gcc/ChangeLog

   1 2018-05-30  Martin Sebor  <msebor@redhat.com>
   2
   3         PR middle-end/85369
   4         * builtins.c (expand_builtin_stpcpy_1): New function.
   5         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
   6         only if the former succeeds.
   7
   8 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
   9
  10         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
  11         in saphira.
  12
  13 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
  14
  15         * doc/invoke.texi (-flinker-output): Document
  16
  17 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
  18
  19         * passes.c (ipa_write_summaries): Only modify statements if body
  20         is in memory.
  21         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
  22         incrementally linking.
  23         (ipa_passes): Likewise.
  24         * lto-cgraph.c (lto_output_node): When incrementally linking do not
  25         pass down resolution info.
  26         * common.opt (flag_incremental_link): Update info.
  27         * gcc.c (plugin specs): Turn flinker-output=* to
  28         -plugin-opt=-linker-output-known
  29         * toplev.c (compile_file): Also cut compilation when doing incremental
  30         link.
  31         * flag-types. (enum lto_partition_model): Add
  32         LTO_LINKER_OUTPUT_NOLTOREL.
  33         (invoke.texi): Add -flinker-output docs.
  34         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
  35         link same way as WPA; do not stream in dead initializers.
  36
  37         * dwarf2out.c (dwarf2out_die_ref_for_decl,
  38         darf2out_register_external_decl): Support incremental link.
  39
  40 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
  41
  42         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
  43
  44 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
  45
  46         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
  47         it down to simple_object_copy_lto_debug_sections.
  48         (run_gcc): Determine incremental LTO link time and configure
  49         lto1 into non-wpa mode, disable renaming of debug sections.
  50
  51 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
  52
  53         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
  54         descriptions of various incorrectly documented functions.
  55
  56 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
  57
  58         Revert:
  59         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
  60         address check not strict.
  61
  62 2018-05-30  Richard Biener  <rguenther@suse.de>
  63
  64         PR tree-optimization/85964
  65         * tracer.c (better_p): Drop initialized count check, we only
  66         call the function with initialized counts now.
  67         (find_best_successor): Do find a best edge if one
  68         has uninitialized count.
  69         (find_best_predecessor): Likewise.  Do BB frequency check only
  70         if count is initialized.
  71
  72 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
  73
  74         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
  75         (aarch64_ldrstr_offset_compare): New.
  76         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
  77         load/store orderings.
  78         (aarch64_gen_adjusted_ldpstp): Likewise.
  79
  80 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
  81
  82         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
  83         Check for subset of GENERAL_REGS and FP_REGS.
  84         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
  85         r=w alternative.
  86
  87 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
  88
  89         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
  90         and wi::to_poly_offset.  Add the current offset and then check
  91         whether the sum fits, rather than using an unchecked addition of
  92         a checked term.  Check for a shwi rather than a uhwi.
  93         * expr.c (get_bit_range): Use tree_to_poly_uint64.
  94         (store_constructor): Use poly_int_tree_p.
  95         (expand_expr_real_1): Likewise.
  96         * function.c (assign_temp): Likewise.
  97         * fold-const.c (const_binop): Use poly_int_tree_p and
  98         wi::to_poly_offset.
  99         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
 100         division.
 101         * ipa-icf-gimple.c (func_checker::compare_operand): Use
 102         to_poly_offset for MEM offsets.
 103         * ipa-icf.c (sem_variable::equals): Likewise.
 104         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
 105         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
 106         wi::to_poly_offset for BIT_FIELD_REF offsets.
 107         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
 108         wi::to_poly_offset.
 109         * var-tracking.c (emit_note_insn_var_location): Use
 110         tree_to_poly_uint64.
 111
 112 2018-05-29  Jim Wilson  <jimw@sifive.com>
 113
 114         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
 115
 116 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
 117
 118         PR target/85950
 119         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
 120         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
 121         sequence.
 122         (sse4_1_round<mode>2): Use nonimmediate_operand
 123         for operand 1 predicate.
 124
 125 2018-05-29  Martin Sebor  <msebor@redhat.com>
 126             Richard Biener  <rguenther@suse.de>
 127
 128         PR testsuite/85888
 129         * calls.c (get_size_range): Call determine_value_range instead
 130         of get_value_range..
 131         * tree-vrp.h (determine_value_range): Declared new function.
 132         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
 133
 134 2018-05-29  Richard Biener  <rguenther@suse.de>
 135
 136         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
 137         sure to use non-pattern stmts for get_earlier_stmt arguments.
 138         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
 139         called on pattern stmts.
 140         (get_later_stmt): Likewise.
 141
 142 2018-05-29  Martin Liska  <mliska@suse.cz>
 143
 144         PR gcov-profile/85759
 145         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
 146         env variables.
 147
 148 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
 149
 150         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
 151         VEC_UNPACK_*_EXPR.
 152         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
 153         VEC_PACK_*_EXPR.
 154
 155         PR target/85918
 156         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
 157         VEC_PACK_FLOAT_EXPR): New tree codes.
 158         * tree-pretty-print.c (op_code_prio): Handle
 159         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
 160         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
 161         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
 162         * tree-inline.c (estimate_operator_cost): Likewise.
 163         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
 164         * fold-const.c (const_binop): Likewise.
 165         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
 166         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
 167         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
 168         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
 169         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
 170         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
 171         * expr.c (expand_expr_real_2): Likewise.
 172         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
 173         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
 174         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
 175         optabs.
 176         * optabs.c (expand_widen_pattern_expr): For
 177         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
 178         sign from result type rather than operand's type.
 179         (expand_binop_directly): For vec_packu_float_optab and
 180         vec_packs_float_optab allow result type to be different from operand's
 181         type.
 182         * optabs-tree.c (optab_for_tree_code): Handle
 183         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
 184         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
 185         * tree-vect-generic.c (expand_vector_operations_1):  Handle
 186         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
 187         VEC_PACK_FLOAT_EXPR.
 188         * tree-vect-stmts.c (supportable_widening_operation): Handle
 189         FIX_TRUNC_EXPR.
 190         (supportable_narrowing_operation): Handle FLOAT_EXPR.
 191         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
 192         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
 193         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
 194         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
 195         mode attributes.
 196         (vec_pack<floatprefix>_float_<mode>): New expander.
 197         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
 198         attributes.
 199         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
 200         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
 201         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
 202         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
 203         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
 204         Document.
 205         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
 206         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
 207         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
 208         VEC_PACK_FLOAT_EXPR): Document.
 209
 210 2018-05-29  Richard Biener  <rguenther@suse.de>
 211
 212         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
 213         member.
 214         (stmt_vec_info_vec): Make pointer.
 215         (init_stmt_vec_info_vec): Remove.
 216         (free_stmt_vec_info_vec): Likewise.
 217         (set_stmt_vec_info_vec): New function.
 218         (free_stmt_vec_infos): Likewise.
 219         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
 220         (set_vinfo_for_stmt): Likewise.
 221         (get_earlier_stmt): Likewise.
 222         (get_later_stmt): Likewise.
 223         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
 224         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
 225         (vec_info::~vec_info): Free stmt_vec_infos.
 226         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
 227         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
 228         (pass_slp_vectorize::execute): Likewise.
 229         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
 230         (free_stmt_vec_info_vec): Likewise.
 231         (set_stmt_vec_info_vec): New function.
 232         (free_stmt_vec_infos): Likewise.
 233         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
 234         the global stmt_vec_info_vec.
 235         * tree-parloops.c (gather_scalar_reductions): Use
 236         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
 237         vector.
 238
 239 2018-05-29  Richard Biener  <rguenther@suse.de>
 240
 241         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
 242
 243 2018-05-29  Martin Liska  <mliska@suse.cz>
 244             David Malcolm  <dmalcolm@redhat.com>
 245
 246         * vec.c (test_reverse): New.
 247         (vec_c_tests): Add new test.
 248         * vec.h (vl_ptr>::reverse): New function.
 249
 250 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
 251
 252         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
 253
 254         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
 255         and later.
 256
 257 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
 258
 259         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
 260
 261 2018-05-28  Richard Biener  <rguenther@suse.de>
 262
 263         PR tree-optimization/85933
 264         * tree-vect-data-refs.c (vect_record_base_alignments): Only
 265         look at stmts marked as vectorizable.
 266
 267 2018-05-28  Richard Biener  <rguenther@suse.de>
 268
 269         PR tree-optimization/85934
 270         * tree-vect-generic.c (expand_vector_operations_1): Hoist
 271         vector boolean check before scalar optimization.
 272
 273 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
 274
 275         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
 276         for armv5te.
 277
 278 2018-05-28  Mark Wielaard  <mark@klomp.org>
 279
 280         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
 281         if it is an expression containing a minus sign.
 282
 283 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
 284
 285         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
 286
 287 2018-05-27  Paul Koning  <ni1d@arrl.net>
 288
 289         * config/pdp11/pdp11.md (truncsihi2): Remove.
 290
 291 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
 292             Chung-Ju Wu  <jasonwucj@gmail.com>
 293
 294         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
 295         implementation.
 296         (unaligned_store_dw): Ditto.
 297         * config/nds32/nds32-memory-manipulation.c
 298         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
 299         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
 300         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
 301         (emit_setmem_word_loop): Rename to ...
 302         (emit_setmem_doubleword_loop): ... this.
 303         (nds32_gen_dup_4_byte_to_word_value): New function.
 304         (nds32_gen_dup_8_byte_to_double_word_value): New function.
 305         (nds32_expand_setmem_loop): Refine implementation.
 306         (nds32_expand_setmem_loop_v3m): Ditto.
 307         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
 308         pattern.
 309
 310 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
 311
 312         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
 313
 314 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
 315
 316         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
 317         (nds32_init_machine_status): Initialize machine->attr_naked_p and
 318         machine->attr_no_prologue_p.
 319         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
 320         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
 321         (nds32_expand_epilogue): Consider attr_naked_p.
 322         (nds32_expand_epilogue_v3pop): Likewise.
 323         (nds32_can_use_return_insn): Likewise.
 324         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
 325         attr_no_prologue_p fields.
 326         * config/nds32/nds32.opt (mret-in-naked-func): New option.
 327
 328 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
 329
 330         PR target/85918
 331         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
 332         attributes.
 333         * config/i386/sse.md
 334         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
 335         Rename to ...
 336         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
 337         ... this.
 338         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
 339         Rename to ...
 340         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
 341         ... this.
 342         (*<floatsuffix>floatv2div2sf2): Rename to ...
 343         (*float<floatunssuffix>v2div2sf2): ... this.
 344         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
 345         (float<floatunssuffix>v2div2sf2_mask): ... this.
 346         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
 347         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
 348         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
 349         to ...
 350         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
 351         ... this.
 352         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
 353         Rename to ...
 354         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
 355         ... this.
 356         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
 357         Rename to ...
 358         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
 359         ... this.
 360         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
 361         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
 362         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
 363         gen_ufix_truncv8dfv8si2.
 364         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
 365         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
 366         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
 367         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
 368         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
 369         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
 370         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
 371         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
 372
 373 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
 374
 375         PR target/85900
 376         PR target/85345
 377         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
 378
 379 2018-05-25  Jim Wilson  <jimw@sifive.com>
 380
 381         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
 382         * config/riscv/riscv.c (struct machine_function): Add
 383         interrupt_handler_p and attribute_checked_p fields.
 384         (riscv_attribute_table): Add interrupt.
 385         (riscv_interrupt_type_p): New.
 386         (riscv_save_reg_p): Save extra regs for interrupt handler.
 387         (riscv_use_save_libcall): Return false  for interrupt handler.
 388         (riscv_first_stack_step): Add forward declaration.
 389         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
 390         for interrupt handler with large frame.  Use it for saved reg list.
 391         (riscv_expand_prologue): Move flag_stack_usage_info support to
 392         eliminate duplication.
 393         (riscv_expand_epilogue): Generate mret for interrupt handler.
 394         (riscv_epilogue_uses): New.
 395         (riscv_can_use_return_insn): Return false for interrupt handler.
 396         (riscv_function_ok_for_sibcall): Likewise.
 397         (riscv_set_current_function): Add interrupt handler support.
 398         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
 399         * config/riscv/riscv.md (UNSPECV_MRET): New.
 400         (GP_REGNUM): New.
 401         (riscv_frflags, riscv_fsflags): Use tab after opcode.
 402         (riscv_mret): New.
 403         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
 404
 405 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
 406
 407         PR tree-optimization/85712
 408         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
 409         this candidate has already been replaced in-situ by a copy.
 410
 411 2018-05-25  Jason Merrill  <jason@redhat.com>
 412
 413         PR c++/80485 - inline function non-zero address.
 414         * symtab.c (nonzero_address): Check DECL_COMDAT.
 415
 416 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
 417
 418         PR target/83628
 419         * config/alpha/alpha.md (ashlsi3): New insn pattern.
 420         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
 421         extension of SImode operation.  Use const123_operand predicate.
 422         (*saddsi_1): Remove.
 423         (*saddl_se_1): Ditto.
 424         (*ssubsi_1): Ditto.
 425         (*ssubl_se_1): Ditto.
 426         * config/alpha/predicates.md (const123_operand): New predicate.
 427         * config/alpha/constraints.md (P): Use IN_RANGE.
 428
 429 2018-05-25  Richard Biener  <rguenther@suse.de>
 430
 431         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
 432         defaulted to true.
 433         (ref_maybe_used_by_stmt_p): Likewise.
 434         (stmt_may_clobber_ref_p): Likewise.
 435         (stmt_may_clobber_ref_p_1): Likewise.
 436         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
 437         and pass it along.
 438         (ref_maybe_used_by_stmt_p): Likewise.
 439         (stmt_may_clobber_ref_p): Likewise.
 440         (stmt_may_clobber_ref_p_1): Likewise.
 441         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
 442         the alias oracle to disambiguate DRs with stmts DR analysis
 443         couldn't handle.
 444         (vect_analyze_data_refs): Do not give up on not analyzable
 445         DRs for BB vectorization.  Remove code truncating the dataref
 446         vector.
 447
 448 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
 449
 450         PR target/85832
 451         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
 452         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
 453         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
 454
 455 2018-05-25  Richard Biener  <rguenther@suse.de>
 456
 457         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
 458         function, combining stmt data ref gathering and fatal analysis
 459         parts.
 460         (vect_analyze_data_refs): Remove now redudnant code and simplify.
 461         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
 462         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
 463         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
 464         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
 465
 466 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
 467
 468         PR tree-optimization/85720
 469         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
 470         SCC if all partitions are builtins.
 471         (version_loop_by_alias_check): New parameter.  Generate cancelable
 472         runtime alias check if all partitions are builtins.
 473         (distribute_loop): Update call to above function.
 474
 475 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
 476
 477         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
 478         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
 479         (parm_default_def_partition_arg): Ditto.
 480         (set_parm_default_def_partition): Ditto.
 481         (get_parm_default_def_partitions): Ditto and make it static.
 482         (get_undefined_value_partitions): Ditto and make it static.
 483         (remove_ssa_form): Refactor call to init_var_map here.
 484         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
 485         computation for loop region.
 486         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
 487         (register_default_def): Delete.
 488         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
 489         (parm_default_def_partition_arg): Ditto.
 490         (set_parm_default_def_partition): Ditto.
 491         (get_parm_default_def_partitions): Ditto and make it static.
 492         (get_undefined_value_partitions): Ditto and make it static.
 493         (coalesce_with_default, coalesce_with_default): Update comment.
 494         (create_coalesce_list_for_region): New func factored out from
 495         create_outofssa_var_map.
 496         (populate_coalesce_list_for_outofssa): New func factored out from
 497         create_outofssa_var_map and coalesce_ssa_name.
 498         (create_outofssa_var_map): Delete.
 499         (coalesce_ssa_name): Refactor to support live range computation.
 500         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
 501         (get_parm_default_def_partitions): Delete.
 502         (get_undefined_value_partitions): Ditto.
 503         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
 504         computation for loop region.
 505         (new_tree_live_info, loe_visit_block): Ditto.
 506         (live_worklist, set_var_live_on_entry): Ditto.
 507         (calculate_live_on_exit, verify_live_on_entry): Ditto.
 508         * tree-ssa-live.h (struct _var_map): New fields.
 509         (init_var_map): Change decl.
 510         (region_contains_p): New.
 511
 512 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
 513
 514         * tree-ssa-live.h (live_merge_and_clear): Delete.
 515
 516 2018-05-25  Richard Biener  <rguenther@suse.de>
 517
 518         PR c++/85912
 519         * tree-dump.c (dequeue_and_dump): Remove access to removed
 520         operand 2 of a SWITCH_EXPR.
 521
 522 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 523
 524         * doc/sourcebuild.texi (vect_double_cond_arith): Include
 525         multiplication and division.
 526         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
 527         (cond_udiv@var{m}, cond_umod@var{m}): Document.
 528         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
 529         (cond_udiv_optab, cond_umod_optab): New optabs.
 530         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
 531         (IFN_COND_RDIV): New internal functions.
 532         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
 533         TRUNC_MOD_EXPR and RDIV_EXPR.
 534         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
 535         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
 536         New unspecs.
 537         (SVE_INT_BINARY): Include mult.
 538         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
 539         (optab, sve_int_op): Handle mult.
 540         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
 541         UNSPEC_COND_DIV.
 542         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
 543         for SVE_INT_BINARY_SD.
 544
 545 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 546
 547         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
 548         (optab, sve_int_op): Handle div and udiv.
 549         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
 550         for SVE_INT_BINARY_SD.
 551         (*<optab><mode>3): New insn for the same.
 552
 553 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 554
 555         * tree-vect-patterns.c: Include predict.h.
 556         (vect_recog_divmod_pattern): Restrict check for division support
 557         to when optimizing for size.
 558
 559 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 560
 561         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
 562         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
 563         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
 564         (gimple_match_op::set_op): Likewise.
 565         (gimple_resimplify4): Declare.
 566         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
 567         (expr::gen_transform): Likewise.
 568         (decision_tree::gen): Generate a simplification routine for 4 operands.
 569         * gimple-match-head.c (gimple_simplify): Add an overload for
 570         4 operands.  In the top-level function, handle up to 4 call
 571         arguments and call gimple_resimplify4.
 572         (gimple_resimplify4): New function.
 573         (build_call_internal): Pass a fourth operand.
 574         (maybe_push_to_seq): Likewise.
 575         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
 576         Fold VEC_COND_EXPRs of an operation and a default value into
 577         an IFN_COND_* function if possible.
 578         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
 579         New unspecs.
 580         (SVE_COND_FP_BINARY): Include them.
 581         (optab, sve_fp_op): Handle them.
 582         (SVE_INT_BINARY_REV): New code iterator.
 583         (SVE_COND_FP_BINARY_REV): New int iterator.
 584         (commutative): New int attribute.
 585         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
 586         Declare.
 587         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
 588         function.
 589         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
 590         (*cond_<optab><mode>): New patterns for reversed operands.
 591
 592 2018-05-25  Richard Biener  <rguenther@suse.de>
 593
 594         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
 595         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
 596         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
 597         (STMT_VINFO_GROUPED_ACCESS): Adjust.
 598         * tree-vect-data-refs.c (everywhere): Adjust users.
 599         * tree-vect-loop.c (everywhere): Likewise.
 600         * tree-vect-slp.c (everywhere): Likewise.
 601         * tree-vect-stmts.c (everywhere): Likewise.
 602         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
 603
 604 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
 605
 606         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
 607         Rename to...
 608         (gcc_cv_as_section_exclude): ... this.
 609         Try Solaris as #exclude syntax.
 610         * configure: Regenerate.
 611         * config.in: Regenerate.
 612         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
 613         SECTION_EXCLUDE.
 614         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
 615         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
 616
 617         * varasm.c (default_elf_asm_named_section): Don't check if
 618         HAVE_GAS_SECTION_EXCLUDE is defined.
 619
 620 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 621
 622         * doc/md.texi: Update the documentation of the cond_* optabs
 623         to mention the new final operand.  Fix GET_MODE_NUNITS call.
 624         Describe the scalar case too.
 625         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
 626         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
 627         instead of 2.
 628         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
 629         (get_conditional_internal_fn): Update comment.
 630         * tree-vect-loop.c (vectorizable_reduction): Pass the original
 631         accumulator value as a final argument to conditional functions.
 632         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
 633         a define_expand and add an "else" operand.  Assert for now that
 634         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
 635         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
 636         (*cond_<optab><mode>): New patterns.
 637         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
 638         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
 639         (UNSPEC_COND_EOR): Delete.
 640         (optab): Remove associated mappings.
 641         (SVE_INT_BINARY): New code iterator.
 642         (sve_int_op): Remove int attribute and add "minus" to the code
 643         attribute.
 644         (SVE_COND_INT_OP): Delete.
 645         (SVE_COND_FP_OP): Rename to...
 646         (SVE_COND_FP_BINARY): ...this.
 647
 648 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 649
 650         * optabs.c (can_reuse_operands_p): New function.
 651         (maybe_legitimize_operands): Try to reuse the results for
 652         earlier operands.
 653
 654 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
 655
 656         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
 657         Add {q} suffix to insn mnemonic.
 658
 659 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
 660
 661         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
 662         (msp430_warn_func_return): New.
 663
 664 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
 665
 666         * fold-const.c (tree_nonzero_bits): New function.
 667         * fold-const.h (tree_nonzero_bits): Likewise.
 668         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
 669         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
 670
 671 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
 672
 673         PR target/85900
 674         PR target/85345
 675         * varasm.c (assemble_alias): Check ifunc_resolver only on
 676         FUNCTION_DECL.
 677
 678 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
 679
 680         PR target/85903
 681         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
 682         when memory input operand is handled.
 683
 684 2018-05-24  Luis Machado  <luis.machado@linaro.org>
 685
 686         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
 687         global.
 688         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
 689
 690 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
 691
 692         * match.pd: Delay FMA folds until after vectorization.
 693
 694 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
 695
 696         PR target/83009
 697         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
 698         address check not strict.
 699
 700 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
 701
 702         * gimple-match.h (gimple_match_op): New class.
 703         (mprts_hook): Replace parameters with a gimple_match_op *.
 704         (maybe_build_generic_op): Likewise.
 705         (gimple_simplified_result_is_gimple_val): Replace parameters with
 706         a const gimple_match_op *.
 707         (gimple_simplify): Replace code_helper * and tree * parameters with
 708         a gimple_match_op * parameter.
 709         (gimple_resimplify1): Replace code_helper *, tree and tree *
 710         parameters with a gimple_match_op * parameter.
 711         (gimple_resimplify2): Likewise.
 712         (gimple_resimplify3): Likewise.
 713         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
 714         parameters with a gimple_match_op * parameter.
 715         * gimple-match-head.c (gimple_simplify): Change prototypes of
 716         auto-generated functions to take a gimple_match_op * instead of
 717         separate code_helper * and tree * parameters.  Make the same
 718         change in the top-level overload and update calls to the
 719         gimple_resimplify routines.  Update calls to the auto-generated
 720         functions and to maybe_push_res_to_seq in the publicly-facing
 721         operation-specific gimple_simplify overloads.
 722         (gimple_match_op::MAX_NUM_OPS): Define.
 723         (gimple_resimplify1): Replace rcode and ops with a single res_op
 724         parameter.  Update call to gimple_simplify.
 725         (gimple_resimplify2): Likewise.
 726         (gimple_resimplify3): Likewise.
 727         (mprts_hook): Replace parameters with a gimple_match_op *.
 728         (maybe_build_generic_op): Likewise.
 729         (build_call_internal): Replace type, nargs and ops with
 730         a gimple_match_op *.
 731         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
 732         with a single gimple_match_op *.  Update calls to mprts_hook,
 733         build_call_internal and gimple_simplified_result_is_gimple_val.
 734         Factor out code that is common to the tree_code and combined_fn cases.
 735         * genmatch.c (expr::gen_transform): Replace tem_code and
 736         tem_ops with a gimple_match_op called tem_op.  Update calls
 737         to the gimple_resimplify functions and maybe_push_res_to_seq.
 738         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
 739         res_ops.  Update call to the gimple_resimplify functions.
 740         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
 741         (decision_tree::gen): Make the functions take a gimple_match_op *
 742         called res_op instead of separate res_code and res_ops parameters.
 743         Update call accordingly.
 744         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
 745         and ops with a single res_op parameter.  Update calls to
 746         maybe_build_generic_op and maybe_push_res_to_seq.
 747         (fold_stmt_1): Update calls to gimple_simplify and
 748         replace_stmt_with_simplification.
 749         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
 750         and gimple_simplified_result_is_gimple_val.
 751         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
 752         gimple_simplify.
 753         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
 754         with a gimple_match_op *.
 755         (vn_nary_build_or_lookup): Likewise.  Update call to
 756         vn_nary_build_or_lookup_1.
 757         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
 758         gimple_match_op *.  Update calls to the gimple_resimplify routines
 759         and to gimple_simplified_result_is_gimple_val.
 760         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
 761         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
 762         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
 763         (visit_nary_op): Likewise.
 764         (visit_reference_op_load): Likewise.
 765
 766 2018-05-23  Luis Machado  <luis.machado@linaro.org>
 767
 768         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
 769         modifier for printing the step amount.
 770
 771 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
 772
 773         PR target/78849
 774         * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
 775         types.
 776
 777 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
 778
 779         * doc/sourcebuild.texi (Endianness): New subsubsection.
 780
 781 2018-05-23  Luis Machado  <luis.machado@linaro.org>
 782
 783         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
 784         <prefetch_dynamic_strides>: New const bool field.
 785         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
 786         prefetch_dynamic_strides.
 787         (exynosm1_prefetch_tune): Likewise.
 788         (thunderxt88_prefetch_tune): Likewise.
 789         (thunderx_prefetch_tune): Likewise.
 790         (thunderx2t99_prefetch_tune): Likewise.
 791         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
 792         false.
 793         (aarch64_override_options_internal): Update to set
 794         PARAM_PREFETCH_DYNAMIC_STRIDES.
 795         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
 796         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
 797         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
 798         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
 799         prefetch-dynamic-strides setting.
 800
 801 2018-05-23  Luis Machado  <luis.machado@linaro.org>
 802
 803         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
 804         <minimum_stride>: New const int field.
 805         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
 806         minimum_stride field defaulting to -1.
 807         (exynosm1_prefetch_tune): Likewise.
 808         (thunderxt88_prefetch_tune): Likewise.
 809         (thunderx_prefetch_tune): Likewise.
 810         (thunderx2t99_prefetch_tune): Likewise.
 811         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
 812         <default_opt_level>: Set to 3.
 813         (aarch64_override_options_internal): Update to set
 814         PARAM_PREFETCH_MINIMUM_STRIDE.
 815         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
 816         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
 817         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
 818         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
 819         stride is constant and is below the minimum stride threshold.
 820
 821 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 822
 823         * config/arm/arm-cpus.in (mode26): Delete.
 824         (armv4): Delete mode26 reference.
 825         * config/arm/arm.c (arm_configure_build_target): Delete use of
 826         isa_bit_mode26.
 827
 828 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
 829
 830         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
 831         New insn pattern.
 832         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
 833         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
 834         for non-SSE modes.
 835         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
 836         (floatunsdidf2): Ditto.
 837
 838 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
 839
 840         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
 841         (fixuns_trunc<mode>si2_avx512f): Ditto.
 842         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
 843         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
 844         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
 845
 846 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
 847
 848         PR rtl-optimization/79985
 849         * df-scan.c (df_insn_refs_collect): Remove special case for
 850         global registers and asm statements.
 851
 852 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
 853
 854         * extend.texi (Global Register Variables): Rewrite the bullet list.
 855         Note that the register is available for allocation. Note that access
 856         via inline asm must use constraints. Add note about async-signal
 857         handlers. Remove paragraph about automagic register selection.
 858
 859 2018-05-23  Richard Biener  <rguenther@suse.de>
 860
 861         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
 862         of fixed offset from memset VN.
 863
 864 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
 865
 866         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
 867         first_interp field.
 868         (alloc_cand_and_find_basis): Initialize first_interp field.
 869         (slsr_process_mul): Modify first_interp field.
 870         (slsr_process_add): Likewise.
 871         (slsr_process_cast): Modify first_interp field for each new
 872         interpretation.
 873         (slsr_process_copy): Likewise.
 874         (dump_candidate): Dump first_interp field.
 875         (replace_mult_candidate): Process all interpretations, not just
 876         subsequent ones.
 877         (replace_rhs_if_not_dup): Likewise.
 878         (replace_one_candidate): Likewise.
 879
 880 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
 881
 882         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
 883         Add new boolean.
 884         (aarch64_needs_frame_chain): New function.
 885         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
 886
 887 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
 888
 889         PR target/84882
 890         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
 891         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
 892         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
 893         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
 894         as true for strict-align.
 895         (aarch64_can_inline_p): Perform checks even when callee has no
 896         attributes to check for strict alignment.
 897         * doc/extend.texi (AArch64 Function Attributes): Document
 898         no-strict-align.
 899         * doc/invoke.texi: (AArch64 Options): Likewise.
 900
 901 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
 902
 903         PR tree-optimization/85853
 904         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
 905         the handling of the root of the node to...
 906         (vect_slp_analyze_node_operations_1): ...this new function,
 907         and run the whole thing with the child nodes' def types
 908         set according to their SLP node's def type.
 909
 910 2018-05-23  Richard Biener  <rguenther@suse.de>
 911
 912         PR middle-end/85874
 913         * tree-data-ref.c (create_runtime_alias_checks): Defer
 914         and ignore overflow warnings.
 915
 916 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
 917
 918         PR tree-optimization/85822
 919         * tree-vrp.c (is_masked_range_test): Fix handling of negative
 920         constants.
 921
 922 2018-05-23  Richard Biener  <rguenther@suse.de>
 923
 924         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
 925         memset constants via native_interpret_expr.
 926
 927 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
 928
 929         PR target/85345
 930         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
 931         attribute.
 932         (cgraph_node::create_alias): Likewise.
 933         (cgraph_node::get_availability): Check ifunc_resolver instead
 934         of looking up ifunc attribute.
 935         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
 936         * varasm.c (do_assemble_alias): Likewise.
 937         (assemble_alias): Likewise.
 938         (default_binds_local_p_3): Likewise.
 939         * cgraph.h (cgraph_node): Add ifunc_resolver.
 940         (cgraph_node::only_called_directly_or_aliased_p): Return false
 941         for IFUNC resolver.
 942         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
 943         attribute.
 944         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
 945         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
 946         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
 947         instead of looking up ifunc attribute.
 948
 949 2018-05-22  Luis Machado  <luis.machado@linaro.org>
 950
 951         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
 952
 953 2018-05-22  Martin Sebor  <msebor@redhat.com>
 954
 955         PR middle-end/85359
 956         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
 957         only when expasion succeeds.
 958         (expand_builtin_strcmp): Same.
 959         (expand_builtin_strncmp): Same.
 960
 961 2018-05-22  Martin Sebor  <msebor@redhat.com>
 962
 963         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
 964
 965 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
 966             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 967
 968         * config/aarch64/aarch64-ldpstp.md: Replace uses of
 969         aarch64_mem_pair_operand with memory_operand and delete operand swapping
 970         code.
 971         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
 972         Add check for legitimate_address.
 973         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
 974         (aarch64_swap_ldrstr_operands): New.
 975         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
 976         Define prototype.
 977
 978 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
 979             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 980
 981         * config/aarch64/aarch64.md: New patterns to generate stp
 982         and ldp.
 983         (store_pair_sw, store_pair_dw): New patterns to generate stp for
 984         single words and double words.
 985         (load_pair_sw, load_pair_dw): Likewise.
 986         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
 987         Delete.
 988         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
 989         Delete.
 990         * config/aarch64/aarch64-ldpstp.md: Modify peephole
 991         for different mode ldpstp and add peephole for merged zero stores.
 992         Likewise for loads.
 993         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
 994         Add size check.
 995         (aarch64_gen_store_pair): Rename calls to match new patterns.
 996         (aarch64_gen_load_pair): Rename calls to match new patterns.
 997         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
 998         (load_pair<DREG:mode><DREG2:mode>): ... This.
 999         (store_pair<mode>): Rename to...
1000         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1001         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1002         New mode iterators.
1003         (V_INT_EQUIV): Handle SImode.
1004         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1005         New predicate.
1006
1007 2018-05-22  Martin Sebor  <msebor@redhat.com>
1008
1009         PR c/85623
1010         * calls.c (maybe_warn_nonstring_arg): Use string length to set
1011         or ajust the presumed bound on an operation to avoid unnecessary
1012         warnings.
1013
1014 2018-05-22  Martin Sebor  <msebor@redhat.com>
1015
1016         PR tree-optimization/85826
1017         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1018         assuming that a DECL necesarily has a constant size.
1019
1020 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
1021
1022         PR middle-end/85862
1023         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1024
1025 2018-05-22  Richard Biener  <rguenther@suse.de>
1026
1027         PR tree-optimization/85834
1028         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1029         non-constant and non-zero memset arguments.
1030
1031 2018-05-22  Martin Liska  <mliska@suse.cz>
1032
1033         PR ipa/85607
1034         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1035
1036 2018-05-22  Richard Biener  <rguenther@suse.de>
1037
1038         PR tree-optimization/85863
1039         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1040         comparisons when vectype is specified.
1041         (vectorizable_condition): Do not specify vectype for
1042         vect_is_simple_cond when SLP vectorizing.
1043
1044 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
1045
1046         PR target/85657
1047         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1048         define __ibm128 as long double.
1049         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1050         as a distinct type when IEEE 128-bit support is enabled.
1051         (init_float128_ieee): Fix up conversions between IFmode and IEEE
1052         128-bit types to use the correct functions.
1053         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1054         convert between 128-bit floating point types that have different
1055         modes but the same representation, instead of using gen_lowpart to
1056         makean alias.
1057         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1058         KFmode.
1059         (IFKF_reg): New attributes to give the register constraints for
1060         IFmode and KFmode.
1061         (extend<mode>tf2_internal): New insns to mark an explicit
1062         conversion between 128-bit floating point types that have a
1063         different mode but share the same representation.
1064
1065 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
1066
1067         PR tree-optimization/85814
1068         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1069         a null return from get_strinfo when unsharing the next
1070         strinfo in the chain.
1071
1072 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
1073
1074         PR gcc/84923
1075         * varasm.c (weak_finish): Clean up weak_decls.
1076
1077 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1078
1079         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1080         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1081         UNSPEC_UADALP values.
1082         * config/aarch64/iterators.md (ABAL): New int iterator.
1083         (ABDL2): Likewise.
1084         (ADALP): Likewise.
1085         (sur): Add mappings for the above.
1086         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1087         New define_insn.
1088         (aarch64_<sur>abal<mode>_4): Likewise.
1089         (aarch64_<sur>adalp<mode>_3): Likewise.
1090         (<sur>sadv16qi): New define_expand.
1091
1092 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
1093
1094         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1095         (*movdf_internal): Ditto.
1096         (*rcpsf2_sse): Ditto.
1097         (*rsqrtsf2_sse): Ditto.
1098         (*sqrt<mode>2_sse): Ditto.
1099
1100 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
1101
1102         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1103         eor3q<mode>4.
1104         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1105         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1106         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1107         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1108         vbcaxq_s64): New.
1109         * config/aarch64/arm_neon.h: Likewise.
1110         * config/aarch64/iterators.md (VQ_I): New.
1111
1112 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
1113
1114         * config.gcc: Add arc/t-multilib-linux to tmake_file for
1115         arc*-*-linux*.
1116         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1117         MULTILIB_DIRNAMES
1118
1119 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
1120
1121         * config/nds32/constraints.md (S): New constraint.
1122         * config/nds32/nds32.md (call_internal): Use constraint S.
1123         (call_value_internal): Likewise.
1124         (sibcall_internal): Likewise.
1125         (sibcall_value_internal): Likewise.
1126
1127 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
1128             Chung-Ju Wu  <jasonwucj@gmail.com>
1129
1130         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1131         into consideration.
1132
1133 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
1134             Chung-Ju Wu  <jasonwucj@gmail.com>
1135
1136         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1137         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1138         (nds32_rtx_costs_impl): Simplify.
1139         (nds32_address_cost_impl): Simplify.
1140         (nds32_init_rtx_costs): New function.
1141         (nds32_rtx_costs_speed_prefer): Likewise.
1142         (nds32_rtx_costs_size_prefer): Likewise.
1143         (nds32_address_cost_speed_prefer): Likewise.
1144         (nds32_address_cost_speed_fwprop): Likewise.
1145         (nds32_address_cost_size_prefer): Likewise.
1146         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1147         * config/nds32/nds32.c (nds32_option_override): Use
1148         nds32_init_rtx_costs function.
1149
1150 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
1151
1152         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1153         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1154         (TARGET_PIPELINE_N8): Likewise.
1155         (TARGET_PIPELINE_N10): Likewise.
1156         (TARGET_PIPELINE_N13): Likewise.
1157         (TARGET_PIPELINE_GRAYWOLF): Likewise.
1158
1159 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
1160
1161         * config/nds32/nds32-fpu.md: Update copyright year.
1162
1163 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1164
1165         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1166
1167 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1168
1169         * config/nds32/nds32.c
1170         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1171         * config/nds32/nds32.opt (minline-asm-r15): New option.
1172
1173 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1174
1175         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1176         MASK_HW_ABS.
1177         * config/nds32/nds32.md (abssi2): New pattern.
1178
1179 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
1180
1181         * config/i386/i386.md (rex64namesuffix): New mode attribute.
1182         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1183         Merge insn pattern from sse_cvtsi2ss<round_name> and
1184         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1185         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1186         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1187         using SWI48 mode iterator.
1188         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1189         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1190         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1191         pattern from sse_cvttss2si<round_saeonly_name>
1192         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1193         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1194         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1195         using SWI48 mode iterator.
1196         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1197         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1198         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1199         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1200         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1201         using SWI48 mode iterator.
1202         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1203         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1204         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1205         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1206         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1207         SWI48 mode iterator.
1208         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1209         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1210         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1211         pattern from sse_cvttsd2si<round_saeonly_name>
1212         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1213
1214 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1215
1216         * config/nds32/nds32-md-auxiliary.c
1217         (nds32_valid_smw_lwm_base_p): Refine.
1218         (nds32_output_smw_single_word): Refine.
1219         (nds32_output_smw_double_word): New.
1220         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1221
1222 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1223
1224         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1225         (nds32_output_stack_pop): Refine.
1226         (nds32_expand_unaligned_load): Refine.
1227         (nds32_expand_unaligned_store): Refine.
1228
1229 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
1230             Chung-Ju Wu  <jasonwucj@gmail.com>
1231
1232         * config/nds32/constants.md: Add TP_REGNUM constant.
1233         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1234         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1235         UNSPEC_ADD32.
1236         * config/nds32/nds32-doubleword.md: Consider flag_pic.
1237         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1238         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1239         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1240         and PIC code generation.
1241         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1242         code generation.
1243         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1244         optimization.
1245         * config/nds32/nds32.md: Support TLS and PIC.
1246         * config/nds32/nds32.c: Support TLS and PIC.
1247         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1248         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1249         predicate.
1250
1251 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1252
1253         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1254         mode with E_ prefix.
1255
1256 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
1257             Chung-Ju Wu  <jasonwucj@gmail.com>
1258
1259         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1260         * config/nds32/nds32-md-auxiliary.c
1261         (symbolic_reference_mentioned_p): New.
1262         (nds32_legitimize_ict_address): New.
1263         (nds32_expand_ict_move): New.
1264         (nds32_indirect_call_referenced_p): New.
1265         (nds32_symbol_binds_local_p): Delete.
1266         (nds32_long_call_p): Modify.
1267         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1268         * config/nds32/nds32-protos.h
1269         (symbolic_reference_mentioned_p): Declare.
1270         (nds32_legitimize_ict_address): Declare.
1271         (nds32_expand_ict_move): Declare.
1272         (nds32_indirect_call_referenced_p): Declare.
1273         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1274         (nds32_relax_group): Use nds32_ict_const_p as condition.
1275         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1276         (nds32_asm_file_start): Output ict_model directive in asm code.
1277         (nds32_legitimate_address_p): Consider indirect call.
1278         (nds32_print_operand): Consider indirect call.
1279         (nds32_print_operand_address): Consider indirect call.
1280         (nds32_insert_attributes): Handle "indirect_call" attribute.
1281         (TARGET_LEGITIMATE_ADDRESS_P): Define.
1282         (TARGET_LEGITIMATE_CONSTANT_P): Define.
1283         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1284         (TARGET_DELEGITIMIZE_ADDRESS): Define.
1285         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1286         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1287         (TARGET_ICT_MODEL_SMALL): Define.
1288         (TARGET_ICT_MODEL_LARGE): Define.
1289         * config/nds32/nds32.md (movsi): Consider ict model.
1290         (call, call_value): Consider ict model.
1291         (sibcall, sibcall_value): Consider ict model.
1292         * config/nds32/nds32.opt (mict-model): New option.
1293         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1294         model.
1295
1296 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
1297             Monk Chiang  <sh.chiang04@gmail.com>
1298             Jim Wilson <jimw@sifive.com>
1299
1300         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1301         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
1302         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1303         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1304         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
1305         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1306         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1307         compute save_libcall_adjustment properly.
1308         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1309         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1310         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1311         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1312         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1313         (ABI_SPEC): Handle mabi=ilp32e.
1314         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1315         (RVE): Add RVE mask.
1316         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1317         <-march>: Add rv32e as an example.
1318
1319 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
1320
1321         PR c++/82899
1322         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1323         (intra_create_variable_infos): Handle C++ constructors.
1324
1325 2018-05-18  Martin Liska  <mliska@suse.cz>
1326
1327         * passes.def: Remove a redundant pass.
1328
1329 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
1330
1331         PR bootstrap/85838
1332         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1333
1334 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1335
1336         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1337         (ARMv4): Update.
1338         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1339         (ARMv6m): Update.
1340         (armv2, armv2a, armv3, armv3m): Delete architectures.
1341         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1342         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1343         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1344         Delete cpus.
1345         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1346         (*mulsidi3adddi): Likewise.
1347         (mulsidi3): Likewise.
1348         (*mulsidi3_nov6): Likewise.
1349         (umulsidi3): Likewise.
1350         (umulsidi3_nov6): Likewise.
1351         (umaddsidi4): Likewise.
1352         (*umulsidi3adddi): Likewise.
1353         (smulsi3_highpart): Likewise.
1354         (*smulsi3_highpart_nov6): Likewise.
1355         (umulsi3_highpart): Likewise.
1356         (*umulsi3_highpart_nov6): Likewise.
1357         * config/arm/arm.h (arm_arch3m): Delete.
1358         * config/arm/arm.c (arm_arch3m): Delete.
1359         (arm_option_override_internal): Update armv3-related comment.
1360         (arm_configure_build_target): Delete use of isa_bit_mode32.
1361         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1362         (arm_rtx_costs_internal): Delete check of arm_arch3m.
1363         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1364         (mulsa3): Likewise.
1365         (mulusa3): Likewise.
1366         * config/arm/arm-protos.h (arm_arch3m): Delete.
1367         * config/arm/arm-tables.opt: Regenerate.
1368         * config/arm/arm-tune.md: Likewise.
1369         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1370         deleted architectures.
1371
1372 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1373
1374         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1375         (armv5t, armv5te): New features.
1376         (ARMv5, ARMv5e): Delete fgroups.
1377         (ARMv5t, ARMv5te): Adjust for above changes.
1378         (ARMv6m): Likewise.
1379         (armv5, armv5e): Delete arches.
1380         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1381         arm_arch5.
1382         (*call_reg_arm): Likewise.
1383         (*call_value_reg_armv5): Likewise.
1384         (*call_value_reg_arm): Likewise.
1385         (*call_symbol): Likewise.
1386         (*call_value_symbol): Likewise.
1387         (*sibcall_insn): Likewise.
1388         (*sibcall_value_insn): Likewise.
1389         (clzsi2): Likewise.
1390         (prefetch): Likewise.
1391         (define_split and define_peephole2 dependent on arm_arch5):
1392         Likewise.
1393         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1394         arm_arch5e.
1395         (TARGET_ARM_QBIT): Likewise.
1396         (TARGET_DSP_MULTIPLY): Likewise.
1397         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1398         (arm_arch5, arm_arch5e): Delete.
1399         (arm_arch5t, arm_arch5te): Declare.
1400         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1401         (arm_arch5t): Declare.
1402         (arm_option_reconfigure_globals): Update for the above.
1403         (arm_options_perform_arch_sanity_checks): Update comment, replace
1404         use of arm_arch5 with arm_arch5t.
1405         (use_return_insn): Likewise.
1406         (arm_emit_call_insn): Likewise.
1407         (output_return_instruction): Likewise.
1408         (arm_final_prescan_insn): Likewise.
1409         (arm_coproc_builtin_available): Likewise.
1410         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1411         arm_arch5e with arm_arch5t and arm_arch5te.
1412         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1413         (arm_arch5t, arm_arch5te): Declare.
1414         * config/arm/arm-tables.opt: Regenerate.
1415         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1416         * config/arm/t-multilib: Likewise.
1417         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1418         instead of arm_arch5.
1419         (*call_reg_thumb1): Likewise.
1420         (*call_value_reg_thumb1_v5): Likewise.
1421         (*call_value_reg_thumb1): Likewise.
1422         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1423         unreachable path.
1424         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1425
1426 2018-05-18  Martin Liska  <mliska@suse.cz>
1427
1428         PR gcov-profile/84846
1429         * doc/gcov.texi: Document -t option of gcov tool.
1430
1431 2018-05-18  Martin Liska  <mliska@suse.cz>
1432
1433         PR gcov-profile/84846
1434         * gcov.c (print_usage): Add new -t option.
1435         (process_args): Handle the option.
1436         (generate_results): Use stdout as output when requested by
1437         the option.
