads_subreg_address): Likewise.
11709         * reload1.c (eliminate_regs_1): Likewise.
11710         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11711         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11712         (simplify_binary_operation_1, simplify_subreg): Likewise.
11713         * targhooks.c (default_function_arg_padding): Likewise.
11714         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11715         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11716         (verify_gimple_assign_ternary): Likewise.
11717         * tree-inline.c (estimate_move_cost): Likewise.
11718         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11719         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11720         (get_address_cost_ainc): Likewise.
11721         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11722         (vect_supportable_dr_alignment): Likewise.
11723         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11724         (vectorizable_reduction): Likewise.
11725         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11726         (vectorizable_operation, vectorizable_load): Likewise.
11727         * tree.c (build_same_sized_truth_vector_type): Likewise.
11728         * valtrack.c (cleanup_auto_inc_dec): Likewise.
11729         * var-tracking.c (emit_note_insn_var_location): Likewise.
11730         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11731         (ADDR_VEC_ALIGN): Likewise.
11732
11733 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11734             Alan Hayward  <alan.hayward@arm.com>
11735             David Sherwood  <david.sherwood@arm.com>
11736
11737         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11738         unsigned short.
11739         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11740         or if measurement_type is polynomial.
11741         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11742         * combine.c (make_extraction): Likewise.
11743         * dse.c (find_shift_sequence): Likewise.
11744         * dwarf2out.c (mem_loc_descriptor): Likewise.
11745         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11746         (extract_bit_field, extract_low_bits): Likewise.
11747         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11748         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11749         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11750         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11751         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11752         * reload.c (find_reloads): Likewise.
11753         * reload1.c (alter_reg): Likewise.
11754         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11755         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11756         * tree-if-conv.c (predicate_mem_writes): Likewise.
11757         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11758         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11759         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11760         * valtrack.c (dead_debug_insert_temp): Likewise.
11761         * varasm.c (mergeable_constant_section): Likewise.
11762         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11763
11764 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11765             Alan Hayward  <alan.hayward@arm.com>
11766             David Sherwood  <david.sherwood@arm.com>
11767
11768         * expr.c (expand_assignment): Cope with polynomial mode sizes
11769         when assigning to a CONCAT.
11770
11771 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11772             Alan Hayward  <alan.hayward@arm.com>
11773             David Sherwood  <david.sherwood@arm.com>
11774
11775         * machmode.h (mode_precision): Change from unsigned short to
11776         poly_uint16_pod.
11777         (mode_to_precision): Return a poly_uint16 rather than an unsigned
11778         short.
11779         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11780         or if measurement_type is not polynomial.
11781         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
11782         in which the mode is already known to be a scalar_int_mode.
11783         * genmodes.c (emit_mode_precision): Change the type of mode_precision
11784         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
11785         initializer.
11786         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11787         for GET_MODE_PRECISION.
11788         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11789         for GET_MODE_PRECISION.
11790         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11791         as polynomial.
11792         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11793         (expand_field_assignment, make_extraction): Likewise.
11794         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11795         (get_last_value): Likewise.
11796         * convert.c (convert_to_integer_1): Likewise.
11797         * cse.c (cse_insn): Likewise.
11798         * expr.c (expand_expr_real_1): Likewise.
11799         * lra-constraints.c (simplify_operand_subreg): Likewise.
11800         * optabs-query.c (can_atomic_load_p): Likewise.
11801         * optabs.c (expand_atomic_load): Likewise.
11802         (expand_atomic_store): Likewise.
11803         * ree.c (combine_reaching_defs): Likewise.
11804         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11805         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11806         * tree.h (type_has_mode_precision_p): Likewise.
11807         * ubsan.c (instrument_si_overflow): Likewise.
11808
11809 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11810             Alan Hayward  <alan.hayward@arm.com>
11811             David Sherwood  <david.sherwood@arm.com>
11812
11813         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11814         polynomial numbers of units.
11815         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11816         (valid_vector_subparts_p): New function.
11817         (build_vector_type): Remove temporary shim and take the number
11818         of units as a poly_uint64 rather than an int.
11819         (build_opaque_vector_type): Take the number of units as a
11820         poly_uint64 rather than an int.
11821         * tree.c (build_vector_from_ctor): Handle polynomial
11822         TYPE_VECTOR_SUBPARTS.
11823         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11824         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11825         (build_vector_from_val): If the number of units is variable,
11826         use build_vec_duplicate_cst for constant operands and
11827         VEC_DUPLICATE_EXPR otherwise.
11828         (make_vector_type): Remove temporary is_constant ().
11829         (build_vector_type, build_opaque_vector_type): Take the number of
11830         units as a poly_uint64 rather than an int.
11831         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11832         VECTOR_CST_NELTS.
11833         * cfgexpand.c (expand_debug_expr): Likewise.
11834         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11835         (store_constructor, expand_expr_real_1): Likewise.
11836         (const_scalar_mask_from_tree): Likewise.
11837         * fold-const-call.c (fold_const_reduction): Likewise.
11838         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11839         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11840         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11841         (fold_relational_const): Likewise.
11842         (native_interpret_vector): Likewise.  Change the size from an
11843         int to an unsigned int.
11844         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11845         TYPE_VECTOR_SUBPARTS.
11846         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11847         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11848         duplicating a non-constant operand into a variable-length vector.
11849         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11850         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11851         * ipa-icf.c (sem_variable::equals): Likewise.
11852         * match.pd: Likewise.
11853         * omp-simd-clone.c (simd_clone_subparts): Likewise.
11854         * print-tree.c (print_node): Likewise.
11855         * stor-layout.c (layout_type): Likewise.
11856         * targhooks.c (default_builtin_vectorization_cost): Likewise.
11857         * tree-cfg.c (verify_gimple_comparison): Likewise.
11858         (verify_gimple_assign_binary): Likewise.
11859         (verify_gimple_assign_ternary): Likewise.
11860         (verify_gimple_assign_single): Likewise.
11861         * tree-pretty-print.c (dump_generic_node): Likewise.
11862         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11863         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11864         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11865         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11866         (vect_shift_permute_load_chain): Likewise.
11867         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11868         (expand_vector_condition, optimize_vector_constructor): Likewise.
11869         (lower_vec_perm, get_compute_type): Likewise.
11870         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11871         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11872         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11873         (vect_recog_mask_conversion_pattern): Likewise.
11874         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11875         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11876         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11877         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11878         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11879         (vectorizable_shift, vectorizable_operation, vectorizable_store)
11880         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11881         (supportable_widening_operation): Likewise.
11882         (supportable_narrowing_operation): Likewise.
11883         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11884         Likewise.
11885         * varasm.c (output_constant): Likewise.
11886
11887 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11888             Alan Hayward  <alan.hayward@arm.com>
11889             David Sherwood  <david.sherwood@arm.com>
11890
11891         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11892         so that both the length == 3 and length != 3 cases set up their
11893         own permute vectors.  Add comments explaining why we know the
11894         number of elements is constant.
11895         (vect_permute_load_chain): Likewise.
11896
11897 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11898             Alan Hayward  <alan.hayward@arm.com>
11899             David Sherwood  <david.sherwood@arm.com>
11900
11901         * machmode.h (mode_nunits): Change from unsigned char to
11902         poly_uint16_pod.
11903         (ONLY_FIXED_SIZE_MODES): New macro.
11904         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11905         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11906         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11907         New typedefs.
11908         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11909         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11910         or if measurement_type is not polynomial.
11911         * genmodes.c (ZERO_COEFFS): New macro.
11912         (emit_mode_nunits_inline): Make mode_nunits_inline return a
11913         poly_uint16.
11914         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11915         Use ZERO_COEFFS when emitting initializers.
11916         * data-streamer.h (bp_pack_poly_value): New function.
11917         (bp_unpack_poly_value): Likewise.
11918         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11919         for GET_MODE_NUNITS.
11920         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11921         for GET_MODE_NUNITS.
11922         * tree.c (make_vector_type): Remove temporary shim and make
11923         the real function take the number of units as a poly_uint64
11924         rather than an int.
11925         (build_vector_type_for_mode): Handle polynomial nunits.
11926         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11927         * emit-rtl.c (const_vec_series_p_1): Likewise.
11928         (gen_rtx_CONST_VECTOR): Likewise.
11929         * fold-const.c (test_vec_duplicate_folding): Likewise.
11930         * genrecog.c (validate_pattern): Likewise.
11931         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11932         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11933         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11934         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11935         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11936         * rtlanal.c (subreg_get_info): Likewise.
11937         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11938         (vect_grouped_load_supported): Likewise.
11939         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11940         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11941         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11942         (simplify_const_unary_operation, simplify_binary_operation_1)
11943         (simplify_const_binary_operation, simplify_ternary_operation)
11944         (test_vector_ops_duplicate, test_vector_ops): Likewise.
11945         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11946         instead of CONST_VECTOR_NUNITS.
11947         * varasm.c (output_constant_pool_2): Likewise.
11948         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11949         explicit-encoded elements in the XVEC for variable-length vectors.
11950
11951 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11952
11953         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11954
11955 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11956             Alan Hayward  <alan.hayward@arm.com>
11957             David Sherwood  <david.sherwood@arm.com>
11958
11959         * coretypes.h (fixed_size_mode): Declare.
11960         (fixed_size_mode_pod): New typedef.
11961         * builtins.h (target_builtins::x_apply_args_mode)
11962         (target_builtins::x_apply_result_mode): Change type to
11963         fixed_size_mode_pod.
11964         * builtins.c (apply_args_size, apply_result_size, result_vector)
11965         (expand_builtin_apply_args_1, expand_builtin_apply)
11966         (expand_builtin_return): Update accordingly.
11967
11968 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11969
11970         * cse.c (hash_rtx_cb): Hash only the encoded elements.
11971         * cselib.c (cselib_hash_rtx): Likewise.
11972         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11973         CONST_VECTOR encoding.
11974
11975 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
11976             Jeff Law  <law@redhat.com>
11977
11978         PR target/83641
11979         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11980         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11981         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11982         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11983
11984         PR target/83641
11985         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11986         explicitly probe *sp in a noreturn function if there were any callee
11987         register saves or frame pointer is needed.
11988
11989 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
11990
11991         PR debug/83621
11992         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11993         BLKmode for ternary, binary or unary expressions.
11994
11995         PR debug/83645
11996         * var-tracking.c (delete_vta_debug_insn): New inline function.
11997         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11998         insns from get_insns () to NULL instead of each bb separately.
11999         Use delete_vta_debug_insn.  No longer static.
12000         (vt_debug_insns_local, variable_tracking_main_1): Adjust
12001         delete_vta_debug_insns callers.
12002         * rtl.h (delete_vta_debug_insns): Declare.
12003         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12004         instead of variable_tracking_main.
12005
12006 2018-01-03  Martin Sebor  <msebor@redhat.com>
12007
12008         PR tree-optimization/83603
12009         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12010         arguments past the endof the argument list in functions declared
12011         without a prototype.
12012         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12013         Avoid checking when arguments are null.
12014
12015 2018-01-03  Martin Sebor  <msebor@redhat.com>
12016
12017         PR c/83559
12018         * doc/extend.texi (attribute const): Fix a typo.
12019         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12020         issuing -Wsuggest-attribute for void functions.
12021
12022 2018-01-03  Martin Sebor  <msebor@redhat.com>
12023
12024         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12025         offset_int::from instead of wide_int::to_shwi.
12026         (maybe_diag_overlap): Remove assertion.
12027         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12028         * gimple-ssa-sprintf.c (format_directive): Same.
12029         (parse_directive): Same.
12030         (sprintf_dom_walker::compute_format_length): Same.
12031         (try_substitute_return_value): Same.
12032
12033 2018-01-03  Jeff Law  <law@redhat.com>
12034
12035         PR middle-end/83654
12036         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12037         non-constant residual for zero at runtime and avoid probing in
12038         that case.  Reorganize code for trailing problem to mirror handling
12039         of the residual.
12040
12041 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12042
12043         PR tree-optimization/83501
12044         * tree-ssa-strlen.c (get_string_cst): New.
12045         (handle_char_store): Call get_string_cst.