1438
1439 2018-05-18  Martin Liska  <mliska@suse.cz>
1440
1441         PR gcov-profile/84846
1442         * coverage.c (coverage_init): Write PWD to .gcno file.
1443         * doc/gcov.texi: Document how working directory is printed.
1444         * gcov-dump.c (dump_gcov_file): Print PWD.
1445         * gcov.c (output_intermediate_file): Likewise.
1446         (read_graph_file): Read PWD string.
1447         (output_lines): Print PWD.
1448
1449 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1450
1451         PR middle-end/85817
1452         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1453         for retval and return false if all args to phi are zero.
1454
1455 2018-05-18  Richard Biener  <rguenther@suse.de>
1456
1457         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1458         method.
1459         (evrp_dom_walker::before_dom_children): Call it.
1460
1461 2018-05-18  Richard Biener  <rguenther@suse.de>
1462
1463         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1464         results when processing array refs with variable index.
1465
1466 2018-05-18  Toon Moene  <toon@moene.org>
1467
1468         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1469         directly after that of -floop-interchange. Indicate that both
1470         options are enabled by default when specifying -O3.
1471
1472 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1473
1474         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1475         iterator.  Delete separate integer-mode vec_set<mode> expander.
1476         (aarch64_simd_vec_setv2di): Delete.
1477         (vec_setv2di): Delete.
1478         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1479         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
1480         the "w, r" alternative.
1481
1482 2018-05-18  Martin Liska  <mliska@suse.cz>
1483
1484         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1485         * tree-pass.h (make_pass_lower_switch_O0): New function.
1486         * tree-switch-conversion.c (node_has_low_bound): Remove.
1487         (node_has_high_bound): Likewise.
1488         (node_is_bounded): Likewise.
1489         (class pass_lower_switch): Make it a template type and create
1490         two instances.
1491         (pass_lower_switch::execute): Add template argument.
1492         (make_pass_lower_switch): New function.
1493         (make_pass_lower_switch_O0): New function.
1494         (do_jump_if_equal): Remove.
1495         (emit_case_nodes): Simplify to just handle all 3 cases and leave
1496         all the hard work to tree optimization passes.
1497
1498 2018-05-18  Martin Liska  <mliska@suse.cz>
1499
1500         * dbgcnt.c (limit_low): Renamed from limit.
1501         (limit_high): New variable.
1502         (dbg_cnt_is_enabled): Check for upper limit.
1503         (dbg_cnt): Adjust dumping.
1504         (dbg_cnt_set_limit_by_index): Add new argument for high
1505         value.
1506         (dbg_cnt_set_limit_by_name): Likewise.
1507         (dbg_cnt_process_single_pair): Parse new format.
1508         (dbg_cnt_process_opt): Use strtok.
1509         (dbg_cnt_list_all_counters): Remove 'value' and add
1510         'limit_high'.
1511         * doc/invoke.texi: Document changes.
1512
1513 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
1514
1515         * doc/sourcebuild.texi (scalar_all_fma): Document.
1516         * tree.def (FMA_EXPR): Delete.
1517         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1518         * internal-fn.c (ternary_direct): New macro.
1519         (expand_ternary_optab_fn): Likewise.
1520         (direct_ternary_optab_supported_p): Likewise.
1521         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1522         * builtins.c (fold_builtin_fma): Delete.
1523         (fold_builtin_3): Don't call it.
1524         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1525         * expr.c (expand_expr_real_2): Likewise.
1526         * fold-const.c (operand_equal_p): Likewise.
1527         (fold_ternary_loc): Likewise.
1528         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1529         * gimple.c (DEFTREECODE): Likewise.
1530         * gimplify.c (gimplify_expr): Likewise.
1531         * optabs-tree.c (optab_for_tree_code): Likewise.
1532         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1533         * tree-eh.c (operation_could_trap_p): Likewise.
1534         (stmt_could_throw_1_p): Likewise.
1535         * tree-inline.c (estimate_operator_cost): Likewise.
1536         * tree-pretty-print.c (dump_generic_node): Likewise.
1537         (op_code_prio): Likewise.
1538         * tree-ssa-loop-im.c (stmt_cost): Likewise.
1539         * tree-ssa-operands.c (get_expr_operands): Likewise.
1540         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1541         * fold-const-call.h (fold_fma): Delete.
1542         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1543         CFN_FNMA and CFN_FNMS.
1544         (fold_fma): Delete.
1545         * genmatch.c (combined_fn): New enum.
1546         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1547         (commutative_op): New function.
1548         (commutate): Use it.  Handle more than 2 operands.
1549         (dt_operand::gen_gimple_expr): Use commutative_op.
1550         (parser::parse_expr): Allow :c to be used with non-binary
1551         operators if the commutative operand is known.
1552         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1553         CFN_FMS, CFN_FNMA and CFN_FNMS.
1554         (backprop::process_assign_use): Remove FMA_EXPR handling.
1555         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1556         (gen_hsa_fma): New function.
1557         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1558         IFN_FNMA and IFN_FNMS.
1559         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1560         * gimple-fold.h (follow_all_ssa_edges): Declare.
1561         * gimple-fold.c (follow_all_ssa_edges): New function.
1562         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1563         gimple_build interface and use follow_all_ssa_edges to fold the result.
1564         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1565         instead of checking for optabs directly.
1566         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1567         rather than FMA_EXPRs.
1568         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1569         call to IFN_FMA instead of an FMA_EXPR.
1570
1571 2018-05-17  Jim Wilson  <jimw@sifive.com>
1572
1573         * expr.c (do_tablejump): When converting index to Pmode, if we have a
1574         sign extended promoted subreg, and the range does not have the sign bit
1575         set, then do a sign extend.
1576
1577         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1578         test, check for sign extended subreg and/or constant operands, and
1579         do a sign extend in that case.
1580
1581 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
1582
1583         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1584         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1585         Add untyped.
1586         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1587         Change logics_shift_reg to logics_shift_imm.
1588         (thunderx2t99_fp_loadpair_basic): Delete.
1589         (thunderx2t99_fp_storepair_basic): Delete.
1590         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1591         (thunderx2t99_asimd_polynomial): Delete.
1592         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1593         and neon_fp_mul_d_scalar_q.
1594         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1595         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1596         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1597         (thunderx2t99_asimd_lut): Add missing tbl types.
1598         (thunderx2t99_asimd_ext): Delete.
1599         (thunderx2t99_asimd_load1_1_mult): Delete.
1600         (thunderx2t99_asimd_load1_2_mult): Delete.
1601         (thunderx2t99_asimd_load1_ldp): New.
1602         (thunderx2t99_asimd_load1): New.
1603         (thunderx2t99_asimd_load2): Add missing *load2* types.
1604         (thunderx2t99_asimd_load3): New.
1605         (thunderx2t99_asimd_load4): New.
1606         (thunderx2t99_asimd_store1_1_mult): Delete.
1607         (thunderx2t99_asimd_store1_2_mult): Delete.
1608         (thunderx2t99_asimd_store2_mult): Delete.
1609         (thunderx2t99_asimd_store2_onelane): Delete.
1610         (thunderx2t99_asimd_store_stp): New.
1611         (thunderx2t99_asimd_store1): New.
1612         (thunderx2t99_asimd_store2): New.
1613         (thunderx2t99_asimd_store3): New.
1614         (thunderx2t99_asimd_store4): New.
1615
1616 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
1617
1618         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1619         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
1620
1621 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
1622             Segher Boessenkool  <segher@kernel.crashing.org>
1623
1624         PR target/85698
1625         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1626         operand.
1627
1628 2018-05-17  Richard Biener  <rguenther@suse.de>
1629
1630         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1631         for pruning loop and prune defs feeding only already visited PHIs.
1632
1633 2018-05-17  Richard Biener  <rguenther@suse.de>
1634
1635         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1636
1637 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
1638             Richard Biener  <rguenther@suse.de>
1639
1640         PR tree-optimization/85793
1641         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1642         for VMAT_ELEMENTWISE.
1643
1644 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
1645
1646         * internal-fn.h (lookup_internal_fn): Declare
1647         * internal-fn.c (lookup_internal_fn): New function.
1648         * gimple.c (gimple_build_call_from_tree): Handle calls to
1649         internal functions.
1650         * gimple-pretty-print.c (dump_gimple_call): Print "." before
1651         internal function names.
1652         * tree-pretty-print.c (dump_generic_node): Likewise.
1653         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1654
1655 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
1656
1657         * gimple-fold.h (gimple_build): Make the function forms take
1658         combined_fn rather than built_in_function.
1659         (gimple_simplify): Likewise.
1660         * gimple-match-head.c (gimple_simplify): Likewise.
1661         * gimple-fold.c (gimple_build): Likewise.
1662         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1663         rather than gimple_build_call_internal.
1664         (get_initial_defs_for_reduction): Likewise.
1665         (vect_create_epilog_for_reduction): Likewise.
1666         (vectorizable_live_operation): Likewise.
1667
1668 2018-05-17  Martin Liska  <mliska@suse.cz>
1669
1670         * gimple-ssa-sprintf.c (format_directive): Do not use
1671         space in between 'G_' and '('.
1672
1673 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
1674
1675         PR target/85323
1676         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1677         even if the mask is not all ones.
1678
1679         PR target/85323
1680         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1681         vector.
1682         (ix86_gimple_fold_builtin): Likewise.
1683
1684         PR target/85323
1685         * config/i386/i386.c: Include tree-vector-builder.h.
1686         (ix86_vector_shift_count): New function.
1687         (ix86_fold_builtin): Fold shift builtins by scalar count.
1688         (ix86_gimple_fold_builtin): Likewise.
1689
1690         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1691         _mm512_setzero): New intrinsics.
1692
1693 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
1694             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1695
1696         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1697         code generation for cases where splatting a value is not useful.
1698         * simplify-rtx.c (simplify_ternary_operation): Simplify
1699         vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1700         mode to a vec_concat.
1701
1702 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
1703
1704         * config.gcc: Support "goldmont-plus".
1705         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1706         "goldmont-plus".
1707         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1708         PROCESSOR_GOLDMONT_PLUS.
1709         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1710         (processor_target_table): Add "goldmont-plus".
1711         (PTA_GOLDMONT_PLUS): Define.
1712         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1713         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1714         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1715         (fold_builtin_cpu): Add "goldmont-plus".
1716         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1717         (ix86_option_override_internal): Add "goldmont-plus".
1718         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1719         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1720         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1721         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1722
1723 2018-05-17  Richard Biener  <rguenther@suse.de>
1724
1725         PR tree-optimization/85757
1726         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1727         remove defs that only feed that PHI from further processing.
1728
1729 2018-05-16  Jim Wilson  <jimw@sifive.com>
1730
1731         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1732         asterisk to name.
1733         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1734
1735 2018-05-16  Mark Wielaard  <mark@klomp.org>
1736
1737         * dwarf2out.c (count_index_strings): New function.
1738         (output_indirect_strings): Call count_index_strings and generate
1739         header for dwarf_version >= 5.
1740
1741 2018-05-16  Mark Wielaard  <mark@klomp.org>
1742
1743         * dwarf2out.c (dwarf_FORM): New function.
1744         (set_indirect_string): Use dwarf_FORM.
1745         (reset_indirect_string): Likewise.
1746         (size_of_die): Likewise.
1747         (value_format): Likewise.
1748         (output_die): Likewise.
1749         (add_skeleton_AT_string): Likewise.
1750         (output_macinfo_op): Likewise.
1751         (index_string): Likewise.
1752         (output_index_string_offset): Likewise.
1753         (output_index_string): Likewise.
1754         (count_index_strings): Likewise.
1755
1756 2018-05-16  Carl Love  <cel@us.ibm.com>
1757
1758         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1759         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1760
1761 2018-05-16  Martin Jambor  <mjambor@suse.cz>
1762
1763         * ipa-prop.c (ipa_free_all_edge_args): Remove.
1764         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1765
1766 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
1767
1768         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1769         (fnma<mode>4): Likewise.
1770         (fms<mode>4): Likewise.
1771         (fnms<mode>4): Likewise.
1772         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1773         (aarch64_fnma<mode>4): Likewise.
1774         (aarch64_fms<mode>4): Likewise.
1775         (aarch64_fnms<mode>4): Likewise.
1776         (aarch64_fnmadd<mode>4): Likewise.
1777
1778 2018-05-16  Jason Merrill  <jason@redhat.com>
1779
1780         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
1781
1782 2018-05-16  Richard Biener  <rguenther@suse.de>
1783
1784         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1785         (dump_stmt_cost): Declare.
1786         (add_stmt_cost): Dump cost we add.
1787         (add_stmt_costs): New function.
1788         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1789         No longer exported.
1790         (vect_analyze_stmt): Adjust prototype.
1791         (vectorizable_condition): Likewise.
1792         (vectorizable_live_operation): Likewise.
1793         (vectorizable_reduction): Likewise.
1794         (vectorizable_induction): Likewise.
1795         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1796         cost vector to pass to vectorizable_ and record afterwards.
1797         (vect_model_reduction_cost): Take cost vector argument and adjust.
1798         (vect_model_induction_cost): Likewise.
1799         (vectorizable_reduction): Likewise.
1800         (vectorizable_induction): Likewise.
1801         (vectorizable_live_operation): Likewise.
1802         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1803         SLP_TREE_NUMBER_OF_VEC_STMTS.
1804         (vect_analyze_slp_cost_1): Remove.
1805         (vect_analyze_slp_cost): Likewise.
1806         (vect_slp_analyze_node_operations): Take visited args and
1807         a target cost vector.  Avoid processing already visited stmt sets.
1808         (vect_slp_analyze_operations): Use a local cost vector to gather
1809         costs and register those of non-discarded instances.
1810         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1811         (vect_schedule_slp_instance): Remove copying of
1812         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
1813         zero.
1814         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1815         adding cost.  Record cost entry location.
1816         (vect_prologue_cost_for_slp_op): Function to compute cost of
1817         a constant or invariant generated for SLP vect in the prologue,
1818         split out from vect_analyze_slp_cost_1.
1819         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
1820         (vect_model_promotion_demotion_cost): Likewise.
1821         (vect_model_store_cost): Likewise, make static.
1822         (vect_model_load_cost): Likewise.
1823         (vectorizable_bswap): Add cost vector arg and adjust.
1824         (vectorizable_call): Likewise.
1825         (vectorizable_simd_clone_call): Likewise.
1826         (vectorizable_conversion): Likewise.
1827         (vectorizable_assignment): Likewise.
1828         (vectorizable_shift): Likewise.
1829         (vectorizable_operation): Likewise.
1830         (vectorizable_store): Likewise.
1831         (vectorizable_load): Likewise.
1832         (vectorizable_condition): Likewise.
1833         (vectorizable_comparison): Likewise.
1834         (can_vectorize_live_stmts): Likewise.
1835         (vect_analyze_stmt): Likewise.
1836         (vect_transform_stmt): Adjust calls to vectorizable_*.
1837         * tree-vectorizer.c: Include gimple-pretty-print.h.
1838         (dump_stmt_cost): New function.
1839
1840 2018-05-16  Richard Biener  <rguenther@suse.de>
1841
1842         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1843         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1844         * tree-ssa-dse.c: Include tree-ssa-loop.h.
1845         (check_name): New callback.
1846         (dse_classify_store): Track cycles via a visited bitmap of PHI
1847         defs and simplify handling of in-loop and across loop dead stores
1848         and properly fail for loop-variant refs.  Handle byte-tracking with
1849         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1850         limiting the walk.
1851
1852 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
1853
1854         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1855         (vect_get_mask_type_for_stmt): Likewise.
1856         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1857         split out from...
1858         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
1859         to determine the statement's vector type and the vector type that
1860         should be used for calculating nunits.  Deal with cases in which
1861         the type has to be deferred.
1862         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1863         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1864         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1865         (vect_determine_vf_for_stmt): New functions, split out from...
1866         (vect_determine_vectorization_factor): ...here.
1867         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1868         (vect_get_mask_type_for_stmt): New functions, split out from
1869         vect_determine_vectorization_factor.
1870
1871 2018-05-16  Richard Biener  <rguenther@suse.de>
1872
1873         * tree-cfg.c (verify_gimple_assign_ternary): Properly
1874         verify the [VEC_]COND_EXPR embedded comparison.
1875
1876 2018-05-15  Martin Sebor  <msebor@redhat.com>
1877
1878         PR tree-optimization/85753
1879         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1880         RECORD_TYPE in addition to ARRAY_TYPE.
1881
1882 2018-05-15  Martin Sebor  <msebor@redhat.com>
1883
1884         PR middle-end/85643
1885         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1886
1887 2018-05-15  Richard Biener  <rguenther@suse.de>
1888
1889         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1890         add by_clobber_p one.  Change algorithm to collect all defs
1891         representing uses we need to walk and try reducing them to
1892         a single one before failing.
1893         (dse_dom_walker::dse_optimize_stmt): Adjust.
1894
1895 2018-05-13  Mark Wielaard  <mark@klomp.org>
1896
1897         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1898         (size_of_loc_descr): Likewise.
1899         (output_loc_operands): Likewise.
1900         (output_loc_operands_raw): Likewise.
1901         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1902         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1903         (hash_loc_operands): Likewise.
1904         (compare_loc_operands): Likewise.
1905
1906 2018-05-14  Mark Wielaard  <mark@klomp.org>
1907
1908         * dwarf2out.c (count_index_addrs): New function.
1909         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1910
1911 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1912
1913         PR tree-optimization/83648
1914         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1915         return value as malloc candidate.
1916
1917 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1918
1919         PR ipa/85734
1920         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1921         as true in call to suggest_attribute.
1922
1923 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
1924
1925         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1926         -mreadonly-in-sdata.
1927
1928 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1929
1930         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1931         New pattern.
1932         (aarch64_crypto_aesd_fused): Likewise.
1933
1934 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
1935
1936         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1937         (movsi_aarch64): Likewise.
1938         (load_pairsi): Likewise.
1939         (load_pairdi): Likewise.
1940         (store_pairsi): Likewise.
1941         (store_pairdi): Likewise.
1942         (load_pairsf): Likewise.
1943         (load_pairdf): Likewise.
1944         (store_pairsf): Likewise.
1945         (store_pairdf): Likewise.
1946         (zero_extend): Likewise.
1947         (trunc): Swap alternatives.
1948         (fcvt_target): Add '?' to prefer w over r.
1949
1950 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
1951
1952         PR target/85756
1953         * config/i386/i386.md: Disallow non-commutative arithmetics in
1954         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1955         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
1956         in the peephole2 before it.
1957
1958 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
1959
1960         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1961         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1962         (ix86_handle_option): Handle -mcldemote.
1963         * config.gcc: New header.
1964         * config/i386/cldemoteintrin.h: New file.
1965         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1966         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1967         -mcldemote.
1968         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1969         OPTION_MASK_ISA_CLDEMOTE.
1970         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1971         (ix86_valid_target_attribute_inner_p): Ditto.
1972         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1973         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1974         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1975         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1976         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1977         (cldemote): New.
1978         * config/i386/i386.opt: Add -mcldemote.
1979         * config/i386/x86intrin.h: New header.
1980         * doc/invoke.texi: Add -mcldemote.
1981
1982 2018-05-14  Richard Biener  <rguenther@suse.de>
1983
1984         * doc/match-and-simplify.texi: Adjust :s documentation.
1985
1986 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
1987
1988         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1989         intended memcpy size.
1990         (REORDER_45): Likewise.
1991
1992 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
1993
1994         * sort.cc: New file.
1995         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1996         * vec.c (qsort_chk): Use gcc_qsort.
1997         * Makefile.in (OBJS-libcommon): Add sort.o.
1998         (build/sort.o): New target.  Use it...
1999         (BUILD_RTL): ... here, and...
2000         (build/gencfn-macros): ... here, and...
2001         (build/genmatch): ... here.
2002
2003 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2004             Chung-Ju Wu  <jasonwucj@gmail.com>
2005
2006         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2007         * config/nds32/nds32-graywolf.md: New file.
2008         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2009         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2010         pipeline.
2011         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2012         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2013         * config/nds32/nds32.md (pipeline_model): Add graywolf.
2014         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2015         * config/nds32/pipelines.md: Include n15 settings.
2016
2017 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2018             Chung-Ju Wu  <jasonwucj@gmail.com>
2019
2020         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2021         * config/nds32/nds32-n13.md: New file.
2022         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2023         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2024         pipeline.
2025         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2026         * config/nds32/nds32.md (pipeline_model): Add n13.
2027         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2028         * config/nds32/pipelines.md: Include n13 settings.
2029
2030 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2031             Chung-Ju Wu  <jasonwucj@gmail.com>
2032
2033         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2034         * config/nds32/nds32-n10.md: New file.
2035         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2036         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2037         pipeline.
2038         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2039         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2040         * config/nds32/nds32.md (pipeline_model): Add n10.
2041         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2042         * config/nds32/pipelines.md: Include n10 settings.
2043
2044 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
2045             Kito Cheng  <kito.cheng@gmail.com>
2046             Chung-Ju Wu  <jasonwucj@gmail.com>
2047
2048         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2049         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2050         Add enum values for DSP extension instructions.
2051         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2052         New constraints.
2053         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2054         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2055         New code iterators.
2056         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2057         * config/nds32/nds32-dspext.md: New file for DSP implementation.
2058         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2059         * config/nds32/nds32-intrinsic.md: Likewise.
2060         * config/nds32/nds32_intrinsic.h: Likewise.
2061         * config/nds32/nds32-md-auxiliary.c: Likewise.
2062         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2063         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2064         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2065         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2066         * config/nds32/nds32-protos.h: New declarations for DSP extension.
2067         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2068         TYPE_DMAC in switch statement.
2069         * config/nds32/nds32.c: New checking and implementation for DSP
2070         extension instructions.
2071         * config/nds32/nds32.h: Likewise.
2072         * config/nds32/nds32.md: Likewise.
2073         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2074         * config/nds32/predicates.md: Implement new predicates for DSP
2075         extension.
2076
2077 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2078
2079         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2080         Reformat alternatives and attributes so it is easier to identify
2081         which constraints/attributes go with which instruction.
2082         (mov<mode>_hardfloat32, FMOVE64): Likewise.
2083         (mov<mode>_softfloat32, FMOVE64): Likewise.
2084         (mov<mode>_hardfloat64, FMOVE64): Likewise.
2085         (mov<mode>_softfloat64, FMOVE64): Likewise.
2086
2087 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2088
2089         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2090         subsection.
2091         (Basic PowerPC Built-in Functions): The new name of the
2092         subsection previously known as "PowerPC Built-in Functions".
2093         (Basic PowerPC Built-in Functions Available on all Configurations):
2094         New subsubsection.
2095         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2096         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2097         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2098         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2099
2100 2018-05-11  Martin Jambor  <mjambor@suse.cz>
2101
2102         PR ipa/85655
2103         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2104         single const.
2105
2106 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
2107
2108         PR target/85733
2109         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2110
2111 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
2112
2113         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2114         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2115         (ix86_handle_option): Handle -mwaitpkg.
2116         * config.gcc: New header.
2117         * config/i386/cpuid.h (bit_WAITPKG): New bit.
2118         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2119         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2120         function type.
2121         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2122         OPTION_MASK_ISA_WAITPKG.
2123         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2124         (ix86_option_override_internal): Add PTA_WAITPKG.
2125         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2126         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2127         IX86_BUILTIN_TPAUSE.
2128         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2129         __builtin_ia32_umwait and __builtin_ia32_tpause.
2130         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2131         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2132         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2133         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2134         UNSPECV_TPAUSE): New.
2135         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2136         * config/i386/i386.opt: Add -mwaitpkg.
2137         * config/i386/waitpkgintrin.h: New file.
2138         * config/i386/x86intrin.h: New header.
2139         * doc/invoke.texi: Add -mwaitpkg.
2140
2141 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
2142
2143         PR target/85606
2144         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2145         equivalent.
2146         (cortex-m0): Use armv6s-m isa.
2147         (cortex-m0plus): Likewise.
2148         (cortex-m1): Likewise.
2149         (cortex-m0.small-multiply): Likewise.
2150         (cortex-m0plus.small-multiply): Likewise.
2151         (cortex-m1.small-multiply): Likewise.
2152
2153 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
2154             Jakub Jelinek  <jakub@redhat.com>
2155
2156         PR tree-optimization/85692
2157         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2158         source permute as well.
2159
2160 2018-05-11  Martin Liska  <mliska@suse.cz>
2161
2162         PR sanitizer/85556
2163         * doc/extend.texi: Document LLVM style format for no_sanitize
2164         attribute.
2165
2166 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
2167
2168         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2169         mode_supports_vsx_dform_quad to mode_supports_dq_form.
2170         (mode_supports_vsx_dform_quad): Likewise.
2171         (mode_supports_vmx_dform): Move these functions to be next to the
2172         other mode_supports functions.
2173         (mode_supports_dq_form): Likewise.
2174         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2175         mode_supports_dq_form.
2176         (reg_offset_addressing_ok_p): Likewise.
2177         (offsettable_ok_by_alignment): Likewise.
2178         (rs6000_legitimate_offset_address_p): Likewise.
2179         (legitimate_lo_sum_address_p): Likewise.
2180         (rs6000_legitimize_address): Likewise.
2181         (rs6000_legitimize_reload_address): Likewise.
2182         (rs6000_secondary_reload_inner): Likewise.
2183         (rs6000_preferred_reload_class): Likewise.
2184         (rs6000_output_move_128bit): Likewise.
2185
2186 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
2187
2188         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2189         Generate SImode target register for null target.
2190         <case IX86_BUILTIN_XGETBV>: Ditto.
2191         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2192         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2193
2194 2018-05-10  Carl Love  <cel@us.ibm.com>
2195
2196         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2197         dcbtt and dcbtstt if operands[2] is 0.
2198
2199 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
2200
2201         PR target/85693
2202         * config/i386/sse.md (usadv64qi): New expander.
2203
2204 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
2205
2206         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2207         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2208         -maltivec=be support.
2209         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2210         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2211         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2212         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2213         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2214         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2215         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2216         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2217         altivec_vsumsws): Adjust.
2218         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2219         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2220         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2221         support.
2222         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2223         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2224         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2225         (altivec_lve<VI_char>x): Delete expand.
2226         (*altivec_lve<VI_char>x_internal): Rename to...
2227         (altivec_lve<VI_char>x): ... this.
2228         (altivec_lvxl_<mode>): Delete expand.
2229         (*altivec_lvxl_<mode>_internal): Rename to ...
2230         (altivec_lvxl_<mode>): ... this.
2231         (altivec_stvxl_<mode>): Delete expand.
2232         (*altivec_stvxl_<mode>_internal): Rename to ...
2233         (altivec_stvxl_<mode>): ... this.
2234         (altivec_stve<VI_char>x): Delete expand.
2235         (*altivec_stve<VI_char>x_internal): Rename to ...
2236         (altivec_stve<VI_char>x): ... this.
2237         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2238         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2239         reduc_plus_scal_<mode>): Adjust.
2240         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2241         comment.
2242         (rs6000_cpu_cpp_builtins): Adjust.
2243         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2244         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2245         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2246         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2247         -maltivec=be support.
2248         (rs6000_split_vec_extract_var): Adjust.
2249         (rs6000_split_v4si_init): Adjust.
2250         (swap_selector_for_mode): Delete.
2251         (altivec_expand_lvx_be, altivec_expand_stvx_be,
2252         altivec_expand_stvex_be): Delete.
2253         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2254         -maltivec=be support.
2255         (rs6000_gimple_fold_builtin): Ditto.
2256         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2257         Adjust.
2258         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2259         (TARGET_DIRECT_MOVE_64BIT): Adjust.
2260         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2261         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2262         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2263         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2264         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2265         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2266         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2267         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2268         anonymous split): Adjust.
2269         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2270         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2271
2272 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
2273
2274         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2275         when --with-gxx-include-dir is also specified.
2276         * configure: Regenerate.
2277
2278 2018-05-09  Jim Wilson  <jimw@sifive.com>
2279
2280         PR target/84797
2281         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2282         * config/riscv/t-withmultilib: New.
2283         * config/riscv/withmultilib.h: New.
2284         * doc/install.texi: Document RISC-V --with-multilib-list support.
2285
2286 2018-05-09  Richard Biener  <rguenther@suse.de>
2287
2288         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2289         vector.
2290         (vect_bb_vectorization_profitable_p): Adjust.  Compute
2291         actual scalar cost using the cost vector and the add_stmt_cost
2292         machinery.
2293
2294 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2295
2296         PR rtl-optimization/85645
2297         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2298         in the REG_CFA_REGISTER note for LR, don't leave it empty.
2299
2300 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2301
2302         PR rtl-optimization/85645
2303         * shrink-wrap.c (spread_components): Return a boolean saying if
2304         anything was changed.
2305         (try_shrink_wrapping_separate): Iterate spread_components until
2306         nothing changes anymore.
2307
2308 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2309
2310         PR rtl-optimization/85645
2311         * regrename.c (build_def_use): Also kill the chains that include the
2312         destination of a REG_CFA_REGISTER note.
2313
2314 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2315
2316         PR rtl-optimization/85645
2317         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2318         insn that has a REG_CFA_REGISTER note.
2319
2320 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
2321
2322         * cfgexpand.c (expand_clobber): New function.
2323         (expand_gimple_stmt_1): Use it.
2324         * tree-vect-stmts.c (vect_clobber_variable): New function,
2325         split out from...
2326         (vectorizable_simd_clone_call): ...here.
2327         (vectorizable_store): Emit a clobber either side of an
2328         IFN_STORE_LANES sequence.
2329         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2330
2331 2018-05-09  Tom de Vries  <tom@codesourcery.com>
2332
2333         PR target/85626
2334         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2335         (define_insn "trap_if_false"): Add exit after trap.
2336
2337 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
2338
2339         PR rtl-optimization/85638
2340         * bb-reorder.c: Include common/common-target.h.
2341         (create_forwarder_block): New function extracted from...
2342         (fix_up_crossing_landing_pad): ...here.  Rename into...
2343         (dw2_fix_up_crossing_landing_pad): ...this.
2344         (sjlj_fix_up_crossing_landing_pad): New function.
2345         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2346         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2347         from both partitions and exit the loop after one iteration.
2348
2349 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2350
2351         Revert:
2352         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2353         subsection.
2354         (Basic PowerPC Built-in Functions): The new name of the
2355         subsection previously known as "PowerPC Built-in Functions".
2356         (Basic PowerPC Built-in Functions Available on all Configurations):
2357         New subsubsection.
2358         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2359         subsubsection.
2360         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2361         subsubsection.
2362         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2363         subsubsection.
2364         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2365         subsubsection.
2366
2367 2018-05-08  Jim Wilson  <jimw@sifive.com>
2368
2369         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2370         (LD_EMUL_SUFFIX): New.
2371         (LINK_SPEC): Use it.
2372
2373 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2374
2375         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2376         subsection.
2377         (Basic PowerPC Built-in Functions): The new name of the
2378         subsection previously known as "PowerPC Built-in Functions".
2379         (Basic PowerPC Built-in Functions Available on all Configurations):
2380         New subsubsection.
2381         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2382         subsubsection.
2383         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2384         subsubsection.
2385         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2386         subsubsection.
2387         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2388         subsubsection.
2389
2390 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
2391
2392         PR target/85683
2393         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2394         after cmpelim optimization.
2395
2396 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
2397
2398         * config.gcc: Support "goldmont".
2399         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2400         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2401         PROCESSOR_GOLDMONT.
2402         * config/i386/i386.c (m_GOLDMONT): Define.
2403         (processor_target_table): Add "goldmont".
2404         (PTA_GOLDMONT): Define.
2405         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2406         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2407         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2408         (fold_builtin_cpu): Add "goldmont".
2409         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2410         (ix86_option_override_internal): Add "goldmont".
2411         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2412         (processor_type): Add PROCESSOR_GOLDMONT.
2413         * config/i386/i386.md: Add CPU "glm".
2414         * config/i386/glm.md: New file.
2415         * config/i386/x86-tune.def: Add m_GOLDMONT.
2416         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2417
2418 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
2419
2420         PR target/85572
2421         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2422         E_V4DImode.
2423         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2424         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
2425         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
2426
2427         PR target/85317
2428         * config/i386/i386.c (ix86_fold_builtin): Handle
2429         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2430
2431         PR target/85480
2432         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2433         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2434
2435 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
2436
2437         PR target/85658
2438         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2439         (check_arch): Likewise.
2440         (check_fpu): Return the result rather than printing it.
2441         (end arch): Fix operator precedence.
2442         (end cpu): Likewise.
2443         (END): Print the result from check_fpu.
2444
2445 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
2446             Alan Hayward  <alan.hayward@arm.com>
2447             David Sherwood  <david.sherwood@arm.com>
2448
2449         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2450         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2451         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2452         (*fcmuo<mode>_and): New patterns.
2453
2454 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
2455
2456         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2457         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2458         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2459         (cmp_op, sve_imm_con): New code attributes.
2460         (SVE_COND_INT_CMP, imm_con): Delete.
2461         (cmp_op): Remove above unspecs from int attribute.
2462         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2463         to...
2464         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
2465         comparison-specific unspecs.
2466         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2467         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2468         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2469         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2470         (*vec_fcm<cmp_op><mode>): Rename to...
2471         (*fcm<cmp_op><mode>): ...this and adjust likewise.
2472         (*vec_fcmuo<mode>): Rename to...
2473         (*fcmuo<mode>): ...this and adjust likewise.
2474         (*pred_fcm<cmp_op><mode>): New pattern.
2475         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2476         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2477         functions.
2478         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2479         and UNORDERED.
2480         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2481         (aarch64_emit_sve_predicated_cond): New function.
2482         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2483         (aarch64_emit_unspec_cond_or): Replace with...
2484         (aarch64_emit_sve_or_conds): ...this new function.  Use
2485         aarch64_emit_sve_ptrue_op for the individual comparisons and
2486         aarch64_emit_binop to OR them together.
2487         (aarch64_emit_inverted_unspec_cond): Replace with...
2488         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
2489         aarch64_emit_sve_ptrue_op for the comparison and
2490         aarch64_emit_unop to invert the result.
2491         (aarch64_expand_sve_vec_cmp_float): Update after the above
2492         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
2493
2494 2018-05-07  Nathan Sidwell  <nathan@acm.org>
2495
2496         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2497         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2498         (Backwards Compatibility): Likewise.
2499
2500 2018-05-07  Luis Machado  <luis.machado@linaro.org>
2501
2502         PR bootstrap/85681
2503         Revert:
2504         2018-05-07  Luis Machado  <luis.machado@linaro.org>
2505
2506         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2507         <prefetch_dynamic_strides>: New const bool field.
2508         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2509         prefetch_dynamic_strides.
2510         (exynosm1_prefetch_tune): Likewise.
2511         (thunderxt88_prefetch_tune): Likewise.
2512         (thunderx_prefetch_tune): Likewise.
2513         (thunderx2t99_prefetch_tune): Likewise.
2514         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2515         (aarch64_override_options_internal): Update to set
2516         PARAM_PREFETCH_DYNAMIC_STRIDES.
2517         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2518         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2519         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2520         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2521         prefetch-dynamic-strides setting.
2522
2523         2018-05-07  Luis Machado  <luis.machado@linaro.org>
2524
2525         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2526         <minimum_stride>: New const int field.
2527         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2528         minimum_stride field.
2529         (exynosm1_prefetch_tune): Likewise.
2530         (thunderxt88_prefetch_tune): Likewise.
2531         (thunderx_prefetch_tune): Likewise.
2532         (thunderx2t99_prefetch_tune): Likewise.
2533         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2534         (aarch64_override_options_internal): Update to set
2535         PARAM_PREFETCH_MINIMUM_STRIDE.
2536         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2537         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2538         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2539         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2540         stride is constant and is below the minimum stride threshold.
2541
2542 2018-05-07  Luis Machado  <luis.machado@linaro.org>
2543
2544         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2545         to 512.
2546
2547 2018-05-07  Luis Machado  <luis.machado@linaro.org>
2548
2549         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2550         <prefetch_dynamic_strides>: New const bool field.
2551         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2552         prefetch_dynamic_strides.
2553         (exynosm1_prefetch_tune): Likewise.
2554         (thunderxt88_prefetch_tune): Likewise.
2555         (thunderx_prefetch_tune): Likewise.
2556         (thunderx2t99_prefetch_tune): Likewise.
2557         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2558         (aarch64_override_options_internal): Update to set
2559         PARAM_PREFETCH_DYNAMIC_STRIDES.
2560         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2561         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2562         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2563         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2564         prefetch-dynamic-strides setting.
2565
2566 2018-05-07  Luis Machado  <luis.machado@linaro.org>
2567
2568         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2569         <minimum_stride>: New const int field.
2570         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2571         minimum_stride field.
2572         (exynosm1_prefetch_tune): Likewise.
2573         (thunderxt88_prefetch_tune): Likewise.
2574         (thunderx_prefetch_tune): Likewise.
2575         (thunderx2t99_prefetch_tune): Likewise.
2576         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2577         (aarch64_override_options_internal): Update to set
2578         PARAM_PREFETCH_MINIMUM_STRIDE.
2579         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2580         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2581         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2582         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2583         stride is constant and is below the minimum stride threshold.
2584
2585 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
2586
2587         PR c++/85659
2588         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2589         the type is addressable.  Don't force op into register if it has
2590         BLKmode.
2591
2592 2018-05-05  Roland McGrath  <mcgrathr@google.com>
2593
2594         PR other/77609
2595         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2596         any section for which we don't know a specific type it should have,
2597         regardless of name.  Previously this was done only for the exact
2598         names ".init_array", ".fini_array", and ".preinit_array".
2599         (default_elf_asm_named_section): Add comment about
2600         relationship with default_section_type_flags and SECTION_NOTYPE.
2601         (get_section): Don't consider it a type conflict if one side has
2602         SECTION_NOTYPE and the other doesn't, as long as neither has the
2603         SECTION_BSS et al used in the default_section_type_flags logic.
2604
2605 2018-05-05  Tom de Vries  <tom@codesourcery.com>
2606
2607         PR target/85653
2608         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2609         (workaround_barsyncs): New function.
2610         (nvptx_reorg): Use workaround_barsyncs.
2611         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2612         (define_expand "nvptx_membar_cta"): New define_expand.
2613         (define_insn "*nvptx_membar_cta"): New insn.
2614
2615 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
2616
2617         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2618         To improve optimization opportunities.
2619         * builtin-types.def: The new needed builtin types for the above.
2620
2621 2018-05-04  Richard Biener  <rguenther@suse.de>
2622
2623         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2624         * gimple-ssa-store-merging.c
2625         (imm_store_chain_info::output_merged_store): Remove redundant create,
2626         release split_store vector contents on failure.
2627         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2628         scalar stmt vector on cache hit.
2629
2630 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
2631
2632         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2633         Xilinx FP support.
2634         * config.gcc (powerpc-xilinx-eabi*): Remove.
2635         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2636         support.
2637         (fusion_addis_mem_combo_load): Ditto.
2638         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2639         FP support.
2640         (rs6000_cpu_cpp_builtins): Ditto.
2641         * config/rs6000/rs6000-linux.c
2642         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2643         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2644         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2645         support.
2646         (rs6000_setup_reg_addr_masks): Ditto.
2647         (rs6000_init_hard_regno_mode_ok): Ditto.
2648         (rs6000_option_override_internal): Ditto.
2649         (legitimate_lo_sum_address_p): Ditto.
2650         (rs6000_legitimize_address): Ditto.
2651         (rs6000_legitimize_reload_address): Ditto.
2652         (rs6000_legitimate_address_p): Ditto.
2653         (abi_v4_pass_in_fpr): Ditto.
2654         (setup_incoming_varargs): Ditto.
2655         (rs6000_gimplify_va_arg): Ditto.
2656         (rs6000_split_multireg_move): Ditto.
2657         (rs6000_savres_strategy): Ditto.
2658         (rs6000_emit_prologue_components): Ditto.
2659         (rs6000_emit_epilogue_components): Ditto.
2660         (rs6000_emit_prologue): Ditto.
2661         (rs6000_emit_epilogue): Ditto.
2662         (rs6000_elf_file_end): Ditto.
2663         (rs6000_function_value): Ditto.
2664         (rs6000_libcall_value): Ditto.
2665         * config/rs6000/rs6000.h: Ditto.
2666         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2667         (TARGET_MINMAX): ... this.  New.
2668         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2669         * config/rs6000/rs6000.md: Remove Xilinx FP support.
2670         (*movsi_internal1_single): Delete.
2671         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2672         mfpu=, mxilinx-fpu): Delete.
2673         * config/rs6000/singlefp.h: Delete.
2674         * config/rs6000/sysv4.h: Remove Xilinx FP support.
2675         * config/rs6000/t-rs6000: Ditto.
2676         * config/rs6000/t-xilinx: Delete.
2677         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2678         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2679         (VStype_simple): Delete.
2680         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2681         * config/rs6000/xfpu.h: Delete.
2682         * config/rs6000/xfpu.md: Delete.
2683         * config/rs6000/xilinx.h: Delete.
2684         * config/rs6000/xilinx.opt: Delete.
2685         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2686         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2687
2688 2018-05-04  Tom de Vries  <tom@codesourcery.com>
2689
2690         PR libgomp/85639
2691         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2692         if ignore == 0.
2693
2694 2018-05-04  Richard Biener  <rguenther@suse.de>
2695
2696         PR middle-end/85627
2697         * tree-complex.c (update_complex_assignment): We are always in SSA form.
2698         (expand_complex_div_wide): Likewise.
2699         (expand_complex_operations_1): Likewise.
2700         (expand_complex_libcall): Preserve EH info of the original stmt.
2701         (tree_lower_complex): Handle removed blocks.
2702         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2703         on complex multiplication and division libcall builtins.
2704
2705 2018-05-04  Richard Biener  <rguenther@suse.de>
2706
2707         PR middle-end/85574
2708         * fold-const.c (negate_expr_p): Restrict negation of operand
2709         zero of a division to when we know that can happen without
2710         overflow.
2711         (fold_negate_expr_1): Likewise.
2712
2713 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
2714
2715         PR libstdc++/85466
2716         * real.h (real_nextafter): Declare.
2717         * real.c (real_nextafter): New function.
2718         * fold-const-call.c (fold_const_nextafter): New function.
2719         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2720         CASE_CFN_NEXTTOWARD.
2721         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2722         even when arg1_mode is different from arg0_mode.
2723
2724 2018-05-03  Nathan Sidwell  <nathan@acm.org>
2725
2726         * doc/extend.texi (Deprecated Features): Remove
2727         -ffriend-injection.
2728         (Backwards Compatibility): Likewise.
2729         * doc/invoke.texi (C++ Language Options): Likewise.
2730         (C++ Dialect Options): Likewise.
2731
2732 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
2733
2734         PR target/85530
2735         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2736         _mm512_mask_mullox_epi64): New intrinsics.
2737
2738 2018-05-03  Tom de Vries  <tom@codesourcery.com>
2739
2740         PR testsuite/85106
2741         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2742         dump files): Add offload-tree.
2743
2744 2018-05-03  Richard Biener  <rguenther@suse.de>
2745
2746         PR tree-optimization/85615
2747         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2748         to loops not nested in BBs loop father to avoid creating multi-entry
2749         loops.
2750
2751 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2752
2753         PR tree-optimization/70291
2754         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2755         arguments.  Change return type to tree.  Emit libcall as a new
2756         statement rather than replacing existing one when inplace_p is true.
2757         (expand_complex_multiplication_components): New function.
2758         (expand_complex_multiplication): Expand floating-point complex
2759         multiplication using the above.
2760         (expand_complex_division): Rename inner_type parameter to type.
2761         Update expand_complex_libcall call-site.
2762         (expand_complex_operations_1): Update expand_complex_multiplication
2763         and expand_complex_division call-sites.
2764
2765 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
2766
2767         PR target/85582
2768         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2769         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2770         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2771         the highest significant bit of the shift count mask is clear.  In
2772         check whether and[sq]i3 is needed verify that all significant bits
2773         of the shift count other than the highest are set.
2774
2775 2018-05-02  Tom de Vries  <tom@codesourcery.com>
2776
2777         PR libgomp/82428
2778         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2779         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2780         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2781         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2782         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2783         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2784         __builtin_goacc_parlevel_size.
2785
2786 2018-05-02  Richard Biener  <rguenther@suse.de>
2787
2788         PR tree-optimization/85597
2789         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2790         do not use split vect_get_vec_defs call but call vect_get_slp_defs
2791         directly.
2792
2793 2018-05-02  Tom de Vries  <tom@codesourcery.com>
2794
2795         PR testsuite/85106
2796         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2797         dump files): Add ltrans-tree.
2798
2799 2018-05-02  Tom de Vries  <tom@codesourcery.com>
2800
2801         PR testsuite/85106
2802         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2803         dump files): Add wpa-ipa.
2804
2805 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
2806
2807         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
2808         powerpc*-*-linux*paired* target.
2809         * config/rs6000/750cl.h: Delete.
2810         * config/rs6000/paired.h: Delete.
2811         * config/rs6000/paired.md: Delete.
2812         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2813         float support.
2814         * config/rs6000/rs6000-builtin.def: Remove paired float support.
2815         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2816         comment.  Remove paired float support.
2817         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2818         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2819         VECTOR_PAIRED.
2820         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2821         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2822         declarations.
2823         * config/rs6000/rs6000.c: Remove paired float support.
2824         (paired_expand_vector_init, paired_expand_vector_move,
2825         paired_emit_vector_compare, paired_emit_vector_cond_expr,
2826         (paired_expand_lv_builtin, paired_expand_stv_builtin,
2827         paired_expand_builtin, paired_expand_predicate_builtin,
2828         paired_init_builtins): Delete.
2829         * config/rs6000/rs6000.h: Remove paired float support.
2830         * config/rs6000/rs6000.md: Remove paired float support.
2831         (move_from_CR_ov_bit): Delete.