12046
12047 2018-01-03  Martin Liska  <mliska@suse.cz>
12048
12049         PR tree-optimization/83593
12050         * tree-ssa-strlen.c: Include tree-cfg.h.
12051         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12052         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12053         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12054         to false.
12055         (strlen_dom_walker::before_dom_children): Call
12056         gimple_purge_dead_eh_edges. Dump tranformation with details
12057         dump flags.
12058         (strlen_dom_walker::before_dom_children): Update call by adding
12059         new argument cleanup_eh.
12060         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12061
12062 2018-01-03  Martin Liska  <mliska@suse.cz>
12063
12064         PR ipa/83549
12065         * cif-code.def (VARIADIC_THUNK): New enum value.
12066         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12067         thunks.
12068
12069 2018-01-03  Jan Beulich  <jbeulich@suse.com>
12070
12071         * sse.md (mov<mode>_internal): Tighten condition for when to use
12072         vmovdqu<ssescalarsize> for TI and OI modes.
12073
12074 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
12075
12076         Update copyright years.
12077
12078 2018-01-03  Martin Liska  <mliska@suse.cz>
12079
12080         PR ipa/83594
12081         * ipa-visibility.c (function_and_variable_visibility): Skip
12082         functions with noipa attribure.
12083
12084 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
12085
12086         * gcc.c (process_command): Update copyright notice dates.
12087         * gcov-dump.c (print_version): Ditto.
12088         * gcov.c (print_version): Ditto.
12089         * gcov-tool.c (print_version): Ditto.
12090         * gengtype.c (create_file): Ditto.
12091         * doc/cpp.texi: Bump @copying's copyright year.
12092         * doc/cppinternals.texi: Ditto.
12093         * doc/gcc.texi: Ditto.
12094         * doc/gccint.texi: Ditto.
12095         * doc/gcov.texi: Ditto.
12096         * doc/install.texi: Ditto.
12097         * doc/invoke.texi: Ditto.
12098
12099 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12100
12101         * vector-builder.h (vector_builder::m_full_nelts): Change from
12102         unsigned int to poly_uint64.
12103         (vector_builder::full_nelts): Update prototype accordingly.
12104         (vector_builder::new_vector): Likewise.
12105         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12106         (vector_builder::operator ==): Likewise.
12107         (vector_builder::finalize): Likewise.
12108         * int-vector-builder.h (int_vector_builder::int_vector_builder):
12109         Take the number of elements as a poly_uint64 rather than an
12110         unsigned int.
12111         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12112         from unsigned int to poly_uint64.
12113         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12114         (vec_perm_indices::new_vector): Likewise.
12115         (vec_perm_indices::length): Likewise.
12116         (vec_perm_indices::nelts_per_input): Likewise.
12117         (vec_perm_indices::input_nelts): Likewise.
12118         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12119         number of elements per input as a poly_uint64 rather than an
12120         unsigned int.  Use the original encoding for variable-length
12121         vectors, rather than clamping each individual element.
12122         For the second and subsequent elements in each pattern,
12123         clamp the step and base before clamping their sum.
12124         (vec_perm_indices::series_p): Handle polynomial element counts.
12125         (vec_perm_indices::all_in_range_p): Likewise.
12126         (vec_perm_indices_to_tree): Likewise.
12127         (vec_perm_indices_to_rtx): Likewise.
12128         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12129         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12130         (tree_vector_builder::new_binary_operation): Handle polynomial
12131         element counts.  Return false if we need to know the number
12132         of elements at compile time.
12133         * fold-const.c (fold_vec_perm): Punt if the number of elements
12134         isn't known at compile time.
12135
12136 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12137
12138         * vec-perm-indices.h (vec_perm_builder): Change element type
12139         from HOST_WIDE_INT to poly_int64.
12140         (vec_perm_indices::element_type): Update accordingly.
12141         (vec_perm_indices::clamp): Handle polynomial element_types.
12142         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12143         (vec_perm_indices::all_in_range_p): Likewise.
12144         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12145         than shwi trees.
12146         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12147         polynomial vec_perm_indices element types.
12148         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12149         * fold-const.c (fold_vec_perm): Likewise.
12150         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12151         * tree-vect-generic.c (lower_vec_perm): Likewise.
12152         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12153         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12154         element type to HOST_WIDE_INT.
12155
12156 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12157             Alan Hayward  <alan.hayward@arm.com>
12158             David Sherwood  <david.sherwood@arm.com>
12159
12160         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12161         rather than an int.  Use plus_constant.
12162         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12163         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12164
12165 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12166             Alan Hayward  <alan.hayward@arm.com>
12167             David Sherwood  <david.sherwood@arm.com>
12168
12169         * calls.c (emit_call_1, expand_call): Change struct_value_size from
12170         a HOST_WIDE_INT to a poly_int64.
12171
12172 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12173             Alan Hayward  <alan.hayward@arm.com>
12174             David Sherwood  <david.sherwood@arm.com>
12175
12176         * calls.c (load_register_parameters): Cope with polynomial
12177         mode sizes.  Require a constant size for BLKmode parameters
12178         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
12179         forces a parameter to be padded at the lsb end in order to
12180         fill a complete number of words, require the parameter size
12181         to be ordered wrt UNITS_PER_WORD.
12182
12183 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12184             Alan Hayward  <alan.hayward@arm.com>
12185             David Sherwood  <david.sherwood@arm.com>
12186
12187         * reload1.c (spill_stack_slot_width): Change element type
12188         from unsigned int to poly_uint64_pod.
12189         (alter_reg): Treat mode sizes as polynomial.
12190
12191 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12192             Alan Hayward  <alan.hayward@arm.com>
12193             David Sherwood  <david.sherwood@arm.com>
12194
12195         * reload.c (complex_word_subreg_p): New function.
12196         (reload_inner_reg_of_subreg, push_reload): Use it.
12197
12198 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12199             Alan Hayward  <alan.hayward@arm.com>
12200             David Sherwood  <david.sherwood@arm.com>
12201
12202         * lra-constraints.c (process_alt_operands): Reject matched
12203         operands whose sizes aren't ordered.
12204         (match_reload): Refer to this check here.