2832         * config/rs6000/rs6000.opt (mpaired): Delete.
2833         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2834         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2835
2836 2018-05-02  Richard Biener  <rguenther@suse.de>
2837
2838         PR middle-end/85567
2839         * gimplify.c (gimplify_save_expr): When in SSA form allow
2840         SAVE_EXPRs to compute to SSA vars.
2841
2842 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
2843
2844         PR target/85582
2845         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2846         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2847         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2848         clobber operands[2], instead use a new pseudo.  Formatting fixes.
2849
2850 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
2851
2852         PR tree-optimization/85586
2853         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2854         exit early for statements in the same group if the accesses are
2855         not strided.
2856
2857 2018-05-02  Tom de Vries  <tom@codesourcery.com>
2858
2859         PR lto/85451
2860         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2861         error message.
2862
2863 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
2864
2865         PR tree-optimization/85143
2866         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2867
2868 2018-05-01  Tom de Vries  <tom@codesourcery.com>
2869
2870         PR lto/85451
2871         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2872         not found" error message.
2873
2874 2018-05-01  Tom de Vries  <tom@codesourcery.com>
2875
2876         PR other/83786
2877         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2878         * vec.c (test_ordered_remove_if): New function.
2879         (vec_c_tests): Call test_ordered_remove_if.
2880         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2881         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2882         * tree-vect-patterns.c (vect_pattern_recog_1): Use
2883         VEC_ORDERED_REMOVE_IF.
2884
2885 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2886
2887         PR tree-optimization/82665
2888         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2889         pointer subtraction where arguments come from a memchr call.
2890
2891 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
2892
2893         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2894         --push-state --as-needed and --pop-state instead of --as-needed and
2895         --no-as-needed if ld supports it.
2896         * configure: Regenerated.
2897
2898         PR web/85578
2899         * doc/install.texi2html: Replace _002d with - and _002a with * in
2900         generated html files using sed.
2901
2902 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2903
2904         PR c++/85523
2905         * gcc-rich-location.c (blank_line_before_p): New function.
2906         (use_new_line): New function.
2907         (gcc_rich_location::add_fixit_insert_formatted): New function.
2908         * gcc-rich-location.h
2909         (gcc_rich_location::add_fixit_insert_formatted): New function.
2910
2911 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2912
2913         * selftest.c (assert_streq): Rename "expected" and "actual" to
2914         "val1" and "val2".  Extend NULL-handling to cover both inputs
2915         symmetrically, while still requiring both to be non-NULL for a pass.
2916         * selftest.h (assert_streq): Rename "expected" and "actual" to
2917         "val1" and "val2".
2918         (ASSERT_EQ): Likewise.
2919         (ASSERT_EQ_AT): Likewise.
2920         (ASSERT_KNOWN_EQ): Likewise.
2921         (ASSERT_KNOWN_EQ_AT): Likewise.
2922         (ASSERT_NE): Likewise.
2923         (ASSERT_MAYBE_NE): Likewise.
2924         (ASSERT_MAYBE_NE_AT): Likewise.
2925         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
2926         the assertion to pass.
2927         (ASSERT_STREQ_AT): Likewise.
2928
2929 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
2930
2931         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2932         interaction with -pie.
2933
2934 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2935
2936         * selftest.h: Fix alphabetization of per-source-file selftest
2937         declarations.
2938
2939 2018-04-30  Jason Merrill  <jason@redhat.com>
2940
2941         PR c++/61982 - dead stores to destroyed objects.
2942         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2943         of clobber.
2944
2945 2018-04-30  Jason Merrill  <jason@redhat.com>
2946
2947         * tree.c (build_clobber): New.
2948         * tree.h: Declare it.
2949         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2950
2951 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2952
2953         * diagnostic-show-locus.c (layout::layout): Update for
2954         location_get_source_line returning a char_span.
2955         (struct char_span): Move to input.h.
2956         (struct correction): Update for fields in char_span becoming
2957         private.
2958         (struct source_line): Update for location_get_source_line
2959         returning a char_span.
2960         (layout::print_line): Likewise.
2961         * edit-context.c (edited_file::print_content): Likewise.
2962         (edited_file::print_diff_hunk): Likewise.
2963         (edited_file::print_run_of_changed_lines): Likewise.
2964         (edited_file::get_num_lines): Likewise.
2965         (edited_line::edited_line): Likewise.
2966         * final.c (asm_show_source): Likewise.
2967         * input.c (location_get_source_line): Convert return type
2968         from const char * to char_span, losing the final "line_len"
2969         param.
2970         (dump_location_info): Update for the above.
2971         (get_substring_ranges_for_loc): Likewise.  Use a char_span
2972         when handling the literal within the line.
2973         (test_reading_source_line): Update for location_get_source_line
2974         returning a char_span.
2975         * input.h (class char_span): Move here from
2976         diagnostic-show-locus.c, converting from a struct to a class.
2977         Make data members private.
2978         (char_span::operator bool): New.
2979         (char_span::length): New.
2980         (char_span::get_buffer): New.
2981         (char_span::operator[]): New.
2982         (char_span::subspan): Make const.
2983         (char_span::xstrdup): New.
2984         (location_get_source_line): Convert return type from const char *
2985         to char_span, losing the final "line_size" param.
2986
2987 2018-04-30  Jan Hubicka  <jh@suse.cz>
2988
2989         * lto-wrapper.c (ltrans_priorities): New static var.
2990         (cmp_priority): New.
2991         (run_gcc): Read priorities and if doing parallel build order
2992         the Makefile by them.
2993
2994 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2995
2996         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2997
2998 2018-04-30  Richard Biener  <rguenther@suse.de>
2999
3000         * tree-cfg.c (verify_address): Remove base argument, add
3001         flag whether to check TREE_ADDRESSABLE and do that.
3002         (verify_expr): Remove.
3003         (verify_types_in_gimple_reference): Add pieces from verify_expr.
3004         (verify_gimple_assign_single): Likewise.
3005         (verify_gimple_switch): Likewise.
3006         (verify_expr_location_1): Dereference tp once.  Add (disabled)
3007         piece from verify_expr.
3008         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3009
3010 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3011
3012         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3013
3014 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3015
3016         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3017         (small_data_pattern): Likewise.
3018         (arc_rewrite_small_data): Likewise.
3019         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3020         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3021         (get_symbol_alignment): New function.
3022         (legitimate_small_data_address_p): Likewise.
3023         (legitimate_scaled_address): Update, call
3024         legitimate_small_data_address_p.
3025         (output_sdata): New static variable.
3026         (arc_print_operand): Update how we handle small data operands.
3027         (arc_print_operand_address): Likewise.
3028         (arc_legitimate_address_p): Update, use
3029         legitimate_small_data_address_p.
3030         (arc_rewrite_small_data_p): Remove.
3031         (arc_rewrite_small_data_1): Likewise.
3032         (arc_rewrite_small_data): Likewise.
3033         (small_data_pattern): Likewise.
3034         (compact_sda_memory_operand): Update to use
3035         legitimate_small_data_address_p and get_symbol_alignment.
3036         (prepare_move_operands): Don't rewite sdata pattern.
3037         (prepare_extend_operands): Remove.
3038         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3039         pattern.
3040         (zero_extendqisi2): Likewise.
3041         (zero_extendhisi2): Likewise.
3042         (extendqihi2): Likewise.
3043         (extendqisi2): Likewise.
3044         (extendhisi2): Likewise.
3045         (addsi3): Likewise.
3046         (subsi3): Likewise.
3047         (andsi3): Likewise.
3048         * config/arc/constraints.md (Usd): Change it to memory constraint.
3049
3050 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3051
3052         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3053         as source of std instructions.
3054         * config/arc/arc.md (movsi_insn): Update pattern predicate to
3055         allow 6-bit constants as source for store instructions.
3056         (movdi_insn): Update instruction pattern to allow 6-bit constants
3057         as source for store instructions.
3058
3059 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
3060
3061         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3062
3063 2018-04-30  Nathan Sidwell  <nathan@acm.org>
3064             Sandra Loosemore <sandra@codesourcery.com>
3065
3066         * dumpfile.c (dump_open): Allow '-' for stdout.
3067         * doc/invoke.texi (Developer Options): Document dump filename
3068         determination early.  Document stdin/stdout selection.
3069
3070 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
3071
3072         Microblaze Target: PIC data text relative
3073
3074         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3075         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3076         Add declaration.
3077         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3078         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3079         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3080         New addressing mode for data-text relative position indepenedent code.
3081         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3082         'ADDRESS_SYMBOLIC_TXT_REL'.
3083         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3084         (microblaze_legitimate_pic_operand): Exclude function calls from
3085         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3086         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3087         addresses cases.
3088         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3089         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3090         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3091         for 'address + offset'.
3092         (microblaze_expand_prologue): Add new function prologue call for
3093         'r20' assignation.
3094         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3095         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3096         table in case of TARGET_PIC_DATA_TEXT_REL.
3097         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3098         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3099         Add new macros 'UNSPEC_TEXT',
3100         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3101         + exclude function calls from 'UNSPEC_PLT' in case of data text
3102         relative mode.
3103         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3104         new target hook for generating address diff vector tables in case of
3105         flag_pic.
3106         * doc/tm.texi : Regenerate.
3107         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3108         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3109         of addr diff vector generation.
3110         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3111         target hook definition.
3112         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3113         Add default function for generate_pic_addr_diff_vec -> flag_pic.
3114         * doc/invoke.texi (Add new pic option): Add new microblaze pic
3115         option for data text relative.
3116
3117 2018-04-30  Richard Biener  <rguenther@suse.de>
3118
3119         * tree-chrec.h (evolution_function_is_constant_p): Remove
3120         redundant check.
3121         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3122
3123 2018-04-30  Richard Biener  <rguenther@suse.de>
3124
3125         PR bootstrap/85571
3126         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3127
3128 2018-04-30  Richard Biener  <rguenther@suse.de>
3129
3130         PR tree-optimization/28364
3131         PR tree-optimization/85275
3132         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3133         copying first exit test.
3134
3135 2018-04-28  Mark Wielaard  <mark@klomp.org>
3136
3137         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3138         dwarf_version >= 5.
3139         (dwarf_AT): Handle DW_AT_addr_base.
3140         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3141
3142 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
3143
3144         PR target/84431
3145         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3146         (*ashl<dwi>3_doubleword_mask_1): Ditto.
3147         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3148         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3149
3150 2018-04-28  Richard Biener  <rguenther@suse.de>
3151
3152         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3153         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3154         to reflect use.  Only add interesting stmts.
3155
3156 2018-04-27  Martin Jambor  <mjambor@suse.cz>
3157
3158         PR ipa/85549
3159         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3160         the jump function allows for passing through aggregate values.
3161
3162 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
3163
3164         * input.h (in_system_header_at): Convert from macro to inline
3165         function.
3166         (from_macro_expansion_at): Likewise.
3167         (from_macro_definition_at): Likewise.
3168
3169 2018-04-27  Jeff Law  <law@redhat.com>
3170
3171         * config.gcc: Mark tile* targets as deprecated/obsolete.
3172
3173 2018-04-27  Richard Biener  <rguenther@suse.de>
3174
3175         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3176         fix for ILP32.
3177
3178 2018-04-27  Richard Biener  <rguenther@suse.de>
3179
3180         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3181
3182 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
3183
3184         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3185         with Yd constraint. Set "preferred_for_speed" attribute from
3186         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3187         with Yd constraint.
3188         (*movdi_internal): Ditto.
3189         (movti_interunit splitters): Remove
3190         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3191         (movdi_interunit splitters): Ditto.
3192         * config/i386/constraints.md (Ye): Remove.
3193         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3194
3195 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3196
3197         PR target/85512
3198         * config/aarch64/constraints.md (Usg): Limit to 31.
3199         (Usj): Limit to 63.
3200
3201 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
3202
3203         PR tree-optimization/85529
3204         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3205         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3206         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3207         zero extension or masking of the MSB bit.
3208         (optimize_range_tests): Add FIRST_BB argument, pass it through
3209         to optimize_range_tests_var_bound.
3210         (maybe_optimize_range_tests, reassociate_bb): Adjust
3211         optimize_range_tests callers.
3212
3213 2018-04-26  Richard Biener  <rguenther@suse.de>
3214             Jakub Jelinek  <jakub@redhat.com>
3215
3216         * cgraph.h (symbol_table): Just declare debug method here.
3217         * symtab.c (symbol_table::debug): Define.
3218
3219 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
3220
3221         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3222
3223 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
3224
3225         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3226         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3227         (*movdi_internal): Substitute Yi and Yj constraint with x
3228         and Ym and Yn constraint with y constraint.  Update "isa"
3229         attribute and set "preferred_for_speed" attribute from
3230         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3231         (*movsi_internal): Ditto.
3232         (*movdf_internal): Ditto.
3233         (*movsf_internal): Ditto.
3234         (*zero_extendsidi2): Ditto.
3235         * config/i386/sse.md (vec_set<mode>_0): Ditto.
3236         (sse2_loadld): Ditto.
3237         (*vec_extract<ssevecmodelower>_0): Ditto.
3238         (*vec_extractv4si_0_zext_sse4): Ditto.
3239         (vec_concatv2di): Ditto.
3240         (*vec_dup<mode>): Ditto.
3241         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3242         * config/i386/constraints.md (Yi): Remove.
3243         (Yj): Remove.
3244         (Ym): Remove.
3245         (Yn): Remove.
3246
3247 2018-04-26  Nathan Sidwell  <nathan@acm.org>
3248
3249         * dumpfile.c (dump_open): New.
3250         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3251         (dump_finish): Detect stdio/stderr by value not name.
3252
3253 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
3254
3255         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3256
3257 2018-04-26  Tom de Vries  <tom@codesourcery.com>
3258
3259         PR target/84952
3260         * config/nvptx/nvptx.c (verify_neutering_jumps)
3261         (verify_neutering_labels): New function
3262         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3263
3264 2018-04-26  Tom de Vries  <tom@codesourcery.com>
3265
3266         PR target/84025
3267         * config/nvptx/nvptx.c (needs_neutering_p): New function.
3268         (nvptx_single): Use needs_neutering_p to skip over insns that do not
3269         need neutering.
3270
3271 2018-04-26  Richard Biener <rguenther@suse.de>
3272             Tom de Vries  <tom@codesourcery.com>
3273
3274         PR lto/85422
3275         * lto-streamer-out.c (output_function): Fixup loops if required to match
3276         discovery done in the reader.
3277
3278 2018-04-26  Richard Biener  <rguenther@suse.de>
3279
3280         PR tree-optimization/85116
3281         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3282         have a loop exit from the single latch predecessor.  Remove
3283         case of header with just condition.
3284         (ch_base::copy_headers): Exclude infinite loops from any
3285         processing.
3286         (pass_ch::execute): Record exits.
3287
3288 2018-04-26  Richard Biener  <rguenther@suse.de>
3289
3290         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3291         prologue cost vector and pass it to vect_get_load_cost.
3292         (vect_get_peeling_costs_all_drs): Likewise.
3293         (vect_peeling_hash_get_lowest_cost): Likewise.
3294         (vect_enhance_data_refs_alignment): Likewise.
3295
3296 2018-04-26  Richard Biener  <rguenther@suse.de>
3297
3298         PR middle-end/85450
3299         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3300         checking of integer<->pointer conversions.
3301         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3302         sign-/zero-extending pointer types.
3303         (expand_omp_for_static_chunk): Likewise.
3304
3305 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
3306             Jean Lee  <xiaoyur347@gmail.com>
3307
3308         * config/mips/mips.c (mips_asan_shadow_offset): New function.
3309         (TARGET_ASAN_SHADOW_OFFSET): Define.
3310         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3311         true for -fsanitize=address.
3312
3313 2018-04-25  Mark Wielaard  <mark@klomp.org>
3314
3315         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3316         shorter ones.
3317
3318 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3319
3320         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3321         than "alu", remove explicit "memory" and "imm_disp" attributes.
3322         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3323
3324         PR middle-end/85414
3325         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3326         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3327         gen_lowpart_no_emit.
3328
3329 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
3330
3331         PR target/85473
3332         * config/i386/i386.c (ix86_expand_builtin): Change memory
3333         operand to XI, extend p0 to Pmode.
3334         * config/i386/i386.md: Change unspec volatile and operand
3335         1 mode to XI, change operand 0 mode to P.
3336
3337 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3338
3339         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3340         GET_MODE_MASK before any checking.
3341         (nds32_can_use_bset_p): Likewise.
3342         (nds32_can_use_btgl_p): Likewise.
3343
3344 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3345
3346         * config/nds32/nds32-doubleword.md: New define_split pattern for
3347         illegal register number.
3348
3349 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3350
3351         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3352
3353 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3354
3355         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3356
3357 2018-04-25  Richard Biener  <rguenther@suse.de>
3358
3359         * lto-streamer.h (LTO_major_version): Bump to 8.
3360
3361 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3362
3363         * BASE-VER: Set to 9.0.0.
3364
3365 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
3366
3367         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3368         in __abskf2 and __powikf2.
3369
3370 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3371
3372         PR target/85512
3373         * config/aarch64/constraints.md (Usg, Usj): New constraints.
3374         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3375         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3376         Use the above on operand 2.  Reindent.
3377         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3378
3379 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
3380
3381         PR target/85485
3382         * common/config/i386/i386-common.c (ix86_handle_option): Don't
3383         handle OPT_mcet.
3384         * config/i386/i386.opt (mcet): Removed.
3385         * doc/install.texi: Remove -mcet documentation.
3386         * doc/invoke.texi: Likewise.
3387
3388 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
3389
3390         PR target/85485
3391         * doc/install.texi: Remove -mcet from bootstrap-cet.
3392
3393 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3394
3395         PR target/85511
3396         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3397         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3398         if TARGET_64BIT.
3399
3400         PR target/85503
3401         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3402         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3403         containing a CONST_VECTOR.
3404
3405 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
3406
3407         * doc/install.texi: Update newlib dependency for nvptx.
3408
3409 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3410
3411         PR target/85508
3412         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3413         instead of INTVAL when shifting x left.
3414
3415 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
3416
3417         PR tree-optimization/85478
3418         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3419         vect_grouped_store_supported for single element vectors.
3420
3421 2018-04-24  Richard Biener  <rguenther@suse.de>
3422
3423         PR target/85491
3424         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3425         load cost increase to the case of non-constant step.
3426
3427 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3428
3429         PR target/84828
3430         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3431         destination if any_malformed_asm.
3432
3433 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
3434
3435         PR middle-end/85496
3436         * expr.c (store_field): In the bitfield case, if the value comes from
3437         a function call and is returned in registers by means of a PARALLEL,
3438         do not change the mode of the temporary unless BLKmode and VOIDmode.
3439
3440 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
3441
3442         PR rtl-optimization/85423
3443         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3444         dependencies to debug insns when the previous insn is non-debug.
3445
3446 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
3447
3448         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3449         enums into a single definition.
3450         (fls): Fix predicates and printing.
3451         (seti): Likewise.
3452
3453 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
3454
3455         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3456         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3457         and short u6 immediate.
3458         (check_if_valid_sleep_operand): Remove.
3459         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3460
3461 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3462
3463         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3464         flag_always_save_lp condition.
3465         * config/nds32/nds32.opt (malways-save-lp): New option.
3466
3467 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
3468
3469         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3470         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3471         * config/nds32/nds32.h
3472         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3473         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3474
3475 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
3476
3477         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3478         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3479
3480 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
3481             Chung-Ju Wu  <jasonwucj@gmail.com>
3482
3483         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3484         Declare.
3485         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3486         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3487
3488 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3489
3490         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3491
3492 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3493
3494         * config/nds32/nds32-protos.h (nds32_data_alignment,
3495         nds32_local_alignment): Declare.
3496         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3497         nds32_local_alignment): New functions.
3498         (TARGET_CONSTANT_ALIGNMENT): Define.
3499         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3500
3501 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3502
3503         * config/nds32/nds32.c
3504         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3505         (TARGET_MODES_TIEABLE_P): Likewise.
3506
3507 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3508
3509         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3510         level Ofast and Og.
3511
3512 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
3513             Chung-Ju Wu  <jasonwucj@gmail.com>
3514
3515         * config/nds32/constants.md (unspec_volatile_element): Add enum values
3516         for unaligned access.
3517         * config/nds32/nds32-intrinsic.c: Implementation of expanding
3518         unaligned access.
3519         * config/nds32/nds32-intrinsic.md: Likewise.
3520         * config/nds32/nds32_intrinsic.h: Likewise.
3521         * config/nds32/nds32.h (nds32_builtins): Likewise.
3522         * config/nds32/nds32.opt (munaligned-access): New option.
3523         * config/nds32/nds32.c (nds32_asm_file_start): Display
3524         flag_unaligned_access status.
3525
3526 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
3527
3528         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3529         -mno-relax is present.
3530         * config/riscv/linux.h (LINK_SPEC): Ditto.
3531
3532 2018-04-20  Martin Sebor  <msebor@redhat.com>
3533
3534         PR c/85365
3535         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3536         for null pointers.
3537         (gimple_fold_builtin_stxcpy_chk): Same.
3538         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3539
3540 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
3541
3542         PR target/85456
3543         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3544         __powikf2 when long double is IEEE 128-bit.
3545
3546 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
3547
3548         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3549         step to make sure stack always aligned.
3550
3551 2018-04-20  Carl Love  <cel@us.ibm.com>
3552
3553         PR target/83402
3554         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3555         size check for arg0.
3556
3557 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
3558             Tom de Vries  <tom@codesourcery.com>
3559
3560         PR target/85445
3561         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3562         Emit insns for calls too.
3563         (nvptx_find_par): Always look for worker-level predecessor insn.
3564         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
3565         calls.
3566         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3567         (nvptx_process_pars): Propagate frames for calls.
3568
3569 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
3570
3571         PR target/85469
3572         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3573         Removed.
3574         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3575         (ix86_handle_option): Don't handle OPT_mibt.
3576         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3577         __SHSTK__.
3578         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3579         has_ibt and ibt.
3580         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3581         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3582         (ix86_target_macros): Define __CET__ with flag_cf_protection
3583         for -fcf-protection.
3584         * config/i386/i386.c (isa2_opts): Remove -mibt.
3585         * config/i386/i386.h (TARGET_IBT): Removed.
3586         (TARGET_IBT_P): Likewise.
3587         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3588         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3589         * config/i386/i386.opt (mcet): Update help message.
3590         (mshstk): Likewise.
3591         (mibt): Removed.
3592         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
3593         -mcet as an alias for -mshstk.
3594
3595 2018-04-20  Richard Biener <rguenther@suse.de>
3596
3597         PR middle-end/85475
3598         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3599         complexity by forcing a single use of the multiply operand.
3600
3601 2018-04-20  Martin Jambor  <mjambor@suse.cz>
3602
3603         ipa/85449
3604         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3605         recursion dependency to only apply to non-clones.
3606
3607 2018-04-20  Martin Jambor  <mjambor@suse.cz>
3608
3609         ipa/85447
3610         * ipa-cp.c (create_specialized_node): Check that clones of
3611         self-recursive edges exist during IPA-CP.
3612
3613 2018-04-19  Toon Moene  <toon@moene.org>
3614
3615         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3616         by -O3.
3617
3618 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
3619
3620         PR tree-optimization/85467
3621         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3622         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
3623         VECTOR_CST element to type.
3624
3625 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
3626
3627         PR target/85397
3628         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3629         * config/i386/i386.md (builtin_setjmp_setup): Removed.
3630         (builtin_longjmp): Likewise.
3631         (save_stack_nonlocal): New pattern.
3632         (restore_stack_nonlocal): Likewise.
3633
3634 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
3635
3636         PR target/85404
3637         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3638         Replace ASM_OUTPUT_LABEL with fprintf.
3639
3640 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
3641
3642         PR target/85417
3643         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3644         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3645         * config/i386/i386-c.c (ix86_target_macros_internal): Also
3646         define __IBT__ and __SHSTK__ for -fcf-protection.
3647         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3648         TARGET_IBT.
3649         (ix86_trampoline_init): Likewise.
3650         (x86_output_mi_thunk): Likewise.
3651         (ix86_notrack_prefixed_insn_p): Likewise.
3652         (ix86_option_override_internal): Don't disallow -fcf-protection.
3653         * config/i386/i386.md (rdssp<mode>): Also enable for
3654         -fcf-protection.
3655         (incssp<mode>): Likewise.
3656         (nop_endbr): Likewise.
3657         * config/i386/i386.opt (mcet): Change help message to built-in
3658         functions only.
3659         (mibt): Likewise.
3660         (mshstk): Likewise.
3661         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3662         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
3663         enable CET built-in functions.
3664
3665 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
3666
3667         * common/config/i386/i386-common.c
3668         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3669         OPTION_MASK_ISA_MOVDIRI_UNSET,
3670         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3671         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3672         * config.gcc (movdirintrin.h): New header.
3673         * config/i386/cpuid.h (bit_MOVDIRI,
3674         bit_MOVDIR64B): New bits.
3675         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3676         and -mmvodir64b.
3677         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3678         (VOID, PVOID, PCVOID)): New function types.
3679         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3680         __builtin_ia32_directstoreu_u64,
3681          __builtin_ia32_movdir64b): New builtins.
3682         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3683         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3684         and -mmovdiri.
3685         (ix86_valid_target_attribute_inner_p): Ditto.
3686         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3687         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3688         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3689         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3690         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3691         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3692         (movdiri<mode>, movdir64b_<mode>): New.
3693         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3694         * config/i386/immintrin.h: Include movdirintrin.h.
3695         * config/i386/movdirintrin.h: New file.
3696         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3697
3698 2018-04-19  Richard Biener  <rguenther@suse.de>
3699
3700         PR middle-end/85455
3701         * cfg.c (clear_bb_flags): When loop state says we have
3702         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3703
3704 2018-04-19  Richard Biener  <rguenther@suse.de>
3705
3706         PR tree-optimization/84737
3707         * tree-vect-data-refs.c (vect_copy_ref_info): New function
3708         copying restrict info.
3709         (vect_setup_realignment): Use it.
3710         * tree-vectorizer.h (vect_copy_ref_info): Declare.
3711         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3712         the first DR to all generated stores.
3713         (vectorizable_load): Likewise for loads.
3714
3715 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
3716
3717         PR tree-optimization/85446
3718         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3719         the integral and pointer types to have the same precision.
3720
3721         * doc/install.texi: Document --disable-cet being the default and
3722         --enable-cet=auto.
3723
3724 2018-04-18  Martin Liska  <mliska@suse.cz>
3725
3726         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3727         style.
3728
3729 2018-04-18  Martin Liska  <mliska@suse.cz>
3730
3731         Revert
3732         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
3733
3734         PR ipa/83983
3735         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3736         arguments if they are comparable.
3737
3738 2018-04-18  Martin Liska  <mliska@suse.cz>
3739
3740         Revert
3741         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
3742
3743         PR lto/84805
3744         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3745         incomplete types.
3746
3747 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
3748
3749         PR target/85388
3750         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3751         ENDBR after calling __morestack.
3752
3753 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
3754
3755         PR jit/85384
3756         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3757         by using gcc_base_ver to generate a gcc_driver_version, and use
3758         it when generating GCC_DRIVER_NAME.
3759         * configure: Regenerate.
3760
3761 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
3762
3763         PR target/81084
3764         * config.gcc: Obsolete powerpc*-*-*spe*.
3765
3766 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
3767
3768         PR debug/84637
3769         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3770         (stabstr_D): Change type of unum from unsigned int to
3771         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
3772         type.
3773
3774 2018-04-17  Jim Wilson  <jimw@sifive.com>
3775
3776         PR 84856
3777         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3778         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3779         Set arg_pointer_offset after using pretend_args_size.
3780
3781 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
3782
3783         PR rtl-optimization/85431
3784         * dse.c (record_store): Ignore zero width stores.
3785
3786         PR sanitizer/85230
3787         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
3788         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3789         __builtin_stack_restore rather than after it.
3790         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3791         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3792         argument instead of virtual_dynamic_stack_rtx.
3793
3794 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3795
3796         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3797         New prototype.
3798         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3799         Add note to error message to explain internal mapping of overloaded
3800         built-in function name to non-overloaded built-in function name.
3801         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3802         function.
3803
3804 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
3805
3806         PR target/85424
3807         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3808         where the inputs overlap with the output.
3809
3810 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
3811
3812         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3813         (=v, v) alternative and explicit "memory" attribute.
3814         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
3815         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3816         attributes.
3817         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3818         "sselog1" type instead of "sselog".
3819         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3820         "sselog".  Remove explicit "memory" attribute.
3821         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3822         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3823         attributes.
3824         (vec_extract_hi_v32hi): Merge all alternatives into one, use
3825         "sselog1" type instead of "sselog".  Remove explicit "memory"
3826         attribute.
3827         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3828         use "sselog1" type instead of "sselog".  Remove explicit "memory"
3829         attribute.
3830         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3831         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3832         attributes.
3833         (vec_extract_hi_v64qi): Merge all alternatives into one, use
3834         "sselog1" type instead of "sselog".  Remove explicit "memory"
3835         attribute.
3836         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3837         use "sselog1" type instead of "sselog".  Remove explicit "memory"
3838         attribute.
3839
3840         PR target/85430
3841         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3842
3843         PR middle-end/85414
3844         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3845         on a SUBREG.
3846
3847 2018-04-17  Martin Jambor  <mjambor@suse.cz>
3848
3849         PR ipa/85421
3850         * ipa-cp.c (create_specialized_node): Call
3851         expand_all_artificial_thunks if necessary.
3852
3853 2018-04-17  Martin Liska  <mliska@suse.cz>
3854
3855         PR lto/85405
3856         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3857         in message, remote space in between '_G' and '('.
3858
3859 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
3860
3861         PR target/85281
3862         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3863         avx512f_vmcmp<mode>3<round_saeonly_name>,
3864         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3865         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3866         avx512f_rndscale<mode><round_saeonly_name>,
3867         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3868         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3869         Use %<iptr>2 instead of %2 for -masm=intel.
3870         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3871         avx512f_vcvttss2usi<round_saeonly_name>,
3872         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3873         -masm=intel.
3874         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3875         avx512f_vcvttsd2usi<round_saeonly_name>,
3876         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3877         Use %q1 instead of %1 for -masm=intel.
3878         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3879         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3880         of %3 for -masm=intel.
3881         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3882         -masm=intel.
3883         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3884         -masm=intel.
3885         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3886         -masm=intel.
3887         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3888         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3889         %g1.
3890         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3891         -masm=intel.
3892         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3893         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3894         %g1 and one with %0 and %1.
3895         (avx512er_vmrcp28<mode><round_saeonly_name>,
3896         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3897         %1 for -masm=intel.
3898         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3899         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3900         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3901         of %0 and %{%4%} for -masm=intel.
3902         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3903         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3904         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3905         order of %0 and %{%5%}%{z%} for -masm=intel.
3906
3907 2018-04-17  Jan Hubicka  <jh@suse.cz>
3908
3909         PR lto/85405
3910         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3911
3912 2018-04-17  Martin Liska  <mliska@suse.cz>
3913
3914         PR ipa/85329
3915         * multiple_target.c (create_dispatcher_calls): Set apostrophes
3916         for target_clone error message.  Make default implementation
3917         clone to be a local declaration.
3918         (separate_attrs): Add new argument and check for an empty
3919         string.
3920         (expand_target_clones): Handle it.
3921         (ipa_target_clone): Make redirection just for target_clones
3922         functions.
3923
3924 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
3925             Tom de Vries  <tom@codesourcery.com>
3926
3927         PR middle-end/84955
3928         * omp-expand.c (expand_oacc_for): Add dummy false branch for
3929         tiled basic blocks without omp continue statements.
3930
3931 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
3932
3933         PR target/83660
3934         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3935         vec_extract expression as having side effects to make sure it gets
3936         a cleanup point.
3937
3938 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
3939
3940         PR target/85403
3941         * config/i386/i386.c (get_builtin_code_for_version): Check
3942         error_mark_node.
3943
3944 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
3945
3946         PR target/84331
3947         * gcc/config.gcc: Support "skylake".
3948         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3949         PROCESSOR_SKYLAKE.
3950         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3951         (processor_target_table): Add "skylake".
3952         (ix86_option_override_internal): Add "skylake".
3953         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3954         PROCESSOR_CANNONLAKE.
3955         (get_builtin_code_for_version): Fix priority for
3956         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3957         PROCESSOR_SKYLAKE-AVX512.
3958         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3959         (processor_type): Add PROCESSOR_SKYLAKE.
3960
3961 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
3962             Jason Merrill  <jason@redhat.com>
3963
3964         PR c++/85112
3965         * convert.c (convert_to_integer_1): Use direct recursion for
3966         enumeral types and types with a precision less than the number
3967         of bits in their mode.
3968
3969 2018-04-16  Julia Koval  <julia.koval@intel.com>
3970
3971         PR target/84413
3972         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3973         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3974
3975 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
3976
3977         PR target/85293
3978         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3979         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3980         and -mno-direct-move.
3981
3982 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
3983
3984         PR target/83402
3985         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3986         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3987         Ensure negative shifts result in {0}.
3988
3989 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
3990
3991         PR rtl-optimization/79916
3992         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3993         regs (if any) to define how to gnerate SD moves when LRA is in
3994         progress.
3995
3996 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
3997
3998         PR rtl-optimization/85393
3999         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4000         * except.c (expand_dw2_landing_pad_for_region): Make static.
4001         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4002         a label and unconditional jump to old_bb, rather than
4003         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4004         basic block.
4005
4006         PR rtl-optimization/85376
4007         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4008         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4009         instead of a specific value.
4010
4011 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
4012             Bin Cheng  <bin.cheng@arm.com>
4013
4014         PR tree-optimization/82965
4015         PR tree-optimization/83991
4016         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4017         by_profile_only parameter.
4018         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4019         information if the loop was predicted to iterate too many times.
4020         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4021
4022 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
4023
4024         PR lto/71991
4025         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4026         always inline.
4027
4028 2018-04-13  Martin Liska  <mliska@suse.cz>
4029             Jakub Jelinek  <jakub@redhat.com>
4030
4031         PR middle-end/81657
4032         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4033         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4034         * builtins.c (expand_builtin_memory_copy_args): Use
4035         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4036         handle dest_addr == pc_rtx.
4037
4038 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
4039
4040         PR target/85291
4041         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4042         asked to not generate direct moves.
4043         (fix_trunc<mode>si2_stfiwx): Similar.
4044         (fix_trunc<mode>si2_internal): Similar.
4045
4046 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4047
4048         PR debug/83157
4049         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4050         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4051         lookup if dest in some wider mode is known to be const0_rtx and
4052         if so, record permanent equivalence for it to be ZERO_EXTEND of
4053         the narrower mode destination.
4054
4055 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
4056
4057         * lto-streamer-out.c (output_function): Revert 259346.
4058         * omp-expand.c (expand_oacc_for): Likewise.
4059
4060 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
4061
4062         PR rtl-optimization/85354
4063         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4064         * sel-sched.c (sel_global_init): ... here.
4065
4066 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
4067
4068         PR target/85238
4069         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4070         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4071         mode for PE-COFF targets.
4072         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4073         (i386_pe_asm_lto_end): Likewise.
4074         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4075         (TARGET_ASM_LTO_END): Likewise.
4076         * config/i386/winnt.c (saved_debug_info_level): New static variable.
4077         (i386_pe_asm_lto_start): New function.
4078         (i386_pe_asm_lto_end): Likewise.
4079
4080 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
4081             Richard Biener  <rguenther@suse.de>
4082
4083         PR middle-end/84955
4084         * lto-streamer-out.c (output_function): Fix CFG loop state before
4085         streaming out.
4086         * omp-expand.c (expand_oacc_for): Handle calls to internal
4087         functions like regular functions.
4088
4089 2018-04-12  Richard Biener  <rguenther@suse.de>
4090
4091         PR lto/85371
4092         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4093         for the early LTO debug to properly generate references to it
4094         during DIE emission.  Do not re-use that for the skeleton for
4095         split-dwarf.
4096         (dwarf2out_early_finish): Likewise.
4097
4098 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4099
4100         PR target/85328
4101         * config/i386/sse.md
4102         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4103         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4104         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4105         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4106         and output is a reg, avoid creating invalid lowpart subreg, but
4107         instead split into a 512-bit move.  Don't split if not AVX512VL,
4108         input is xmm16+ reg and output is a mem.
4109         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4110         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4111         xmm16+ reg and output is a mem.
4112
4113 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4114
4115         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4116         also for flag_dwarf2_cfi_asm.
4117
4118 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4119
4120         PR rtl-optimization/85342
4121         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4122         a bool scalar var inside of the loop instead.  Don't try to update
4123         recog_data.operand after failed apply_change_group.
4124
4125 2018-04-12  Tom de Vries  <tom@codesourcery.com>
4126
4127         PR target/85296
4128         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4129         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
4130         array with flexible array member as array without given dimension.
4131         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4132         argument for undefined param to true.
4133
4134 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
4135
4136         PR target/85321
4137         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4138         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4139         from PowerPC section.
4140         * config/rs6000/sysv4.opt (mcall-): Improve help text.
4141         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4142         help text that is too long.
4143         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4144         help text that is too long.
4145         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4146         help text that is too long.
4147
4148 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
4149
4150         * config/alpha/alpha.md (stack_probe_internal): Rename
4151         from "probe_stack".  Update all callers.
4152
4153 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4154
4155         PR rtl-optimization/84566
4156         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4157         sched_macro_fuse_insns.
4158
4159 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4160
4161         PR target/84301
4162         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4163         (compute_block_dependences): ... from here.
4164
4165 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4166
4167         PR tree-optimization/85331
4168         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4169         from int to HOST_WIDE_INT.
4170
4171 2018-04-11  Martin Jambor  <mjambor@suse.cz>
4172
4173         PR ipa/84149
4174         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4175         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4176         not the same as the source val.
4177         (cgraph_edge_brings_value_p): New parameter.
4178         (gather_edges_for_value): Pass destination value to
4179         cgraph_edge_brings_value_p.
4180         (perhaps_add_new_callers): Likewise.
4181         (get_info_about_necessary_edges): Likewise and exclude values brought
4182         only by self-recursive edges.
4183         (create_specialized_node): Redirect only clones of self-calling edges.
4184         (+self_recursive_pass_through_p): New function.
4185         (find_more_scalar_values_for_callers_subset): Use it.
4186         (find_aggregate_values_for_callers_subset): Likewise.
4187         (known_aggs_to_agg_replacement_list): Removed.
4188         (decide_whether_version_node): Re-calculate known constants for all
4189         remaining context clones.
4190
4191 2018-04-11  Richard Biener  <rguenther@suse.de>
4192
4193         PR lto/85339
4194         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4195         from early DWARF output.
4196         (dwarf2out_early_finish): Output line info unconditionally into
4197         early DWARF and add reference to it.
4198
4199 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4200
4201         PR target/85281
4202         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4203         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4204         other than V2DFmode using iptr mode attribute.
4205         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4206
4207 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4208
4209         PR rtl-optimization/84659
4210         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4211
4212 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4213
4214         PR debug/85302
4215         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4216         SIZEP is NULL.
4217         (output_loc_list): Pass address of a dummy size variable even in the
4218         locview handling loop.
4219         (index_location_lists): Add comment on why skip_loc_list_entry can't
4220         call size_of_locs.
4221
4222 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
4223
4224         PR target/85261
4225         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4226         into register.
4227
4228 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
4229
4230         PR target/85321
4231         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4232         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4233         and -mstring-compare-inline-limit.
4234
4235 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4236
4237         PR target/85287
4238         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4239         for stack clash protection in a register whenever we need it to be in
4240         a register.
4241
4242 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4243
4244         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4245         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4246
4247 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4248
4249         PR target/85321
4250         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4251         the help text.
4252         (mlong-double-): Ditto.
4253         * config/rs6000/sysv4.opt (msdata=): Ditto.
4254         (mtls-size=): Ditto.
4255
4256 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4257
4258         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4259         erroneous entries for
4260         "vector int vec_ldl (int, long int *)", and
4261         "vector unsigned int vec_ldl (int, unsigned long int *)".
4262         Add comments and entries for
4263         "vector bool char vec_ldl (int, bool char *)",
4264         "vector bool short vec_ldl (int, bool short *)",
4265         "vector bool int vec_ldl (int, bool int *)",
4266         "vector bool long long vec_ldl (int, bool long long *)",
4267         "vector pixel vec_ldl (int, pixel *)",
4268         "vector long long vec_ldl (int, long long *)",
4269         "vector unsigned long long vec_ldl (int, unsigned long long *)".
4270         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4271         type tree bool_long_long_type_node and correct definition of
4272         bool_V2DI_type_node to make reference to this new type tree.
4273         (rs6000_mangle_type): Replace erroneous reference to
4274         bool_long_type_node with bool_long_long_type_node.
4275         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4276         comments to emphasize sign distinctions for char and int types and
4277         replace RS6000_BTI_bool_long constant with
4278         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
4279         use of RS6000_BTI_pixel.
4280         (bool_long_type_node): Remove this macro definition.
4281         (bool_long_long_type_node): New macro definition
4282
4283 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4284
4285         PR rtl-optimization/85300
4286         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4287         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4288         simplify_unary_operation fails.
4289
4290 2018-04-10  Martin Liska  <mliska@suse.cz>
4291
4292         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4293         cgraph_edge and ipa_ref.
4294
4295 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4296
4297         PR target/85177
4298         PR target/85255
4299         * config/i386/sse.md
4300         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4301         computation of the VEC_MERGE selector from mask.
4302         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4303         Fix decoding of the VEC_MERGE selector into mask.
4304
4305 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4306
4307         PR tree-optimization/85286
4308         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4309
4310 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4311
4312         * final.c (final_1): Set insn_last_address as well as
4313         insn_current_address.
4314
4315 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4316
4317         PR target/85173
4318         * explow.c (emit_stack_probe): Call validize_mem on memory location
4319         before passing it to gen_probe_stack.  Create address operand and
4320         legitimize it for the probe_stack_address case.
4321
4322 2018-04-09  Jan Hubicka  <jh@suse.cz>
4323
4324         PR lto/85078
4325         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4326         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4327         * tree.c (free_lang_data_in_type): Fix handling of binfos;
4328         walk basetypes.
4329         (free_lang_data): Rebuild type inheritance graph.
4330
4331 2018-04-09  Martin Sebor  <msebor@redhat.com>
4332
4333         * invoke.texi (-finline-small-functions): Mention other optimization
4334         options.
4335         (-findirect-inlining, -fpartial-inlining): Same.
4336         (-finline-functions-called-once): Same.
4337         (-freorder-blocks-and-partition): Same.
4338
4339 2018-04-09  Jan Hubicka  <jh@suse.cz>
4340
4341         PR rtl/84058
4342         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4343         jumps; choose last target that matches the criteria (i.e.
4344         no partition changes for non-crossing jumps).
4345         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4346         support for redirecting crossing jumps to non-crossing.
4347
4348 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
4349
4350         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4351         also for naked functions.
4352
4353 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
4354
4355         * config/arc/arc.md (add_shift): New pattern.
4356         (add_shift2): Likewise.
4357         (sub_shift): Likewise.
4358         (sub_shift_cmp0_noout): Likewise.
4359         (compare_si_ashiftsi): Likewise.
4360         (xbfu_cmp0_noout): New combine pattern.
4361         (xbfu_cmp0"): Likewise.
4362         (movsi_set_cc_insn): Place the predicable variant first.
4363         (commutative_binary_cmp0_noout): Remove clobber.
4364         (commutative_binary_cmp0): New pattern.
4365         (noncommutative_binary_cmp0): Likewise.
4366         (noncommutative_binary_cmp0_noout): Likewise.
4367         (noncommutative_binary_comparison_result_used): Removed.
4368         (rsub_cmp0): New pattern.
4369         (rsub_cmp0_noout): Likewise.
4370         (extzvsi): Changed, keep only meaningful variants.
4371         (SQH, SEZ): New iterators.
4372         (SQH_postfix): New mode attribute.
4373         (SEZ_prefix): New code attribute.
4374         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4375         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4376         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4377         of numerical value.
4378         (noncommutative_operator): Check the availability of barrel
4379         shifter option.
4380
4381 2018-04-09  Richard Biener  <rguenther@suse.de>
4382
4383         PR tree-optimization/85284
4384         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4385         Only use the niter constraining form of simple_iv when the exit
4386         is always executed.
4387
4388 2018-04-09  Tom de Vries  <tom@codesourcery.com>
4389
4390         PR target/84041
4391         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4392         (define_expand "*memory_barrier"): New define_expand.
4393         (define_insn "memory_barrier"): New insn.
4394
4395 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4396
4397         PR rtl-optimization/80463
4398         PR rtl-optimization/83972
4399         PR rtl-optimization/83480
4400
4401         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4402         correct producer for the insn.
4403         (tidy_control_flow): Fixup seqnos in case of debug insns.
4404
4405 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4406
4407         PR rtl-optimization/83913
4408
4409         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4410         different sched-times when merging exprs.
4411
4412 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4413
4414         PR rtl-optimization/83962
4415
4416         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4417         tidy_fallthru_edge and tidy_control_flow.
4418
4419 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4420
4421         PR rtl-optimization/83530
4422
4423         * sel-sched.c (force_next_insn): New global variable.
4424         (remove_insn_for_debug): When force_next_insn is true, also leave only
4425         next insn in the ready list.
4426         (sel_sched_region): When the region wasn't scheduled, make another pass
4427         over it with force_next_insn set to 1.
4428
4429 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
4430
4431         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4432         into tm_file.
4433         * config/nds32/constants.md (unspec_volatile_element): Add enum values
4434         for interrupt control.
4435         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4436         functions for interrupt control.
4437         * config/nds32/nds32-intrinsic.md: Likewise.
4438         * config/nds32/nds32_intrinsic.h: Likewise.