12205
12206 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12207             Alan Hayward  <alan.hayward@arm.com>
12208             David Sherwood  <david.sherwood@arm.com>
12209
12210         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12211         that the mode size is in the set {1, 2, 4, 8, 16}.
12212
12213 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12214             Alan Hayward  <alan.hayward@arm.com>
12215             David Sherwood  <david.sherwood@arm.com>
12216
12217         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12218         Use plus_constant instead of gen_rtx_PLUS.
12219
12220 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12221             Alan Hayward  <alan.hayward@arm.com>
12222             David Sherwood  <david.sherwood@arm.com>
12223
12224         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12225         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12226         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12227         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12228         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12229         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12230         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12231         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12232         * config/i386/i386.c (ix86_push_rounding): ...this new function.
12233         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12234         a poly_int64.
12235         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12236         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12237         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12238         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12239         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12240         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12241         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12242         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12243         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12244         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12245         function.
12246         * expr.c (emit_move_resolve_push): Treat the input and result
12247         of PUSH_ROUNDING as a poly_int64.
12248         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12249         (emit_push_insn): Likewise.
12250         * lra-eliminations.c (mark_not_eliminable): Likewise.
12251         * recog.c (push_operand): Likewise.
12252         * reload1.c (elimination_effects): Likewise.
12253         * rtlanal.c (nonzero_bits1): Likewise.
12254         * calls.c (store_one_arg): Likewise.  Require the padding to be
12255         known at compile time.
12256
12257 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12258             Alan Hayward  <alan.hayward@arm.com>
12259             David Sherwood  <david.sherwood@arm.com>
12260
12261         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12262         Use plus_constant instead of gen_rtx_PLUS.
12263
12264 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12265             Alan Hayward  <alan.hayward@arm.com>
12266             David Sherwood  <david.sherwood@arm.com>
12267
12268         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12269         rather than an int.
12270
12271 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12272             Alan Hayward  <alan.hayward@arm.com>
12273             David Sherwood  <david.sherwood@arm.com>
12274
12275         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12276         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12277         via stack temporaries.  Treat the mode size as polynomial too.
12278
12279 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12280             Alan Hayward  <alan.hayward@arm.com>
12281             David Sherwood  <david.sherwood@arm.com>
12282
12283         * expr.c (expand_expr_real_2): When handling conversions involving
12284         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12285         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
12286         as a poly_uint64 too.
12287
12288 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12289             Alan Hayward  <alan.hayward@arm.com>
12290             David Sherwood  <david.sherwood@arm.com>
12291
12292         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12293
12294 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12295             Alan Hayward  <alan.hayward@arm.com>
12296             David Sherwood  <david.sherwood@arm.com>
12297
12298         * combine.c (can_change_dest_mode): Handle polynomial
12299         REGMODE_NATURAL_SIZE.
12300         * expmed.c (store_bit_field_1): Likewise.
12301         * expr.c (store_constructor): Likewise.
12302         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12303         and polynomial REGMODE_NATURAL_SIZE.
12304         (gen_lowpart_common): Likewise.
12305         * reginfo.c (record_subregs_of_mode): Likewise.
12306         * rtlanal.c (read_modify_subreg_p): Likewise.
12307
12308 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12309             Alan Hayward  <alan.hayward@arm.com>
12310             David Sherwood  <david.sherwood@arm.com>
12311
12312         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12313         numbers of elements.
12314
12315 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12316             Alan Hayward  <alan.hayward@arm.com>
12317             David Sherwood  <david.sherwood@arm.com>
12318
12319         * match.pd: Cope with polynomial numbers of vector elements.
12320
12321 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12322             Alan Hayward  <alan.hayward@arm.com>
12323             David Sherwood  <david.sherwood@arm.com>
12324
12325         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12326         in a POINTER_PLUS_EXPR.
12327
12328 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12329             Alan Hayward  <alan.hayward@arm.com>
12330             David Sherwood  <david.sherwood@arm.com>
12331
12332         * omp-simd-clone.c (simd_clone_subparts): New function.
12333         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12334         (ipa_simd_modify_function_body): Likewise.
12335
12336 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12337             Alan Hayward  <alan.hayward@arm.com>
12338             David Sherwood  <david.sherwood@arm.com>
12339
12340         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12341         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12342         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12343         (expand_vector_condition, vector_element): Likewise.
12344         (subparts_gt): New function.
12345         (get_compute_type): Use subparts_gt.
12346         (count_type_subparts): Delete.
12347         (expand_vector_operations_1): Use subparts_gt instead of
12348         count_type_subparts.
12349
12350 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12351             Alan Hayward  <alan.hayward@arm.com>
12352             David Sherwood  <david.sherwood@arm.com>
12353
12354         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12355         (vect_compile_time_alias): ...this new function.  Do the calculation
12356         on poly_ints rather than trees.
12357         (vect_prune_runtime_alias_test_list): Update call accordingly.
12358
12359 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12360             Alan Hayward  <alan.hayward@arm.com>
12361             David Sherwood  <david.sherwood@arm.com>
12362
12363         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12364         numbers of units.
12365         (vect_schedule_slp_instance): Likewise.
12366
12367 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12368             Alan Hayward  <alan.hayward@arm.com>
12369             David Sherwood  <david.sherwood@arm.com>
12370
12371         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12372         constant and extern definitions for variable-length vectors.
12373         (vect_get_constant_vectors): Note that the number of units
12374         is known to be constant.
12375
12376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12377             Alan Hayward  <alan.hayward@arm.com>
12378             David Sherwood  <david.sherwood@arm.com>
12379
12380         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12381         of units as polynomial.  Choose between WIDE and NARROW based
12382         on multiple_p.
12383
12384 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12385             Alan Hayward  <alan.hayward@arm.com>
12386             David Sherwood  <david.sherwood@arm.com>
12387
12388         * tree-vect-stmts.c (simd_clone_subparts): New function.
12389         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12390
12391 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12392             Alan Hayward  <alan.hayward@arm.com>
12393             David Sherwood  <david.sherwood@arm.com>
12394
12395         * tree-vect-stmts.c (vectorizable_call): Treat the number of
12396         vectors as polynomial.  Use build_index_vector for
12397         IFN_GOMP_SIMD_LANE.
12398
12399 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12400             Alan Hayward  <alan.hayward@arm.com>
12401             David Sherwood  <david.sherwood@arm.com>
12402
12403         * tree-vect-stmts.c (get_load_store_type): Treat the number of
12404         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12405         for variable-length vectors.
12406         (vectorizable_mask_load_store): Treat the number of units as
12407         polynomial, asserting that it is constant if the condition has
12408         already been enforced.
12409         (vectorizable_store, vectorizable_load): Likewise.
12410
12411 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12412             Alan Hayward  <alan.hayward@arm.com>
12413             David Sherwood  <david.sherwood@arm.com>
12414
12415         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12416         of units as polynomial.  Punt if we can't tell at compile time
12417         which vector contains the final result.
12418
12419 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12420             Alan Hayward  <alan.hayward@arm.com>
12421             David Sherwood  <david.sherwood@arm.com>
12422
12423         * tree-vect-loop.c (vectorizable_induction): Treat the number
12424         of units as polynomial.  Punt on SLP inductions.  Use an integer
12425         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
12426         cast of such a series for variable-length floating-point
12427         reductions.
12428
12429 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12430             Alan Hayward  <alan.hayward@arm.com>
12431             David Sherwood  <david.sherwood@arm.com>
12432
12433         * tree.h (build_index_vector): Declare.
12434         * tree.c (build_index_vector): New function.
12435         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12436         of units as polynomial, forcibly converting it to a constant if
12437         vectorizable_reduction has already enforced the condition.
12438         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
12439         to create a {1,2,3,...} vector.
12440         (vectorizable_reduction): Treat the number of units as polynomial.
12441         Choose vectype_in based on the largest scalar element size rather
12442         than the smallest number of units.  Enforce the restrictions
12443         relied on above.
12444
12445 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12446             Alan Hayward  <alan.hayward@arm.com>
12447             David Sherwood  <david.sherwood@arm.com>
12448
12449         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12450         number of units as polynomial.
12451
12452 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12453             Alan Hayward  <alan.hayward@arm.com>
12454             David Sherwood  <david.sherwood@arm.com>
12455
12456         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12457         * target.def (autovectorize_vector_sizes): Return the vector sizes
12458         by pointer, using vector_sizes rather than a bitmask.
12459         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12460         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12461         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12462         Likewise.
12463         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12464         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12465         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12466         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12467         * omp-general.c (omp_max_vf): Likewise.
12468         * omp-low.c (omp_clause_aligned_alignment): Likewise.
12469         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12470         * tree-vect-loop.c (vect_analyze_loop): Likewise.
12471         * tree-vect-slp.c (vect_slp_bb): Likewise.
12472         * doc/tm.texi: Regenerate.
12473         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12474         to a poly_uint64.
12475         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12476         the vector size as a poly_uint64 rather than an unsigned int.
12477         (current_vector_size): Change from an unsigned int to a poly_uint64.
12478         (get_vectype_for_scalar_type): Update accordingly.
12479         * tree.h (build_truth_vector_type): Take the size and number of
12480         units as a poly_uint64 rather than an unsigned int.
12481         (build_vector_type): Add a temporary overload that takes
12482         the number of units as a poly_uint64 rather than an unsigned int.
12483         * tree.c (make_vector_type): Likewise.
12484         (build_truth_vector_type): Take the number of units as a poly_uint64
12485         rather than an unsigned int.
12486
12487 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12488             Alan Hayward  <alan.hayward@arm.com>
12489             David Sherwood  <david.sherwood@arm.com>
12490
12491         * target.def (get_mask_mode): Take the number of units and length
12492         as poly_uint64s rather than unsigned ints.
12493         * targhooks.h (default_get_mask_mode): Update accordingly.
12494         * targhooks.c (default_get_mask_mode): Likewise.
12495         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12496         * doc/tm.texi: Regenerate.
12497
12498 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12499             Alan Hayward  <alan.hayward@arm.com>
12500             David Sherwood  <david.sherwood@arm.com>
12501
12502         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12503         * omp-general.c (omp_max_vf): Likewise.
12504         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12505         (expand_omp_simd): Handle polynomial safelen.
12506         * omp-low.c (omplow_simd_context): Add a default constructor.
12507         (omplow_simd_context::max_vf): Change from int to poly_uint64.
12508         (lower_rec_simd_input_clauses): Update accordingly.
12509         (lower_rec_input_clauses): Likewise.
12510
12511 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12512             Alan Hayward  <alan.hayward@arm.com>
12513             David Sherwood  <david.sherwood@arm.com>
12514
12515         * tree-vectorizer.h (vect_nunits_for_cost): New function.
12516         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12517         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12518         (vect_analyze_slp_cost): Likewise.
12519         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12520         (vect_model_load_cost): Likewise.
12521
12522 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12523             Alan Hayward  <alan.hayward@arm.com>
12524             David Sherwood  <david.sherwood@arm.com>
12525
12526         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12527         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12528         from an unsigned int * to a poly_uint64_pod *.
12529         (calculate_unrolling_factor): New function.
12530         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
12531
12532 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12533             Alan Hayward  <alan.hayward@arm.com>
12534             David Sherwood  <david.sherwood@arm.com>
12535
12536         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12537         from an unsigned int to a poly_uint64.
12538         (_loop_vec_info::slp_unrolling_factor): Likewise.
12539         (_loop_vec_info::vectorization_factor): Change from an int
12540         to a poly_uint64.
12541         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12542         (vect_get_num_vectors): New function.
12543         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12544         (vect_get_num_copies): Use vect_get_num_vectors.
12545         (vect_analyze_data_ref_dependences): Change max_vf from an int *
12546         to an unsigned int *.
12547         (vect_analyze_data_refs): Change min_vf from an int * to a
12548         poly_uint64 *.
12549         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12550         than an unsigned HOST_WIDE_INT.
12551         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12552         (vect_analyze_data_ref_dependence): Change max_vf from an int *
12553         to an unsigned int *.
12554         (vect_analyze_data_ref_dependences): Likewise.
12555         (vect_compute_data_ref_alignment): Handle polynomial vf.
12556         (vect_enhance_data_refs_alignment): Likewise.