4439         * config/nds32/nds32.h (nds32_builtins): Likewise.
4440
4441 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
4442
4443         * config/nds32/nds32.c (nds32_init_machine_status,
4444         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4445         strict_aligned_p field.
4446         (nds32_expand_to_rtl_hook): New function.
4447         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4448         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4449
4450 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4451             Chung-Ju Wu  <jasonwucj@gmail.com>
4452
4453         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4454         * config/nds32/nds32-n7.md: New file.
4455         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4456         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4457         pipeline.
4458         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4459         * config/nds32/nds32.md (pipeline_model): Add n7.
4460         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4461         * config/nds32/pipelines.md: Include n7 settings.
4462
4463 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4464             Chung-Ju Wu  <jasonwucj@gmail.com>
4465
4466         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4467         * config/nds32/nds32-e8.md: New file.
4468         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4469         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4470         pipeline.
4471         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4472         * config/nds32/nds32.md (pipeline_model): Add e8.
4473         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4474         * config/nds32/pipelines.md: Include e8 settings.
4475
4476 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4477             Chung-Ju Wu  <jasonwucj@gmail.com>
4478
4479         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4480         * config/nds32/nds32-n8.md: New file.
4481         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4482         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4483         pipeline.
4484         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4485         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4486         * config/nds32/nds32.md (pipeline_model): Add n8.
4487         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4488         * config/nds32/pipelines.md: Include n8 settings.
4489
4490 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4491             Chung-Ju Wu  <jasonwucj@gmail.com>
4492
4493         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4494         * config/nds32/nds32-n9-2r1w.md: New file.
4495         * config/nds32/nds32-n9-3r2w.md: New file.
4496         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4497         nds32_register_ports): New or modify for cpu n9.
4498         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4499         pipeline.
4500         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4501         * config/nds32/nds32-utils.c: New file.
4502         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4503         TARGET_MUL_SLOW): Define.
4504         * config/nds32/nds32.md (pipeline_model): New attribute.
4505         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4506         New options that support cpu n9.
4507         * config/nds32/pipelines.md: Include n9 settings.
4508         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4509
4510 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
4511
4512         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4513         information if necessary.
4514         (output_cond_branch_compare_zero): Likewise.
4515         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4516         (nds32_target_alignment): Refine for alignment.
4517         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4518         (FUNCTION_BOUNDARY): Modify.
4519         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4520         align case.
4521         * config/nds32/nds32.opt (malways-align, malign-functions): New.
4522
4523 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
4524
4525         * config/nds32/constants.md (unspec_volatile_element): Add values for
4526         TLB operation and data prefetch.
4527         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4528         functions for TLB operation and data prefetch.
4529         * config/nds32/nds32-intrinsic.md: Likewise.
4530         * config/nds32/nds32_intrinsic.h: Likewise.
4531         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4532         (nds32_print_operand): Likewise.
4533         * config/nds32/nds32.h (nds32_builtins): Likewise.
4534
4535 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
4536         Andrew Pinski <pinsika@gcc.gnu.org>
4537
4538         PR middle-end/82976
4539         * match.pd: Use constant_boolean_node of correct type instead of
4540         boolean_true_node or boolean_false_node for simplifying
4541         pointer comparisons to zero.
4542
4543 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
4544
4545         PR tree-optimization/80021
4546         * tree.c (verify_type_variant): Make error call in verify_variant_match
4547         translatable and remove final full stop.
4548
4549 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
4550
4551         * config/nds32/constants.md (unspec_volatile_element): Add
4552         UNSPEC_VOLATILE_EH_RETURN.
4553         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4554         nds32_output_stack_pop): Support dwarf exception handling process.
4555         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4556         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4557         exception handling process.
4558         (nds32_compute_stack_frame): Likewise.
4559         (nds32_return_addr_rtx): Likewise.
4560         (nds32_initial_elimination_offset): Likewise.
4561         (nds32_expand_prologue): Likewise.
4562         (nds32_expand_epilogue): Likewise.
4563         (nds32_dynamic_chain_address): New function.
4564         * config/nds32/nds32.h (machine_function): Add fields for dwarf
4565         exception handling.
4566         (DYNAMIC_CHAIN_ADDRESS): Define.
4567         (EH_RETURN_DATA_REGNO): Define.
4568         (EH_RETURN_STACKADJ_RTX): Define.
4569         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4570         patterns for dwarf exception handling.
4571
4572 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
4573
4574         * config/nds32/nds32.h: Clean up obsolete macros.
4575
4576 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4577
4578         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4579         Add enum values for particular instructions.
4580         * config/nds32/nds32-intrinsic.c: Implementation of expanding
4581         particular intrinsic functions.
4582         * config/nds32/nds32-intrinsic.md: Likewise.
4583         * config/nds32/nds32_intrinsic.h: Likewise.
4584         * config/nds32/nds32.h (nds32_builtins): Likewise.
4585         * config/nds32/nds32.md (type): Add pbsad and pbsada.
4586         (btst, ave): New patterns for particular instructions.
4587
4588 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4589
4590         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4591         Add enum values for atomic load/store and memory sync.
4592         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4593         and memory sync.
4594         * config/nds32/nds32-intrinsic.md: Likewise.
4595         * config/nds32/nds32_intrinsic.h: Likewise.
4596         * config/nds32/nds32.h (nds32_builtins): Likewise.
4597
4598 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
4599
4600         PR tree-optimization/85257
4601         * fold-const.c (native_encode_vector): If not all elts could fit
4602         and off is -1, return 0 rather than offset.
4603         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4604         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4605         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
4606         adjust buffer in native_interpret_expr call.
4607
4608 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4609
4610         * config/nds32/constants.md (unspec_volatile_element): Add cache
4611         control enum values.
4612         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4613         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4614         * config/nds32/nds32.c (nds32_cctl_names): New.
4615         (nds32_print_operand): Handle cache control register names.
4616         * config/nds32/nds32.h (nds32_builtins): New enum values.
4617         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4618         macros.
4619         * config/nds32/nds32.md (type): Add mmu.
4620         * config/nds32/pipelines.md (simple_insn): Add mmu.
4621
4622 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
4623
4624         * config/nds32/nds32.md (type): Remove call.
4625         * config/nds32/pipelines.md (simple_insn): Likewise.
4626
4627 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4628
4629         * config/nds32/constants.md (unspec_volatile_element): Add
4630         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4631         UNSPEC_VOLATILE_FMFCFG.
4632         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4633         description for fmfcfg and fmfcsr.
4634         (bdesc_1arg): Add fmtcsr.
4635         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4636         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4637         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4638         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4639         unspec_fmfcfg): New patterns.
4640         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4641         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4642         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4643         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4644         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4645         __nds32__fmfcfg): Define.
4646
4647 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4648
4649         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4650         intrinsic register names.
4651         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4652         intrinsic register enum values and macros.
4653
4654 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
4655
4656         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4657         for load/store addressing form.
4658         (nds32_print_operand_address): Likewise.
4659
4660 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
4661
4662         PR target/85196
4663         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4664         based on LABEL_REF.  Remove useless assertion.
4665         (pic_address_needs_scratch): Fix formatting.
4666         (sparc_legitimize_pic_address): Minor tweaks.
4667         (sparc_delegitimize_address): Adjust assertion accordingly.
4668         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4669         into symbolic_operand.
4670         (movsi_high_pic_label_ref): Likewise.
4671         (movsi_lo_sum_pic_label_ref): Likewise.
4672         (movdi_pic_label_ref): Likewise.
4673         (movdi_high_pic_label_ref): Likewise.
4674         (movdi_lo_sum_pic_label_ref): Likewise.
4675
4676 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
4677
4678         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4679         custom LIB_SPEC setup.
4680
4681 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
4682             Kito Cheng  <kito.cheng@gmail.com>
4683
4684         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4685         * config/riscv/freebsd.h: New.
4686
4687 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
4688
4689         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4690         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4691         file.
4692
4693 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
4694             Kito Cheng  <kito.cheng@gmail.com>
4695
4696         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4697         nds32_output_call, nds32_symbol_binds_local_p): New functions.
4698         * config/nds32/nds32-protos.h (nds32_output_call,
4699         nds32_output_return): Declare.
4700         * config/nds32/nds32.md: Refine all the call and return patterns.
4701
4702 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
4703
4704         PR debug/85252
4705         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4706         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4707
4708         PR rtl-optimization/84872
4709         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4710         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4711         EDGE_CROSSING edge.
4712
4713 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
4714
4715         * expr.c (copy_blkmode_to_reg): Revert 254862.
4716         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4717
4718 2018-04-06  Richard Biener  <rguenther@suse.de>
4719
4720         PR middle-end/85244
4721         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4722         after seeing a component reference with an adjacent field.  Treat
4723         refs to arrays at struct end of external decls similar to
4724         refs to unconstrained commons.
4725
4726 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
4727
4728         PR sanitizer/85213
4729         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4730         look through SAVE_EXPRs with non-side-effects argument.  Adjust
4731         recursive calls.
4732         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4733         save_p here.
4734
4735 2018-04-06  Richard Biener  <rguenther@suse.de>
4736
4737         PR middle-end/85180
4738         * alias.c (find_base_term): New wrapper around find_base_term
4739         unwinding CSELIB_VAL_PTR changes.
4740         (find_base_term): Do not restore CSELIB_VAL_PTR during the
4741         recursion.
4742
4743 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4744
4745         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4746         instructions.
4747         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4748         constant definitions.
4749         ("nop"): lr 0,0 -> nopr r0
4750         ("nop_lr0", "nop_lr1"): New insn definitions.
4751
4752 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
4753
4754         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4755         NDS32_V3PUSH_AVAILABLE_P macro.
4756
4757 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
4758             Chung-Ju Wu  <jasonwucj@gmail.com>
4759
4760         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4761         (nds32*-*-*): Add float and fpu_config into supported_defaults.
4762         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4763         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4764         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4765         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4766         * config/nds32/constraints.md: New constraints and checking for hard
4767         float configuration.
4768         * config/nds32/iterators.md: New mode iterator and attribute for hard
4769         float configuration.
4770         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4771         patterns.
4772         * config/nds32/nds32-fpu.md: New file.
4773         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4774         deal with hard float code generation.
4775         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4776         ARCH_V3S.
4777         (abi_type, float_reg_number): New enum type.
4778         * config/nds32/nds32-predicates.c: New predicates for hard float.
4779         * config/nds32/nds32-protos.h: Declare functions for hard float.
4780         * config/nds32/nds32.c: Implementation for hard float configuration.
4781         * config/nds32/nds32.h: Definitions for hard float configuration.
4782         * config/nds32/nds32.md: Include hard float machine description and
4783         modify patterns for hard float configuration.
4784         * config/nds32/nds32.opt: New options for hard float configuration.
4785         * config/nds32/predicates.md: New predicates for hard float
4786         configuration.
4787
4788 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
4789
4790         * common/config/nds32/nds32-common.c
4791         (nds32_option_optimization_table): Enable -mreleax-hint by default.
4792
4793 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
4794
4795         PR middle-end/85195
4796         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4797         CONSTRUCTOR_ELT (ctor, ...)->value.
4798
4799 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
4800
4801         PR target/85193
4802         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4803
4804 2018-04-05  Tom de Vries  <tom@codesourcery.com>
4805
4806         PR target/85204
4807         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4808         cond jump.
4809
4810 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
4811             Kito Cheng  <kito.cheng@gmail.com>
4812
4813         * config/nds32/constraints.md (U33): Fine-tune checking condition.
4814         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4815         * config/nds32/nds32.h (nds32_16bit_address_type): Add
4816         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4817
4818 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
4819             Kito Cheng  <kito.cheng@gmail.com>
4820
4821         * config/nds32/constraints.md (Ufe): New memory constraint.
4822         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4823         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4824         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4825         operands.
4826         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4827         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4828
4829 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4830
4831         * config/nds32/nds32.md: Use optimize_size in the condition for
4832         alu-shift instructions.
4833
4834 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4835
4836         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4837
4838 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4839
4840         * config/nds32/nds32.md (negsi2): Refine pattern.
4841
4842 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
4843             Chung-Ju Wu  <jasonwucj@gmail.com>
4844
4845         * config/nds32/iterators.md (shift_rotate): New code iterator.
4846         (shift): New code attribute.
4847         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4848         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4849         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4850         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4851         bit-wise operations.
4852         (andsi3, *andsi3): Ditto.
4853         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4854         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4855         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4856         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4857         nds32_ior_operand, nds32_xor_operand): New predicates.
4858
4859 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4860
4861         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4862         (addsi3, subsi3): ... this.
4863
4864 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4865
4866         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4867
4868 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4869
4870         * config/nds32/nds32.md: Adjust indention.
4871
4872 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
4873
4874         * config/nds32/nds32.md (feature): New attribute.
4875
4876 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4877
4878         * config/nds32/nds32.md (subtype): New attribute.
4879
4880 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
4881
4882         PR target/85203
4883         * config/arm/arm-builtins.c (arm_expand_builtin): Change
4884         expansion to perform a bitwise AND of the argument followed by a
4885         boolean negation of the result.
4886
4887 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
4888
4889         PR rtl-optimization/84878
4890         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4891         the basic block.  Assert the use reference is not artificial and that
4892         it has an associated insn.
4893
4894 2018-04-04  Michael Matz  <matz@suse.de>
4895
4896         * builtins.c (compute_objsize): Pass correct operand
4897         to array_at_struct_end_p.
4898
4899 2018-04-04  Richard Biener  <rguenther@suse.de>
4900
4901         PR lto/85176
4902         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4903         from contexts for DINFO_LEVEL_TERSE and below.
4904
4905 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
4906
4907         * config/nds32/nds32-doubleword.md (move_<mode>): Require
4908         resiter_operand condition.
4909         * config/nds32/nds32.md (*move<mode>): Ditto.
4910
4911 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
4912             Monk Chiang  <sh.chiang04@gmail.com>
4913
4914         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4915
4916 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
4917
4918         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4919
4920 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
4921             Kito Cheng  <kito.cheng@gmail.com>
4922
4923         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4924         nds32_cond_code_str, output_cond_branch,
4925         output_cond_branch_compare_zero, nds32_expand_cbranch,
4926         nds32_expand_cstore, nds32_expand_movcc,
4927         nds32_output_cbranchsi4_equality_zero,
4928         nds32_output_cbranchsi4_equality_reg,
4929         nds32_output_cbranchsi4_equality_reg_or_const_int,
4930         nds32_output_cbranchsi4_greater_less_zero: New functions.
4931         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4932         nds32_expand_cstore, nds32_expand_movcc,
4933         nds32_output_cbranchsi4_equality_zero,
4934         nds32_output_cbranchsi4_equality_reg,
4935         nds32_output_cbranchsi4_equality_reg_or_const_int,
4936         nds32_output_cbranchsi4_greater_less_zero): Declare.
4937         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4938         nds32_rimm11s_operand): New predicates.
4939         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4940         * config/nds32/nds32.md: Rewrite all the branch and conditional move
4941         patterns.
4942
4943 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
4944
4945         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4946         * config/nds32/nds32.md: Ditto.
4947         * config/nds32/pipelines.md: Ditto.
4948
4949 2018-04-04  Richard Biener  <rguenther@suse.de>
4950
4951         PR tree-optimization/85168
4952         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4953         propagating abnormals.
4954
4955 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
4956
4957         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4958
4959 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
4960             Kito Cheng  <kito.cheng@gmail.com>
4961
4962         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4963         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4964         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4965         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4966         * config/nds32/nds32.md (sibcall_internal): New.
4967         (sibcall_register): Remove.
4968         (sibcall_immediate): Remove.
4969         (sibcall_value_internal): New.
4970         (sibcall_value_register): Remove.
4971         (sibcall_value_immediate): Remove.
4972         * config/nds32/predicates.md (nds32_general_register_operand): New.
4973         (nds32_call_address_operand): New.
4974
4975 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
4976
4977         PR rtl-optimization/85167
4978         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4979         bb_defs if *split_p, instead preinitialize it to NULL.
4980
4981         PR tree-optimization/85156
4982         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4983         evaluating the argument multiple times.
4984
4985 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
4986
4987         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4988         than vector.
4989         (_mm_cvtpd_ps): Likewise.
4990         (_mm_cvttpd_epi32): Likewise.
4991         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4992         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4993         vector, pixel, and bool following altivec.h include.
4994
4995 2018-04-03  Martin Sebor  <msebor@redhat.com>
4996
4997         * doc/extend.texi (Common Function Attributes): Clarify.
4998         (const attribute): Likewise.
4999         (pure attribute): Likewise.
5000
5001 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
5002
5003         PR target/85169
5004         * config/i386/i386.c (ix86_expand_vector_set): Use
5005         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
5006
5007 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
5008
5009         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5010         instructions when changing rounding bits to preserve precision bits
5011         in the x87 control word.
5012
5013 2018-04-03  Martin Liska  <mliska@suse.cz>
5014
5015         PR tree-optimization/82491
5016         * rtl.h (strip_offset_and_add): Replace += suboffset with
5017         poly_uint64 () + suboffset.
5018
5019 2018-03-29  Martin Liska  <mliska@suse.cz>
5020             Martin Jambor  <mjambor@suse.cz>
5021
5022         PR ipa/84947
5023         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5024         param_type is not an integral or pointer type.
5025
5026 2018-04-03  Richard Biener  <rguenther@suse.de>
5027
5028         * sese.h (recompute_all_dominators): Remove.
5029
5030 2018-04-02  Martin Sebor  <msebor@redhat.com>
5031
5032         * doc/invoke.texi (-Wrestrict): Fix typos.
5033
5034 2018-04-02  Jim Wilson  <jimw@sifive.com>
5035
5036         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5037         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5038         (<optab>di3, <optab>si3_extend): Likewise.
5039         (<optab>si3_mask, <optab>si3_mask_1): New.
5040         (<optab>di3_mask, <optab>di3_mask_1): New.
5041         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5042         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5043         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5044
5045 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
5046
5047         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5048         example.
5049
5050 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
5051
5052         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5053         (nds32_canonicalize_comparison): New function.
5054
5055 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5056             Kito Cheng  <kito.cheng@gmail.com>
5057             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
5058
5059         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5060         * config/nds32/constants.md (unspec_volatile_element): Add
5061         UNSPEC_VOLATILE_RELAX_GROUP.
5062         * config/nds32/nds32-relax-opt.c: New file.
5063         * config/nds32/nds32-predicates.c
5064         (nds32_symbol_load_store_p): New function.
5065         * config/nds32/nds32-protos.h
5066         (nds32_symbol_load_store_p): Declare function.
5067         (make_pass_nds32_relax_opt): Declare new rtl pass function.
5068         * config/nds32/nds32.c
5069         (nds32_register_pass): New function to register pass.
5070         (nds32_register_passes): New function to register passes.
5071         * config/nds32/nds32.md (relax_group): New pattern.
5072         * config/nds32/nds32.opt (mrelax-hint): New option.
5073         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5074
5075 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
5076
5077         * config/nds32/t-nds32: Modify files dependency.
5078
5079 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5080
5081         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5082         (PROFILE_HOOK): Define its implementation.
5083
5084 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5085
5086         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5087         type and 32-bit size.
5088
5089 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
5090
5091         PR middle-end/85090
5092         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5093         (V_128_256): New mode iterator.
5094         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5095         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5096         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5097         of V.
5098         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5099         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5100
5101 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
5102
5103         PR target/83315
5104         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5105         NaN inputs correctly.
5106
5107 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
5108
5109         PR target/80546
5110         * config/rs6000/vsx.md (??r): New mode attribute.
5111         (*vsx_mov<mode>_64bit): Use it.
5112         (*vsx_mov<mode>_32bit): Likewise.
5113
5114 2018-03-30  Martin Sebor  <msebor@redhat.com>
5115
5116         PR tree-optimization/84818
5117         * builtins.c (check_access): Use warning_n.
5118
5119 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
5120
5121         PR target/83822
5122         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5123         condition.
5124         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5125         condition.
5126
5127 2018-03-30  Julia Koval  <julia.koval@intel.com>
5128
5129         PR target/84413
5130         * x86-tune.def (movx, partial_reg_dependency): Enable for
5131         m_SKYLAKE_AVX512.
5132
5133 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
5134
5135         PR inline-asm/84985
5136         * lra-constraints.c (process_alt_operands): Move setting
5137         this_alternative_matches below.
5138
5139 2018-03-29  Martin Liska  <mliska@suse.cz>
5140
5141         PR lto/84995.
5142         * doc/invoke.texi: Document how LTO works with debug info.
5143         Describe auto-load support of binutils.  Mention 'x86-64'
5144         as valid option value of -march option.
5145
5146 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
5147
5148         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5149
5150         PR c/85094
5151         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5152         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5153         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5154         checking.
5155
5156 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
5157
5158         PR target/84912
5159         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5160         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5161         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5162         for RS6000_BTM_POWERPC64.
5163         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5164         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5165         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5166         definition.
5167         (DIVDE): Use it.
5168         (DIVDEU): Likewise.
5169
5170 2018-03-28 Carl Love  <cel@us.ibm.com>
5171
5172         Revert
5173         2017-09-27  Carl Love  <cel@us.ibm.com>
5174
5175         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5176         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5177         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5178         fctiw instruction.
5179
5180 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5181
5182         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5183         instead of __vector bool.
5184         (_mm_max_pu8): Likewise.
5185         (_mm_min_pi16): Likewise.
5186
5187 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
5188
5189         PR target/84912
5190         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5191         (DIVWEUO): Likewise.
5192         (DIVDEO): Likewise.
5193         (DIVDEUO): Likewise.
5194         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5195         DIVWEUO and DIVDEUO.
5196         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5197         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5198         (div_extend): Likewise.
5199         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5200         builtin function.
5201         (__builtin_divweuo): Likewise.
5202         (__builtin_divdeo): Likewise.
5203         (__builtin_divdeuo): Likewise.
5204
5205 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
5206
5207         PR target/85095
5208         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5209         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5210
5211         PR tree-optimization/82004
5212         * gimple-match-head.c (optimize_pow_to_exp): New function.
5213         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5214         Don't fold to exp if optimize_pow_to_exp is false.
5215
5216 2018-03-28  Martin Liska  <mliska@suse.cz>
5217
5218         PR other/84819
5219         * calls.c (initialize_argument_information): Fix trailing space.
5220         * common.opt: Fix typo and provide better explanation for
5221         -fsanitize-coverage option.
5222         * config/i386/i386.opt: Fix typo.
5223
5224 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
5225             Martin Liska  <mliska@suse.cz>
5226
5227         PR sanitizer/85081
5228         * gimplify.c (asan_poison_variable): Don't do the check for
5229         gimplify_omp_ctxp here.
5230         (gimplify_decl_expr): Do it here.
5231         (gimplify_target_expr): Likewise.
5232
5233 2018-03-28  Martin Liska  <mliska@suse.cz>
5234
5235         PR target/84988
5236         * config/i386/i386.c (ix86_function_arg_advance): Do not call
5237         chkp_type_bounds_count if MPX is not enabled.
5238
5239 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
5240
5241         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5242
5243 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5244
5245         PR target/84914
5246         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5247         function to create the function decl for complex long double
5248         multiply and divide for -mabi=ieeelongdouble.
5249         (init_float128_ieee): Call it.
5250
5251 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
5252
5253         PR target/85044
5254         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5255         -fcf-protection=branch -mibt.
5256         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5257
5258 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5259
5260         PR target/81863
5261         * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5262
5263 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
5264
5265         PR target/85056
5266         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5267         extern array declarations.
5268
5269 2018-03-27  Richard Biener  <rguenther@suse.de>
5270
5271         PR middle-end/84067
5272         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5273         explicit single_use checks.
5274
5275 2018-03-27  Richard Biener  <rguenther@suse.de>
5276
5277         PR tree-optimization/85082
5278         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5279         Valueize the VUSE.
5280
5281 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5282
5283         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5284         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5285         Turn on fasynchronous-unwind-tables and funwind-tables.
5286
5287 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
5288
5289         PR target/85073
5290         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5291         (*bmi_blsr_<mode>_ccz): Ditto.
5292
5293 2018-03-26  Tom de Vries  <tom@codesourcery.com>
5294
5295         PR tree-optimization/85063
5296         * omp-general.c (offloading_function_p): New function.  Factor out
5297         of ...
5298         * omp-offload.c (pass_omp_target_link::gate): ... here.
5299         * omp-general.h (offloading_function_p): Declare.
5300         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5301         with attribute omp declare target for offloading functions.
5302
5303 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
5304
5305         PR tree-optimization/84005
5306         * tree-data-ref.h (get_base_for_alignment): Declare.
5307         * tree-data-ref.c (get_base_for_alignment_1): New function.
5308         (get_base_for_alignment): Likewise.
5309         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5310         get_base_for_alignment to find a suitable base object, instead
5311         of always using drb->base_address.
5312
5313 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5314
5315         PR inline-asm/85022
5316         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5317         known size by default.
5318
5319 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
5320
5321         PR inline-asm/85030
5322         * lra-constraints.c (process_alt_operands): Don't match BLKmode
5323         and non BLKmode operands.
5324
5325 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5326
5327         PR target/85026
5328         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5329         Clean up attributes.
5330
5331 2018-03-23  Richard Biener  <rguenther@suse.de>
5332
5333         PR debug/85020
5334         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5335         we are going to emit early debug for LTO.
5336
5337 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5338
5339         PR inline-asm/85034
5340         * function.c (match_asm_constraints_1): Don't optimize if input
5341         doesn't satisfy general_operand predicate for output's mode.
5342
5343         PR inline-asm/85022
5344         * alias.c (write_dependence_p): Don't require for x_canonicalized
5345         non-VOIDmode if x has VOIDmode.
5346
5347         PR sanitizer/85029
5348         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5349         just don't try to optimize it rather than assert it never happens.
5350
5351 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5352
5353         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5354         macro expansions for definition of ST_INTERNAL_<mode> and
5355         LD_INTERNAL_<mode> builtins.
5356         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5357         Remove prototype.
5358         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5359         function.
5360         (altivec_expand_st_builtin): Likewise.
5361         (altivec_expand_builtin): Remove calls to deleted functions.
5362         (rs6000_address_for_altivec): Delete this function.
5363         * config/rs6000/vector.md: Remove expands for
5364         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5365
5366 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
5367
5368         PR target/84826
5369         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5370         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5371         re-computing once computed.
5372         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5373         (arm_init_machine_status): Initialize
5374         machine->static_chain_stack_bytes.
5375
5376 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5377
5378         PR target/84760
5379         * doc/extend.texi: Add four new prototypes for vec_ld.
5380         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5381         definitions for more logical presentation.
5382         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5383         entries for V1TI variants of __builtin_altivec_ld builtin.
5384         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5385         handling of V1TI variant of LVX icode pattern.
5386         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5387         (rs6000_gimple_fold_builtin): Likewise.
5388         (altivec_init_builtins): Add code to define
5389         __builtin_altivec_lvx_v1ti function.
5390
5391 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
5392
5393         PR inline-asm/84941
5394         * function.c (match_asm_constraints_1): Don't do the optimization
5395         if input isn't a REG, SUBREG, MEM or constant.
5396
5397 2018-03-22  Tom de Vries  <tom@codesourcery.com>
5398
5399         PR tree-optimization/84956
5400         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5401         bb_has_abnormal_pred.
5402
5403 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
5404
5405         PR sanitizer/85018
5406         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5407         DECL_INITIAL (decl) to decl at the end.
5408         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5409         adjust the comment.
5410
5411 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
5412
5413         * doc/extend.texi (__builtin_tgmath): Document when complex
5414         integer types are treated as _Complex _Float64.
5415
5416 2018-03-21  Tom de Vries  <tom@codesourcery.com>
5417
5418         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5419
5420 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5421
5422         PR tree-optimization/84960
5423         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5424         if it is ENTRY block, move them into single succ of ENTRY in that case.
5425
5426 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
5427
5428         PR tree-optimization/84811
5429         * poly-int.h (poly_span_traits): Remove the T3 parameter and
5430         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5431         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5432         (known_subrange_p): Update accordingly.  Cast each value involved
5433         in the size comparison, rather than casting the result of the
5434         subtraction.
5435
5436 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5437
5438         PR tree-optimization/84982
5439         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5440         by flipping the least significant bit rather than all bits from
5441         bitpos to bitpos + bitsize - 1.
5442
5443 2018-03-21  Nathan Sidwell  <nathan@acm.org>
5444
5445         * doc/extend.texi (Deprecated Features): Remove mention of
5446         long-deleted deprecations.
5447
5448 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5449
5450         PR jit/84288
5451         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5452         * configure: Regenerate.
5453
5454 2018-03-21  Tom de Vries  <tom@codesourcery.com>
5455
5456         PR tree-optimization/83126
5457         * tree-parloops.c (num_phis): New function.
5458         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5459
5460 2018-03-21  Nathan Sidwell  <nathan@acm.org>
5461
5462         * doc/extend.texi (Deprecated Features): Update deprecated flags,
5463         mention anon-struct/union members and trailing attributes.
5464
5465 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
5466
5467         PR tree-optimization/84969
5468         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5469         builtin memset partitions if they set different rhs values.
5470
5471 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5472
5473         PR rtl-optimization/84989
5474         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5475         VEC_DUPLICATE with scalar result mode.
5476
5477 2018-03-21  Martin Liska  <mliska@suse.cz>
5478
5479         PR ipa/84963
5480         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5481         not intended return statement.
5482
5483 2018-03-21  Martin Liska  <mliska@suse.cz>
5484
5485         PR target/84988
5486         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5487         (chkp_find_bound_slots_1): Limit number of iterations.
5488
5489 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
5490
5491         PR target/84838
5492         * Minor grammar fixes for x86 options.
5493
5494 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
5495
5496         PR debug/84875
5497         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5498         holding REG_CFA_RESTORE notes, instead turn them into a USE.
5499
5500 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
5501
5502         PR target/83789
5503         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5504         (altivec_lvx_<mode>_1op): Likewise.
5505         (altivec_stvx_<mode>_2op): Likewise.
5506         (altivec_stvx_<mode>_1op): Likewise.
5507         (altivec_lvx_<VM2:mode>): New define_expand.
5508         (altivec_stvx_<VM2:mode>): Likewise.
5509         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5510         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5511         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5512         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5513         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5514         (rs6000_gen_lvx): Likewise.
5515         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5516         (altivec_expand_stv_builtin): Likewise.
5517         (altivec_expand_builtin): Likewise.
5518         * config/rs6000/vector.md: Likewise.
5519
5520 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5521
5522         PR target/82518
5523         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5524         BYTES_BIG_ENDIAN.
5525
5526 2018-03-20  Richard Biener  <rguenther@suse.de>
5527
5528         PR target/84986
5529         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5530         sign-conversions as zero, fall back to standard scalar_stmt
5531         cost for the rest.
5532
5533 2018-03-20  Martin Liska  <mliska@suse.cz>
5534
5535         PR ipa/84825
5536         * predict.c (rebuild_frequencies): Handle case when we have
5537         PROFILE_ABSENT, but flag_guess_branch_prob is false.
5538
5539 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
5540
5541         PR target/84990
5542         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5543         flag_section_anchors.
5544         * varasm.c (use_blocks_for_decl_p): Remove hack for
5545         dw2_force_const_mem.
5546
5547         PR target/84845
5548         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5549         to ...
5550         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
5551         be created, use lowpart_subreg of operands[0] rather than operands[0]
5552         itself.
5553         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5554         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5555         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5556         and n constraint instead of aarch64_shift_imm_di and Usd.
5557         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5558         (*aarch64_<optab>_reg_minus<mode>3): ... this.
5559
5560 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
5561
5562         PR target/82989
5563         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5564         to favor GPR over NEON registers.
5565         (<shift>di3_neon): Likewise.
5566
5567 2018-03-20  Tom de Vries  <tom@codesourcery.com>
5568
5569         PR target/84952
5570         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5571         (nvptx_process_pars): Emit bar.sync asap and alap.
5572
5573 2018-03-20  Tom de Vries  <tom@codesourcery.com>
5574
5575         PR target/84954
5576         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5577         seen_label if seen_label is already set.
5578
5579 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
5580
5581         PR target/84945
5582         * config/i386/i386.c (fold_builtin_cpu): For features above 31
5583         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5584         Use 1U instead of 1.  Formatting fixes.
5585
5586         PR c/84953
5587         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5588         instead of TREE_TYPE (s1) for the return value.
5589
5590 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
5591
5592         PR tree-optimization/84946
5593         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5594         bitsize + bitsize in poly_uint64 rather than poly_int64.
5595
5596         PR sanitizer/78651
5597         * dwarf2asm.c: Include fold-const.c.
5598         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5599         of decl rather than decl itself.
5600
5601         PR rtl-optimization/84643
5602         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5603
5604 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
5605
5606         PR sanitizer/78651
5607         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5608         calling assemble_variable.
5609
5610 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
5611
5612         PR target/81647
5613         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5614         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5615
5616 2018-03-19  Jim Wilson  <jimw@sifive.com>
5617
5618         PR bootstrap/84856
5619         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5620         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5621         (riscv_first_stack_step): Likewise.
5622         (riscv_option_override): Use STACK_BOUNDARY instead of
5623         MIN_STACK_BOUNDARY.
5624         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5625         MIN_STACK_BOUNDARY.
5626         (BIGGEST_ALIGNMENT): Set to 128.
5627         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5628         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5629         STACK_BOUNDARY.
5630
5631 2018-03-19  Richard Biener  <rguenther@suse.de>
5632
5633         PR tree-optimization/84933
5634         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5635         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5636
5637 2018-03-19  Richard Biener  <rguenther@suse.de>
5638
5639         PR tree-optimization/84859
5640         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5641         (cond_if_else_store_replacement): Perform sinking operation on
5642         single-store BBs regardless of MAX_STORES_TO_SINK setting.
5643         Generalize what a BB with a single eligible store is.
5644
5645 2018-03-19  Richard Biener  <rguenther@suse.de>
5646
5647         PR tree-optimization/84929
5648         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5649         chrec_is_positive against non-chrec arg.
5650
5651 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
5652
5653         PR target/84711
5654         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5655
5656 2018-03-18  Martin Liska  <mliska@suse.cz>
5657
5658         PR rtl-optimization/84635
5659         * regrename.c (build_def_use): Use matches_mode only when
5660         matches >= 0.
5661
5662 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
5663
5664         PR tree-optimization/84913
5665         * tree-vect-loop.c (vectorizable_reduction): Don't try to
5666         vectorize chains of COND_EXPRs.
5667
5668 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
5669
5670         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5671
5672 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
5673
5674         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5675
5676 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
5677
5678         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5679
5680 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
5681             Kito Cheng  <kito.cheng@gmail.com>
5682
5683         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5684         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5685         (nds32_adjust_reg_alloc_order): New function.
5686         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5687
5688 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
5689
5690         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5691         nds32_print_operand, nds32_print_operand_address): Use
5692         HOST_WIDE_INT_PRINT_DEC instead.
5693
5694 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
5695
5696         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5697
5698 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
5699
5700         PR target/84902
5701         * config/i386/i386.c (initial_ix86_tune_features,
5702         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5703         unsigned long long.
5704         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5705         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5706         rather than 1u << ix86_tune.  Formatting fix.
5707         (ix86_option_override_internal): Change ix86_arch_mask from
5708         unsigned int to unsigned HOST_WIDE_INT, initialize to
5709         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5710         (ix86_function_specific_restore): Likewise.
5711
5712 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
5713
5714         PR target/84899
5715         * postreload.c (reload_combine_recognize_pattern): Perform
5716         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5717         truncate_int_for_mode the result for the destination's mode.
5718
5719         PR c/84909
5720         * hsa-gen.c (mem_type_for_type): Fix comment typo.
5721         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5722         Likewise.
5723         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5724         Likewise.
5725
5726 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
5727
5728         PR target/84876
5729         * lra-assigns.c (lra_split_hard_reg_for): Don't use
5730         regno_allocno_class_array and sorted_pseudos.
5731         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5732         insns where regno is used.
5733
5734 2018-03-16  Martin Liska  <mliska@suse.cz>
5735
5736         PR ipa/84833
5737         * multiple_target.c (create_dispatcher_calls): Redirect
5738         reference in the symbol table.
5739
5740 2018-03-16  Martin Liska  <mliska@suse.cz>
5741
5742         PR ipa/84722
5743         * multiple_target.c (create_dispatcher_calls): Redirect also
5744         an alias.
5745
5746 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
5747
5748         PR c++/79937
5749         PR c++/82410
5750         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5751         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5752         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5753
5754 2018-03-16  Julia Koval  <julia.koval@intel.com>
5755
5756         * doc/invoke.texi (Skylake Server): Add CLWB.
5757         Cannonlake): Remove CLWB.
5758
5759 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
5760
5761         PR tree-optimization/84841
5762         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5763         1 << 3.
5764         (FLOAT_ONE_CONST_TYPE): Define.
5765         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5766         (sort_by_operand_rank): Put entries with higher constant_type last
5767         rather than first to match comments.
5768
5769 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
5770
5771         * config/nios2/nios2.md (movsi_internal): Fix thinko in
5772         split predicate.
5773
5774 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
5775
5776         PR c++/79085
5777         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5778         check and use address of target always.
5779
5780 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
5781
5782         PR target/84574
5783         * config/i386/i386.c (indirect_thunk_needed): Update comments.
5784         (indirect_thunk_bnd_needed): Likewise.
5785         (indirect_thunks_used): Likewise.
5786         (indirect_thunks_bnd_used): Likewise.
5787         (indirect_return_needed): New.
5788         (indirect_return_bnd_needed): Likewise.
5789         (output_indirect_thunk_function): Add a bool argument for
5790         function return.
5791         (output_indirect_thunk_function): Don't generate alias for
5792         function return thunk.
5793         (ix86_code_end): Call output_indirect_thunk_function to generate
5794         function return thunks.
5795         (ix86_output_function_return): Set indirect_return_bnd_needed
5796         and indirect_return_needed instead of indirect_thunk_bnd_needed
5797         and indirect_thunk_needed.
5798
5799 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
5800
5801         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5802         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5803         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5804
5805 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
5806             Paul Hua <paul.hua.gm@gmail.com>
5807
5808         PR c/84852
5809         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5810
5811 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
5812
5813         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5814         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5815         resp. SFmode cases.
5816
5817 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
5818
5819         PR target/84711
5820         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5821         instead of GET_MODE_SIZE when comparing Units.
5822
5823 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
5824
5825         PR target/68256
5826         * varasm.c (hash_section): Return an unchangeble hash value
5827         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5828         Return !aarch64_can_use_per_function_literal_pools_p ().
5829
5830 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
5831
5832         PR target/84860
5833         * optabs.c (emit_conditional_move): Pass address of cmode's copy
5834         rather than address of cmode as last argument to prepare_cmp_insn.
5835
5836 2018-03-15  Julia Koval  <julia.koval@intel.com>
5837
5838         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5839         F_AVX512VNNI, F_AVX512BITALG): New.
5840
5841 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
5842
5843         PR target/83451
5844         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
5845         insn for floating-point loads and stores.
5846
5847 2018-03-14  Carl Love  <cel@us.ibm.com>
5848
5849         * config/rs6000/rs6000-c.c: Add macro definitions for
5850         ALTIVEC_BUILTIN_VEC_PERMXOR.
5851         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5852         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5853         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5854         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5855         UNSPEC_VPERMXOR.
5856         * config/doc/extend.texi: Add prototypes for vec_permxor.
5857
5858 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
5859
5860         PR c/84852
5861         * diagnostic-show-locus.c (class layout_point): Convert m_line
5862         from int to linenum_type.
5863         (line_span::comparator): Use linenum "compare" function when
5864         comparing line numbers.
5865         (test_line_span): New function.
5866         (layout_range::contains_point): Convert param "row" from int to
5867         linenum_type.
5868         (layout_range::intersects_line_p): Likewise.
5869         (layout::will_show_line_p): Likewise.
5870         (layout::print_source_line): Likewise.
5871         (layout::should_print_annotation_line_p): Likewise.
5872         (layout::print_annotation_line): Likewise.
5873         (layout::print_leading_fixits): Likewise.
5874         (layout::annotation_line_showed_range_p): Likewise.
5875         (struct line_corrections): Likewise for field m_row.
5876         (line_corrections::line_corrections): Likewise for param "row".
5877         (layout::print_trailing_fixits): Likewise.
5878         (layout::get_state_at_point): Likewise.
5879         (layout::get_x_bound_for_row): Likewise.
5880         (layout::print_line): Likewise.
5881         (diagnostic_show_locus): Likewise for locals "last_line" and
5882         "row".
5883         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5884         * input.c (selftest::test_linenum_comparisons): New function.
5885         (selftest::input_c_tests): Call it.
5886         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5887         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5888         * selftest.h (ASSERT_GT): New macro.
5889         (ASSERT_GT_AT): New macro.
5890         (ASSERT_LT): New macro.
5891         (ASSERT_LT_AT): New macro.
5892
5893 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
5894
5895         PR rtl-optimization/84780
5896         * combine.c (distribute_links): Don't make a link based on pc_rtx.
5897
5898 2018-03-14  Martin Liska  <mliska@suse.cz>
5899
5900         * tree.c (record_node_allocation_statistics): Use
5901         get_stats_node_kind.
5902         (get_stats_node_kind): New function extracted from
5903         record_node_allocation_statistics.
5904         (free_node): Use get_stats_node_kind.
5905
5906 2018-03-14  Richard Biener  <rguenther@suse.de>
5907
5908         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5909         that the value-set of ANTIC_IN doesn't grow.
5910
5911         Revert
5912         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5913         member.
5914         (BB_VISITED_WITH_VISITED_SUCCS): New define.
5915         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5916
5917 2018-03-14  Julia Koval  <julia.koval@intel.com>
5918
5919         * config.gcc (icelake-client, icelake-server): New.
5920         (icelake): Remove.
5921         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5922         (initial_ix86_arch_features): Ditto.
5923         (PTA_SKYLAKE): Add SGX.
5924         (PTA_ICELAKE): Remove.
5925         (PTA_ICELAKE_CLIENT): New.
5926         (PTA_ICELAKE_SERVER): New.
5927         (ix86_option_override_internal): Split up icelake on icelake client and
5928         icelake server.
5929         (get_builtin_code_for_version): Ditto.
5930         (fold_builtin_cpu): Ditto.
5931         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5932         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5933         * config/i386/i386.h (processor_type): Ditto.
5934         * doc/invoke.texi: Ditto.
5935
5936 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
5937
5938         PR sanitizer/83392
5939         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5940         INTEGER_CST offset, add it together with bitpos / 8 and
5941         sign extend based on POINTER_SIZE.
5942
5943         PR target/84844
5944         Revert
5945         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
5946
5947         PR target/78090
5948         * config/i386/constraints.md (Yc): New register constraint.
5949         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5950         Use Yc constraint for alternative 2 of operand 0.  Remove
5951         preferred_for_speed attribute.
5952
5953 2018-03-14  Richard Biener  <rguenther@suse.de>
5954
5955         PR tree-optimization/84830
5956         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5957         with the old one to avoid oscillations.
5958
5959 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
5960
5961         PR target/83712
5962         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5963         pseudos.
5964         (assign_by_spills): Return a flag of reload assignment failure.
5965         Do not process the reload assignment failures.  Do not spill other
5966         reload pseudos if they has the same reg class.  Update n if
5967         necessary.
5968         (lra_assign): Add a return arg.  Set up from the result of
5969         assign_by_spills call.
5970         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5971         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
5972         usage_insns if it is not NULL.
5973         (spill_hard_reg_in_range): New function.
5974         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5975         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5976         function prototypes.
5977         (lra_assign): Change prototype.
5978         * lra.c (lra): Add code to deal with fails by splitting hard reg
5979         live ranges.
5980
5981 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
5982
5983         * config/riscv/riscv.opt (mrelax): New option.
5984         * config/riscv/riscv.c (riscv_file_start): Emit ".option
5985         "norelax" when riscv_mrelax is disabled.
5986         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5987
5988 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
5989
5990         PR target/84743
5991         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5992         reassociation for int modes.
5993
5994 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
5995
5996         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5997         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5998         for big-endian.
5999         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6000         * config/aarch64/aarch64-sve.md
6001         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6002         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6003         (*extend<mode><Vwide>2): Rename to...
6004         (aarch64_sve_extend<mode><Vwide>2): ...this.
6005         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6006         renaming the old pattern to...
6007         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
6008         unsigned packs.
6009         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6010         define_expand, renaming the old pattern to...
6011         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6012         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6013         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6014         account when deciding which SVE instruction the optab should use.
6015         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6016
6017 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6018
6019         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6020         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6021         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6022         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6023         (tlsdesc_small_<mode>): Turn a define_expand and use
6024         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
6025         (tlsdesc_small_advsimd_<mode>): ...this.
6026         (tlsdesc_small_sve_<mode>): New pattern.
6027
6028 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6029
6030         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6031         (UNSPEC_UMUL_HIGHPART): New constants.
6032         (MUL_HIGHPART): New int iteraor.
6033         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6034         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6035         define_expand.
6036         (*<su>mul<mode>3_highpart): New define_insn.
6037
6038 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
6039
6040         PR lto/84805
6041         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6042         incomplete types.
6043
6044 2018-03-13  Martin Liska  <mliska@suse.cz>
6045
6046         PR ipa/84658.
6047         * (sem_item_optimizer::sem_item_optimizer): Initialize new
6048         vector.
6049         (sem_item_optimizer::~sem_item_optimizer): Release it.
6050         (sem_item_optimizer::merge_classes): Register variable aliases.
6051         (sem_item_optimizer::fixup_pt_set): New function.
6052         (sem_item_optimizer::fixup_points_to_sets): Likewise.
6053         * ipa-icf.h: Declare new variables and functions.
6054
6055 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
6056
6057         PR middle-end/84834
6058         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6059         integer_pow2p@2 and test integer_pow2p in condition.