12557         (vect_prune_runtime_alias_test_list): Likewise.
12558         (vect_shift_permute_load_chain): Likewise.
12559         (vect_supportable_dr_alignment): Likewise.
12560         (dependence_distance_ge_vf): Take the vectorization factor as a
12561         poly_uint64 rather than an unsigned HOST_WIDE_INT.
12562         (vect_analyze_data_refs): Change min_vf from an int * to a
12563         poly_uint64 *.
12564         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12565         vfm1 as a poly_uint64 rather than an int.  Make the same change
12566         for the returned bound_scalar.
12567         (vect_gen_vector_loop_niters): Handle polynomial vf.
12568         (vect_do_peeling): Likewise.  Update call to
12569         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12570         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12571         be constant.
12572         * tree-vect-loop.c (vect_determine_vectorization_factor)
12573         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12574         (vect_get_known_peeling_cost): Likewise.
12575         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12576         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12577         (vect_transform_loop): Likewise.  Use the lowest possible VF when
12578         updating the upper bounds of the loop.
12579         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
12580         rather than an int.
12581         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12582         polynomial unroll factors.
12583         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12584         (vect_make_slp_decision): Likewise.
12585         (vect_supported_load_permutation_p): Likewise, and polynomial
12586         vf too.
12587         (vect_analyze_slp_cost): Handle polynomial vf.
12588         (vect_slp_analyze_node_operations): Likewise.
12589         (vect_slp_analyze_bb_1): Likewise.
12590         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12591         than an unsigned HOST_WIDE_INT.
12592         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12593         (vectorizable_load): Handle polynomial vf.
12594         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12595         a poly_uint64.
12596         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12597
12598 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12599             Alan Hayward  <alan.hayward@arm.com>
12600             David Sherwood  <david.sherwood@arm.com>
12601
12602         * match.pd: Handle bit operations involving three constants
12603         and try to fold one pair.
12604
12605 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12606
12607         * tree-vect-loop-manip.c: Include gimple-fold.h.
12608         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12609         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
12610         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12611         Add a path that uses a step of VF instead of 1, but disable it
12612         for now.
12613         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12614         and niters_no_overflow parameters.  Update calls to
12615         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12616         Create a new SSA name if the latter choses to use a ste other
12617         than zero, and return it via niters_vector_mult_vf_var.
12618         * tree-vect-loop.c (vect_transform_loop): Update calls to
12619         vect_do_peeling, vect_gen_vector_loop_niters and
12620         slpeel_make_loop_iterate_ntimes.
12621         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12622         (vect_gen_vector_loop_niters): Update declarations after above changes.
12623
12624 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
12625
12626         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12627         128-bit round to integer instructions.
12628         (ceil<mode>2): Likewise.
12629         (btrunc<mode>2): Likewise.
12630         (round<mode>2): Likewise.
12631
12632 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
12633
12634         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12635         unaligned VSX load/store on P8/P9.
12636         (expand_block_clear): Allow the use of unaligned VSX
12637         load/store on P8/P9.
12638
12639 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
12640
12641         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12642         New function.
12643         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12644         swap associated with both a load and a store.
12645
12646 2018-01-02  Andrew Waterman  <andrew@sifive.com>
12647
12648         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12649         * config/riscv/riscv.md (clear_cache): Use it.
12650
12651 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
12652
12653         * web.c: Remove out-of-date comment.
12654
12655 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12656
12657         * expr.c (fixup_args_size_notes): Check that any existing
12658         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12659         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12660         (emit_single_push_insn): ...here.
12661
12662 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12663
12664         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12665         (const_vector_encoded_nelts): New function.
12666         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12667         (const_vector_int_elt, const_vector_elt): Declare.
12668         * emit-rtl.c (const_vector_int_elt_1): New function.
12669         (const_vector_elt): Likewise.
12670         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12671         of CONST_VECTOR_ELT.
12672
12673 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12674
12675         * expr.c: Include rtx-vector-builder.h.
12676         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12677         directly on the tree encoding.
12678         (const_vector_from_tree): Likewise.
12679         * optabs.c: Include rtx-vector-builder.h.
12680         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12681         sequence of "u" values.
12682         * vec-perm-indices.c: Include rtx-vector-builder.h.
12683         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12684         directly on the vec_perm_indices encoding.
12685
12686 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12687
12688         * doc/rtl.texi (const_vector): Describe new encoding scheme.
12689         * Makefile.in (OBJS): Add rtx-vector-builder.o.
12690         * rtx-vector-builder.h: New file.
12691         * rtx-vector-builder.c: Likewise.
12692         * rtl.h (rtx_def::u2): Add a const_vector field.
12693         (CONST_VECTOR_NPATTERNS): New macro.
12694         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12695         (CONST_VECTOR_DUPLICATE_P): Likewise.
12696         (CONST_VECTOR_STEPPED_P): Likewise.
12697         (CONST_VECTOR_ENCODED_ELT): Likewise.
12698         (const_vec_duplicate_p): Check for a duplicated vector encoding.
12699         (unwrap_const_vec_duplicate): Likewise.
12700         (const_vec_series_p): Check for a non-duplicated vector encoding.
12701         Say that the function only returns true for integer vectors.
12702         * emit-rtl.c: Include rtx-vector-builder.h.
12703         (gen_const_vec_duplicate_1): Delete.
12704         (gen_const_vector): Call gen_const_vec_duplicate instead of
12705         gen_const_vec_duplicate_1.
12706         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12707         (gen_const_vec_duplicate): Use rtx_vector_builder.
12708         (gen_const_vec_series): Likewise.
12709         (gen_rtx_CONST_VECTOR): Likewise.
12710         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12711         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12712         Build a new vector rather than modifying a CONST_VECTOR in-place.
12713         (handle_special_swappables): Update call accordingly.
12714         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12715         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12716         Build a new vector rather than modifying a CONST_VECTOR in-place.
12717         (handle_special_swappables): Update call accordingly.
12718
12719 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12720
12721         * simplify-rtx.c (simplify_const_binary_operation): Use
12722         CONST_VECTOR_ELT instead of XVECEXP.
12723
12724 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12725
12726         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12727         the selector elements to be different from the data elements
12728         if the selector is a VECTOR_CST.
12729         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12730         ssizetype for the selector.
12731
12732 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12733
12734         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12735         before testing each element individually.
12736         * tree-vect-generic.c (lower_vec_perm): Likewise.
12737
12738 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12739
12740         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12741         * selftest-run-tests.c (selftest::run_tests): Call it.
12742         * vector-builder.h (vector_builder::operator ==): New function.
12743         (vector_builder::operator !=): Likewise.
12744         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12745         (vec_perm_indices::all_from_input_p): New function.
12746         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12747         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12748         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12749         instead of reading the VECTOR_CST directly.  Detect whether both
12750         vector inputs are the same before constructing the vec_perm_indices,
12751         and update the number of inputs argument accordingly.  Use the
12752         utility functions added above.  Only construct sel2 if we need to.
12753
12754 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12755
12756         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12757         the broadcast of the low byte.
12758         (expand_mult_highpart): Use an explicit encoding for the permutes.
12759         * optabs-query.c (can_mult_highpart_p): Likewise.
12760         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12761         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12762         (vectorizable_bswap): Likewise.
12763         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12764         explicit encoding for the power-of-2 permutes.
12765         (vect_permute_store_chain): Likewise.
12766         (vect_grouped_load_supported): Likewise.
12767         (vect_permute_load_chain): Likewise.
12768
12769 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12770
12771         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12772         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12773         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12774         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12775         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12776         (vect_gen_perm_mask_any): Likewise.
12777
12778 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12779
12780         * int-vector-builder.h: New file.
12781         * vec-perm-indices.h: Include int-vector-builder.h.
12782         (vec_perm_indices): Redefine as an int_vector_builder.
12783         (auto_vec_perm_indices): Delete.
12784         (vec_perm_builder): Redefine as a stand-alone class.
12785         (vec_perm_indices::vec_perm_indices): New function.
12786         (vec_perm_indices::clamp): Likewise.
12787         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12788         (vec_perm_indices::new_vector): New function.
12789         (vec_perm_indices::new_expanded_vector): Update for new
12790         vec_perm_indices class.
12791         (vec_perm_indices::rotate_inputs): New function.
12792         (vec_perm_indices::all_in_range_p): Operate directly on the
12793         encoded form, without computing elided elements.
12794         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12795         encoding.  Update for new vec_perm_indices class.
12796         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12797         the given vec_perm_builder.
12798         (expand_vec_perm_var): Update vec_perm_builder constructor.
12799         (expand_mult_highpart): Use vec_perm_builder instead of
12800         auto_vec_perm_indices.
12801         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12802         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
12803         or double series encoding as appropriate.
12804         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12805         vec_perm_indices instead of auto_vec_perm_indices.
12806         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12807         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12808         (vect_permute_store_chain): Likewise.
12809         (vect_grouped_load_supported): Likewise.
12810         (vect_permute_load_chain): Likewise.
12811         (vect_shift_permute_load_chain): Likewise.
12812         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12813         (vect_transform_slp_perm_load): Likewise.
12814         (vect_schedule_slp_instance): Likewise.
12815         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12816         (vectorizable_mask_load_store): Likewise.
12817         (vectorizable_bswap): Likewise.
12818         (vectorizable_store): Likewise.
12819         (vectorizable_load): Likewise.
12820         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12821         vec_perm_indices instead of auto_vec_perm_indices.  Use
12822         tree_to_vec_perm_builder to read the vector from a tree.
12823         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12824         vec_perm_builder instead of a vec_perm_indices.
12825         (have_whole_vector_shift): Use vec_perm_builder and
12826         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
12827         truncation to calc_vec_perm_mask_for_shift.
12828         (vect_create_epilog_for_reduction): Likewise.
12829         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12830         from auto_vec_perm_indices to vec_perm_indices.
12831         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12832         instead of changing individual elements.
12833         (aarch64_vectorize_vec_perm_const): Use new_vector to install
12834         the vector in d.perm.
12835         * config/arm/arm.c (expand_vec_perm_d::perm): Change
12836         from auto_vec_perm_indices to vec_perm_indices.
12837         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12838         instead of changing individual elements.
12839         (arm_vectorize_vec_perm_const): Use new_vector to install
12840         the vector in d.perm.
12841         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12842         Update vec_perm_builder constructor.
12843         (rs6000_expand_interleave): Likewise.
12844         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12845         (rs6000_expand_interleave): Likewise.
12846
12847 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12848
12849         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12850         to qimode could truncate the indices.
12851         * optabs.c (expand_vec_perm_var): Likewise.
12852
12853 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12854
12855         * Makefile.in (OBJS): Add vec-perm-indices.o.
12856         * vec-perm-indices.h: New file.
12857         * vec-perm-indices.c: Likewise.
12858         * target.h (vec_perm_indices): Replace with a forward class
12859         declaration.
12860         (auto_vec_perm_indices): Move to vec-perm-indices.h.
12861         * optabs.h: Include vec-perm-indices.h.
12862         (expand_vec_perm): Delete.
12863         (selector_fits_mode_p, expand_vec_perm_var): Declare.
12864         (expand_vec_perm_const): Declare.
12865         * target.def (vec_perm_const_ok): Replace with...
12866         (vec_perm_const): ...this new hook.
12867         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12868         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12869         * doc/tm.texi: Regenerate.
12870         * optabs.def (vec_perm_const): Delete.
12871         * doc/md.texi (vec_perm_const): Likewise.
12872         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12873         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12874         expand_vec_perm for constant permutation vectors.  Assert that
12875         the mode of variable permutation vectors is the integer equivalent
12876         of the mode that is being permuted.
12877         * optabs-query.h (selector_fits_mode_p): Declare.
12878         * optabs-query.c: Include vec-perm-indices.h.
12879         (selector_fits_mode_p): New function.
12880         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12881         is defined, instead of checking whether the vec_perm_const_optab
12882         exists.  Use targetm.vectorize.vec_perm_const instead of
12883         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
12884         fit in the vector mode before using a variable permute.
12885         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12886         vec_perm_indices instead of an rtx.
12887         (expand_vec_perm): Replace with...