6060         (A < 0 ? C : 0): Similarly for @1.
6061
6062         PR middle-end/84831
6063         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6064         characters starting at p contain '\0' character, don't look beyond
6065         that.
6066
6067         PR target/84827
6068         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6069         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6070
6071         PR target/84828
6072         * reg-stack.c (change_stack): Change update_end var from int to
6073         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6074         also call set_block_for_insn on the newly added insns and rescan.
6075
6076         PR target/84786
6077         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6078         on the last operand.
6079
6080         PR c++/84704
6081         * tree.c (stabilize_reference_1): Return save_expr (e) for
6082         STATEMENT_LIST even if it doesn't have side-effects.
6083
6084 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
6085
6086         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6087
6088 2018-03-12  Renlin Li  <renlin.li@arm.com>
6089
6090         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6091         aarch64_output_scalar_simd_mov_immediate.
6092
6093 2018-03-12  Martin Sebor  <msebor@redhat.com>
6094
6095         PR tree-optimization/83456
6096         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6097         for perfectly overlapping calls to memcpy.
6098         (gimple_fold_builtin_memory_chk): Same.
6099         (gimple_fold_builtin_strcpy): Handle no-warning.
6100         (gimple_fold_builtin_stxcpy_chk): Same.
6101         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6102
6103 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
6104
6105         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6106         parameter.  Use it for SFmode.
6107         (rs6000_function_arg_advance_1): Adjust.
6108         (rs6000_function_arg): Adjust.
6109         (rs6000_gimplify_va_arg): Pass false for that new parameter.
6110
6111 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
6112
6113         PR rtl-optimization/84169
6114         PR rtl-optimization/84780
6115         * combine.c (can_combine_p): Check for a 2-insn combination whether
6116         the destination register is used between the two insns, too.
6117
6118 2018-03-12  Richard Biener  <rguenther@suse.de>
6119
6120         PR tree-optimization/84803
6121         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6122         for refs DR analysis didn't process.
6123
6124 2018-03-12  Richard Biener  <rguenther@suse.de>
6125
6126         PR tree-optimization/84777
6127         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6128         force-vectorize loops ignore whether we are optimizing for size.
6129
6130 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6131
6132         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6133         (TARGET_MD_ASM_ADJUST): Define.
6134
6135 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
6136             Kito Cheng  <kito.cheng@gmail.com>
6137             Chung-Ju Wu  <jasonwucj@gmail.com>
6138
6139         * config/nds32/nds32.c (nds32_compute_stack_frame,
6140         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6141         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6142         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6143         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6144         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6145         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6146         * config/nds32/nds32.md (prologue, epilogue): Use macro
6147         NDS32_V3PUSH_AVAILABLE_P to do checking.
6148
6149 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
6150
6151         PR debug/58150
6152         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6153         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6154         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
6155         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
6156         addition of most attributes on !orig_type_die or the attribute not
6157         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6158
6159 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6160             Chung-Ju Wu  <jasonwucj@gmail.com>
6161
6162         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6163         __NDS32_VH__ macro.
6164         * config/nds32/nds32.opt (mvh): New option.
6165
6166 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6167             Chung-Ju Wu  <jasonwucj@gmail.com>
6168
6169         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6170         function.
6171         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6172         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6173         definition.
6174
6175 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6176             Chung-Ju Wu  <jasonwucj@gmail.com>
6177
6178         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6179         function.
6180         * config/nds32/nds32-multiple.md (strlensi): New pattern.
6181         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6182
6183 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
6184             Kito Cheng  <kito.cheng@gmail.com>
6185             Chung-Ju Wu  <jasonwucj@gmail.com>
6186
6187         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6188         UNSPEC_FFMISM and UNSPEC_FLMISM.
6189         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6190         for ffb, ffmism and flmism.
6191         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6192         (unspec_ffmism): Ditto.
6193         (unspec_flmism): Ditto.
6194         (nds32_expand_builtin_impl): Check if string extension is available.
6195         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6196         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6197
6198 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
6199
6200         Reverting patch:
6201         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6202
6203         PR target/83712
6204         * lra-assigns.c (assign_by_spills): Return a flag of reload
6205         assignment failure.  Do not process the reload assignment
6206         failures.  Do not spill other reload pseudos if they has the same
6207         reg class.
6208         (lra_assign): Add a return arg.  Set up from the result of
6209         assign_by_spills call.
6210         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6211         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6212         usage_insns if it is not NULL.
6213         (spill_hard_reg_in_range): New function.
6214         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6215         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6216         function prototypes.
6217         (lra_assign): Change prototype.
6218         * lra.c (lra): Add code to deal with fails by splitting hard reg
6219         live ranges.
6220
6221 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
6222
6223         PR target/84807
6224         * config/i386/i386.opt: Replace Enforcment with Enforcement.
6225
6226 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
6227
6228         PR debug/84620
6229         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6230         (dw_val_node): Add val_symbolic_view.
6231         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6232         (symview_upper_bound): New.
6233         (new_line_info_table): Initialize symviews_since_reset.
6234         (dwarf2out_source_line): Count symviews_since_reset and set
6235         symview_upper_bound.
6236         (dw_val_equal_p): Handle symview.
6237         (add_AT_symview): New.
6238         (print_dw_val): Handle symview.
6239         (attr_checksum, attr_checksum_ordered): Likewise.
6240         (same_dw_val_p, size_of_die): Likewise.
6241         (value_format, output_die): Likewise.
6242         (add_high_low_attributes): Use add_AT_symview for entry_view.
6243         (dwarf2out_finish): Reset symview_upper_bound, clear
6244         zero_view_p.
6245
6246 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
6247
6248         PR target/83969
6249         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6250         Add strict argument and use it.
6251         (rs6000_split_multireg_move): Update for new strict argument.
6252         (mem_operand_gpr): Disallow all non-offsettable addresses.
6253         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6254
6255 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
6256
6257         PR target/84772
6258         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6259         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6260         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6261
6262         PR c++/84767
6263         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6264         decl, use remap_type if we want to use the type.
6265
6266 2018-03-09  Martin Sebor  <msebor@redhat.com>
6267
6268         PR tree-optimization/84526
6269         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6270         Remove dead code.
6271         (builtin_access::generic_overlap): Be prepared to handle non-array
6272         base objects.
6273
6274 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
6275
6276         PR rtl-optimization/84682
6277         * lra-constraints.c (process_address_1): Check is_address flag
6278         for address constraints.
6279         (process_alt_operands): Likewise.
6280         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6281         preprocess_constraints.
6282         * recog.h (preprocess_constraints): Add oploc parameter.
6283         Adjust callers.
6284         * recog.c (preprocess_constraints): Test address_operand for
6285         CT_ADDRESS constraints.
6286
6287 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6288
6289         PR target/83712
6290         * lra-assigns.c (assign_by_spills): Return a flag of reload
6291         assignment failure.  Do not process the reload assignment
6292         failures.  Do not spill other reload pseudos if they has the same
6293         reg class.
6294         (lra_assign): Add a return arg.  Set up from the result of
6295         assign_by_spills call.
6296         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6297         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6298         usage_insns if it is not NULL.
6299         (spill_hard_reg_in_range): New function.
6300         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6301         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6302         function prototypes.
6303         (lra_assign): Change prototype.
6304         * lra.c (lra): Add code to deal with fails by splitting hard reg
6305         live ranges.
6306
6307 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6308
6309         PR target/83193
6310         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6311         Accept complain bool parameter.  Only emit errors if it is true.
6312         (arm_parse_cpu_option_name): Likewise.
6313         (arm_target_thumb_only): Adjust callers of the above.
6314         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6315         prototype to take a default true bool parameter.
6316         (arm_parse_arch_option_name): Likewise.
6317
6318 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
6319             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
6320
6321         PR jit/64089
6322         PR jit/84288
6323         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6324         * configure: Regenerate.
6325         * configure.ac ("linker --version-script option"): New.
6326         ("linker soname option"): New.
6327
6328 2018-03-09  Richard Biener  <rguenther@suse.de>
6329
6330         PR tree-optimization/84775
6331         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6332         immediate uses of predicate stmts and mark them modified.
6333
6334         Revert
6335         PR tree-optimization/84178
6336         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6337         to caller.
6338         (version_loop_for_if_conversion): Delay update_ssa call.
6339         (tree_if_conversion): Delay update_ssa until after predicate
6340         insertion.
6341
6342 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
6343
6344         PR target/84763
6345         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6346         when the function accesses prior frames.
6347
6348 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6349
6350         PR debug/84456
6351         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6352         gen_llsym, otherwise call maybe_gen_llsym.
6353
6354         PR inline-asm/84742
6355         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6356         has ',' character inside of it.
6357
6358 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6359
6360         PR target/84748
6361         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6362         as clobbering CC_REGNUM.
6363
6364 2018-03-08  Richard Biener  <rguenther@suse.de>
6365
6366         PR middle-end/84552
6367         * tree-scalar-evolution.c: Include tree-into-ssa.h.
6368         (follow_copies_to_constant): Do not follow SSA names registered
6369         for update.
6370
6371 2018-03-08  Richard Biener  <rguenther@suse.de>
6372
6373         PR tree-optimization/84178
6374         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6375         to caller.
6376         (version_loop_for_if_conversion): Delay update_ssa call.
6377         (tree_if_conversion): Delay update_ssa until after predicate
6378         insertion.
6379
6380 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
6381
6382         PR tree-optimization/84178
6383         * tree-if-conv.c (release_bb_predicate): Remove the
6384         the assertion that the stmts have NULL use_ops.
6385         Discard the statements, asserting that they haven't
6386         yet been added to a BB.
6387
6388 2018-03-08  Richard Biener  <rguenther@suse.de>
6389
6390         PR tree-optimization/84746
6391         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6392         (phi_translate): Pass in destination ANTIC_OUT set.
6393         (phi_translate_1): Likewise.  For a simplified result lookup
6394         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6395         (phi_translate_set): Adjust.
6396         (do_pre_regular_insertion): Likewise.
6397         (do_pre_partial_partial_insertion): Likewise.
6398
6399 2018-03-08  Martin Liska  <mliska@suse.cz>
6400
6401         PR gcov-profile/84735
6402         * doc/gcov.texi: Document usage of profile files.
6403         * gcov-io.h: Document changes in the format.
6404
6405 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
6406
6407         PR debug/84404
6408         PR debug/84408
6409         * dwarf2out.c (struct dw_line_info_table): Update comments for
6410         view == -1.
6411         (FORCE_RESET_NEXT_VIEW): New.
6412         (FORCE_RESETTING_VIEW_P): New.
6413         (RESETTING_VIEW_P): Check for -1 too.
6414         (ZERO_VIEW_P): Likewise.
6415         (new_line_info_table): Force-reset next view.
6416         (dwarf2out_begin_function): Likewise.
6417         (dwarf2out_source_line): Simplify zero_view_p initialization.
6418         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6419         view directly.  Omit view when omitting .loc at line 0.
6420
6421 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6422
6423         PR tree-optimization/84740
6424         * tree-switch-conversion.c (process_switch): Call build_constructors
6425         only if info.phi_count is non-zero.
6426
6427         PR tree-optimization/84739
6428         * tree-tailcall.c (find_tail_calls): Check call arguments against
6429         DECL_ARGUMENTS (current_function_decl) rather than
6430         DECL_ARGUMENTS (func) when checking for tail recursion.
6431
6432 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
6433
6434         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6435         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6436         Volker Reichelt's entry and add entries for people that perform
6437         GCC fuzzy testing and report numerous bugs.
6438
6439 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
6440
6441         PR target/82411
6442         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6443         readonly data in sdata, if that is disabled.
6444         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6445         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6446         -mreadonly-in-sdata option.
6447
6448 2018-03-07  Martin Sebor  <msebor@redhat.com>
6449
6450         PR tree-optimization/84468
6451         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6452         basic block when looking for nul assignment.
6453
6454 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
6455
6456         PR target/84277
6457         * except.h (output_function_exception_table): Adjust prototype.
6458         * except.c (output_function_exception_table): Remove FNNAME parameter
6459         and add SECTION parameter.  Ouput one part of the table at a time.
6460         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6461         the first part of the exception table and emit unwind directives.
6462         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6463         (i386_pe_seh_cold_init): Likewise.
6464         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6465         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6466         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6467         (ix86_output_call_insn): Emit a nop in one more case for SEH.
6468         * config/i386/winnt.c: Include except.h.
6469         (struct seh_frame_state): Add reg_offset, after_prologue and
6470         in_cold_section fields.
6471         (i386_pe_seh_end_prologue): Set seh->after_prologue.
6472         (i386_pe_seh_cold_init): New function.
6473         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6474         to seh->in_cold_section.
6475         (seh_emit_push): Record the offset of the push.
6476         (seh_emit_save): Record the offet of the save.
6477         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6478         Test seh->after_prologue to disregard the epilogue.
6479         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6480         (i386_pe_end_cold_function): New function.
6481
6482 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
6483
6484         PR fortran/84565
6485         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6486         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6487
6488         PR c++/84704
6489         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6490         on tmp_var.
6491         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6492         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6493
6494         PR middle-end/84723
6495         * multiple_target.c: Include tree-inline.h and intl.h.
6496         (expand_target_clones): Diagnose and fail if node->definition and
6497         !tree_versionable_function_p (node->decl).
6498
6499 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
6500
6501         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6502         sprint_ul.
6503         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6504         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6505         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6506
6507 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
6508
6509         PR target/84710
6510         * combine.c (try_combine): Use reg_or_subregno instead of handling
6511         just paradoxical SUBREGs and REGs.
6512
6513 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
6514
6515          * config/arc/arc.c (arc_finalize_pic): Remove function.
6516          (arc_must_save_register): We use single base PIC register, remove
6517          checks to save/restore the PIC register.
6518          (arc_expand_prologue): Likewise.
6519          * config/arc/arc-protos.h (arc_set_default_type_attributes):
6520          Remove.
6521          (arc_verify_short): Likewise.
6522          (arc_attr_type): Likewise.
6523          * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6524          (walk_stores): Likewise.
6525          (arc_address_cost): Make it static.
6526          (arc_verify_short): Likewise.
6527          (branch_dest): Likewise.
6528          (arc_attr_type): Likewise.
6529          * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6530          (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6531          (arc_final_prescan_insn): Remove inserting the nops due to
6532          hardware hazards.  It is done in reorg step.
6533          (insn_length_variant_t): Remove.
6534          (insn_length_parameters_t): Likewise.
6535          (arc_insn_length_parameters): Likewise.
6536          (arc_get_insn_variants): Likewise.
6537          * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6538
6539 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
6540
6541         PR inline-asm/84683
6542         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6543         assertion failure.
6544
6545         PR tree-optimization/84687
6546         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6547         on new_node->decl.
6548         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6549
6550 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6551
6552         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6553         Rename to ppc_speculation_barrier.
6554         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6555         __builtin_ppc_speculation_barrier.
6556
6557 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
6558
6559         PR target/84700
6560         * combine.c (combine_simplify_rtx): Don't try to simplify if
6561         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6562         are equal to x.
6563
6564 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
6565
6566         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6567         to 32 bytes when compiling for POWER9.
6568
6569 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
6570
6571         PR target/84564
6572         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6573         regparm >= 3 with no arg reg available also for calls with
6574         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
6575
6576         PR target/84524
6577         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6578         orig,vex.
6579         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
6580
6581 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
6582
6583         PR target/84264
6584         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6585
6586 2018-03-05  Richard Biener  <rguenther@suse.de>
6587
6588         PR tree-optimization/84486
6589         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6590         When inserting a __builtin_assume_aligned call set the LHS
6591         SSA name alignment info accordingly.
6592
6593 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
6594
6595         PR tree-optimization/84114
6596         * config/aarch64/aarch64.c (aarch64_reassociation_width)
6597         Avoid reassociation of FLOAT_MODE addition.
6598
6599 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
6600
6601         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6602         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6603         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6604         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6605         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6606         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6607         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6608         and -mwbnoinvd.
6609         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6610         __builtin_ia32_wbinvd): New builtins.
6611         (SPECIAL_ARGS2): New.
6612         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6613         (SPECIAL_ARGS2): New.
6614         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6615         (ix86_valid_target_attribute_inner_p): Ditto.
6616         (ix86_init_mmx_sse_builtins): Add special_args2.
6617         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6618         TARGET_WBNOINVD_P): New.
6619         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6620         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6621         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6622         * config/i386/immintrin.h (_wbinvd): New intrinsic.
6623         * config/i386/pconfigintrin.h: New file.
6624         * config/i386/wbnoinvdintrin.h: Ditto.
6625         * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6626         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6627
6628 2018-03-05  Richard Biener  <rguenther@suse.de>
6629
6630         PR tree-optimization/84670
6631         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6632         member.
6633         (BB_VISITED_WITH_VISITED_SUCCS): New define.
6634         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6635         (compute_antic_aux): Only assert the number of values in ANTIC_IN
6636         doesn't grow if all successors (recursively) were visited at least
6637         once.
6638
6639 2018-03-05  Richard Biener  <rguenther@suse.de>
6640
6641         PR tree-optimization/84650
6642         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6643         if executed in the loop pipeline.
6644
6645 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
6646
6647         * doc/configfiles.texi (Configuration Files): Move info about
6648         conditionalizing $target-protos.h to...
6649         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
6650         differs from $target-protos.h.
6651
6652 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
6653             Chung-Ju Wu  <jasonwucj@gmail.com>
6654
6655         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6656         * config/nds32/nds32-multiple.md (setmemsi): Define.
6657         * config/nds32/nds32-memory-manipulation.c
6658         (nds32_gen_dup_4_byte_to_word_value): New.
6659         (emit_setmem_word_loop): New.
6660         (emit_setmem_byte_loop): New.
6661         (nds32_expand_setmem_loop): New.
6662         (nds32_expand_setmem_loop_v3m): New.
6663         (nds32_expand_setmem_unroll): New.
6664         (nds32_expand_setmem): New.
6665
6666 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
6667             Chung-Ju Wu  <jasonwucj@gmail.com>
6668
6669         * config/nds32/nds32-memory-manipulation.c
6670         (nds32_emit_load_store): New.
6671         (nds32_emit_post_inc_load_store): New.
6672         (nds32_emit_mem_move): New.
6673         (nds32_emit_mem_move_block): New.
6674         (nds32_expand_movmemsi_loop_unknown_size): New.
6675         (nds32_expand_movmemsi_loop_known_size): New.
6676         (nds32_expand_movmemsi_loop): New.
6677         (nds32_expand_movmemsi_unroll): New.
6678         (nds32_expand_movmemqi): Rename ...
6679         (nds32_expand_movmemsi): ... to this.
6680         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6681         (movmemsi): ... to this.
6682         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6683         (nds32_expand_movmemsi): ... to this.
6684
6685 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
6686             Monk Chiang  <sh.chiang04@gmail.com>
6687             Chung-Ju Wu  <jasonwucj@gmail.com>
6688
6689         * config/nds32/nds32-protos.h
6690         (nds32_expand_load_multiple): New arguments.
6691         (nds32_expand_store_multiple): Ditto.
6692         (nds32_valid_multiple_load_store): Rename ...
6693         (nds32_valid_multiple_load_store_p): ... to this.
6694         * config/nds32/nds32-memory-manipulation.c
6695         (nds32_expand_load_multiple): Refine implementation.
6696         (nds32_expand_store_multiple): Ditto.
6697         * config/nds32/nds32-multiple.md
6698         (load_multiple): Update nds32_expand_load_multiple interface.
6699         (store_multiple): Update nds32_expand_store_multiple interface.
6700         * config/nds32/nds32-predicates.c
6701         (nds32_valid_multiple_load_store): Rename ...
6702         (nds32_valid_multiple_load_store_p): ... to this and refine
6703         implementation.
6704         * config/nds32/predicates.md
6705         (nds32_load_multiple_and_update_address_operation): New predicate.
6706         (nds32_store_multiple_and_update_address_operation): New predicate.
6707
6708 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
6709             Chung-Ju Wu  <jasonwucj@gmail.com>
6710
6711         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6712         (combo): New attribute.
6713         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6714
6715 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
6716
6717         * config/nds32/nds32.opt: Change -mcmodel= default value.
6718
6719 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
6720             Monk Chiang  <sh.chiang04@gmail.com>
6721             Chung-Ju Wu  <jasonwucj@gmail.com>
6722
6723         * config/nds32/constants.md (unspec_element): New enum.
6724         * config/nds32/constraints.md (Umw): New constraint.
6725         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6726         * config/nds32/nds32-intrinsic.md: Likewise.
6727         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6728         (nds32_valid_smw_lwm_base_p): New.
6729         (nds32_output_smw_single_word): New.
6730         (nds32_output_lmw_single_word): New.
6731         (nds32_expand_unaligned_load): New.
6732         (nds32_expand_unaligned_store): New.
6733         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6734         (nds32_output_smw_single_word): Declare.
6735         (nds32_output_lmw_single_word): Declare.
6736         (nds32_expand_unaligned_load): Declare.
6737         (nds32_expand_unaligned_store): Declare.
6738         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6739         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6740         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6741         NDS32_BUILTIN_UASTORE_DW.
6742         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6743         predicate.
6744
6745 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
6746             Kito Cheng  <kito.cheng@gmail.com>
6747             Chung-Ju Wu  <jasonwucj@gmail.com>
6748
6749         * config/nds32/nds32-intrinsic.c
6750         (nds32_expand_builtin_null_ftype_reg): Delete.
6751         (nds32_expand_builtin_reg_ftype_imm): Ditto.
6752         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6753         (nds32_read_argument): New.
6754         (nds32_legitimize_target): Ditto.
6755         (nds32_legitimize_argument): Ditto.
6756         (nds32_check_constant_argument): Ditto.
6757         (nds32_expand_unop_builtin): Ditto.
6758         (nds32_expand_unopimm_builtin): Ditto.
6759         (nds32_expand_binop_builtin): Ditto.
6760         (nds32_builtin_decl_impl): Ditto.
6761         (builtin_description): Ditto.
6762         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6763         (nds32_init_builtins_impl): Ditto.
6764         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6765         (nds32_builtin_decl): New.
6766         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6767         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6768
6769 2018-03-02  Jeff Law  <law@redhat.com>
6770
6771         * reorg.c (stop_search_p): Handle DEBUG_INSN.
6772         (redundant_insn, fill_simple_delay_slots): Likewise.
6773         (fill_slots_from_thread): Likewise.
6774         * resource.c (mark_referenced_resources): Likewise.
6775         (mark_set_resources, find_dead_or_set_registers): Likewise.
6776
6777 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
6778
6779         * substring-locations.h (format_warning_va): Formatting fix for
6780         ATTRIBUTE_GCC_DIAG.
6781         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6782         argument.
6783         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6784         * substring-locations.c: Include intl.h.
6785         (format_warning_va): Turned into small wrapper around
6786         format_warning_n_va, renamed to ...
6787         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6788         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6789         use ngettext.
6790         (format_warning_at_substring_n): New function.
6791         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6792         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
6793         format_warning_at_substring with just a shorter name instead of
6794         const function pointer.
6795         (fmtwarn_n): New function.
6796         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6797         appropriate, get rid of all the fmtstr temporaries, move conditionals
6798         with G_() wrapped string literals directly into fmtwarn arguments,
6799         cast dir.len to (int), formatting fixes.
6800
6801 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
6802
6803         * doc/invoke.texi: Remove "Cilk Plus" references.
6804
6805 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
6806             Richard Biener  <rguenther@suse.de>
6807
6808         PR ipa/84628
6809         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6810         for error or warning attributes if CALL_FROM_THUNK_P is set.
6811         Formatting fixes.
6812
6813 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
6814
6815         PR target/56540
6816         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6817         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6818
6819         PR target/56540
6820         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6821         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6822
6823         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6824         instead of -1U in last predictors element's probability member.
6825
6826 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
6827
6828         PR ipa/83983
6829         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6830         arguments if they are comparable.
6831
6832 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
6833
6834         PR tree-optimization/84634
6835         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6836         masks and masked_loop_p with a single loop_masks, making sure it's
6837         null for bb vectorization.
6838
6839 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
6840
6841         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6842         (vect_analyze_data_ref_access): Use loop->safe_len rather than
6843         loop->force_vectorize to check whether there is no alias.
6844
6845 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
6846
6847         PR target/84614
6848         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6849         prototypes.
6850         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6851         comments.
6852         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6853         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6854         instead of a loop around prev_real_insn.
6855         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6856         prev_real_insn.
6857
6858         PR inline-asm/84625
6859         * config/i386/i386.c (ix86_print_operand): Use conditional
6860         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6861         zero vector.
6862
6863 2018-03-02  Richard Biener  <rguenther@suse.de>
6864
6865         PR tree-optimization/84427
6866         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6867         (bitmap_set_subtract_values): Rewrite to handle multiple
6868         exprs per value.
6869         (clean): Likewise.
6870         (prune_clobbered_mems): Likewise.
6871         (phi_translate): Take edge instead of pred/phiblock.
6872         (phi_translate_1): Likewise.
6873         (phi_translate_set): Likewise.  Insert all translated
6874         exprs for a value into the set, keeping possibly multiple
6875         expressions per value.
6876         (compute_antic_aux): Adjust for phi_translate changes.
6877         When intersecting union the expressions and prune those
6878         not in the final value set, keeping possibly multiple
6879         expressions per value.  Do not use value-insertion
6880         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6881         all expressions.  Add verification that the value-sets
6882         only shrink during iteration.
6883         (compute_partial_antic_aux): Adjust for the phi_translate changes.
6884         (do_pre_regular_insertion): Likewise.
6885         (do_pre_partial_partial_insertion): Likewise.
6886
6887 2018-03-02  Richard Biener  <rguenther@suse.de>
6888
6889         PR target/82005
6890         * config/darwin.c (saved_debug_info_level): New static global.
6891         (darwin_asm_lto_start): Disable debug info generation for LTO out.
6892         (darwin_asm_lto_end): Restore debug info generation settings.
6893
6894 2018-03-01  Martin Liska  <mliska@suse.cz>
6895
6896         PR sanitizer/82484
6897         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6898         volatile arguments.
6899
6900 2018-03-01  Richard Biener  <rguenther@suse.de>
6901
6902         PR debug/84645
6903         * dwarf2out.c (gen_variable_die): Properly handle late VLA
6904         type annotation with LTO when debug was disabled at compile-time.
6905
6906 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
6907
6908         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6909         XINT with INTVAL.
6910         (mips_final_postscan_insn): Likewise.
6911
6912 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
6913
6914         PR rtl-optimization/84528
6915         * alias.c (init_alias_target): Add commentary.
6916         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6917         a unique base value if the frame pointer is not eliminated
6918         to the stack pointer.
6919
6920 2018-03-01  Tom de Vries  <tom@codesourcery.com>
6921
6922         PR rtl-optimization/83327
6923         * lra-int.h (hard_regs_spilled_into): Declare.
6924         * lra.c (hard_regs_spilled_into): Define.
6925         (init_reg_info): Init hard_regs_spilled_into.
6926         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6927         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6928         (process_bb_lives): Handle hard_regs_spilled_into.
6929         (lra_create_live_ranges_1): Before doing liveness propagation, clear
6930         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6931
6932 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
6933
6934         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6935         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6936         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6937         * config/rs6000/aix72.h: New file.
6938
6939 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
6940
6941         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6942         instead of warning_at with conditional singular and plural messages
6943         where possible.
6944
6945         PR target/52991
6946         * stor-layout.c (update_alignment_for_field): For
6947         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6948         && !DECL_PACKED (field), do the alignment update, just use
6949         only desired_align instead of MAX (type_align, desired_align)
6950         as the alignment.
6951         (place_field): Don't do known_align < desired_align handling
6952         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6953         is non-NULL, instead do it after rli->prev_field handling and
6954         only if not within a bitfield word.  For DECL_PACKED (field)
6955         use type_align of BITS_PER_UNIT.
6956
6957 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
6958
6959         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6960         superfluous parentheses and trailing spaces.
6961
6962 2018-02-28  Richard Biener  <rguenther@suse.de>
6963
6964         PR tree-optimization/84584
6965         * graphite-scop-detection.c (scop_detection::add_scop): Discard
6966         SCoPs with fake exit edge.
6967
6968 2018-02-28  Martin Liska  <mliska@suse.cz>
6969
6970         PR testsuite/84597
6971         * timevar.c (timer::print): Fix format to properly print 100%
6972         values.
6973
6974 2018-02-28  Richard Biener  <rguenther@suse.de>
6975
6976         PR middle-end/84607
6977         * genmatch.c (capture_info::walk_match): Do not mark
6978         captured expressions without operands as expr_p given
6979         they act more like predicates and should be subject to
6980         "lost tail" side-effect preserving.
6981
6982 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
6983
6984         PR rtl-optimization/81611
6985         * auto-inc-dec.c (attempt_change): Move dead note from
6986         mem_insn if it's the next use of regno
6987         (find_address): Take address use of reg holding
6988         non-incremented value.  Add parm to limit search to the named
6989         reg only.
6990         (merge_in_block): Attempt to use a mem insn that is the next
6991         use of the original regno.
6992
6993 2018-02-27  Martin Sebor  <msebor@redhat.com>
6994
6995         PR c++/83871
6996         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6997         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6998
6999 2018-02-27  Martin Sebor  <msebor@redhat.com>
7000
7001         PR translation/84207
7002         * diagnostic-core.h (warning_n, error_n, inform_n): Change
7003         n argument to unsigned HOST_WIDE_INT.
7004         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7005         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
7006         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7007         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7008
7009 2018-02-27  Richard Biener  <rguenther@suse.de>
7010
7011         PR tree-optimization/84512
7012         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7013         Do not use the estimate returned from record_stmt_cost for
7014         the scalar iteration cost but sum properly using add_stmt_cost.
7015
7016 2018-02-27  Richard Biener  <rguenther@suse.de>
7017
7018         PR tree-optimization/84466
7019         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7020         Adjust last change to less strictly validate use operands.
7021
7022 2018-02-27  Martin Liska  <mliska@suse.cz>
7023
7024         PR gcov-profile/84548
7025         * gcov.c (process_file): Allow partial overlap and consider it
7026         also as group functions.
7027         (output_lines): Properly calculate range of lines for a group.
7028
7029 2018-02-27  Martin Liska  <mliska@suse.cz>
7030
7031         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7032         'ggc' suffixes.  Change first column width.
7033         (timer::print): Fix formatting of the column.
7034
7035 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
7036
7037         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7038         preserve inline entry blocks for the sake of debug inline
7039         entry point markers alone.
7040         (remove_unused_locals): Suggest in comments a better place to
7041         force the preservation of inline entry blocks that are
7042         otherwise unused, but do not preserve them.
7043
7044 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7045
7046         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7047
7048 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7049
7050         PR target/84039
7051         * config/i386/constraints.md (Bs): Replace
7052         ix86_indirect_branch_register with
7053         TARGET_INDIRECT_BRANCH_REGISTER.
7054         (Bw): Likewise.
7055         * config/i386/i386.md (indirect_jump): Likewise.
7056         (tablejump): Likewise.
7057         (*sibcall_memory): Likewise.
7058         (*sibcall_value_memory): Likewise.
7059         Peepholes of indirect call and jump via memory: Likewise.
7060         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7061         (*sibcall_value_GOT_32): Likewise.
7062         * config/i386/predicates.md (indirect_branch_operand): Likewise.
7063         (GOT_memory_operand): Likewise.
7064         (call_insn_operand): Likewise.
7065         (sibcall_insn_operand): Likewise.
7066         (GOT32_symbol_operand): Likewise.
7067         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7068
7069 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
7070
7071         PR rtl-optimization/83496
7072         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7073         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
7074         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7075         redundant insn, if any.
7076         (relax_delay_slots): Likewise.
7077         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7078
7079 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
7080
7081         PR tree-optimization/83965
7082         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7083         that grouped statements are part of a reduction chain.  Return
7084         true if the statement is not marked as a reduction itself but
7085         is part of a group.
7086         (vect_recog_dot_prod_pattern): Don't check whether the statement
7087         is part of a group here.
7088         (vect_recog_sad_pattern): Likewise.
7089         (vect_recog_widen_sum_pattern): Likewise.
7090
7091 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
7092
7093         PR debug/84545
7094         * final.c (rest_of_clean_state): Also look for calls inside sequences.
7095
7096 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7097
7098         PR target/84530
7099         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7100         the bool argument.
7101         (ix86_output_indirect_function_return): New prototype.
7102         (ix86_split_simple_return_pop_internal): Likewise.
7103         * config/i386/i386.c (indirect_return_via_cx): New.
7104         (indirect_return_via_cx_bnd): Likewise.
7105         (indirect_thunk_name): Handle return va CX_REG.
7106         (output_indirect_thunk_function): Create alias for
7107         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7108         (ix86_output_indirect_jmp): Remove the bool argument.
7109         (ix86_output_indirect_function_return): New function.
7110         (ix86_split_simple_return_pop_internal): Likewise.
7111         * config/i386/i386.md (*indirect_jump): Don't pass false
7112         to ix86_output_indirect_jmp.
7113         (*tablejump_1): Likewise.
7114         (simple_return_pop_internal): Change it to define_insn_and_split.
7115         Call ix86_split_simple_return_pop_internal to split it for
7116         -mfunction-return=.
7117         (simple_return_indirect_internal): Call
7118         ix86_output_indirect_function_return instead of
7119         ix86_output_indirect_jmp.
7120
7121 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
7122
7123         PR bootstrap/84405
7124         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7125         memset and value initialization afterwards.
7126
7127 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
7128
7129         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7130
7131 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
7132
7133         PR target/84521
7134         * common/config/aarch64/aarch64-common.c
7135         (aarch_option_optimization_table[]): Switch
7136         off fomit-frame-pointer
7137
7138 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
7139             Chung-Ju Wu  <jasonwucj@gmail.com>
7140
7141         * config/nds32/nds32-multiple.md (load_multiple): Disallow
7142         volatile memory.
7143         (store_multiple): Ditto.
7144
7145 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
7146
7147         * config.gcc: Add --with-cpu support for nds32 target.
7148         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7149         * config/nds32/nds32.opt: Add -mcpu= option.
7150
7151 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
7152
7153         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7154         isel=yes): Warn for these deprecated options.
7155
7156 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
7157
7158         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7159         ISA_2_5_MASKS_EMBEDDED.
7160
7161 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
7162
7163         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7164         p->max as pointers rather than using iterative_hash_expr.
7165
7166 2018-02-23  Carl Love  <cel@us.ibm.com>
7167
7168         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7169         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7170         BU_P8V_OVERLOAD_2.
7171         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7172         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7173         P8V_BUILTIN_VEC_VUNSIGNED2.
7174
7175 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
7176
7177         PR target/81572
7178         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7179         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7180         LRA_UNKNOWN_ALT.
7181         * lra-constraints.c (curr_insn_transform): Set up
7182         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
7183         LRA_UNKNOWN_ALT.
7184         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7185         * lra-eliminations.c (spill_pseudos): Ditto.
7186         (process_insn_for_elimination): Ditto.
7187         * lra-lives.c (reg_early_clobber_p): Use the new macros.
7188         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7189         LRA_NON_CLOBBERED_ALT.
7190
7191 2018-02-22  Martin Sebor  <msebor@redhat.com>
7192
7193         PR tree-optimization/84480
7194         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7195         to maybe_diag_stxncpy_trunc.  Call it.
7196         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7197         from gimple_fold_builtin_strcpy.  Print inlining stack.
7198         (handle_builtin_stxncpy): Print inlining stack.
7199         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7200
7201 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
7202
7203         PR target/84176
7204         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7205         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7206         and -fcheck-pointer-bounds are used together.
7207         (indirect_thunk_prefix): New enum.
7208         (indirect_thunk_need_prefix): New function.
7209         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
7210         "_nt" instead of "_bnd" for NOTRACK prefix.
7211         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7212         (output_indirect_thunk_function): Likewise.
7213         (): Likewise.
7214         (ix86_code_end): Update output_indirect_thunk_function calls.
7215         (ix86_output_indirect_branch_via_reg): Replace
7216         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7217         (ix86_output_indirect_branch_via_push): Likewise.
7218         (ix86_output_function_return): Likewise.
7219         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7220         incompatible with -fcf-protection=branch and
7221         -fcheck-pointer-bounds.
7222
7223 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
7224
7225         PR target/83335
7226         * config/aarch64/aarch64.c (aarch64_print_address_internal):
7227         Change gcc_assert call to output_operand_lossage.
7228
7229 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
7230
7231         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7232
7233 2018-02-22  DJ Delorie  <dj@redhat.com>
7234             Sebastian Perta  <sebastian.perta@renesas.com>
7235             Oleg Endo  <olegendo@gcc.gnu.org>
7236
7237         * config/rx/rx.c (rx_rtx_costs): New function.
7238         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7239
7240 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7241
7242         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7243
7244 2018-02-22  Martin Liska  <mliska@suse.cz>
7245
7246         PR driver/83193
7247         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7248         Add "native" as a possible value.
7249
7250 2018-02-22  Martin Liska  <mliska@suse.cz>
7251
7252         PR driver/83193
7253         * config/i386/i386.c (ix86_option_override_internal):
7254         Add "native" as a possible value for -march and -mtune.
7255
7256 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
7257
7258         PR target/84502
7259         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7260         to all type variants.
7261
7262         PR tree-optimization/84503
7263         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7264         width as info->bitpos + info->bitsize - start.
7265         (merged_store_group::merge_overlapping): Simplify width computation.
7266         (check_no_overlap): New function.
7267         (imm_store_chain_info::try_coalesce_bswap): Compute expected
7268         start + width and last_order of the group, fail if check_no_overlap
7269         fails.
7270         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7271         to group if check_no_overlap fails.
7272
7273 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7274
7275         * config/rs6000/altivec.md: Delete contraint arguments to
7276         define_expand, define_split, and define_peephole2, and in
7277         define_insn_and_split if always unused.
7278         * config/rs6000/darwin.md: Ditto.
7279         * config/rs6000/dfp.md: Ditto.
7280         * config/rs6000/rs6000.md: Ditto.
7281         * config/rs6000/sync.md: Ditto.
7282         * config/rs6000/vector.md: Ditto.
7283         * config/rs6000/vsx.md: Ditto.
7284
7285 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7286
7287         * config/rs6000/altivec.md: Write output control strings as braced
7288         blocks instead of double-quoted strings.
7289         * config/rs6000/darwin.md: Ditto.
7290         * config/rs6000/rs6000.md: Ditto.
7291         * config/rs6000/vector.md: Ditto.
7292         * config/rs6000/vsx.md: Ditto.
7293
7294 2018-02-21  Jason Merrill  <jason@redhat.com>
7295
7296         PR c++/84314 - ICE with templates and fastcall attribute.
7297         * attribs.c (build_type_attribute_qual_variant): Remove assert.
7298
7299 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7300
7301         * ipa-cp.c (determine_versionability): Fix comment typos.
7302
7303 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7304
7305         PR c/84229
7306         * ipa-cp.c (determine_versionability): Do not version functions caling
7307         va_arg_pack.
7308
7309 2018-02-21  Martin Liska  <mliska@suse.cz>
7310
7311         PR driver/83193
7312         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7313         Add "native" as a possible value.
7314         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
7315         the macro when native cpu detection is available.
7316
7317 2018-02-21  Martin Liska  <mliska@suse.cz>
7318
7319         PR driver/83193
7320         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7321         Add "native" as a possible value.
7322         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7323         when native cpu detection is available.
7324
7325 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
7326             Martin Sebor  <msebor@redhat.com>
7327
7328         PR tree-optimization/84478
7329         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7330         false.
7331         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7332         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7333         support which is conservatively correct, for 2 only stay conservative
7334         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
7335         argument to the 2 argument get_range_strlen, adjust 6 arg
7336         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7337         false.
7338         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7339         (gimple_fold_builtin_strlen): Pass true as last argument to
7340         get_range_strlen.
7341
7342 2018-02-20  Martin Sebor  <msebor@redhat.com>
7343
7344         PR middle-end/84095
7345         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7346         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
7347         (builtin_memref::builtin_memref): Factor out parts into
7348         set_base_and_offset and call it.
7349
7350 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7351
7352         PR middle-end/84406
7353         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7354         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7355         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
7356         search at the associated MODE_INT.
7357
7358 2018-02-20  Jeff Law  <law@redhat.com>
7359
7360         PR middle-end/82123
7361         PR tree-optimization/81592
7362         PR middle-end/79257
7363         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7364         for range data rather than using global data.
7365         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7366         range data rather than using global data.
7367         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7368         pass it to children as needed.
7369         (struct directive::fmtresult): Similarly.
7370         (struct directive::set_width): Similarly.
7371         (struct directive::set_precision): Similarly.
7372         (format_integer, format_directive, parse_directive): Similarly.
7373         (format_none): Accept unnamed vr_values parameter.
7374         (format_percent, format_floating, format_character): Similarly.
7375         (format_string, format_plain): Similarly.
7376         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7377         the EVRP range analyzer for range data rather than using global data.
7378         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7379         gimple-ssa-evrp-analyze.h
7380         (class sprintf_dom_walker): Add after_dom_children member function.
7381         Add evrp_range_analyzer member.
7382         (sprintf_dom_walker::before_dom_children): Call into the EVRP
7383         range analyzer as needed.
7384         (sprintf_dom_walker::after_dom_children): New member function.
7385         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7386         if not optimizing.
7387         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7388         (evrp_range_analyzer::pop_to_marker): Likewise.
7389
7390 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7391
7392         PR tree-optimization/84419
7393         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7394         with the required type if its current type is compatible but
7395         different.
7396
7397 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
7398
7399         PR middle-end/82004
7400         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7401         after vectorization.
7402
7403 2018-02-20  Martin Liska  <mliska@suse.cz>
7404
7405         PR driver/83193
7406         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7407         possible values if we don't have a hint.
7408
7409 2018-02-20  Martin Liska  <mliska@suse.cz>
7410
7411         PR c/84310
7412         PR target/79747
7413         * final.c (shorten_branches): Build align_tab array with one
7414         more element.
7415         * opts.c (finish_options): Add alignment option limit check.
7416         (MAX_CODE_ALIGN): Likewise.
7417         (MAX_CODE_ALIGN_VALUE): Likewise.
7418         * doc/invoke.texi: Document maximum allowed option value for
7419         all -falign-* options.
7420
7421 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
7422
7423         PR target/84146
7424         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7425         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7426         * var-tracking.c (emit_note_insn_var_location): Remove all references
7427         to NOTE_INSN_CALL_ARG_LOCATION.
7428         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7429         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7430         Use copy_rtx_if_shared.
7431         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7432         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7433         (dwarf2out_var_location): Remove handling of
7434         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7435         on call_insn.
7436         * final.c (final_scan_insn): Remove all references to
7437         NOTE_INSN_CALL_ARG_LOCATION.
7438         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
7439         before dumping final insns.
7440         * except.c (emit_note_eh_region_end): Remove all references to
7441         NOTE_INSN_CALL_ARG_LOCATION.
7442         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7443         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7444         * config/arc/arc.c (hwloop_optimize): Likewise.
7445         * config/arm/arm.c (create_fix_barrier): Likewise.
7446         * config/s390/s390.c (s390_chunkify_start): Likewise.
7447         * config/sh/sh.c (find_barrier): Likewise.
7448         * config/i386/i386.c (rest_of_insert_endbranch,
7449         ix86_seh_fixup_eh_fallthru): Likewise.
7450         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7451         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7452         * config/frv/frv.c (frv_function_prologue): Likewise.
7453         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
7454         reg note.
7455         (note_outside_basic_block_p): Remove all references to
7456         NOTE_INSN_CALL_ARG_LOCATION.
7457         * gengtype.c (adjust_field_rtx_def): Likewise.
7458         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7459         Likewise.
7460         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7461         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7462
7463         PR c++/84444
7464         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7465         is ADDR_EXPR.
7466
7467         PR tree-optimization/84452
7468         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7469         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7470         is NULL.
7471
7472 2018-02-19  Martin Liska  <mliska@suse.cz>
7473
7474         PR sanitizer/82183
7475         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7476
7477 2018-02-19  Martin Liska  <mliska@suse.cz>
7478             Richard Sandiford  <richard.sandiford@linaro.org>
7479
7480         PR tree-optimization/82491
7481         * gimple-fold.c (get_base_constructor): Make earlier bail out
7482         to prevent ubsan.
7483
7484 2018-02-19  Carl Love  <cel@us.ibm.com>
7485
7486         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7487         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7488         BU_P8V_OVERLOAD_1.
7489         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7490         P8V_BUILTIN_VEC_NEG.
7491
7492 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
7493
7494         * config/rl78/rl78.md (movdf): New define expand.
7495
7496 2018-02-19  Martin Liska  <mliska@suse.cz>
7497
7498         PR other/80589
7499         * doc/invoke.texi: Fix typo.
7500         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7501
7502 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
7503
7504         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7505         handle rs6000_single_float and rs6000_double_float specially for
7506         e500 family CPUs.
7507
7508 2018-02-16  Jeff Law  <law@redhat.com>
7509
7510         * config/rx/rx.c (add_pop_cfi_notes): New function.;
7511         (pop_regs): Use it.
7512
7513 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
7514
7515         PR ipa/84425
7516         * ipa-inline.c (inline_small_functions): Fix a typo.
7517
7518 2018-02-16  Nathan Sidwell  <nathan@acm.org>
7519
7520         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7521
7522 2018-02-16  Carl Love  <cel@us.ibm.com>
7523
7524         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7525         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7526         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7527         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7528         expansion to P8V_BUILTIN_VEC_FLOAT2.
7529
7530 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
7531
7532         PR rtl-optimization/70023
7533         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7534         src_regno into account.
7535
7536 2018-02-16  Carl Love  <cel@us.ibm.com>
7537
7538         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7539         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7540         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7541         * config/rs6000/rs6000.c: Remove case statements for
7542         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7543         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7544         and P9V_BUILTIN_VEC_VINSERT4B.
7545         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7546         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7547         * config/rs6000/vsx.md:
7548         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7549         vec_insert4b.
7550
7551 2018-02-16  Carl Love  <cel@us.ibm.com>
7552
7553         * config/rs6000/altivec.h: Add builtin names vec_extract4b
7554         vec_insert4b.
7555         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7556         definitions.
7557         * config/rs6000/rs6000-c.c: Add the definitions for
7558         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7559         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7560         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7561         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
7562         definition for insert4b and define insn *insert3b_internal.
7563         * doc/extend.texi: Add documentation for vec_extract4b.
7564
7565 2018-02-16  Nathan Sidwell  <nathan@acm.org>
7566
7567         * doc/extend.texi (Backwards Compatibility): Mention friend
7568         injection.  Note for-scope is deprecated.
7569         * doc/invoke.texi (-ffriend-injection): Deprecate.
7570
7571 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
7572
7573         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7574         that moved to I2, also allow destinations that are a paradoxical
7575         subreg (instead of a normal reg).
7576
7577 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
7578
7579         PR target/83831
7580         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7581         to QImode.
7582
7583 2018-02-16  Richard Biener  <rguenther@suse.de>
7584
7585         PR tree-optimization/84037
7586         PR tree-optimization/84016
7587         PR target/82862
7588         * config/i386/i386.c (ix86_builtin_vectorization_cost):
7589         Adjust vec_construct for the fact we need additional higher latency
7590         128bit inserts for AVX256 and AVX512 vector builds.
7591         (ix86_add_stmt_cost): Scale vector construction cost for
7592         elementwise loads.
7593
7594 2018-02-16  Richard Biener  <rguenther@suse.de>
7595
7596         PR tree-optimization/84417
7597         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7598         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7599         (non_rewritable_lvalue_p): Likewise, use poly-ints.
7600
7601 2018-02-16  Martin Liska  <mliska@suse.cz>
7602
7603         PR sanitizer/84307
7604         * internal-fn.def (ASAN_CHECK): Set proper flags.
7605         (ASAN_MARK): Likewise.
7606
7607 2018-02-16  Julia Koval  <julia.koval@intel.com>
7608
7609         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7610         from PTA_CANNONLAKE.
7611
7612 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
7613
7614         PR target/84272
7615         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7616         Use ++iter rather than iter++ for std::list iterators.
7617         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
7618         defer deleting them until all nodes in the forest are processed.  Do
7619         free even leaf nodes.  Change to_process into auto_vec.
7620
7621         PR bootstrap/84405
7622         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7623         * vec.h (vec_default_construct): Use memset instead of placement new
7624         if BROKEN_VALUE_INITIALIZATION is defined.
7625         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7626         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7627         is defined.
7628
7629         PR rtl-optimization/83723
7630         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7631         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
7632         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
7633         recursive calls.
7634         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7635         callers.
7636         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7637
7638 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
7639
7640         PR rtl-optimization/81443
7641         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7642         from inner REGs to paradoxical SUBREGs.
7643
7644 2018-02-16  Richard Biener  <rguenther@suse.de>
7645
7646         PR tree-optimization/84399
7647         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7648         For operands we can analyze at their definition make sure we can
7649         analyze them at each use as well.
7650
7651 2018-02-16  Richard Biener  <rguenther@suse.de>
7652
7653         PR tree-optimization/84190
7654         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7655         volatile accesses if the decl isn't volatile.
7656
7657 2018-02-15  Jason Merrill  <jason@redhat.com>
7658
7659         PR c++/84314 - ICE with templates and fastcall attribute.
7660         * attribs.c (build_type_attribute_qual_variant): Don't clobber
7661         TYPE_CANONICAL on an existing type.
7662
7663 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
7664
7665         PR tree-optimization/84383
7666         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7667         dstoff nor call operand_equal_p if dstbase is NULL.
7668
7669         PR tree-optimization/84334
7670         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7671         also a CONSTANT_CLASS_P, punt.
7672
7673 2018-02-14  Jim Wilson  <jimw@sifive.com>
7674
7675         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7676         first SMALL_OPERAND check.  New local min_second_step.  Move assert
7677         to where locals are set.  Add TARGET_RVC support.
7678         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7679
7680 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
7681
7682         * doc/invoke.texi: Correct -Wformat-overflow code sample.
7683
7684 2018-02-14  Martin Sebor  <msebor@redhat.com>
7685
7686         PR tree-optimization/83698
7687         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7688         arrays constrain the offset range to their bounds.
7689         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7690         (builtin_access::overlap): Avoid setting the size of overlap if it's
7691         already been set.
7692         (maybe_diag_overlap): Also consider arrays when deciding what values
7693         of offsets to include in diagnostics.
7694
7695 2018-02-14  Martin Sebor  <msebor@redhat.com>
7696
7697         PR c/84108
7698         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7699         that correspond to the kind of a declaration.
7700
7701 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
7702
7703         PR target/83984
7704         * config/pa/pa.md: Load address of PIC label using the linkage table
7705         if the label is nonlocal.
7706
7707 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7708
7709         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7710         warning message if user requests -maltivec=be.
7711         * doc/invoke.texi: Document deprecation of -maltivec=be.
7712
7713 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
7714
7715         PR target/84220
7716         * config/rs6000/rs6000-c.c: Update definitions for
7717         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7718         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7719
7720 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7721
7722         PR target/84239
7723         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7724         add _get_ssp intrinsics. Remove argument from
7725         __builtin_ia32_rdssp[d|q].
7726         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7727         * config/i386/i386-builtin.def: Remove argument from
7728         __builtin_ia32_rdssp[d|q].
7729         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7730         ix86_expand_special_args_builtin for _rdssp[d|q].
7731         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7732         Clear register before usage.
7733         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7734         Add documentation for new _get_ssp and _inc_ssp intrinsics.
7735
7736 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
7737
7738         PR tree-optimization/84357
7739         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7740         operand 1 of an ARRAY_REF too.
7741
7742 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
7743
7744         PR target/83831
7745         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7746         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7747         declarations.
7748         (set_of_reg): New struct.
7749         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7750         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7751         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7752         functions.
7753         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7754         Split into bitclr, bitset, bitinvert patterns if appropriate.
7755         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7756         use rx_fuse_in_memory_bitop.
7757         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7758         to named insn, correct maximum insn length.
7759
7760 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
7761
7762         PR target/79242
7763         * machmode.def: Define a complex mode for PARTIAL_INT.
7764         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7765         MODE_PARTIAL_INT.
7766         * doc/rtl.texi: Document CSPImode.
7767         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7768         handling.
7769         (msp430_hard_regno_nregs_with_padding): Likewise.
7770
7771 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
7772
7773         PR target/84279
7774         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7775
7776 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
7777
7778         PR rtl-optimization/84169
7779         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
7780         we generated a parallel as new i3 and we split that to new i2 and i3
7781         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
7782         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
7783         those to i2, not i1.  Partially rewrite this scan code.
7784
7785 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
7786
7787         PR c/82210
7788         * stor-layout.c (place_field): For variable length fields, adjust
7789         offset_align afterwards not just based on the field's alignment,
7790         but also on the size.
7791
7792         PR middle-end/84309
7793         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7794         of exps and logs in the use_exp2 case.
7795
7796 2018-02-13  Jeff Law  <law@redhat.com>
7797
7798         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7799         entry for "vector".
7800
7801         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7802         ARGS as unused.
7803
7804 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
7805
7806         PR debug/84342
7807         PR debug/84319
7808         * common.opt (gas-loc-support, gas-locview-support): New.
7809         (ginline-points, ginternal-reset-location-views): New.
7810         * doc/invoke.texi: Document them.  Use @itemx where intended.
7811         (gvariable-location-views): Adjust.
7812         * target.def (reset_location_view): New.
7813         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7814         (TARGET_RESET_LOCATION_VIEW): New.
7815         * doc/tm.texi: Rebuilt.
7816         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7817         (dwarf2out_default_as_locview_support): New.
7818         (output_asm_line_debug_info): Use option variables.
7819         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7820         (output_loc_list): Likewise.
7821         (add_high_low_attributes): Check option variables.
7822         Don't output entry view attribute in strict mode.
7823         (gen_inlined_subroutine_die): Check option variables.
7824         (dwarf2out_inline_entry): Likewise.
7825         (init_sections_and_labels): Likewise.
7826         (dwarf2out_early_finish): Likewise.
7827         (maybe_reset_location_view): New, from...
7828         (dwarf2out_var_location): ... here.  Call it.
7829         * debug.h (dwarf2out_default_as_loc_support): Declare.
7830         (dwarf2out_default_as_locview_support): Declare.
7831         * hooks.c (hook_int_rtx_insn_0): New.
7832         * hooks.h (hook_int_rtx_insn_0): Declare.
7833         * toplev.c (process_options): Take -gas-loc-support and
7834         -gas-locview-support from dwarf2out.  Enable
7835         -gvariable-location-views by default only with locview
7836         assembler support.  Enable -ginternal-reset-location-views by
7837         default only if the target defines the corresponding hook.
7838         Enable -ginline-points by default if location views are
7839         enabled; force it disabled if statement frontiers are
7840         disabled.
7841         * tree-inline.c (expand_call_inline): Check option variables.
7842         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7843
7844 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
7845
7846         PR tree-optimization/84321
7847         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7848         handling.  Also check whether the anti-range contains any values
7849         that satisfy the mask; switch to a VR_RANGE if not.
7850
7851 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
7852
7853         PR sanitizer/84340
7854         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7855
7856 2018-02-13  Martin Jambor  <mjambor@suse.cz>
7857
7858         PR c++/83990
7859         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7860         of call statements, also set location of a load to a temporary.
7861
7862 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
7863
7864         * config/rl78/rl78.c (add_vector_labels): New function.
7865         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7866         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7867         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7868         which checks that no arguments are passed.
7869         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7870         * doc/extend.texi: Documentation for the new attribute.
7871
7872 2018-02-13  Andreas Schwab  <schwab@suse.de>
7873
7874         * config/riscv/linux.h (CPP_SPEC): Define.
7875
7876 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
7877
7878         PR target/84335
7879         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7880         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7881         OPTION_MASK_ISA_AES as first argument to def_builtin_const
7882         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7883         instead of OPTION_MASK_ISA_PCLMUL as first argument to
7884         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7885         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7886         temporarily for AES and PCLMUL builtins.
7887
7888         PR tree-optimization/84339
7889         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7890         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7891         Formatting fixes.
7892
7893         PR middle-end/84309
7894         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7895         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7896         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7897         inline function.
7898         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7899         inline function.
7900         * omp-simd-clone.h: New file.
7901         * omp-simd-clone.c: Include omp-simd-clone.h.
7902         (expand_simd_clones): No longer static.
7903         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7904         cgraph.h and omp-simd-clone.h.
7905         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7906         (vect_recog_widen_shift_pattern): Formatting fix.
7907         (vect_pattern_recog_1): Don't check optab for calls.
7908
7909         PR target/84336
7910         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7911         operands[2] into a REG before using gen_lowpart on it.
7912
7913 2018-02-12  Jeff Law  <law@redhat.com>
7914
7915         PR target/83760
7916         * config/sh/sh.c (find_barrier): Consider a sibling call
7917         a barrier as well.
7918
7919         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7920         successfully back substituting a reg.
7921
7922 2018-02-12  Richard Biener  <rguenther@suse.de>
7923
7924         PR tree-optimization/84037
7925         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7926         parameter, move visited init to caller.
7927         (vect_slp_analyze_operations): Separate cost from validity
7928         check, initialize visited once for all instances.
7929         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7930         for all instances.
7931         * tree-vect-stmts.c (vect_model_simple_cost): Make early
7932         out an assert.
7933         (vect_model_promotion_demotion_cost): Likewise.
7934         (vectorizable_bswap): Guard cost modeling with !slp_node
7935         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7936         SLP stmts.
7937         (vectorizable_call): Likewise.
7938         (vectorizable_conversion): Likewise.
7939         (vectorizable_assignment): Likewise.
7940         (vectorizable_shift): Likewise.
7941         (vectorizable_operation): Likewise.
7942         (vectorizable_store): Likewise.
7943         (vectorizable_load): Likewise.
7944         (vectorizable_condition): Likewise.
7945         (vectorizable_comparison): Likewise.
7946
7947 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
7948
7949         PR sanitizer/84307
7950         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7951         (ASAN_MARK): Fix fnspec to account for return value, change pointer
7952         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7953
7954 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
7955
7956         PR middle-end/83665
7957         * params.def (inline-min-speedup): Increase from 8 to 15.
7958         (max-inline-insns-auto): Decrease from 40 to 30.
7959         * ipa-split.c (consider_split): Add some buffer for function to
7960         be considered inlining candidate.
7961         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7962         default values.
7963
7964 2018-02-12  Richard Biener  <rguenther@suse.de>
7965
7966         PR tree-optimization/84037
7967         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7968         matched stmts if we cannot swap the non-matched ones.
7969
7970 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
7971
7972         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7973         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7974         _mm_maskz_scalef_round_ss): New intrinsics.
7975         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7976         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7977         __builtin_ia32_scalefss_round): Remove.
7978         (__builtin_ia32_scalefsd_mask_round,
7979         __builtin_ia32_scalefss_mask_round): New intrinsics.
7980         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7981         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7982         ((match_operand:VF_128 2 "<round_nimm_predicate>"
7983         "<round_constraint>")): Changed to ...
7984         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7985         "<round_scalar_constraint>")): ... this.
7986         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7987         %0, %1, %2<round_op3>}"): Changed to ...
7988         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7989         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7990         %2<round_scalar_mask_op3>}"): ... this.
7991         * config/i386/subst.md (round_scalar_nimm_predicate): New.
7992
7993 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
7994
7995         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7996         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7997         (_mm_maskz_sqrt_round_ss): New intrinsics.
7998         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7999         (__builtin_ia32_sqrtsd_mask_round)
8000         (__builtin_ia32_sqrtss_mask_round): New builtins.
8001         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8002         (__builtin_ia32_sqrtss_round): Remove.
8003         (__builtin_ia32_sqrtsd_mask_round)
8004         (__builtin_ia32_sqrtss_mask_round): New builtins.
8005         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8006         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8007         ((match_operand:VF_128 1 "vector_operand"
8008         "xBm,<round_constraint>")): Changed to ...
8009         ((match_operand:VF_128 1 "vector_operand"
8010         "xBm,<round_scalar_constraint>")): ... this.
8011         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8012         %0, %2, %<iptr>1<round_op3>}): Changed to ...
8013         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8014         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8015         %<iptr>1<round_scalar_mask_op3>}): ... this.
8016         ((set_attr "prefix" "<round_prefix>")): Changed to ...
8017         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8018
8019 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
8020
8021         PR target/84266
8022         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8023         Cast vec_cmpeq result to correct type.
8024         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8025         Cast vec_cmpgt result to correct type.
8026
8027 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
8028
8029         * final.c (final_scan_insn_1): Renamed from...
8030         (final_scan_insn): ... this.  New wrapper, to recover
8031         seen from the outermost call in recursive ones.
8032         * config/sparc/sparc.c (output_return): Drop seen from call.
8033         (output_sibcall): Likewise.
8034         * config/visium/visium.c (output_branch): Likewise.
8035
8036 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
8037
8038         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8039         function label.
8040
8041 2018-02-10  Alan Modra  <amodra@gmail.com>
8042
8043         PR target/84300
8044         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8045         Specify LR as an input.
8046
8047 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
8048
8049         PR sanitizer/83987
8050         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8051         remove_member_access_dummy_vars): New functions.
8052         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8053         lower_omp_1, execute_lower_omp): Use them.
8054
8055         PR rtl-optimization/84308
8056         * shrink-wrap.c (spread_components): Release todo vector.
8057
8058 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
8059
8060         PR rtl-optimization/57193
8061         * ira-color.c (struct allocno_color_data): Add member
8062         conflict_allocno_hard_prefs.
8063         (update_conflict_allocno_hard_prefs): New.
8064         (bucket_allocno_compare_func): Add a preference based on
8065         conflict_allocno_hard_prefs.
8066         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8067         (color_allocnos): Remove a dead code.  Initiate
8068         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
8069
8070 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
8071
8072         PR target/84226
8073         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8074         constraint from =wa to wa.  Avoid a subreg on the output operand,
8075         instead use a pseudo and subreg it in a move.
8076         (p9_xxbrd_<mode>): Changed to ...
8077         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8078         (p9_xxbrd_v2df): New expander.
8079         (p9_xxbrw_<mode>): Changed to ...
8080         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8081         (p9_xxbrw_v4sf): New expander.
8082
8083 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
8084
8085         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8086
8087 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
8088
8089         PR target/83926
8090         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8091         multiply in 32-bit mode.
8092         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8093         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8094         mode.
8095
8096 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
8097
8098         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8099         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8100         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8101         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8102
8103 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
8104
8105         PR lto/84213
8106         * dwarf2out.c (is_trivial_indirect_ref): New function.
8107         (dwarf2out_late_global_decl): Do not generate a location
8108         attribute for variables that have a non-trivial DECL_VALUE_EXPR
8109         and that are not defined in the current unit.
8110
8111 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
8112
8113         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8114         instead of a libcall for UNORDERED.
8115
8116 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
8117
8118         PR target/82641
8119         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8120         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8121
8122 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8123
8124         PR target/PR84295
8125         * config/s390/s390.c (s390_set_current_function): Invoke
8126         s390_indirect_branch_settings also if fndecl didn't change.
8127
8128 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
8129
8130         * config/rs6000/rs6000.md (blockage): Set length to zero.
8131
8132 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
8133
8134         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8135
8136 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
8137
8138         PR sanitizer/84285
8139         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8140         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8141         -static-lib*san.
8142
8143         PR debug/84252
8144         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8145         PARALLEL incoming that failed vt_get_decl_and_offset check.
8146
8147         PR middle-end/84237
8148         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8149         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8150         TREE_READONLY bit.
8151         (get_variable_section): For decls in named .bss* sections pass true as
8152         second argument to bss_initializer_p.
8153
8154 2018-02-09  Marek Polacek  <polacek@redhat.com>
8155             Jakub Jelinek  <jakub@redhat.com>
8156
8157         PR c++/83659
8158         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8159         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
8160         Sync some changes from cxx_fold_indirect_ref.
8161
8162 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
8163
8164         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8165         markers.
8166         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8167         (BLOCK_INLINE_ENTRY_LABEL): New.
8168         (dwarf2out_var_location): Disregard inline entry markers.
8169         (inline_entry_data): New struct.
8170         (inline_entry_data_hasher): New hashtable type.
8171         (inline_entry_data_hasher::hash): New.
8172         (inline_entry_data_hasher::equal): New.
8173         (inline_entry_data_table): New variable.
8174         (add_high_low_attributes): Add DW_AT_entry_pc and
8175         DW_AT_GNU_entry_view attributes if a pending entry is found
8176         in inline_entry_data_table.  Add old entry_pc attribute only
8177         if debug nonbinding markers are disabled.
8178         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8179         markers are enabled.
8180         (block_within_block_p, dwarf2out_inline_entry): New.
8181         (dwarf2out_finish): Check that no entries remained in
8182         inline_entry_data_table.
8183         * final.c (reemit_insn_block_notes): Handle inline entry notes.
8184         (final_scan_insn, notice_source_line): Likewise.
8185         (rest_of_clean_state): Skip inline entry markers.
8186         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8187         markers.
8188         * gimple.c (gimple_build_debug_inline_entry): New.
8189         * gimple.h (enum gimple_debug_subcode): Add
8190         GIMPLE_DEBUG_INLINE_ENTRY.
8191         (gimple_build_debug_inline_entry): Declare.
8192         (gimple_debug_inline_entry_p): New.
8193         (gimple_debug_nonbind_marker_p): Adjust.
8194         * insn-notes.def (INLINE_ENTRY): New.
8195         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8196         inline entry marker notes.
8197         (print_insn): Likewise.
8198         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8199         (INSN_DEBUG_MARKER_KIND): Likewise.
8200         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8201         * tree-inline.c (expand_call_inline): Build and insert
8202         debug_inline_entry stmt.
8203         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8204         inline entry blocks early, if nonbind markers are enabled.
8205         (dump_scope_block): Dump fragment info.
8206         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8207         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8208         (gimple_build_debug_inline_entry): New.
8209         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8210         Enable/disable inline entry points too.
8211         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8212         (DEBUG_INSN): Describe inline entry markers.
8213
8214         * common.opt (gvariable-location-views): New.
8215         (gvariable-location-views=incompat5): New.
8216         * config.in: Rebuilt.
8217         * configure: Rebuilt.
8218         * configure.ac: Test assembler for view support.
8219         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8220         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8221         * dwarf2out.c (var_loc_view): New typedef.
8222         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8223         (dwarf2out_locviews_in_attribute): New.
8224         (dwarf2out_locviews_in_loclist): New.
8225         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8226         (enum dw_line_info_opcode): Add LI_adv_address.
8227         (struct dw_line_info_table): Add view.
8228         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8229         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8230         (zero_view_p): New variable.
8231         (ZERO_VIEW_P): New macro.
8232         (output_asm_line_debug_info): New.
8233         (struct var_loc_node): Add view.
8234         (add_AT_view_list, AT_loc_list): New.
8235         (add_var_loc_to_decl): Add view param.  Test it against last.
8236         (new_loc_list): Add view params.  Record them.
8237         (AT_loc_list_ptr): Handle loc and view lists.
8238         (view_list_to_loc_list_val_node): New.
8239         (print_dw_val): Handle dw_val_class_view_list.
8240         (size_of_die): Likewise.
8241         (value_format): Likewise.
8242         (loc_list_has_views): New.
8243         (gen_llsym): Set vl_symbol too.
8244         (maybe_gen_llsym, skip_loc_list_entry): New.
8245         (dwarf2out_maybe_output_loclist_view_pair): New.
8246         (output_loc_list): Output view list or entries too.
8247         (output_view_list_offset): New.
8248         (output_die): Handle dw_val_class_view_list.
8249         (output_dwarf_version): New.
8250         (output_compilation_unit_header): Use it.
8251         (output_skeleton_debug_sections): Likewise.
8252         (output_rnglists, output_line_info): Likewise.
8253         (output_pubnames, output_aranges): Update version comments.
8254         (output_one_line_info_table): Output view numbers in asm comments.
8255         (dw_loc_list): Determine current endview, pass it to new_loc_list.
8256         Call maybe_gen_llsym.
8257         (loc_list_from_tree_1): Adjust.
8258         (add_AT_location_description): Create view list attribute if
8259         needed, check it's absent otherwise.
8260         (convert_cfa_to_fb_loc_list): Adjust.
8261         (maybe_emit_file): Call output_asm_line_debug_info for test.
8262         (dwarf2out_var_location): Reset views as needed.  Precompute
8263         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
8264         attribute.  Set view.
8265         (new_line_info_table): Reset next view.
8266         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8267         (dwarf2out_source_line): Likewise.  Output view resets and labels to
8268         the assembler, or select appropriate line info opcodes.
8269         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8270         (optimize_string_length): Catch it.  Adjust.
8271         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
8272         dw_val_class_view_list, and remove it if no longer needed.
8273         (hash_loc_list): Hash view numbers.
8274         (loc_list_hasher::equal): Compare them.
8275         (optimize_location_lists): Check whether a view list symbol is
8276         needed, and whether the locview attribute is present, and
8277         whether they match.  Remove the locview attribute if no longer
8278         needed.
8279         (index_location_lists): Call skip_loc_list_entry for test.
8280         (dwarf2out_finish): Call output_asm_line_debug_info for test.
8281         Use output_dwarf_version.
8282         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8283         (struct dw_val_node): Add val_view_list.
8284         * final.c (SEEN_NEXT_VIEW): New.
8285         (set_next_view_needed): New.
8286         (clear_next_view_needed): New.
8287         (maybe_output_next_view): New.
8288         (final_start_function): Rename to...
8289         (final_start_function_1): ... this.  Take pointer to FIRST,
8290         add SEEN parameter.  Emit param bindings in the initial view.
8291         (final_start_function): Reintroduce SEEN-less interface.
8292         (final): Rename to...
8293         (final_1): ... this.  Take SEEN parameter.  Output final pending
8294         next view at the end.
8295         (final): Reintroduce seen-less interface.
8296         (final_scan_insn): Output pending next view before switching
8297         sections or ending a block.  Mark the next view as needed when
8298         outputting variable locations.  Notify debug backend of section
8299         changes, and of location view changes.
8300         (rest_of_handle_final): Adjust.
8301         * toplev.c (process_options): Autodetect value for debug variable
8302         location views option.  Warn on incompat5 without -gdwarf-5.
8303         * doc/invoke.texi (gvariable-location-views): New.
8304         (gvariable-location-views=incompat5): New.
8305         (gno-variable-location-views): New.
8306
8307 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
8308
8309         PR tree-optimization/84136
8310         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8311         that the result of find_edge is non-NULL.
8312
8313 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8314
8315         PR target/83008
8316         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8317         storing integer register in SImode.  Fix cost of 256 and 512
8318         byte aligned SSE register store.
8319
8320 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8321
8322         * config/i386/i386.c (ix86_multiplication_cost): Fix
8323         multiplication cost for TARGET_AVX512DQ.
8324
8325 2018-02-08  Marek Polacek  <polacek@redhat.com>
8326
8327         PR tree-optimization/84238
8328         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8329         get_range_strlen.
8330
8331 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8332
8333         PR tree-optimization/84265
8334         * tree-vect-stmts.c (vectorizable_store): Don't treat
8335         VMAT_CONTIGUOUS accesses as grouped.
8336         (vectorizable_load): Likewise.
8337
8338 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8339
8340         PR tree-optimization/81635
8341         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8342         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8343         (test_round_for_mask): New functions.
8344         (wide_int_cc_tests): Call test_round_for_mask.
8345         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8346         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8347         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8348         range returned by get_range_info.
8349
8350 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
8351
8352         PR ipa/81360
8353         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8354         * symtab.c: Include builtins.h
8355         (symtab_node::output_to_lto_symbol_table_p): Move here
8356         from lto-streamer-out.c:output_symbol_p.
8357         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8358         (output_symbol_p): Move all logic to symtab.c
8359         (produce_symtab): Update.
8360
8361 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8362
8363         * config/s390/s390-opts.h (enum indirect_branch): Define.
8364         * config/s390/s390-protos.h (s390_return_addr_from_memory)
8365         (s390_indirect_branch_via_thunk)
8366         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8367         (enum s390_indirect_branch_type): Define.
8368         * config/s390/s390.c (struct s390_frame_layout, struct
8369         machine_function): Remove.
8370         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8371         (indirect_branch_table_label_no, indirect_branch_table_name):
8372         Define variables.
8373         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8374         (enum s390_indirect_branch_option): Define.
8375         (s390_return_addr_from_memory): New function.
8376         (s390_handle_string_attribute): New function.
8377         (s390_attribute_table): Add new attribute handler.
8378         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8379         (s390_indirect_branch_via_thunk): New function.
8380         (s390_indirect_branch_via_inline_thunk): New function.
8381         (s390_function_ok_for_sibcall): When jumping via thunk disallow
8382         sibling call optimization for non z10 compiles.
8383         (s390_emit_call): Force indirect branch target to be a single
8384         register.  Add r1 clobber for non-z10 compiles.
8385         (s390_emit_epilogue): Emit return jump via return_use expander.
8386         (s390_reorg): Handle JUMP_INSNs as execute targets.
8387         (s390_option_override_internal): Perform validity checks for the
8388         new command line options.
8389         (s390_indirect_branch_attrvalue): New function.
8390         (s390_indirect_branch_settings): New function.
8391         (s390_set_current_function): Invoke s390_indirect_branch_settings.
8392         (s390_output_indirect_thunk_function):  New function.
8393         (s390_code_end): Implement target hook.
8394         (s390_case_values_threshold): Implement target hook.
8395         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8396         macros.
8397         * config/s390/s390.h (struct s390_frame_layout)
8398         (struct machine_function): Move here from s390.c.
8399         (TARGET_INDIRECT_BRANCH_NOBP_RET)
8400         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8401         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8402         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8403         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8404         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8405         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8406         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8407         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8408         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8409         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8410         (mnemonic attribute): Add values which aren't recognized
8411         automatically.
8412         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8413         pattern for branch conversion.  Fix mnemonic attribute.
8414         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8415         indirect branch via thunk if requested.
8416         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8417         ("*indirect_jump"): Disable for branch conversion using out of
8418         line thunks.
8419         ("indirect_jump_via_thunk<mode>_z10")
8420         ("indirect_jump_via_thunk<mode>")
8421         ("indirect_jump_via_inlinethunk<mode>_z10")
8422         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8423         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8424         ("casesi_jump_via_inlinethunk<mode>_z10")
8425         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8426         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8427         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8428         ("*indirect2_jump"): Disable for branch conversion.
8429         ("casesi_jump"): Turn into expander and expand patterns for branch
8430         conversion.
8431         ("return_use"): New expander.
8432         ("*return"): Emit return via thunk and rename it to ...
8433         ("*return<mode>"): ... this one.
8434         * config/s390/s390.opt: Add new options and and enum for the
8435         option values.
8436
8437 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8438
8439         * lra-constraints.c (match_reload): Unconditionally use
8440         gen_lowpart_SUBREG, rather than selecting between that
8441         and equivalent gen_rtx_SUBREG code.
8442
8443 2018-02-08  Richard Biener  <rguenther@suse.de>
8444
8445         PR tree-optimization/84233
8446         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8447         changed flag instead of boguously re-using phi_inserted.
8448
8449 2018-02-08  Martin Jambor  <mjambor@suse.cz>
8450
8451         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8452         static local variables.
8453
8454 2018-02-08  Richard Biener  <rguenther@suse.de>
8455
8456         PR tree-optimization/84278
8457         * tree-vect-stmts.c (vectorizable_store): When looking for
8458         smaller vector types to perform grouped strided loads/stores
8459         make sure the mode is supported by the target.
8460         (vectorizable_load): Likewise.
8461
8462 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
8463
8464         * config/aarch64/aarch64.c (aarch64_components_for_bb):
8465         Increase LDP/STP opportunities by adding adjacent callee-saves.
8466
8467 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
8468
8469         PR rtl-optimization/84068
8470         PR rtl-optimization/83459
8471         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8472
8473 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
8474
8475         PR tree-optimization/84224
8476         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8477         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8478         non-zero arguments.
8479
8480 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
8481
8482         PR target/84113
8483         * config/rs6000/altivec.md (*restore_world): Remove LR use.
8484         * config/rs6000/predicates.md (restore_world_operation): Adjust op
8485         count, remove one USE.
8486
8487 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
8488
8489         * doc/install.texi (Configuration): Document the
8490         --with-long-double-format={ibm,ieee} PowerPC configuration
8491         options.
8492
8493         PR target/84154
8494         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8495         Convert from define_expand to be define_insn_and_split.  Rework
8496         float/double/_Float128 conversions to QI/HI/SImode to work with
8497         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
8498         conversions to QI/HImode types did a store and then a load to
8499         truncate the value.  For conversions to VSX registers, don't split
8500         the insn, instead emit the code directly.  Use the code iterator
8501         any_fix to combine signed and unsigned conversions.
8502         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8503         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8504         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8505         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8506         (fix_<mode>di2_hw): Likewise.
8507         (fixuns_<mode>di2_hw): Likewise.
8508         (fix_<mode>si2_hw): Likewise.
8509         (fixuns_<mode>si2_hw): Likewise.
8510         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8511         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8512         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8513         fix<uns>_trunc<SFDF:mode>si2_p8.
8514         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8515         used.
8516         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8517         (fix<uns>_<mode>_mem): Likewise.
8518         (fctiw<u>z_<mode>_mem): Likewise.
8519         (fix<uns>_<mode>_mem): Likewise.
8520         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8521         the register allocator from doing a direct move to the GPRs to do
8522         a store, and instead use the ISA 3.0 store byte/half-word from
8523         vector register instruction.  For IEEE 128-bit floating point,
8524         also optimize stores of 32-bit ints.
8525         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8526
8527 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
8528
8529         * genextract.c (push_pathstr_operand): New function to support
8530         [a-zA-Z].
8531         (walk_rtx): Call push_pathstr_operand.
8532         (print_path): Support [a-zA-Z].
8533
8534 2018-02-07  Richard Biener  <rguenther@suse.de>
8535
8536         PR tree-optimization/84037
8537         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8538         (cse_and_gimplify_to_preheader): Declare.
8539         (vect_get_place_in_interleaving_chain): Likewise.
8540         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8541         ivexpr_map.
8542         (_loop_vec_info::~_loop_vec_info): Delete it.
8543         (cse_and_gimplify_to_preheader): New function.
8544         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8545         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8546         (vectorizable_load): Likewise.  For grouped stores always base
8547         the IV on the first element.
8548         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8549         condition before gimplifying.
8550
8551 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
8552
8553         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8554         *DIV_EXPR and *MOD_EXPR.
8555
8556 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
8557
8558         PR target/84248
8559         * config/i386/i386.c (ix86_option_override_internal): Mask out
8560         the CF_SET bit when checking -fcf-protection.
8561
8562 2018-02-07  Tom de Vries  <tom@codesourcery.com>
8563
8564         PR libgomp/84217
8565         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8566         enough.
8567
8568 2018-02-07  Richard Biener  <rguenther@suse.de>
8569
8570         PR tree-optimization/84204
8571         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8572         this place.
8573
8574         PR tree-optimization/84205
8575         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8576         special-case isl_ast_op_zdiv_r.
8577
8578         PR tree-optimization/84223
8579         * graphite-scop-detection.c (gather_bbs::before_dom_children):
8580         Only add conditions from within the region.
8581         (gather_bbs::after_dom_children): Adjust.
8582
8583 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
8584
8585         PR target/84209
8586         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8587         * config/avr/avr.md: Only post-reload split REG-REG moves if
8588         either register is GENERAL_REG_P.
8589
8590 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
8591
8592         PR tree-optimization/84235
8593         * tree-ssa-scopedtables.c
8594         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8595         if the subtraction is performed in floating point type where NaNs are
8596         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8597         build 1.  Formatting fix.
8598
8599 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
8600
8601         PR target/84146
8602         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8603         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8604         and skip it regardless of bb boundaries.  Use CALL_P macro,
8605         don't test INSN_P (insn) together with CALL_P or JUMP_P check
8606         unnecessarily, formatting fix.
8607
8608 2018-02-06  Michael Collison  <michael.collison@arm.com>
8609
8610         * config/arm/thumb2.md:
8611         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8612         (*thumb_mov_notscc): Ditto.
8613
8614 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8615
8616         PR target/84154
8617         * config/rs6000/rs6000.md (su code attribute): Use "u" for
8618         unsigned_fix, not "s".
8619
8620 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8621
8622         * configure.ac (gcc_fn_eh_frame_ro): New function.
8623         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8624         correct .eh_frame permissions.
8625         * configure: Regenerate.
8626
8627 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
8628
8629         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8630         irrelevant options.
8631
8632 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8633
8634         * config/rs6000/rs6000.c (rs6000_option_override_internal):
8635         Display warning message for -mno-speculate-indirect-jumps.
8636
8637 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
8638
8639         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8640         Undocumented.
8641         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8642
8643 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
8644
8645         PR tree-optimization/84225
8646         * tree-eh.c (find_trapping_overflow): Only call
8647         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8648
8649 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
8650
8651         PR target/84145
8652         * config/i386/i386.c: Reimplement the check of possible options
8653         -mibt/-mshstk conbination. Change error messages.
8654         * doc/invoke.texi: Fix a typo: remove extra '='.
8655
8656 2018-02-06  Marek Polacek  <polacek@redhat.com>
8657
8658         PR tree-optimization/84228
8659         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8660
8661 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
8662
8663         PR target/82641
8664         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8665         emitted arch directives.
8666         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8667         __ARM_FEATURE_COPROC before changing architectures.
8668
8669 2018-02-06  Richard Biener  <rguenther@suse.de>
8670
8671         * config/i386/i386.c (print_reg): Fix typo.
8672         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8673
8674 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
8675
8676         * configure: Regenerate.
8677
8678 2018-02-05  Martin Sebor  <msebor@redhat.com>
8679
8680         PR tree-optimization/83369
8681         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8682         inlining context.
8683
8684 2018-02-05  Martin Liska  <mliska@suse.cz>
8685
8686         * doc/invoke.texi: Cherry-pick upstream r323995.
8687
8688 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
8689
8690         * ira.c (ira_init_register_move_cost): Adjust comment.
8691
8692 2018-02-05  Martin Liska  <mliska@suse.cz>
8693
8694         PR gcov-profile/84137
8695         * doc/gcov.texi: Fix typo in documentation.
8696
8697 2018-02-05  Martin Liska  <mliska@suse.cz>
8698
8699         PR gcov-profile/83879
8700         * doc/gcov.texi: Document necessity of --dynamic-list-data when
8701         using dlopen functionality.
8702
8703 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
8704
8705         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8706         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8707         _mm_maskz_range_ss, _mm_mask_range_round_ss,
8708         _mm_maskz_range_round_ss): New intrinsics.
8709         (__builtin_ia32_rangesd128_round)
8710         (__builtin_ia32_rangess128_round): Remove.
8711         (__builtin_ia32_rangesd128_mask_round,
8712         __builtin_ia32_rangess128_mask_round): New builtins.
8713         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8714         __builtin_ia32_rangess128_round): Remove.
8715         (__builtin_ia32_rangesd128_mask_round,
8716         __builtin_ia32_rangess128_mask_round): New builtins.
8717         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8718         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8719         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8720         "<round_saeonly_constraint>")): Changed to ...
8721         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8722         "<round_saeonly_scalar_constraint>")): ... this.
8723         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8724         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8725         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8726         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8727         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8728
8729 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
8730
8731         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8732         options.
8733         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8734         Remove all values except native, 8540 and 8548.
8735
8736 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
8737
8738         * config/i386/i386.c (ix86_output_function_return): Pass
8739         INVALID_REGNUM, instead of -1, as invalid register number to
8740         indirect_thunk_name and output_indirect_thunk.
8741
8742 2018-02-02  Julia Koval  <julia.koval@intel.com>
8743
8744         * config.gcc: Add -march=icelake.
8745         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8746         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8747         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8748         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8749         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8750         (processor_target_table): Add icelake.
8751         (ix86_option_override_internal): Handle new PTAs.
8752         (get_builtin_code_for_version): Handle icelake.
8753         (M_INTEL_COREI7_ICELAKE): New.
8754         (fold_builtin_cpu): Handle icelake.
8755         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8756         * doc/invoke.texi: Add -march=icelake.
8757
8758 2018-02-02  Julia Koval  <julia.koval@intel.com>
8759
8760         * config/i386/i386.c (ix86_option_override_internal): Change flags type
8761         to wide_int_bitmask.
8762         * wide-int-bitmask.h: New.
8763
8764 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
8765
8766         PR target/84066
8767         * config/i386/i386.md: Replace Pmode with word_mode in
8768         builtin_setjmp_setup and builtin_longjmp to support x32.
8769
8770 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
8771
8772         PR target/56010
8773         PR target/83743
8774         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8775         #include "opts.h".
8776         (rs6000_supported_cpu_names): New static variable.
8777         (linux_cpu_translation_table): Likewise.
8778         (elf_platform) <cpu>: Define new static variable and use it.
8779         Translate kernel AT_PLATFORM name to canonical name if needed.
8780         Error if platform name is unknown.
8781
8782 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
8783
8784         PR target/84089
8785         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8786
8787 2018-02-01  Jeff Law  <law@redhat.com>
8788
8789         PR target/84128
8790         * config/i386/i386.c (release_scratch_register_on_entry): Add new
8791         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
8792         the scratch if RELEASE_VIA_POP is false.
8793         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8794         If we have to save a temporary register, decrement SIZE appropriately.
8795         Pass new arguments to release_scratch_register_on_entry.
8796         (ix86_adjust_stack_and_probe): Likewise.
8797         (ix86_emit_probe_stack_range): Pass new arguments to
8798         release_scratch_register_on_entry.
8799
8800 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
8801
8802         PR rtl-optimization/84157
8803         * combine.c (change_zero_ext): Use REG_P predicate in
8804         front of HARD_REGISTER_P predicate.
8805
8806 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
8807
8808         * config/avr/avr.c (avr_option_override): Move disabling of
8809         -fdelete-null-pointer-checks to...
8810         * common/config/avr/avr-common.c (avr_option_optimization_table):
8811         ...here.
8812
8813 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8814
8815         PR tree-optimization/81635
8816         * tree-data-ref.c (split_constant_offset_1): For types that
8817         wrap on overflow, try to use range info to prove that wrapping
8818         cannot occur.
8819
8820 2018-02-01  Renlin Li  <renlin.li@arm.com>
8821
8822         PR target/83370
8823         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8824         TAILCALL_ADDR_REGS.
8825         (aarch64_register_move_cost): Likewise.
8826         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8827         TAILCALL_ADDR_REGS.
8828         (REG_CLASS_NAMES): Likewise.
8829         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8830         TAILCALL_ADDR_REGS. Remove IP registers.
8831         * config/aarch64/aarch64.md (Ucs): Update register constraint.
8832
8833 2018-02-01  Richard Biener  <rguenther@suse.de>
8834
8835         * domwalk.h (dom_walker::dom_walker): Add additional constructor
8836         for specifying RPO order and allow NULL for that.
8837         * domwalk.c (dom_walker::dom_walker): Likewise.
8838         (dom_walker::walk): Handle NULL RPO order.
8839         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8840         in RPO order.
8841         (rewrite_update_dom_walker): Likewise.
8842         (mark_def_dom_walker): Likewise.
8843
8844 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8845
8846         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8847         (aarch64_maybe_expand_sve_subreg_move): Declare.
8848         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8849         * config/aarch64/predicates.md (aarch64_any_register_operand): New
8850         predicate.
8851         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8852         that are semantically a reverse operation.
8853         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8854         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8855         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8856         functions.
8857         (aarch64_can_change_mode_class): For big-endian, forbid changes
8858         between two SVE modes if they have different element sizes.
8859
8860 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8861
8862         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8863         the TImode handling for big-endian targets.
8864
8865 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8866
8867         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8868         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
8869         not just bytes.
8870         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8871         Remove BSWAP handing for big-endian targets and use the form of
8872         LD1RQ appropariate for the mode.
8873
8874 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8875
8876         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8877         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8878         duplicated element.
8879
8880 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8881
8882         PR tearget/83845
8883         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8884         check for operands that need to go through aarch64_sve_reload_be.
8885
8886 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
8887
8888         PR tree-optimization/81661
8889         PR tree-optimization/84117
8890         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8891         * tree-eh.c: Include gimplify.h.
8892         (find_trapping_overflow, replace_trapping_overflow,
8893         rewrite_to_non_trapping_overflow): New functions.
8894         * tree-vect-loop.c: Include tree-eh.h.
8895         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8896         * tree-data-ref.c: Include tree-eh.h.
8897         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8898
8899 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
8900
8901         PR rtl-optimization/84123
8902         * combine.c (change_zero_ext): Check if hard register satisfies
8903         can_change_dest_mode before calling gen_lowpart_SUBREG.
8904
8905 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
8906
8907         PR target/82444
8908         * ira.c (ira_init_register_move_cost): Remove assert.
8909
8910 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
8911
8912         PR rtl-optimization/84071
8913         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8914         * doc/tm.texi: Regenerate.
8915
8916 2018-01-31  Richard Biener  <rguenther@suse.de>
8917
8918         PR tree-optimization/84132
8919         * tree-data-ref.c (analyze_miv_subscript): Properly
8920         check whether evolution_function_is_affine_multivariate_p
8921         before calling gcd_of_steps_may_divide_p.
8922
8923 2018-01-31  Julia Koval  <julia.koval@intel.com>
8924
8925         PR target/83618
8926         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8927         * config/i386/i386.md (rdpid_rex64) New.
8928         (rdpid): Make 32bit only.
8929
8930 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
8931
8932         PR lto/84105
8933         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8934         an IDENTIFIER_NODE for FUNCTION_TYPE's.
8935
8936 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
8937
8938         Revert
8939         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
8940
8941         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8942
8943 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
8944
8945         PR rtl-optimization/84071
8946         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8947         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8948
8949 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
8950
8951         * config/arc/arc.c (arc_handle_aux_attribute): New function.
8952         (arc_attribute_table): Add 'aux' attribute.
8953         (arc_in_small_data_p): Consider aux like variables.
8954         (arc_is_aux_reg_p): New function.
8955         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8956         (arc_get_aux_arg): New function.
8957         (prepare_move_operands): Handle aux-register access.
8958         (arc_handle_aux_attribute): New function.
8959         * doc/extend.texi (ARC Variable attributes): Add subsection.
8960
8961 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
8962
8963         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8964         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8965         (arc_attribute_table): Add 'uncached' attribute.
8966         (arc_print_operand): Print '.di' flag for uncached memory
8967         accesses.
8968         (arc_in_small_data_p): Do not consider for small data the uncached
8969         types.
8970         (arc_is_uncached_mem_p): New function.
8971         * config/arc/predicates.md (compact_store_memory_operand): Check
8972         for uncached memory accesses.
8973         (nonvol_nonimm_operand): Likewise.
8974         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8975
8976 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
8977
8978         PR c/84100
8979         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8980         falign-loops=): Add Optimization flag.
8981
8982 2018-01-30  Jeff Law  <law@redhat.com>
8983
8984         PR target/84064
8985         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8986         INT_REGISTERS_SAVED.  Check it prior to calling
8987         get_scratch_register_on_entry.
8988         (ix86_adjust_stack_and_probe): Similarly.
8989         (ix86_emit_probe_stack_range): Similarly.
8990         (ix86_expand_prologue): Corresponding changes.