12888         (expand_vec_perm_const): ...this new function.  Take the selector
12889         as a vec_perm_indices rather than an rtx.  Also take the mode of
12890         the selector.  Update call to shift_amt_for_vec_perm_mask.
12891         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12892         Use vec_perm_indices::new_expanded_vector to expand the original
12893         selector into bytes.  Check whether the indices fit in the vector
12894         mode before using a variable permute.
12895         (expand_vec_perm_var): Make global.
12896         (expand_mult_highpart): Use expand_vec_perm_const.
12897         * fold-const.c: Includes vec-perm-indices.h.
12898         * tree-ssa-forwprop.c: Likewise.
12899         * tree-vect-data-refs.c: Likewise.
12900         * tree-vect-generic.c: Likewise.
12901         * tree-vect-loop.c: Likewise.
12902         * tree-vect-slp.c: Likewise.
12903         * tree-vect-stmts.c: Likewise.
12904         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12905         Delete.
12906         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12907         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12908         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12909         (aarch64_vectorize_vec_perm_const): ...this new function.
12910         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12911         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12912         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12913         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12914         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12915         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12916         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12917         into...
12918         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
12919         check for NEON modes.
12920         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12921         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12922         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12923         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12924         into...
12925         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
12926         the old VEC_PERM_CONST conditions.
12927         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12928         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12929         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12930         (ia64_vectorize_vec_perm_const_ok): Merge into...
12931         (ia64_vectorize_vec_perm_const): ...this new function.
12932         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12933         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12934         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12935         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12936         * config/mips/mips.c (mips_expand_vec_perm_const)
12937         (mips_vectorize_vec_perm_const_ok): Merge into...
12938         (mips_vectorize_vec_perm_const): ...this new function.
12939         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12940         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12941         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12942         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12943         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12944         (rs6000_expand_vec_perm_const): Delete.
12945         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12946         Delete.
12947         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12948         (altivec_expand_vec_perm_const_le): Take each operand individually.
12949         Operate on constant selectors rather than rtxes.
12950         (altivec_expand_vec_perm_const): Likewise.  Update call to
12951         altivec_expand_vec_perm_const_le.
12952         (rs6000_expand_vec_perm_const): Delete.
12953         (rs6000_vectorize_vec_perm_const_ok): Delete.
12954         (rs6000_vectorize_vec_perm_const): New function.
12955         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12956         an element count and rtx array.
12957         (rs6000_expand_extract_even): Update call accordingly.
12958         (rs6000_expand_interleave): Likewise.
12959         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12960         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12961         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12962         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12963         (rs6000_expand_vec_perm_const): Delete.
12964         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12965         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12966         (altivec_expand_vec_perm_const_le): Take each operand individually.
12967         Operate on constant selectors rather than rtxes.
12968         (altivec_expand_vec_perm_const): Likewise.  Update call to
12969         altivec_expand_vec_perm_const_le.
12970         (rs6000_expand_vec_perm_const): Delete.
12971         (rs6000_vectorize_vec_perm_const_ok): Delete.
12972         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
12973         reference to the SPE evmerge intructions.
12974         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12975         an element count and rtx array.
12976         (rs6000_expand_extract_even): Update call accordingly.
12977         (rs6000_expand_interleave): Likewise.
12978         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12979         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12980         new function.
12981         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12982
12983 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12984
12985         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12986         vector mode and that that mode matches the mode of the data
12987         being permuted.
12988         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12989         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
12990         directly using expand_vec_perm_1 when forcing selectors into
12991         registers.
12992         (expand_vec_perm_var): New function, split out from expand_vec_perm.
12993
12994 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
12995
12996         * optabs-query.h (can_vec_perm_p): Delete.
12997         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12998         * optabs-query.c (can_vec_perm_p): Split into...
12999         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13000         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13001         particular selector is valid.
13002         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13003         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13004         (vect_grouped_load_supported): Likewise.
13005         (vect_shift_permute_load_chain): Likewise.
13006         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13007         (vect_transform_slp_perm_load): Likewise.
13008         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13009         (vectorizable_bswap): Likewise.
13010         (vect_gen_perm_mask_checked): Likewise.
13011         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
13012         implementations of variable permutation vectors into account
13013         when deciding which selector to use.
13014         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13015         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13016         with a false third argument.
13017         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13018         to test whether the constant selector is valid and can_vec_perm_var_p
13019         to test whether a variable selector is valid.
13020
13021 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13022
13023         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13024         * optabs-query.c (can_vec_perm_p): Likewise.
13025         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13026         instead of vec_perm_indices.
13027         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13028         (vect_gen_perm_mask_checked): Likewise,
13029         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13030         (vect_gen_perm_mask_checked): Likewise,
13031
13032 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13033
13034         * optabs-query.h (qimode_for_vec_perm): Declare.
13035         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13036         (qimode_for_vec_perm): ...this new function.
13037         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13038
13039 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13040
13041         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13042         does not have a conditional at the top.
13043
13044 2018-01-02  Richard Biener  <rguenther@suse.de>
13045
13046         * ipa-inline.c (big_speedup_p): Fix expression.
13047
13048 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
13049
13050         PR target/81616
13051         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13052         for generic 4->6.
13053
13054 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
13055
13056         PR target/81616
13057         Generic tuning.
13058         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13059         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13060         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13061         cond_taken_branch_cost 3->4.
13062
13063 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
13064
13065         PR tree-optimization/83581
13066         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13067         TODO_cleanup_cfg if any changes have been made.
13068
13069         PR middle-end/83608
13070         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13071         convert_modes if target mode has the right side, but different mode
13072         class.
13073
13074         PR middle-end/83609
13075         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13076         last argument when extracting from CONCAT.  If either from_real or
13077         from_imag is NULL, use expansion through memory.  If result is not
13078         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13079         the parts directly to inner mode, if even that fails, use expansion
13080         through memory.
13081
13082         PR middle-end/83623
13083         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13084         check for bswap in mode rather than HImode and use that in expand_unop
13085         too.
13086 \f
13087 Copyright (C) 2018 Free Software Foundation, Inc.
13088
13089 Copying and distribution of this file, with or without modification,
13090 are permitted in any medium without royalty provided the copyright
13091 notice and this notice are preserved.