8991
8992 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8993
8994         PR target/40411
8995         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8996         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8997
8998 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
8999
9000         PR target/84112
9001         * lra-constraints.c (curr_insn_transform): Process AND in the
9002         address.
9003
9004 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
9005
9006         PR rtl-optimization/83986
9007         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9008         dependence against last_pending_memory_flush in addition to
9009         pending_jump_insns.
9010
9011 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
9012
9013         PR tree-optimization/81611
9014         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9015         copies.
9016
9017 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
9018
9019         PR target/83758
9020         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9021         a reg rtx.
9022
9023 2018-01-30  Richard Biener  <rguenther@suse.de>
9024             Jakub Jelinek  <jakub@redhat.com>
9025
9026         PR tree-optimization/84111
9027         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9028         inner loops added during recursion, as they don't have up-to-date
9029         SSA form.
9030
9031 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
9032
9033         PR ipa/81360
9034         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9035         (can_inline_edge_by_limits_p): ... here.
9036         (can_early_inline_edge_p, check_callers,
9037         update_caller_keys, update_callee_keys, recursive_inlining,
9038         add_new_edges_to_heap, speculation_useful_p,
9039         inline_small_functions,
9040         inline_small_functions, flatten_function,
9041         inline_to_all_callers_1): Update.
9042
9043 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
9044
9045         * profile-count.c (profile_count::combine_with_ipa_count): Handle
9046         zeros correctly.
9047
9048 2018-01-30  Richard Biener  <rguenther@suse.de>
9049
9050         PR tree-optimization/83008
9051         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9052         invariant and constant vector uses in stmts when they need
9053         more than one stmt.
9054
9055 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9056
9057         PR bootstrap/84017
9058         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9059         * configure: Regenerate.
9060
9061 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9062
9063         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9064         pattern.
9065         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9066         Use gen_rtx_REG rather than gen_lowpart.
9067
9068 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9069
9070         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9071         rather than 0 when creating partial subregs.
9072
9073 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9074
9075         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9076         of usage.
9077
9078 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
9079
9080         PR target/81550
9081         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9082         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9083         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9084         flags.  This restores the settings used before the 2017-07-24.
9085         Turning off pre increment/decrement/modify allows IVOPTS to
9086         optimize DF/SF loops where the index is an int.
9087
9088 2018-01-29  Richard Biener  <rguenther@suse.de>
9089             Kelvin Nilsen  <kelvin@gcc.gnu.org>
9090
9091         PR bootstrap/80867
9092         * tree-vect-stmts.c (vectorizable_call): Don't call
9093         targetm.vectorize_builtin_md_vectorized_function if callee is
9094         NULL.
9095
9096 2018-01-22  Carl Love  <cel@us.ibm.com>
9097
9098         * doc/extend.tex: Fix typo in second arg in
9099         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9100
9101 2018-01-29  Richard Biener  <rguenther@suse.de>
9102
9103         PR tree-optimization/84086
9104         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9105         (flush_ssaname_freelist): When SSA names were released reset
9106         the SCEV hash table.
9107
9108 2018-01-29  Richard Biener  <rguenther@suse.de>
9109
9110         PR tree-optimization/84057
9111         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9112         removed paths when removing edges.
9113
9114 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
9115
9116         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9117         -mfunction-return=@var{choice}.
9118
9119 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9120
9121         PR diagnostic/84034
9122         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9123         Handle CR like TAB.
9124         (layout::print_source_line): Likewise.
9125         (test_get_line_width_without_trailing_whitespace): Add test cases.
9126
9127 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
9128
9129         PR middle-end/84040
9130         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9131         debug insns.
9132
9133 2018-01-26  Jim Wilson  <jimw@sifive.com>
9134
9135         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9136
9137         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9138         specified.
9139
9140 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9141
9142         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9143         and CMP + SUB-immediate -> SUBS.
9144
9145 2018-01-26  Martin Sebor  <msebor@redhat.com>
9146
9147         PR tree-optimization/83896
9148         * tree-ssa-strlen.c (get_string_len): Rename...
9149         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
9150         Avoid assuming length is constant.
9151         (handle_char_store): Use HOST_WIDE_INT for string length.
9152
9153 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
9154
9155         PR target/81763
9156         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9157         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9158
9159 2018-01-26  Richard Biener  <rguenther@suse.de>
9160
9161         PR rtl-optimization/84003
9162         * dse.c (record_store): Only record redundant stores when
9163         the earlier store aliases at least all accesses the later one does.
9164
9165 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
9166
9167         PR rtl-optimization/83985
9168         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9169         REG_CFA_RESTORE insns.
9170         (delete_unmarked_insns): Don't ignore separate shrink wrapping
9171         REG_CFA_RESTORE insns here.
9172
9173         PR c/83989
9174         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9175         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9176
9177 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9178
9179         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9180         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9181         (arc_init): Likewise.
9182         (arc_override_options): Likewise.
9183         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9184         value.
9185         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9186         support.
9187         * config/arc/arc.h (TARGET_DBNZ): Define.
9188         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9189         properly set the tune attribute.
9190         (dbnz): Use TARGET_DBNZ guard.
9191         * config/arc/arc.opt (mtune): Add core3 option.
9192
9193 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9194
9195         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9196         recognize new pic like addresses.
9197         (arc_delegitimize_address): Clean up.
9198
9199 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9200
9201         * config/arc/arc-arches.def: Option mrf16 valid for all
9202         architectures.
9203         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9204         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9205         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9206         * config/arc/arc-tables.opt: Regenerate.
9207         * config/arc/arc.c (arc_conditional_register_usage): Handle
9208         reduced register file case.
9209         (arc_file_start): Set must have build attributes.
9210         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9211         mrf16 option value.
9212         * config/arc/arc.opt (mrf16): Add new option.
9213         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9214         * config/arc/genmultilib.awk: Handle new mrf16 option.
9215         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9216         * config/arc/t-multilib: Regenerate.
9217         * doc/invoke.texi (ARC Options): Document mrf16 option.
9218
9219 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9220
9221         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9222         * config/arc/arc.c (arc_handle_secure_attribute): New function.
9223         (arc_attribute_table): Add 'secure_call' attribute.
9224         (arc_print_operand): Print secure call operand.
9225         (arc_function_ok_for_sibcall): Don't optimize tail calls when
9226         secure.
9227         (arc_is_secure_call_p): New function.  * config/arc/arc.md
9228         (call_i): Add support for sjli instruction.
9229         (call_value_i): Likewise.
9230         * config/arc/constraints.md (Csc): New constraint.
9231
9232 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9233             John Eric Martin  <John.Martin@emmicro-us.com>
9234
9235         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9236         * config/arc/arc.c (_arc_jli_section): New struct.
9237         (arc_jli_section): New type.
9238         (rc_jli_sections): New static variable.
9239         (arc_handle_jli_attribute): New function.
9240         (arc_attribute_table): Add jli_always and jli_fixed attribute.
9241         (arc_file_end): New function.
9242         (TARGET_ASM_FILE_END): Define.
9243         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9244         (arc_add_jli_section): New function.
9245         (jli_call_scan): Likewise.
9246         (arc_reorg): Call jli_call_scan.
9247         (arc_output_addsi): Remove 'S' from printing asm operand.
9248         (arc_is_jli_call_p): New function.
9249         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9250         operand.
9251         (movhi_insn): Likewise.
9252         (movsi_insn): Likewise.
9253         (movsi_set_cc_insn): Likewise.
9254         (loadqi_update): Likewise.
9255         (load_zeroextendqisi_update): Likewise.
9256         (load_signextendqisi_update): Likewise.
9257         (loadhi_update): Likewise.
9258         (load_zeroextendhisi_update): Likewise.
9259         (load_signextendhisi_update): Likewise.
9260         (loadsi_update): Likewise.
9261         (loadsf_update): Likewise.
9262         (movsicc_insn): Likewise.
9263         (bset_insn): Likewise.
9264         (bxor_insn): Likewise.
9265         (bclr_insn): Likewise.
9266         (bmsk_insn): Likewise.
9267         (bicsi3_insn): Likewise.
9268         (cmpsi_cc_c_insn): Likewise.
9269         (movsi_ne): Likewise.
9270         (movsi_cond_exec): Likewise.
9271         (clrsbsi2): Likewise.
9272         (norm_f): Likewise.
9273         (normw): Likewise.
9274         (swap): Likewise.
9275         (divaw): Likewise.
9276         (flag): Likewise.
9277         (sr): Likewise.
9278         (kflag): Likewise.
9279         (ffs): Likewise.
9280         (ffs_f): Likewise.
9281         (fls): Likewise.
9282         (call_i): Remove 'S' asm letter, add jli instruction.
9283         (call_value_i): Likewise.
9284         * config/arc/arc.op (mjli-always): New option.
9285         * config/arc/constraints.md (Cji): New constraint.
9286         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9287         operand.
9288         (subsf3_fpx): Likewise.
9289         (mulsf3_fpx): Likewise.
9290         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9291         asm operand.
9292         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9293         function attrbutes.
9294         * doc/invoke.texi (ARC): Document mjli-always option.
9295
9296 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
9297
9298         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9299         avoid addition with 0 and use incw and decw where possible.
9300
9301 2018-01-26  Richard Biener  <rguenther@suse.de>
9302
9303         PR tree-optimization/81082
9304         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9305         association if it requires casting to unsigned.
9306         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9307         from fold_plusminus_mult_expr to catch important cases late when
9308         range info is available.
9309
9310 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9311
9312         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9313         * configure.ac (hidden_linkonce): New test.
9314         * configure: Regenerate.
9315         * config.in: Regenerate.
9316
9317 2018-01-26  Julia Koval  <julia.koval@intel.com>
9318
9319         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9320         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9321         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9322         _mm_mask_bitshuffle_epi64_mask): Fix type.
9323         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9324         USI_FTYPE_V4DI_V4DI_USI): Remove.
9325         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9326         __builtin_ia32_vpshufbitqmb256_mask,
9327         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9328         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9329         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9330
9331 2018-01-26  Alan Modra  <amodra@gmail.com>
9332
9333         PR target/84033
9334         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9335         UNSPEC_VBPERMQ.  Sort other unspecs.
9336
9337 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
9338
9339         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9340
9341 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
9342
9343         PR middle-end/83055
9344         * predict.c (drop_profile): Do not push/pop cfun; update also
9345         node->count.
9346         (handle_missing_profiles): Fix logic looking for zero profiles.
9347
9348 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
9349
9350         PR middle-end/83977
9351         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9352         on functions with #pragma omp declare simd or functions with simd
9353         attribute.
9354         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9355         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9356         Remove trailing \n from warning_at calls.
9357
9358 2018-01-25  Tom de Vries  <tom@codesourcery.com>
9359
9360         PR target/84028
9361         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9362         for neutered workers.
9363
9364 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
9365
9366         PR target/68467
9367         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9368         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9369
9370 2018-01-24  Jeff Law  <law@redhat.com>
9371
9372         PR target/83994
9373         * i386.c (get_probe_interval): Move to earlier point.
9374         (ix86_compute_frame_layout): If -fstack-clash-protection and
9375         the frame is larger than the probe interval, then use pushes
9376         to save registers rather than reg->mem moves.
9377         (ix86_expand_prologue): Remove conditional for int_registers_saved
9378         assertion.
9379
9380 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
9381
9382         PR target/84014
9383         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9384         min/max for never referenced object.
9385
9386 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
9387
9388         PR middle-end/83977
9389         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9390         here.
9391         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9392         attributes from DECL_ATTRIBUTES (decl) without affecting
9393         DECL_ATTRIBUTES (current_function_decl).
9394         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9395         functions with non-NULL DECL_ABSTRACT_ORIGIN.
9396
9397 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
9398
9399         PR tree-optimization/83979
9400         * fold-const.c (fold_comparison): Use constant_boolean_node
9401         instead of boolean_{true,false}_node.
9402
9403 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
9404
9405         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9406         with zero counts.
9407
9408 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9409
9410         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9411         Simplify the clause that sets the length attribute.
9412         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9413         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9414         clause that sets the length attribute.
9415         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9416
9417 2018-01-24  Tom de Vries  <tom@codesourcery.com>
9418
9419         PR target/83589
9420         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9421         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9422         Add strict parameter.
9423         (prevent_branch_around_nothing): Insert dummy insn between branch to
9424         label and label with no ptx insn inbetween.
9425         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9426
9427 2018-01-24  Tom de Vries  <tom@codesourcery.com>
9428
9429         PR target/81352
9430         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9431         for neutered threads in warp.
9432         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9433
9434 2018-01-24  Richard Biener  <rguenther@suse.de>
9435
9436         PR tree-optimization/83176
9437         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9438         operands.
9439
9440 2018-01-24  Richard Biener  <rguenther@suse.de>
9441
9442         PR tree-optimization/82819
9443         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9444         code generating pluses that are no-ops in the target precision.
9445
9446 2018-01-24  Richard Biener  <rguenther@suse.de>
9447
9448         PR middle-end/84000
9449         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9450
9451 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
9452
9453         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9454         to merge probabilities.
9455         * predict.c (probably_never_executed): Also mark as cold functions
9456         with global 0 profile and guessed local profile.
9457         * profile-count.c (profile_probability::combine_with_count): New
9458         member function.
9459         * profile-count.h (profile_probability::operator*,
9460         profile_probability::operator*=, profile_probability::operator/,
9461         profile_probability::operator/=): Reduce precision to adjusted
9462         and set value to guessed on contradictory divisions.
9463         (profile_probability::combine_with_freq): Remove.
9464         (profile_probability::combine_wiht_count): Declare.
9465         (profile_count::force_nonzero):: Set to adjusted.
9466         (profile_count::probability_in):: Set quality to adjusted.
9467         * tree-ssa-tail-merge.c (replace_block_by): Use
9468         combine_with_count.
9469
9470 2018-01-23  Andrew Waterman  <andrew@sifive.com>
9471             Jim Wilson  <jimw@sifive.com>
9472
9473         * config/riscv/riscv.c (riscv_stack_boundary): New.
9474         (riscv_option_override): Set riscv_stack_boundary.  Handle
9475         riscv_preferred_stack_boundary_arg.
9476         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9477         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9478         (STACK_BOUNDARY): Set to riscv_stack_boundary.
9479         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9480         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9481         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9482
9483 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
9484
9485         PR target/83905
9486         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9487         of struct ix86_frame.
9488         (ix86_expand_epilogue): Likewise.  Add a local variable for
9489         the reg_save_offset field in struct ix86_frame.
9490
9491 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
9492
9493         PR tree-optimization/82604
9494         * tree-loop-distribution.c (enum partition_kind): New enum item
9495         PKIND_PARTIAL_MEMSET.
9496         (partition_builtin_p): Support above new enum item.
9497         (generate_code_for_partition): Ditto.
9498         (compute_access_range): Differentiate cases that equality can be
9499         proven at all loops, the innermost loops or no loops.
9500         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9501         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
9502         (finalize_partitions, distribute_loop): Don't fuse partition of
9503         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9504         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9505         parloop is enabled.
9506
9507 2018-01-23  Martin Liska  <mliska@suse.cz>
9508
9509         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9510         order to ignore the predictor.
9511         (PRED_POLYMORPHIC_CALL): Likewise.
9512         (PRED_RECURSIVE_CALL): Likewise.
9513
9514 2018-01-23  Martin Liska  <mliska@suse.cz>
9515
9516         * tree-profile.c (tree_profiling): Print function header to
9517         aware reader which function we are working on.
9518         * value-prof.c (gimple_find_values_to_profile): Do not print
9519         not interesting value histograms.
9520
9521 2018-01-23  Martin Liska  <mliska@suse.cz>
9522
9523         * profile-count.h (enum profile_quality): Add
9524         profile_uninitialized as the first value. Do not number values
9525         as they are zero based.
9526         (profile_count::verify): Update sanity check.
9527         (profile_probability::verify): Likewise.
9528
9529 2018-01-23  Nathan Sidwell  <nathan@acm.org>
9530
9531         * doc/invoke.texi (ffor-scope): Deprecate.
9532
9533 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
9534
9535         PR tree-optimization/83510
9536         * domwalk.c (set_all_edges_as_executable): New function.
9537         (dom_walker::dom_walker): Convert bool param
9538         "skip_unreachable_blocks" to enum reachability.  Move setup of
9539         edge flags to set_all_edges_as_executable and only do it when
9540         reachability is REACHABLE_BLOCKS.
9541         * domwalk.h (enum dom_walker::reachability): New enum.
9542         (dom_walker::dom_walker): Convert bool param
9543         "skip_unreachable_blocks" to enum reachability.
9544         (set_all_edges_as_executable): New decl.
9545         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
9546         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9547         "reachability".
9548         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9549         but converting true to REACHABLE_BLOCKS.
9550         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9551         * tree-vrp.c
9552         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9553         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9554         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9555         REACHABLE_BLOCKS.
9556         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9557         if check_all_array_refs will be called.
9558
9559 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
9560
9561         * tree.c (selftest::test_location_wrappers): Add more test
9562         coverage.
9563
9564 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
9565
9566         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9567         (selftest::test_bit_in_range): Likewise.
9568
9569 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
9570
9571         PR testsuite/83888
9572         * doc/sourcebuild.texi (vect_float): Say that the selector
9573         only describes the situation when -funsafe-math-optimizations is on.
9574         (vect_float_strict): Document.
9575
9576 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
9577
9578         PR tree-optimization/83965
9579         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9580         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9581         instead of checking only for a reduction.
9582         (vect_recog_widen_sum_pattern): Likewise.
9583
9584 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
9585
9586         * predict.c (probably_never_executed): Only use precise profile info.
9587         (compute_function_frequency): Skip after inlining hack since we now
9588         have quality checking.
9589
9590 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
9591
9592         * profile-count.h (profile_probability::very_unlikely,
9593         profile_probability::unlikely, profile_probability::even): Set
9594         precision to guessed.
9595
9596 2018-01-23  Richard Biener  <rguenther@suse.de>
9597
9598         PR tree-optimization/83963
9599         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9600         Properly terminate dominator walk when crossing the exit edge not
9601         when visiting its source block.
9602
9603 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
9604
9605         PR c++/83918
9606         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9607         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9608
9609 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
9610
9611         PR tree-optimization/83957
9612         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
9613         semicolon after for body surrounded by braces.
9614
9615         PR tree-optimization/83081
9616         * profile-count.h (profile_probability::split): New method.
9617         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9618         Use profile_probability::split.
9619         (do_compare_rtx_and_jump): Fix adjustment of probabilities
9620         when splitting a single conditional jump into 2.
9621
9622 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
9623
9624         PR tree-optimization/69452
9625         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9626         decl.
9627
9628 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9629
9630         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9631         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9632         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9633
9634 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9635
9636         * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9637         * config/rl78/rl78.md (movdi): New define_expand.
9638         * config/rl78/rl78.c (rl78_split_movdi): New function.
9639
9640 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
9641
9642         PR target/83862
9643         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9644         no longer used.
9645         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9646         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9647         128-bit to produce an UNSPEC move to get the double word with the
9648         signbit and then a shift directly to do signbit.
9649         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9650         implementation with a new version that just does either a direct
9651         move or a regular move.  Move memory interface to separate insns.
9652         Move insns so they are next to the expander.
9653         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9654         with signbit move.  Split big and little endian case.
9655         (signbit<mode>2_dm_mem_le): Likewise.
9656         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9657         (signbit<mode>2_dm2): Likewise.
9658
9659 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9660
9661         * config/rl78/rl78.md (anddi3): New define_expand.
9662
9663 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9664
9665         * config/rl78/rl78.md (umindi3): New define_expand.
9666
9667 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9668
9669         * config/rl78/rl78.md (smindi3): New define_expand.
9670
9671 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9672
9673         * config/rl78/rl78.md (smaxdi3): New define_expand.
9674
9675 2018-01-22  Carl Love  <cel@us.ibm.com>
9676
9677         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9678         LVX_V1TI): Add macro expansion.
9679         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9680         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9681         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9682         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9683         Change check to determine if the instruction is a byte reversing
9684         entry.  Fix typo in comment.
9685         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9686         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9687         Add def_builtin calls for new builtins.
9688         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9689         Add define_insn expansion.
9690
9691 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9692
9693         * config/rl78/rl78.md (umaxdi3): New define_expand.
9694
9695 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9696
9697         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9698         for non-QImode registers.
9699
9700 2018-01-22  Richard Biener  <rguenther@suse.de>
9701
9702         PR tree-optimization/83963
9703         * graphite-scop-detection.c (scop_detection::get_sese): Delay
9704         including the loop exit block.
9705         (scop_detection::merge_sese): Likewise.
9706         (scop_detection::add_scop): Do it here instead.
9707
9708 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9709
9710         * doc/sourcebuild.texi (arm_softfloat): Document.
9711
9712 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
9713
9714         PR gcc/77734
9715         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9716         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9717         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9718
9719 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9720             David Edelsohn  <dje.gcc@gmail.com>
9721
9722         PR target/83946
9723         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9724         Change "crset eq" to "crset 2".
9725         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9726         (*call_indirect_aix<mode>_nospec): Likewise.
9727         (*call_value_indirect_aix<mode>_nospec): Likewise.
9728         (*call_indirect_elfv2<mode>_nospec): Likewise.
9729         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9730         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9731         change assembly output from . to $.
9732         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9733         (indirect_jump<mode>_nospec): Change assembly output from . to $.
9734         (*tablejump<mode>_internal1_nospec): Likewise.
9735
9736 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
9737
9738         PR target/80870
9739         * config/sh/sh_optimize_sett_clrt.cc:
9740         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9741
9742 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
9743
9744         PR tree-optimization/83940
9745         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9746         offset_dt to vect_constant_def rather than vect_unknown_def_type.
9747         (vect_check_load_store_mask): Add a mask_dt_out parameter and
9748         use it to pass back the definition type.
9749         (vect_check_store_rhs): Likewise rhs_dt_out.
9750         (vect_build_gather_load_calls): Add a mask_dt argument and use
9751         it instead of a call to vect_is_simple_use.
9752         (vectorizable_store): Update calls to vect_check_load_store_mask
9753         and vect_check_store_rhs.  Use the dt returned by the latter instead
9754         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
9755         instead of calls to vect_is_simple_use.  Pass the scalar rather
9756         than the vector operand to vect_is_simple_use when handling
9757         second and subsequent copies of an rhs value.
9758         (vectorizable_load): Update calls to vect_check_load_store_mask
9759         and vect_build_gather_load_calls.  Use the cached mask_dt and
9760         gs_info.offset_dt instead of calls to vect_is_simple_use.
9761
9762 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
9763
9764         PR middle-end/83945
9765         * tree-emutls.c: Include gimplify.h.
9766         (lower_emutls_2): New function.
9767         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9768         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9769         it before further processing.
9770
9771         PR target/83930
9772         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9773         UINTVAL (trueop1) instead of INTVAL (op1).
9774
9775 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
9776
9777         PR debug/81570
9778         PR debug/83728
9779         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9780         INCOMING_FRAME_SP_OFFSET if not defined.
9781         (scan_trace): Add ENTRY argument.  If true and
9782         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9783         emit a note to adjust the CFA offset.
9784         (create_cfi_notes): Adjust scan_trace callers.
9785         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9786         INCOMING_FRAME_SP_OFFSET in the CIE.
9787         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9788         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9789         Likewise.
9790         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9791         * doc/tm.texi: Regenerated.
9792
9793 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
9794
9795         PR rtl-optimization/83147
9796         * lra-constraints.c (remove_inheritance_pseudos): Use
9797         lra_substitute_pseudo_within_insn.
9798
9799 2018-01-19  Tom de Vries  <tom@codesourcery.com>
9800             Cesar Philippidis  <cesar@codesourcery.com>
9801
9802         PR target/83920
9803         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9804
9805 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
9806
9807         PR target/83790
9808         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9809         spaces for function labels.
9810
9811 2018-01-19  Martin Liska  <mliska@suse.cz>
9812
9813         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9814         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9815         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9816         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9817         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9818         (PRED_CONST_RETURN): Change from 69 to 65.
9819         (PRED_NULL_RETURN): Change from 91 to 71.
9820         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9821         (PRED_LOOP_GUARD): Change from 66 to 73.
9822
9823 2018-01-19  Martin Liska  <mliska@suse.cz>
9824
9825         * predict.c (predict_insn_def): Add new assert.
9826         (struct branch_predictor): Change type to signed integer.
9827         (test_prediction_value_range): Amend test to cover
9828         PROB_UNINITIALIZED.
9829         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9830         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9831         (PRED_LOOP_ITERATIONS_MAX): Likewise.
9832         (PRED_LOOP_IV_COMPARE): Likewise.
9833         * predict.h (PROB_UNINITIALIZED): Define new constant.
9834
9835 2018-01-19  Martin Liska  <mliska@suse.cz>
9836
9837         * predict.c (dump_prediction): Add new format for
9838         analyze_brprob.py script which is enabled with -details
9839         suboption.
9840         * profile-count.h (precise_p): New function.
9841
9842 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
9843
9844         PR tree-optimization/83922
9845         * tree-vect-loop.c (vect_verify_full_masking): Return false if
9846         there are no statements that need masking.
9847         (vect_active_double_reduction_p): New function.
9848         (vect_analyze_loop_operations): Use it when handling phis that
9849         are not in the loop header.
9850
9851 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
9852
9853         PR tree-optimization/83914
9854         * tree-vect-loop.c (vectorizable_induction): Don't convert
9855         init_expr or apply the peeling adjustment for inductions
9856         that are nested within the vectorized loop.
9857
9858 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9859
9860         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9861         instead of NEG.
9862
9863 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
9864
9865         PR sanitizer/81715
9866         PR testsuite/83882
9867         * function.h (gimplify_parameters): Add gimple_seq * argument.
9868         * function.c: Include gimple.h and options.h.
9869         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9870         for the added local temporaries if needed.
9871         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9872         if there are any parameter cleanups, wrap whole body into a
9873         try/finally with the cleanups.
9874
9875 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
9876
9877         PR target/82964
9878         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9879         Use GET_MODE_CLASS for scalar floating point.
9880
9881 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
9882
9883         PR ipa/82256
9884         patch by PaX Team
9885         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9886         Fix call of call_cgraph_insertion_hooks.
9887
9888 2018-01-18  Martin Sebor  <msebor@redhat.com>
9889
9890         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9891
9892 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
9893
9894         PR ipa/83619
9895         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9896         frequencies.
9897
9898 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
9899
9900         PR other/70268
9901         * common.opt: (-ffile-prefix-map): New option.
9902         * opts.c (common_handle_option): Defer it.
9903         * opts-global.c (handle_common_deferred_options): Handle it.
9904         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9905         * file-prefix-map.h: New file.
9906         (remap_debug_filename, add_debug_prefix_map): ...here.
9907         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9908         * final.c (debug_prefix_map, add_debug_prefix_map
9909         remap_debug_filename): Move to...
9910         * file-prefix-map.c: New file.
9911         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9912         generalize, get rid of alloca(), use strrchr() instead of strchr().
9913         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9914         Implement in terms of add_prefix_map().
9915         (remap_macro_filename, remap_debug_filename): Implement in term of
9916         remap_filename().
9917         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9918         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9919         * dbxout.c: Include file-prefix-map.h.
9920         * varasm.c: Likewise.
9921         * vmsdbgout.c: Likewise.
9922         * xcoffout.c: Likewise.
9923         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9924         * doc/cppopts.texi (-fmacro-prefix-map): Document.
9925         * doc/invoke.texi (-ffile-prefix-map): Document.
9926         (-fdebug-prefix-map): Update description.
9927
9928 2018-01-18  Martin Liska  <mliska@suse.cz>
9929
9930         * config/i386/i386.c (indirect_thunk_name): Document that also
9931         lfence is emitted.
9932         (output_indirect_thunk): Document why both instructions
9933         (pause and lfence) are generated.
9934
9935 2018-01-18  Richard Biener  <rguenther@suse.de>
9936
9937         PR tree-optimization/83887
9938         * graphite-scop-detection.c
9939         (scop_detection::get_nearest_dom_with_single_entry): Remove.
9940         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9941         (scop_detection::merge_sese): Re-implement with a flood-fill
9942         algorithm that properly finds a SESE region if it exists.
9943
9944 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
9945
9946         PR c/61240
9947         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9948         pointer_diff optimizations use view_convert instead of convert.
9949
9950 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9951
9952         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9953         Generate different code for -mno-speculate-indirect-jumps.
9954         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9955         (*call_indirect_aix<mode>): Disable for
9956         -mno-speculate-indirect-jumps.
9957         (*call_indirect_aix<mode>_nospec): New define_insn.
9958         (*call_value_indirect_aix<mode>): Disable for
9959         -mno-speculate-indirect-jumps.
9960         (*call_value_indirect_aix<mode>_nospec): New define_insn.
9961         (*sibcall_nonlocal_sysv<mode>): Generate different code for
9962         -mno-speculate-indirect-jumps.
9963         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9964
9965 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
9966
9967         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9968         long double type, set the flags for noting the default long double
9969         type, even if we don't pass or return a long double type.
9970
9971 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
9972
9973         PR ipa/83051
9974         * ipa-inline.c (flatten_function): Do not overwrite final inlining
9975         failure.
9976
9977 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
9978
9979         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9980         support for merge[hl].
9981         (fold_mergehl_helper): New helper function.
9982         (tree-vector-builder.h): New #include for tree_vector_builder usage.
9983         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9984         (altivec_vmrglw_direct): Add xxmrglw insn.
9985
9986 2018-01-17  Andrew Waterman  <andrew@sifive.com>
9987
9988         * config/riscv/riscv.c (riscv_conditional_register_usage): If
9989         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9990
9991 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
9992
9993         PR lto/83121
9994         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9995         call the lto_location_cache before reading the
9996         DECL_SOURCE_LOCATION of the types.
9997
9998 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
9999             Richard Sandiford  <richard.sandiford@linaro.org>
10000
10001         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10002         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10003         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10004         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10005         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10006         Add declaration.
10007         * config/aarch64/constraints.md (aarch64_movti_operand):
10008         Limit immediates.
10009         * config/aarch64/predicates.md (Uti): Add new constraint.
10010
10011 2018-01-17 Carl Love  <cel@us.ibm.com>
10012
10013         * config/rs6000/vsx.md (define_expand xl_len_r,
10014         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10015         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10016         lxvll.
10017         (define_expand, define_insn): Move the shift left from  the
10018         define_insn to the define_expand for lxvl and stxvl instructions.
10019         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10020         and XL_LEN_R definitions to PURE.
10021
10022 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
10023
10024         * config/i386/i386.c (indirect_thunk_name): Declare regno
10025         as unsigned int.  Compare regno with INVALID_REGNUM.
10026         (output_indirect_thunk): Ditto.
10027         (output_indirect_thunk_function): Ditto.
10028         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
10029         in the call to output_indirect_thunk_function.
10030
10031 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
10032
10033         PR middle-end/83884
10034         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10035         rather than the size of inner_type to determine the stack slot size
10036         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10037
10038 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
10039
10040         PR target/83546
10041         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10042         to PTA_SILVERMONT.
10043
10044 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
10045
10046         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10047         endian Linux systems to optionally enable multilibs for selecting
10048         the long double type if the user configured an explicit type.
10049         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10050         have no long double multilibs if not defined.
10051         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10052         warn if the user used -mabi={ieee,ibm}longdouble and we built
10053         multilibs for long double.
10054         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10055         appropriate multilib option.
10056         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10057         multilib options.
10058         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10059         for building long double multilibs.
10060         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10061
10062 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
10063
10064         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10065         copies.
10066
10067         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10068         64 bits.
10069         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10070         128 bits.
10071
10072         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10073         variables.
10074
10075         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10076         return value.
10077
10078 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
10079
10080         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10081         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10082
10083 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10084
10085         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10086         different rtl trees depending on TARGET_64BIT.
10087         (rs6000_gen_lvx): Likewise.
10088
10089 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
10090
10091         * config/visium/visium.md (nop): Tweak comment.
10092         (hazard_nop): Likewise.
10093
10094 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10095
10096         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10097         -mspeculate-indirect-jumps.
10098         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10099         for -mno-speculate-indirect-jumps.
10100         (*call_indirect_elfv2<mode>_nospec): New define_insn.
10101         (*call_value_indirect_elfv2<mode>): Disable for
10102         -mno-speculate-indirect-jumps.
10103         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10104         (indirect_jump): Emit different RTL for
10105         -mno-speculate-indirect-jumps.
10106         (*indirect_jump<mode>): Disable for
10107         -mno-speculate-indirect-jumps.
10108         (*indirect_jump<mode>_nospec): New define_insn.
10109         (tablejump): Emit different RTL for
10110         -mno-speculate-indirect-jumps.
10111         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10112         (tablejumpsi_nospec): New define_expand.
10113         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10114         (tablejumpdi_nospec): New define_expand.
10115         (*tablejump<mode>_internal1): Disable for
10116         -mno-speculate-indirect-jumps.
10117         (*tablejump<mode>_internal1_nospec): New define_insn.
10118         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10119         option.
10120
10121 2018-01-16  Artyom Skrobov tyomitch@gmail.com
10122
10123         * caller-save.c (insert_save): Drop unnecessary parameter.  All
10124         callers updated.
10125
10126 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10127             Richard Biener  <rguenth@suse.de>
10128
10129         PR libgomp/83590
10130         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10131         return early, inline manually is_gimple_sizepos.  Make sure if we
10132         call gimplify_expr we don't end up with a gimple constant.
10133         * tree.c (variably_modified_type_p): Don't return true for
10134         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
10135         * gimplify.h (is_gimple_sizepos): Remove.
10136
10137 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10138
10139         PR tree-optimization/83857
10140         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10141         vectorizable_live_operation for pure SLP statements.
10142         (vectorizable_live_operation): Handle PHIs.
10143
10144 2018-01-16  Richard Biener  <rguenther@suse.de>
10145
10146         PR tree-optimization/83867
10147         * tree-vect-stmts.c (vect_transform_stmt): Precompute
10148         nested_in_vect_loop_p since the scalar stmt may get invalidated.
10149
10150 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10151
10152         PR c/83844
10153         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10154         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10155         If off is not INTEGER_CST, issue a may not be aligned warning
10156         rather than isn't aligned.  Use isn%'t rather than isn't.
10157         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10158         into MULT_EXPR.
10159         <case MULT_EXPR>: Improve the case when bottom and one of the
10160         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10161         operand, in that case check if the other operand is multiple of
10162         bottom divided by the INTEGER_CST operand.
10163
10164 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10165
10166         PR target/83858
10167         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10168         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10169         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10170         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10171         * config/pa/pa.c (pa_function_arg_advance): Likewise.
10172         (pa_function_arg, pa_arg_partial_bytes): Likewise.
10173         (pa_function_arg_size): New function.
10174
10175 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10176
10177         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10178         in a separate statement.
10179
10180 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10181
10182         PR tree-optimization/83847
10183         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10184         group gathers and scatters.
10185
10186 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10187
10188         PR rtl-optimization/86620
10189         * params.def (max-sched-ready-insns): Bump minimum value to 1.
10190
10191         PR rtl-optimization/83213
10192         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10193         to last if both are JUMP_INSNs.
10194
10195         PR tree-optimization/83843
10196         * gimple-ssa-store-merging.c
10197         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10198         store_immediate_info for bswap/nop orig_stores.
10199
10200 2018-01-15  Andrew Waterman  <andrew@sifive.com>
10201
10202         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10203         !TARGET_MUL.
10204         <UDIV>: Increase cost if !TARGET_DIV.
10205
10206 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
10207
10208         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10209         (define_attr "cr_logical_3op"): New.
10210         (cceq_ior_compare): Adjust.
10211         (cceq_ior_compare_complement): Adjust.
10212         (*cceq_rev_compare): Adjust.
10213         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10214         (is_cracked_insn): Adjust.
10215         (insn_must_be_first_in_group): Adjust.
10216         * config/rs6000/40x.md: Adjust.
10217         * config/rs6000/440.md: Adjust.
10218         * config/rs6000/476.md: Adjust.
10219         * config/rs6000/601.md: Adjust.
10220         * config/rs6000/603.md: Adjust.
10221         * config/rs6000/6xx.md: Adjust.
10222         * config/rs6000/7450.md: Adjust.
10223         * config/rs6000/7xx.md: Adjust.
10224         * config/rs6000/8540.md: Adjust.
10225         * config/rs6000/cell.md: Adjust.
10226         * config/rs6000/e300c2c3.md: Adjust.
10227         * config/rs6000/e500mc.md: Adjust.
10228         * config/rs6000/e500mc64.md: Adjust.
10229         * config/rs6000/e5500.md: Adjust.
10230         * config/rs6000/e6500.md: Adjust.
10231         * config/rs6000/mpc.md: Adjust.
10232         * config/rs6000/power4.md: Adjust.
10233         * config/rs6000/power5.md: Adjust.
10234         * config/rs6000/power6.md: Adjust.
10235         * config/rs6000/power7.md: Adjust.
10236         * config/rs6000/power8.md: Adjust.
10237         * config/rs6000/power9.md: Adjust.
10238         * config/rs6000/rs64.md: Adjust.
10239         * config/rs6000/titan.md: Adjust.
10240
10241 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10242
10243         * config/i386/predicates.md (indirect_branch_operand): Rewrite
10244         ix86_indirect_branch_register logic.
10245
10246 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10247
10248         * config/i386/constraints.md (Bs): Update
10249         ix86_indirect_branch_register check.  Don't check
10250         ix86_indirect_branch_register with GOT_memory_operand.
10251         (Bw): Likewise.
10252         * config/i386/predicates.md (GOT_memory_operand): Don't check
10253         ix86_indirect_branch_register here.
10254         (GOT32_symbol_operand): Likewise.
10255
10256 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10257
10258         * config/i386/predicates.md (constant_call_address_operand):
10259         Rewrite ix86_indirect_branch_register logic.
10260         (sibcall_insn_operand): Likewise.
10261
10262 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10263
10264         * config/i386/constraints.md (Bs): Replace
10265         ix86_indirect_branch_thunk_register with
10266         ix86_indirect_branch_register.
10267         (Bw): Likewise.
10268         * config/i386/i386.md (indirect_jump): Likewise.
10269         (tablejump): Likewise.
10270         (*sibcall_memory): Likewise.
10271         (*sibcall_value_memory): Likewise.
10272         Peepholes of indirect call and jump via memory: Likewise.
10273         * config/i386/i386.opt: Likewise.
10274         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10275         (GOT_memory_operand): Likewise.
10276         (call_insn_operand): Likewise.
10277         (sibcall_insn_operand): Likewise.
10278         (GOT32_symbol_operand): Likewise.
10279
10280 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10281
10282         PR middle-end/83837
10283         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10284         type rather than type addr's type points to.
10285         (expand_omp_atomic_mutex): Likewise.
10286         (expand_omp_atomic): Likewise.
10287
10288 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10289
10290         PR target/83839
10291         * config/i386/i386.c (output_indirect_thunk_function): Use
10292         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10293         for  __x86_return_thunk.
10294
10295 2018-01-15  Richard Biener  <rguenther@suse.de>
10296
10297         PR middle-end/83850
10298         * expmed.c (extract_bit_field_1): Fix typo.
10299
10300 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10301
10302         PR target/83687
10303         * config/arm/iterators.md (VF): New mode iterator.
10304         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10305         Remove integer-related logic from pattern.
10306         (neon_vabd<mode>_3): Likewise.
10307
10308 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10309
10310         PR middle-end/82694
10311         * common.opt (fstrict-overflow): No longer an alias.
10312         (fwrapv-pointer): New option.
10313         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10314         also for pointer types based on flag_wrapv_pointer.
10315         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10316         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10317         opts->x_flag_wrapv got set.
10318         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10319         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10320         POINTER_TYPE_OVERFLOW_UNDEFINED.
10321         * match.pd: Likewise in address comparison pattern.
10322         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10323
10324 2018-01-15  Richard Biener  <rguenther@suse.de>
10325
10326         PR lto/83804
10327         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10328         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
10329         Reset type names to their identifier if their TYPE_DECL doesn't
10330         have linkage (and thus is used for ODR and devirt).
10331         (save_debug_info_for_decl): Remove.
10332         (save_debug_info_for_type): Likewise.
10333         (add_tree_to_fld_list): Adjust.
10334         * tree-pretty-print.c (dump_generic_node): Make dumping of
10335         type names more robust.
10336
10337 2018-01-15  Richard Biener  <rguenther@suse.de>
10338
10339         * BASE-VER: Bump to 8.0.1.
10340
10341 2018-01-14  Martin Sebor  <msebor@redhat.com>
10342
10343         PR other/83508
10344         * builtins.c (check_access): Avoid warning when the no-warning bit
10345         is set.
10346
10347 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
10348
10349         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10350         * ira-color (allocno_hard_regs_compare): Likewise.
10351
10352 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
10353
10354         PR target/83013
10355         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10356         Use .pushsection/.popsection.
10357
10358 2018-01-14  Martin Sebor  <msebor@redhat.com>
10359
10360         PR c++/81327
10361         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10362
10363 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
10364
10365         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10366         entry from extra_headers.
10367         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10368         extra_headers, make the list bitwise identical to the i?86-*-* one.
10369
10370 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10371
10372         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10373         -mcmodel=large with -mindirect-branch=thunk,
10374         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10375         -mfunction-return=thunk-extern.
10376         * doc/invoke.texi: Document -mcmodel=large is incompatible with
10377         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10378         -mfunction-return=thunk and -mfunction-return=thunk-extern.
10379
10380 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10381
10382         * config/i386/i386.c (print_reg): Print the name of the full
10383         integer register without '%'.
10384         (ix86_print_operand): Handle 'V'.
10385          * doc/extend.texi: Document 'V' modifier.
10386
10387 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10388
10389         * config/i386/constraints.md (Bs): Disallow memory operand for
10390         -mindirect-branch-register.
10391         (Bw): Likewise.
10392         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10393         (GOT_memory_operand): Likewise.
10394         (call_insn_operand): Likewise.
10395         (sibcall_insn_operand): Likewise.
10396         (GOT32_symbol_operand): Likewise.
10397         * config/i386/i386.md (indirect_jump): Call convert_memory_address
10398         for -mindirect-branch-register.
10399         (tablejump): Likewise.
10400         (*sibcall_memory): Likewise.
10401         (*sibcall_value_memory): Likewise.
10402         Disallow peepholes of indirect call and jump via memory for
10403         -mindirect-branch-register.
10404         (*call_pop): Replace m with Bw.
10405         (*call_value_pop): Likewise.
10406         (*sibcall_pop_memory): Replace m with Bs.
10407         * config/i386/i386.opt (mindirect-branch-register): New option.
10408         * doc/invoke.texi: Document -mindirect-branch-register option.
10409
10410 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10411
10412         * config/i386/i386-protos.h (ix86_output_function_return): New.
10413         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10414         set function_return_type.
10415         (indirect_thunk_name): Add ret_p to indicate thunk for function
10416         return.
10417         (output_indirect_thunk_function): Pass false to
10418         indirect_thunk_name.
10419         (ix86_output_indirect_branch_via_reg): Likewise.
10420         (ix86_output_indirect_branch_via_push): Likewise.
10421         (output_indirect_thunk_function): Create alias for function
10422         return thunk if regno < 0.
10423         (ix86_output_function_return): New function.
10424         (ix86_handle_fndecl_attribute): Handle function_return.
10425         (ix86_attribute_table): Add function_return.
10426         * config/i386/i386.h (machine_function): Add
10427         function_return_type.
10428         * config/i386/i386.md (simple_return_internal): Use
10429         ix86_output_function_return.
10430         (simple_return_internal_long): Likewise.
10431         * config/i386/i386.opt (mfunction-return=): New option.
10432         (indirect_branch): Mention -mfunction-return=.
10433         * doc/extend.texi: Document function_return function attribute.
10434         * doc/invoke.texi: Document -mfunction-return= option.
10435
10436 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10437
10438         * config/i386/i386-opts.h (indirect_branch): New.
10439         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10440         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10441         with local indirect jump when converting indirect call and jump.
10442         (ix86_set_indirect_branch_type): New.
10443         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10444         (indirectlabelno): New.
10445         (indirect_thunk_needed): Likewise.
10446         (indirect_thunk_bnd_needed): Likewise.
10447         (indirect_thunks_used): Likewise.
10448         (indirect_thunks_bnd_used): Likewise.
10449         (INDIRECT_LABEL): Likewise.
10450         (indirect_thunk_name): Likewise.
10451         (output_indirect_thunk): Likewise.
10452         (output_indirect_thunk_function): Likewise.
10453         (ix86_output_indirect_branch_via_reg): Likewise.
10454         (ix86_output_indirect_branch_via_push): Likewise.
10455         (ix86_output_indirect_branch): Likewise.
10456         (ix86_output_indirect_jmp): Likewise.
10457         (ix86_code_end): Call output_indirect_thunk_function if needed.
10458         (ix86_output_call_insn): Call ix86_output_indirect_branch if
10459         needed.
10460         (ix86_handle_fndecl_attribute): Handle indirect_branch.
10461         (ix86_attribute_table): Add indirect_branch.
10462         * config/i386/i386.h (machine_function): Add indirect_branch_type
10463         and has_local_indirect_jump.
10464         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10465         to true.
10466         (tablejump): Likewise.
10467         (*indirect_jump): Use ix86_output_indirect_jmp.
10468         (*tablejump_1): Likewise.
10469         (simple_return_indirect_internal): Likewise.
10470         * config/i386/i386.opt (mindirect-branch=): New option.
10471         (indirect_branch): New.
10472         (keep): Likewise.
10473         (thunk): Likewise.
10474         (thunk-inline): Likewise.
10475         (thunk-extern): Likewise.
10476         * doc/extend.texi: Document indirect_branch function attribute.
10477         * doc/invoke.texi: Document -mindirect-branch= option.
10478
10479 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
10480
10481         PR ipa/83051
10482         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10483
10484 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
10485
10486         * ipa-inline.c (want_inline_small_function_p): Return false if
10487         inlining has already failed with CIF_FINAL_ERROR.
10488         (update_caller_keys): Call want_inline_small_function_p before
10489         can_inline_edge_p.
10490         (update_callee_keys): Likewise.
10491
10492 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10493
10494         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10495         New function.
10496         (rs6000_quadword_masked_address_p): Likewise.
10497         (quad_aligned_load_p): Likewise.
10498         (quad_aligned_store_p): Likewise.
10499         (const_load_sequence_p): Add comment to describe the outer-most loop.
10500         (mimic_memory_attributes_and_flags): New function.
10501         (rs6000_gen_stvx): Likewise.
10502         (replace_swapped_aligned_store): Likewise.
10503         (rs6000_gen_lvx): Likewise.
10504         (replace_swapped_aligned_load): Likewise.
10505         (replace_swapped_load_constant): Capitalize argument name in
10506         comment describing this function.
10507         (rs6000_analyze_swaps): Add a third pass to search for vector loads
10508         and stores that access quad-word aligned addresses and replace
10509         with stvx or lvx instructions when appropriate.
10510         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10511         New function prototype.
10512         (rs6000_quadword_masked_address_p): Likewise.
10513         (rs6000_gen_lvx): Likewise.
10514         (rs6000_gen_stvx): Likewise.
10515         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10516         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10517         when memory address is aligned.
10518         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10519         this split to select lvx instruction when memory address is aligned.
10520         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10521         instruction when memory address is aligned.
10522         (*vsx_le_perm_load_v16qi): Likewise.
10523         (four unnamed splitters): Modify to select the stvx instruction
10524         when memory is aligned.
10525
10526 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
10527
10528         * predict.c (determine_unlikely_bbs): Handle correctly BBs
10529         which appears in the queue multiple times.
10530
10531 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10532             Alan Hayward  <alan.hayward@arm.com>
10533             David Sherwood  <david.sherwood@arm.com>
10534
10535         * tree-vectorizer.h (vec_lower_bound): New structure.
10536         (_loop_vec_info): Add check_nonzero and lower_bounds.
10537         (LOOP_VINFO_CHECK_NONZERO): New macro.
10538         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10539         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10540         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10541         fields.  Make seg_len the distance travelled, not including the
10542         access size.
10543         (dr_direction_indicator): Declare.
10544         (dr_zero_step_indicator): Likewise.
10545         (dr_known_forward_stride_p): Likewise.
10546         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10547         tree-ssanames.h.
10548         (runtime_alias_check_p): Allow runtime alias checks with
10549         variable strides.
10550         (operator ==): Compare access_size and align.
10551         (prune_runtime_alias_test_list): Rework for new distinction between
10552         the access_size and seg_len.
10553         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
10554         segment lengths.
10555         (get_segment_min_max): New function.
10556         (create_intersect_range_checks): Use it.
10557         (dr_step_indicator): New function.
10558         (dr_direction_indicator): Likewise.
10559         (dr_zero_step_indicator): Likewise.
10560         (dr_known_forward_stride_p): Likewise.
10561         * tree-loop-distribution.c (data_ref_segment_size): Return
10562         DR_STEP * (niters - 1).
10563         (compute_alias_check_pairs): Update call to the dr_with_seg_len
10564         constructor.
10565         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10566         (vect_preserves_scalar_order_p): New function, split out from...
10567         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
10568         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10569         (vect_vfa_access_size): New function.
10570         (vect_vfa_align): Likewise.
10571         (vect_compile_time_alias): Take access_size_a and access_b arguments.
10572         (dump_lower_bound): New function.
10573         (vect_check_lower_bound): Likewise.
10574         (vect_small_gap_p): Likewise.
10575         (vectorizable_with_step_bound_p): Likewise.
10576         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10577         depencies if the vectorization factor is 1.  Convert the checks
10578         for nonzero steps into checks on the bounds of DR_STEP.  Try using
10579         a bunds check for variable steps if the minimum required step is
10580         relatively small. Update calls to the dr_with_seg_len
10581         constructor and to vect_compile_time_alias.
10582         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10583         function.
10584         (vect_loop_versioning): Call it.
10585         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10586         when retrying.
10587         (vect_estimate_min_profitable_iters): Account for any bounds checks.
10588
10589 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10590             Alan Hayward  <alan.hayward@arm.com>
10591             David Sherwood  <david.sherwood@arm.com>
10592
10593         * doc/sourcebuild.texi (vect_scatter_store): Document.
10594         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10595         optabs.
10596         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10597         Document.
10598         * genopinit.c (main): Add supports_vec_scatter_store and
10599         supports_vec_scatter_store_cached to target_optabs.
10600         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10601         IFN_MASK_SCATTER_STORE.
10602         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10603         functions.
10604         * internal-fn.h (internal_store_fn_p): Declare.
10605         (internal_fn_stored_value_index): Likewise.
10606         * internal-fn.c (scatter_store_direct): New macro.
10607         (expand_scatter_store_optab_fn): New function.
10608         (direct_scatter_store_optab_supported_p): New macro.
10609         (internal_store_fn_p): New function.
10610         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10611         IFN_MASK_SCATTER_STORE.
10612         (internal_fn_mask_index): Likewise.
10613         (internal_fn_stored_value_index): New function.
10614         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10615         for scatter stores.
10616         * optabs-query.h (supports_vec_scatter_store_p): Declare.
10617         * optabs-query.c (supports_vec_scatter_store_p): New function.
10618         * tree-vectorizer.h (vect_get_store_rhs): Declare.
10619         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10620         true for scatter stores.
10621         (vect_gather_scatter_fn_p): Handle scatter stores too.
10622         (vect_check_gather_scatter): Consider using scatter stores if
10623         supports_vec_scatter_store_p.
10624         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10625         scatter stores too.
10626         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10627         internal_fn_stored_value_index.
10628         (check_load_store_masking): Handle scatter stores too.
10629         (vect_get_store_rhs): Make public.
10630         (vectorizable_call): Use internal_store_fn_p.
10631         (vectorizable_store): Handle scatter store internal functions.
10632         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10633         when deciding whether the end of the group has been reached.
10634         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10635         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10636         (mask_scatter_store<mode>): New insns.
10637
10638 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10639             Alan Hayward  <alan.hayward@arm.com>
10640             David Sherwood  <david.sherwood@arm.com>
10641
10642         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10643         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10644         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10645         function.
10646         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10647         Use vect_truncate_gather_scatter_offset if we can't treat the
10648         operation as a normal gather load or scatter store.
10649         (get_group_load_store_type): Take the gather_scatter_info
10650         as argument.  Try using a gather load or scatter store for
10651         single-element groups.
10652         (get_load_store_type): Update calls to get_group_load_store_type
10653         and vect_use_strided_gather_scatters_p.
10654
10655 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10656             Alan Hayward  <alan.hayward@arm.com>
10657             David Sherwood  <david.sherwood@arm.com>
10658
10659         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10660         optional tree argument.
10661         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10662         null target hooks.
10663         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10664         but continue to use the current value as a fallback.
10665         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10666         to compare the updates.
10667         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10668         (get_load_store_type): Use it when handling a strided access.
10669         (vect_get_strided_load_store_ops): New function.
10670         (vect_get_data_ptr_increment): Likewise.
10671         (vectorizable_load): Handle strided gather loads.  Always pass
10672         a step to vect_create_data_ref_ptr and bump_vector_ptr.
10673
10674 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10675             Alan Hayward  <alan.hayward@arm.com>
10676             David Sherwood  <david.sherwood@arm.com>
10677
10678         * doc/md.texi (gather_load@var{m}): Document.
10679         (mask_gather_load@var{m}): Likewise.
10680         * genopinit.c (main): Add supports_vec_gather_load and
10681         supports_vec_gather_load_cached to target_optabs.
10682         * optabs-tree.c (init_tree_optimization_optabs): Use
10683         ggc_cleared_alloc to allocate target_optabs.
10684         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10685         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10686         functions.
10687         * internal-fn.h (internal_load_fn_p): Declare.
10688         (internal_gather_scatter_fn_p): Likewise.
10689         (internal_fn_mask_index): Likewise.
10690         (internal_gather_scatter_fn_supported_p): Likewise.
10691         * internal-fn.c (gather_load_direct): New macro.
10692         (expand_gather_load_optab_fn): New function.
10693         (direct_gather_load_optab_supported_p): New macro.
10694         (direct_internal_fn_optab): New function.
10695         (internal_load_fn_p): Likewise.
10696         (internal_gather_scatter_fn_p): Likewise.
10697         (internal_fn_mask_index): Likewise.
10698         (internal_gather_scatter_fn_supported_p): Likewise.
10699         * optabs-query.c (supports_at_least_one_mode_p): New function.
10700         (supports_vec_gather_load_p): Likewise.
10701         * optabs-query.h (supports_vec_gather_load_p): Declare.
10702         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10703         and memory_type field.
10704         (NUM_PATTERNS): Bump to 15.
10705         * tree-vect-data-refs.c: Include internal-fn.h.
10706         (vect_gather_scatter_fn_p): New function.
10707         (vect_describe_gather_scatter_call): Likewise.
10708         (vect_check_gather_scatter): Try using internal functions for
10709         gather loads.  Recognize existing calls to a gather load function.
10710         (vect_analyze_data_refs): Consider using gather loads if
10711         supports_vec_gather_load_p.
10712         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10713         (vect_get_gather_scatter_offset_type): Likewise.
10714         (vect_convert_mask_for_vectype): Likewise.
10715         (vect_add_conversion_to_patterm): Likewise.
10716         (vect_try_gather_scatter_pattern): Likewise.
10717         (vect_recog_gather_scatter_pattern): New pattern recognizer.
10718         (vect_vect_recog_func_ptrs): Add it.
10719         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10720         internal_fn_mask_index and internal_gather_scatter_fn_p.
10721         (check_load_store_masking): Take the gather_scatter_info as an
10722         argument and handle gather loads.
10723         (vect_get_gather_scatter_ops): New function.
10724         (vectorizable_call): Check internal_load_fn_p.
10725         (vectorizable_load): Likewise.  Handle gather load internal
10726         functions.
10727         (vectorizable_store): Update call to check_load_store_masking.
10728         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10729         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10730         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10731         (aarch64_gather_scale_operand_d): New predicates.
10732         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10733         (mask_gather_load<mode>): New insns.
10734
10735 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10736             Alan Hayward  <alan.hayward@arm.com>
10737             David Sherwood  <david.sherwood@arm.com>
10738
10739         * optabs.def (fold_left_plus_optab): New optab.
10740         * doc/md.texi (fold_left_plus_@var{m}): Document.
10741         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10742         * internal-fn.c (fold_left_direct): Define.
10743         (expand_fold_left_optab_fn): Likewise.
10744         (direct_fold_left_optab_supported_p): Likewise.
10745         * fold-const-call.c (fold_const_fold_left): New function.
10746         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10747         * tree-parloops.c (valid_reduction_p): New function.
10748         (gather_scalar_reductions): Use it.
10749         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10750         (vect_finish_replace_stmt): Declare.
10751         * tree-vect-loop.c (fold_left_reduction_fn): New function.
10752         (needs_fold_left_reduction_p): New function, split out from...
10753         (vect_is_simple_reduction): ...here.  Accept reductions that
10754         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10755         (vect_force_simple_reduction): Also store the reduction type in
10756         the assignment's STMT_VINFO_REDUC_TYPE.
10757         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10758         (merge_with_identity): New function.
10759         (vect_expand_fold_left): Likewise.
10760         (vectorize_fold_left_reduction): Likewise.
10761         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
10762         scalar phi in place for it.  Check for target support and reject
10763         cases that would reassociate the operation.  Defer the transform
10764         phase to vectorize_fold_left_reduction.
10765         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10766         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10767         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10768
10769 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10770
10771         * tree-if-conv.c (predicate_mem_writes): Remove redundant
10772         call to ifc_temp_var.
10773
10774 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10775             Alan Hayward  <alan.hayward@arm.com>
10776             David Sherwood  <david.sherwood@arm.com>
10777
10778         * target.def (legitimize_address_displacement): Take the original
10779         offset as a poly_int.
10780         * targhooks.h (default_legitimize_address_displacement): Update
10781         accordingly.
10782         * targhooks.c (default_legitimize_address_displacement): Likewise.
10783         * doc/tm.texi: Regenerate.
10784         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10785         as an argument, moving assert of ad->disp == ad->disp_term to...
10786         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
10787         Try calling targetm.legitimize_address_displacement before expanding
10788         the address rather than afterwards, and adjust for the new interface.
10789         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10790         Match the new hook interface.  Handle SVE addresses.
10791         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10792         new hook interface.
10793
10794 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10795
10796         * Makefile.in (OBJS): Add early-remat.o.
10797         * target.def (select_early_remat_modes): New hook.
10798         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10799         * doc/tm.texi: Regenerate.
10800         * targhooks.h (default_select_early_remat_modes): Declare.
10801         * targhooks.c (default_select_early_remat_modes): New function.
10802         * timevar.def (TV_EARLY_REMAT): New timevar.
10803         * passes.def (pass_early_remat): New pass.
10804         * tree-pass.h (make_pass_early_remat): Declare.
10805         * early-remat.c: New file.
10806         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10807         function.
10808         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10809
10810 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10811             Alan Hayward  <alan.hayward@arm.com>
10812             David Sherwood  <david.sherwood@arm.com>
10813
10814         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10815         vfm1 with a bound_epilog parameter.
10816         (vect_do_peeling): Update calls accordingly, and move the prologue
10817         call earlier in the function.  Treat the base bound_epilog as 0 for
10818         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
10819         this base when peeling for gaps.
10820         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10821         with fully-masked loops.
10822         (vect_estimate_min_profitable_iters): Handle the single peeled
10823         iteration in that case.
10824
10825 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10826             Alan Hayward  <alan.hayward@arm.com>
10827             David Sherwood  <david.sherwood@arm.com>
10828
10829         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10830         single-element interleaving even if the size is not a power of 2.
10831         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10832         accesses for single-element interleaving if the group size is
10833         not a power of 2.
10834
10835 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10836             Alan Hayward  <alan.hayward@arm.com>
10837             David Sherwood  <david.sherwood@arm.com>
10838
10839         * doc/md.texi (fold_extract_last_@var{m}): Document.
10840         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10841         * optabs.def (fold_extract_last_optab): New optab.
10842         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10843         * internal-fn.c (fold_extract_direct): New macro.
10844         (expand_fold_extract_optab_fn): Likewise.
10845         (direct_fold_extract_optab_supported_p): Likewise.
10846         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10847         * tree-vect-loop.c (vect_model_reduction_cost): Handle
10848         EXTRACT_LAST_REDUCTION.
10849         (get_initial_def_for_reduction): Do not create an initial vector
10850         for EXTRACT_LAST_REDUCTION reductions.
10851         (vectorizable_reduction): Leave the scalar phi in place for
10852         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
10853         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
10854         epilogue code for EXTRACT_LAST_REDUCTION and defer the
10855         transform phase to vectorizable_condition.
10856         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10857         split out from...
10858         (vect_finish_stmt_generation): ...here.
10859         (vect_finish_replace_stmt): New function.
10860         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10861         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10862         pattern.
10863         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10864
10865 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10866             Alan Hayward  <alan.hayward@arm.com>
10867             David Sherwood  <david.sherwood@arm.com>
10868
10869         * doc/md.texi (extract_last_@var{m}): Document.
10870         * optabs.def (extract_last_optab): New optab.
10871         * internal-fn.def (EXTRACT_LAST): New internal function.
10872         * internal-fn.c (cond_unary_direct): New macro.
10873         (expand_cond_unary_optab_fn): Likewise.
10874         (direct_cond_unary_optab_supported_p): Likewise.
10875         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10876         loops using EXTRACT_LAST.
10877         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10878         (extract_last_<mode>): ...this optab.
10879         (vec_extract<mode><Vel>): Update accordingly.
10880
10881 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10882             Alan Hayward  <alan.hayward@arm.com>
10883             David Sherwood  <david.sherwood@arm.com>
10884
10885         * target.def (empty_mask_is_expensive): New hook.
10886         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10887         * doc/tm.texi: Regenerate.
10888         * targhooks.h (default_empty_mask_is_expensive): Declare.
10889         * targhooks.c (default_empty_mask_is_expensive): New function.
10890         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10891         if the target says that empty masks are expensive.
10892         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10893         New function.
10894         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10895
10896 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10897             Alan Hayward  <alan.hayward@arm.com>
10898             David Sherwood  <david.sherwood@arm.com>
10899
10900         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10901         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10902         (vect_use_loop_mask_for_alignment_p): New function.
10903         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10904         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10905         niters_skip argument.  Make sure that the first niters_skip elements
10906         of the first iteration are inactive.
10907         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10908         Update call to vect_set_loop_masks_directly.
10909         (get_misalign_in_elems): New function, split out from...
10910         (vect_gen_prolog_loop_niters): ...here.
10911         (vect_update_init_of_dr): Take a code argument that specifies whether
10912         the adjustment should be added or subtracted.
10913         (vect_update_init_of_drs): Likewise.
10914         (vect_prepare_for_masked_peels): New function.
10915         (vect_do_peeling): Skip prologue peeling if we're using a mask
10916         instead.  Update call to vect_update_inits_of_drs.
10917         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10918         mask_skip_niters.
10919         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10920         alignment.  Do not include the number of peeled iterations in
10921         the minimum threshold in that case.
10922         (vectorizable_induction): Adjust the start value down by
10923         LOOP_VINFO_MASK_SKIP_NITERS iterations.
10924         (vect_transform_loop): Call vect_prepare_for_masked_peels.
10925         Take the number of skipped iterations into account when calculating
10926         the loop bounds.
10927         * tree-vect-stmts.c (vect_gen_while_not): New function.
10928
10929 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10930             Alan Hayward  <alan.hayward@arm.com>
10931             David Sherwood  <david.sherwood@arm.com>
10932
10933         * doc/sourcebuild.texi (vect_fully_masked): Document.
10934         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10935         default value to 0.
10936         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10937         split out from...
10938         (vect_analyze_loop_2): ...here. Don't check the vectorization
10939         factor against the number of loop iterations if the loop is
10940         fully-masked.
10941
10942 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10943             Alan Hayward  <alan.hayward@arm.com>
10944             David Sherwood  <david.sherwood@arm.com>
10945
10946         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10947         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10948         (dump_groups): Update accordingly.
10949         (iv_use::mem_type): New member variable.
10950         (address_p): New function.
10951         (record_use): Add a mem_type argument and initialize the new
10952         mem_type field.
10953         (record_group_use): Add a mem_type argument.  Use address_p.
10954         Remove obsolete null checks of base_object.  Update call to record_use.
10955         (find_interesting_uses_op): Update call to record_group_use.
10956         (find_interesting_uses_cond): Likewise.
10957         (find_interesting_uses_address): Likewise.
10958         (get_mem_type_for_internal_fn): New function.
10959         (find_address_like_use): Likewise.
10960         (find_interesting_uses_stmt): Try find_address_like_use before
10961         calling find_interesting_uses_op.
10962         (addr_offset_valid_p): Use the iv mem_type field as the type
10963         of the addressed memory.
10964         (add_autoinc_candidates): Likewise.
10965         (get_address_cost): Likewise.
10966         (split_small_address_groups_p): Use address_p.
10967         (split_address_groups): Likewise.
10968         (add_iv_candidate_for_use): Likewise.
10969         (autoinc_possible_for_pair): Likewise.
10970         (rewrite_groups): Likewise.
10971         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10972         (determine_group_iv_cost): Update after split of USE_ADDRESS.
10973         (get_alias_ptr_type_for_ptr_address): New function.
10974         (rewrite_use_address): Rewrite address uses in calls that were
10975         identified by find_address_like_use.
10976
10977 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10978             Alan Hayward  <alan.hayward@arm.com>
10979             David Sherwood  <david.sherwood@arm.com>
10980
10981         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10982         TARGET_MEM_REFs.
10983         * gimple-expr.h (is_gimple_addressable: Likewise.
10984         * gimple-expr.c (is_gimple_address): Likewise.
10985         * internal-fn.c (expand_call_mem_ref): New function.
10986         (expand_mask_load_optab_fn): Use it.
10987         (expand_mask_store_optab_fn): Likewise.
10988
10989 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10990             Alan Hayward  <alan.hayward@arm.com>
10991             David Sherwood  <david.sherwood@arm.com>
10992
10993         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10994         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10995         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10996         (cond_umax@var{mode}): Document.
10997         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10998         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10999         (cond_umin_optab, cond_umax_optab): New optabs.
11000         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11001         (COND_IOR, COND_XOR): New internal functions.
11002         * internal-fn.h (get_conditional_internal_fn): Declare.
11003         * internal-fn.c (cond_binary_direct): New macro.
11004         (expand_cond_binary_optab_fn): Likewise.
11005         (direct_cond_binary_optab_supported_p): Likewise.
11006         (get_conditional_internal_fn): New function.
11007         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11008         Cope with reduction statements that are vectorized as calls rather
11009         than assignments.
11010         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11011         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11012         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11013         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11014         (UNSPEC_COND_EOR): New unspecs.
11015         (optab): Add mappings for them.
11016         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11017         (sve_int_op, sve_fp_op): New int attributes.
11018
11019 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11020             Alan Hayward  <alan.hayward@arm.com>
11021             David Sherwood  <david.sherwood@arm.com>
11022
11023         * optabs.def (while_ult_optab): New optab.
11024         * doc/md.texi (while_ult@var{m}@var{n}): Document.
11025         * internal-fn.def (WHILE_ULT): New internal function.
11026         * internal-fn.h (direct_internal_fn_supported_p): New override
11027         that takes two types as argument.
11028         * internal-fn.c (while_direct): New macro.
11029         (expand_while_optab_fn): New function.
11030         (convert_optab_supported_p): Likewise.
11031         (direct_while_optab_supported_p): New macro.
11032         * wide-int.h (wi::udiv_ceil): New function.
11033         * tree-vectorizer.h (rgroup_masks): New structure.
11034         (vec_loop_masks): New typedef.
11035         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11036         and fully_masked_p.
11037         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11038         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11039         (vect_max_vf): New function.
11040         (slpeel_make_loop_iterate_ntimes): Delete.
11041         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11042         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11043         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11044         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11045         internal-fn.h, stor-layout.h and optabs-query.h.
11046         (vect_set_loop_mask): New function.
11047         (add_preheader_seq): Likewise.
11048         (add_header_seq): Likewise.
11049         (interleave_supported_p): Likewise.
11050         (vect_maybe_permute_loop_masks): Likewise.
11051         (vect_set_loop_masks_directly): Likewise.
11052         (vect_set_loop_condition_masked): Likewise.
11053         (vect_set_loop_condition_unmasked): New function, split out from
11054         slpeel_make_loop_iterate_ntimes.
11055         (slpeel_make_loop_iterate_ntimes): Rename to..
11056         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
11057         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11058         (vect_do_peeling): Update call accordingly.
11059         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11060         loops.
11061         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11062         mask_compare_type, can_fully_mask_p and fully_masked_p.
11063         (release_vec_loop_masks): New function.
11064         (_loop_vec_info): Use it to free the loop masks.
11065         (can_produce_all_loop_masks_p): New function.
11066         (vect_get_max_nscalars_per_iter): Likewise.
11067         (vect_verify_full_masking): Likewise.
11068         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11069         retries, and free the mask rgroups before retrying.  Check loop-wide
11070         reasons for disallowing fully-masked loops.  Make the final decision
11071         about whether use a fully-masked loop or not.
11072         (vect_estimate_min_profitable_iters): Do not assume that peeling
11073         for the number of iterations will be needed for fully-masked loops.
11074         (vectorizable_reduction): Disable fully-masked loops.
11075         (vectorizable_live_operation): Likewise.
11076         (vect_halve_mask_nunits): New function.
11077         (vect_double_mask_nunits): Likewise.
11078         (vect_record_loop_mask): Likewise.
11079         (vect_get_loop_mask): Likewise.
11080         (vect_transform_loop): Handle the case in which the final loop
11081         iteration might handle a partial vector.  Call vect_set_loop_condition
11082         instead of slpeel_make_loop_iterate_ntimes.
11083         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11084         (check_load_store_masking): New function.
11085         (prepare_load_store_mask): Likewise.
11086         (vectorizable_store): Handle fully-masked loops.
11087         (vectorizable_load): Likewise.
11088         (supportable_widening_operation): Use vect_halve_mask_nunits for
11089         booleans.
11090         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11091         (vect_gen_while): New function.
11092         * config/aarch64/aarch64.md (umax<mode>3): New expander.
11093         (aarch64_uqdec<mode>): New insn.
11094
11095 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11096             Alan Hayward  <alan.hayward@arm.com>
11097             David Sherwood  <david.sherwood@arm.com>
11098
11099         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11100         (reduc_xor_scal_optab): New optabs.
11101         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11102         (reduc_xor_scal_@var{m}): Document.
11103         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11104         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11105         internal functions.
11106         * fold-const-call.c (fold_const_call): Handle them.
11107         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11108         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11109         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11110         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11111         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11112         (UNSPEC_XORV): New unspecs.
11113         (optab): Add entries for them.
11114         (BITWISEV): New int iterator.
11115         (bit_reduc_op): New int attributes.
11116
11117 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11118             Alan Hayward  <alan.hayward@arm.com>
11119             David Sherwood  <david.sherwood@arm.com>
11120
11121         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11122         * internal-fn.def (VEC_SHL_INSERT): New internal function.
11123         * optabs.def (vec_shl_insert_optab): New optab.
11124         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11125         (duplicate_and_interleave): Likewise.
11126         * tree-vect-loop.c: Include internal-fn.h.
11127         (neutral_op_for_slp_reduction): New function, split out from
11128         get_initial_defs_for_reduction.
11129         (get_initial_def_for_reduction): Handle option 2 for variable-length
11130         vectors by loading the neutral value into a vector and then shifting
11131         the initial value into element 0.
11132         (get_initial_defs_for_reduction): Replace the code argument with
11133         the neutral value calculated by neutral_op_for_slp_reduction.
11134         Use gimple_build_vector for constant-length vectors.
11135         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11136         but the first group_size elements have a neutral value.
11137         Use duplicate_and_interleave otherwise.
11138         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11139         Update call to get_initial_defs_for_reduction.  Handle SLP
11140         reductions for variable-length vectors by creating one vector
11141         result for each scalar result, with the elements associated
11142         with other scalar results stubbed out with the neutral value.
11143         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11144         Require IFN_VEC_SHL_INSERT for double reductions on
11145         variable-length vectors, or SLP reductions that have
11146         a neutral value.  Require can_duplicate_and_interleave_p
11147         support for variable-length unchained SLP reductions if there
11148         is no neutral value, such as for MIN/MAX reductions.  Also require
11149         the number of vector elements to be a multiple of the number of
11150         SLP statements when doing variable-length unchained SLP reductions.
11151         Update call to vect_create_epilog_for_reduction.
11152         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11153         and remove initial values.
11154         (duplicate_and_interleave): Make public.
11155         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11156         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11157
11158 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11159             Alan Hayward  <alan.hayward@arm.com>
11160             David Sherwood  <david.sherwood@arm.com>
11161
11162         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11163         (can_duplicate_and_interleave_p): New function.
11164         (vect_get_and_check_slp_defs): Take the vector of statements
11165         rather than just the current one.  Remove excess parentheses.
11166         Restriction rejectinon of vect_constant_def and vect_external_def
11167         for variable-length vectors to boolean types, or types for which
11168         can_duplicate_and_interleave_p is false.
11169         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11170         (duplicate_and_interleave): New function.
11171         (vect_get_constant_vectors): Use gimple_build_vector for
11172         constant-length vectors and suitable variable-length constant
11173         vectors.  Use duplicate_and_interleave for other variable-length
11174         vectors.  Don't defer the update when inserting new statements.
11175
11176 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11177             Alan Hayward  <alan.hayward@arm.com>
11178             David Sherwood  <david.sherwood@arm.com>
11179
11180         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11181         min_profitable_iters doesn't go negative.
11182
11183 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11184             Alan Hayward  <alan.hayward@arm.com>
11185             David Sherwood  <david.sherwood@arm.com>
11186
11187         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11188         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11189         * optabs.def (vec_mask_load_lanes_optab): New optab.
11190         (vec_mask_store_lanes_optab): Likewise.
11191         * internal-fn.def (MASK_LOAD_LANES): New internal function.
11192         (MASK_STORE_LANES): Likewise.
11193         * internal-fn.c (mask_load_lanes_direct): New macro.
11194         (mask_store_lanes_direct): Likewise.
11195         (expand_mask_load_optab_fn): Handle masked operations.
11196         (expand_mask_load_lanes_optab_fn): New macro.
11197         (expand_mask_store_optab_fn): Handle masked operations.
11198         (expand_mask_store_lanes_optab_fn): New macro.
11199         (direct_mask_load_lanes_optab_supported_p): Likewise.
11200         (direct_mask_store_lanes_optab_supported_p): Likewise.
11201         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11202         parameter.
11203         (vect_load_lanes_supported): Likewise.
11204         * tree-vect-data-refs.c (strip_conversion): New function.
11205         (can_group_stmts_p): Likewise.
11206         (vect_analyze_data_ref_accesses): Use it instead of checking
11207         for a pair of assignments.
11208         (vect_store_lanes_supported): Take a masked_p parameter.
11209         (vect_load_lanes_supported): Likewise.
11210         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11211         vect_store_lanes_supported and vect_load_lanes_supported.
11212         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11213         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11214         parameter.  Don't allow gaps for masked accesses.
11215         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
11216         and vect_load_lanes_supported.
11217         (get_load_store_type): Take a masked_p parameter and update
11218         call to get_group_load_store_type.
11219         (vectorizable_store): Update call to get_load_store_type.
11220         Handle IFN_MASK_STORE_LANES.
11221         (vectorizable_load): Update call to get_load_store_type.
11222         Handle IFN_MASK_LOAD_LANES.
11223
11224 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11225             Alan Hayward  <alan.hayward@arm.com>
11226             David Sherwood  <david.sherwood@arm.com>
11227
11228         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11229         modes for SVE.
11230         * config/aarch64/aarch64-protos.h
11231         (aarch64_sve_struct_memory_operand_p): Declare.
11232         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11233         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11234         (VPRED, vpred): Handle SVE structure modes.
11235         * config/aarch64/constraints.md (Utx): New constraint.
11236         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11237         (aarch64_sve_struct_nonimmediate_operand): New predicates.
11238         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11239         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11240         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11241         structure modes.  Split into pieces after RA.
11242         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11243         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11244         New patterns.
11245         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11246         SVE structure modes.
11247         (aarch64_classify_address): Likewise.
11248         (sizetochar): Move earlier in file.
11249         (aarch64_print_operand): Handle SVE register lists.
11250         (aarch64_array_mode): New function.
11251         (aarch64_sve_struct_memory_operand_p): Likewise.
11252         (TARGET_ARRAY_MODE): Redefine.
11253
11254 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11255             Alan Hayward  <alan.hayward@arm.com>
11256             David Sherwood  <david.sherwood@arm.com>
11257
11258         * target.def (array_mode): New target hook.
11259         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11260         * doc/tm.texi: Regenerate.
11261         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11262         * hooks.c (hook_optmode_mode_uhwi_none): New function.
11263         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11264         targetm.array_mode.
11265         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
11266         type sizes.
11267
11268 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11269             Alan Hayward  <alan.hayward@arm.com>
11270             David Sherwood  <david.sherwood@arm.com>
11271
11272         * fold-const.c (fold_binary_loc): Check the argument types
11273         rather than the result type when testing for a vector operation.
11274
11275 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11276
11277         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11278         * doc/tm.texi: Regenerate.
11279
11280 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11281             Alan Hayward  <alan.hayward@arm.com>
11282             David Sherwood  <david.sherwood@arm.com>
11283
11284         * doc/invoke.texi (-msve-vector-bits=): Document new option.
11285         (sve): Document new AArch64 extension.
11286         * doc/md.texi (w): Extend the description of the AArch64
11287         constraint to include SVE vectors.
11288         (Upl, Upa): Document new AArch64 predicate constraints.
11289         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11290         enum.
11291         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11292         (msve-vector-bits=): New option.
11293         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11294         SVE when these are disabled.
11295         (sve): New extension.
11296         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11297         modes.  Adjust their number of units based on aarch64_sve_vg.
11298         (MAX_BITSIZE_MODE_ANY_MODE): Define.
11299         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11300         aarch64_addr_query_type.
11301         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11302         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11303         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11304         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11305         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11306         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11307         (aarch64_simd_imm_zero_p): Delete.
11308         (aarch64_check_zero_based_sve_index_immediate): Declare.
11309         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11310         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11311         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11312         (aarch64_sve_float_mul_immediate_p): Likewise.
11313         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11314         rather than an rtx.
11315         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11316         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11317         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11318         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11319         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11320         (aarch64_regmode_natural_size): Likewise.
11321         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11322         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11323         left one place.
11324         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11325         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11326         for VG and the SVE predicate registers.
11327         (V_ALIASES): Add a "z"-prefixed alias.
11328         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11329         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11330         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11331         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11332         (REG_CLASS_NAMES): Add entries for them.
11333         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
11334         and the predicate registers.
11335         (aarch64_sve_vg): Declare.
11336         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11337         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11338         (REGMODE_NATURAL_SIZE): Define.
11339         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11340         SVE macros.
11341         * config/aarch64/aarch64.c: Include cfgrtl.h.
11342         (simd_immediate_info): Add a constructor for series vectors,
11343         and an associated step field.
11344         (aarch64_sve_vg): New variable.
11345         (aarch64_dbx_register_number): Handle VG and the predicate registers.
11346         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11347         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11348         (VEC_ANY_DATA, VEC_STRUCT): New constants.
11349         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11350         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11351         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11352         (aarch64_get_mask_mode): New functions.
11353         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11354         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11355         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
11356         predicate modes and predicate registers.  Explicitly restrict
11357         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
11358         to store a vector mode if it is recognized by
11359         aarch64_classify_vector_mode.
11360         (aarch64_regmode_natural_size): New function.
11361         (aarch64_hard_regno_caller_save_mode): Return the original mode
11362         for predicates.
11363         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11364         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11365         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11366         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11367         functions.
11368         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
11369         does not overlap dest if the function is frame-related.  Handle
11370         SVE constants.
11371         (aarch64_split_add_offset): New function.
11372         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11373         them aarch64_add_offset.
11374         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11375         and update call to aarch64_sub_sp.
11376         (aarch64_add_cfa_expression): New function.
11377         (aarch64_expand_prologue): Pass extra temporary registers to the
11378         functions above.  Handle the case in which we need to emit new
11379         DW_CFA_expressions for registers that were originally saved
11380         relative to the stack pointer, but now have to be expressed
11381         relative to the frame pointer.
11382         (aarch64_output_mi_thunk): Pass extra temporary registers to the
11383         functions above.
11384         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
11385         IP0 and IP1 values for SVE frames.
11386         (aarch64_expand_vec_series): New function.
11387         (aarch64_expand_sve_widened_duplicate): Likewise.
11388         (aarch64_expand_sve_const_vector): Likewise.
11389         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11390         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
11391         into the register, rather than emitting a SET directly.
11392         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11393         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11394         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11395         (offset_9bit_signed_scaled_p): New functions.
11396         (aarch64_replicate_bitmask_imm): New function.
11397         (aarch64_bitmask_imm): Use it.
11398         (aarch64_cannot_force_const_mem): Reject expressions involving
11399         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
11400         (aarch64_classify_index): Handle SVE indices, by requiring
11401         a plain register index with a scale that matches the element size.
11402         (aarch64_classify_address): Handle SVE addresses.  Assert that
11403         the mode of the address is VOIDmode or an integer mode.
11404         Update call to aarch64_classify_symbol.
11405         (aarch64_classify_symbolic_expression): Update call to
11406         aarch64_classify_symbol.
11407         (aarch64_const_vec_all_in_range_p): New function.
11408         (aarch64_print_vector_float_operand): Likewise.
11409         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
11410         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
11411         and the FP immediates 1.0 and 0.5.
11412         (aarch64_print_address_internal): Handle SVE addresses.
11413         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11414         (aarch64_regno_regclass): Handle predicate registers.
11415         (aarch64_secondary_reload): Handle big-endian reloads of SVE
11416         data modes.
11417         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11418         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11419         (aarch64_convert_sve_vector_bits): New function.
11420         (aarch64_override_options): Use it to handle -msve-vector-bits=.
11421         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11422         rather than an rtx.
11423         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11424         Handle SVE vector and predicate modes.  Accept VL-based constants
11425         that need only one temporary register, and VL offsets that require
11426         no temporary registers.
11427         (aarch64_conditional_register_usage): Mark the predicate registers
11428         as fixed if SVE isn't available.
11429         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11430         Return true for SVE vector and predicate modes.
11431         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11432         rather than an unsigned int.  Handle SVE modes.
11433         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
11434         SVE modes.
11435         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11436         if SVE is enabled.
11437         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11438         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11439         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11440         (aarch64_sve_float_mul_immediate_p): New functions.
11441         (aarch64_sve_valid_immediate): New function.
11442         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11443         Explicitly reject structure modes.  Check for INDEX constants.
11444         Handle PTRUE and PFALSE constants.
11445         (aarch64_check_zero_based_sve_index_immediate): New function.
11446         (aarch64_simd_imm_zero_p): Delete.
11447         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11448         vector modes.  Accept constants in the range of CNT[BHWD].
11449         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11450         ask for an Advanced SIMD mode.
11451         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11452         (aarch64_simd_vector_alignment): Handle SVE predicates.
11453         (aarch64_vectorize_preferred_vector_alignment): New function.
11454         (aarch64_simd_vector_alignment_reachable): Use it instead of
11455         the vector size.
11456         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11457         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11458         functions.
11459         (MAX_VECT_LEN): Delete.
11460         (expand_vec_perm_d): Add a vec_flags field.
11461         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11462         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11463         (aarch64_evpc_ext): Don't apply a big-endian lane correction
11464         for SVE modes.
11465         (aarch64_evpc_rev): Rename to...
11466         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
11467         (aarch64_evpc_rev_global): New function.
11468         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11469         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11470         MAX_VECT_LEN.
11471         (aarch64_evpc_sve_tbl): New function.
11472         (aarch64_expand_vec_perm_const_1): Update after rename of
11473         aarch64_evpc_rev.  Handle SVE permutes too, trying
11474         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11475         than aarch64_evpc_tbl.
11476         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11477         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11478         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11479         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11480         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11481         (aarch64_expand_sve_vcond): New functions.
11482         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11483         of aarch64_vector_mode_p.
11484         (aarch64_dwarf_poly_indeterminate_value): New function.
11485         (aarch64_compute_pressure_classes): Likewise.
11486         (aarch64_can_change_mode_class): Likewise.
11487         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11488         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11489         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11490         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11491         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11492         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11493         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11494         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11495         constraints.
11496         (Dn, Dl, Dr): Accept const as well as const_vector.
11497         (Dz): Likewise.  Compare against CONST0_RTX.
11498         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11499         of "vector" where appropriate.
11500         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11501         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11502         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11503         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11504         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11505         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11506         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11507         (v_int_equiv): Extend to SVE modes.
11508         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11509         mode attributes.
11510         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11511         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11512         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11513         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11514         (SVE_COND_FP_CMP): New int iterators.
11515         (perm_hilo): Handle the new unpack unspecs.
11516         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11517         attributes.
11518         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11519         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11520         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11521         (aarch64_equality_operator, aarch64_constant_vector_operand)
11522         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11523         (aarch64_sve_nonimmediate_operand): Likewise.
11524         (aarch64_sve_general_operand): Likewise.
11525         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11526         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11527         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11528         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11529         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11530         (aarch64_sve_float_arith_immediate): Likewise.
11531         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11532         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11533         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11534         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11535         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11536         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11537         (aarch64_sve_float_arith_operand): Likewise.
11538         (aarch64_sve_float_arith_with_sub_operand): Likewise.
11539         (aarch64_sve_float_mul_operand): Likewise.
11540         (aarch64_sve_vec_perm_operand): Likewise.
11541         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11542         (aarch64_mov_operand): Accept const_poly_int and const_vector.
11543         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11544         as well as const_vector.
11545         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11546         in file.  Use CONST0_RTX and CONSTM1_RTX.
11547         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
11548         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11549         Use aarch64_simd_imm_zero.
11550         * config/aarch64/aarch64-sve.md: New file.
11551         * config/aarch64/aarch64.md: Include it.
11552         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11553         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11554         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11555         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11556         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11557         (sve): New attribute.
11558         (enabled): Disable instructions with the sve attribute unless
11559         TARGET_SVE.
11560         (movqi, movhi): Pass CONST_POLY_INT operaneds through
11561         aarch64_expand_mov_immediate.
11562         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11563         CNT[BHSD] immediates.
11564         (movti): Split CONST_POLY_INT moves into two halves.
11565         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11566         Split additions that need a temporary here if the destination
11567         is the stack pointer.
11568         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11569         (*add<mode>3_poly_1): New instruction.
11570         (set_clobber_cc): New expander.
11571
11572 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11573
11574         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11575         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
11576         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11577         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11578         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
11579         Change innermode from fixed_mode_size to machine_mode.
11580         (simplify_subreg): Update call accordingly.  Handle a constant-sized
11581         subreg of a variable-length CONST_VECTOR.
11582
11583 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11584             Alan Hayward  <alan.hayward@arm.com>
11585             David Sherwood  <david.sherwood@arm.com>
11586
11587         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11588         (add_offset_to_base): New function, split out from...
11589         (create_mem_ref): ...here.  When handling a scale other than 1,
11590         check first whether the address is valid without the offset.
11591         Add it into the base if so, leaving the index and scale as-is.
11592
11593 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
11594
11595         PR c++/83778
11596         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11597         fold_for_warn before checking if arg2 is INTEGER_CST.
11598
11599 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
11600
11601         * config/rs6000/predicates.md (load_multiple_operation): Delete.
11602         (store_multiple_operation): Delete.
11603         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11604         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11605         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11606         guarded by TARGET_STRING.
11607         (rs6000_output_load_multiple): Delete.
11608         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11609         OPTION_MASK_STRING / TARGET_STRING handling.
11610         (print_operand) <'N', 'O'>: Add comment that these are unused now.
11611         (const rs6000_opt_masks) <"string">: Change mask to 0.
11612         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11613         (MASK_STRING): Delete.
11614         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11615         parts.  Simplify.
11616         (load_multiple): Delete.
11617         (*ldmsi8): Delete.
11618         (*ldmsi7): Delete.
11619         (*ldmsi6): Delete.
11620         (*ldmsi5): Delete.
11621         (*ldmsi4): Delete.
11622         (*ldmsi3): Delete.
11623         (store_multiple): Delete.
11624         (*stmsi8): Delete.
11625         (*stmsi7): Delete.
11626         (*stmsi6): Delete.
11627         (*stmsi5): Delete.
11628         (*stmsi4): Delete.
11629         (*stmsi3): Delete.
11630         (movmemsi_8reg): Delete.
11631         (corresponding unnamed define_insn): Delete.
11632         (movmemsi_6reg): Delete.
11633         (corresponding unnamed define_insn): Delete.
11634         (movmemsi_4reg): Delete.
11635         (corresponding unnamed define_insn): Delete.
11636         (movmemsi_2reg): Delete.
11637         (corresponding unnamed define_insn): Delete.
11638         (movmemsi_1reg): Delete.
11639         (corresponding unnamed define_insn): Delete.
11640         * config/rs6000/rs6000.opt (mno-string): New.
11641         (mstring): Replace by deprecation warning stub.
11642         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11643
11644 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
11645
11646         * regrename.c (regrename_do_replace): If replacing the same
11647         reg multiple times, try to reuse last created gen_raw_REG.
11648
11649         PR debug/81155
11650         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11651         main to workaround a bug in GDB.
11652
11653 2018-01-12  Tom de Vries  <tom@codesourcery.com>
11654
11655         PR target/83737
11656         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11657
11658 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
11659
11660         PR rtl-optimization/80481
11661         * ira-color.c (get_cap_member): New function.
11662         (allocnos_conflict_by_live_ranges_p): Use it.
11663         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11664         (setup_slot_coalesced_allocno_live_ranges): Ditto.
11665
11666 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
11667
11668         PR target/83628
11669         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11670         (*saddl_se_1): Ditto.
11671         (*ssubsi_1): Ditto.
11672         (*ssubl_se_1): Ditto.
11673
11674 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
11675
11676         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11677         rather than wi::to_widest for DR_INITs.
11678         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11679         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11680         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11681         INTEGER_CSTs.
11682         (vect_analyze_group_access_1): Note that here.
11683
11684 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
11685
11686         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11687         polynomial type sizes.
11688
11689 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
11690
11691         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11692         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11693         (gimple_add_tmp_var): Likewise.
11694
11695 2018-01-12  Martin Liska  <mliska@suse.cz>
11696
11697         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11698         (gimple_alloc_sizes): Likewise.
11699         (dump_gimple_statistics): Use PRIu64 in printf format.
11700         * gimple.h: Change uint64_t to int.
11701
11702 2018-01-12  Martin Liska  <mliska@suse.cz>
11703
11704         * tree-core.h: Use uint64_t instead of int.
11705         * tree.c (tree_node_counts): Likewise.
11706         (tree_node_sizes): Likewise.
11707         (dump_tree_statistics): Use PRIu64 in printf format.
11708
11709 2018-01-12  Martin Liska  <mliska@suse.cz>
11710
11711         * Makefile.in: As qsort_chk is implemented in vec.c, add
11712         vec.o to linkage of gencfn-macros.
11713         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11714         passing the info to record_node_allocation_statistics.
11715         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11716         and pass the info.
11717         * ggc-common.c (struct ggc_usage): Add operator== and use
11718         it in operator< and compare function.
11719         * mem-stats.h (struct mem_usage): Likewise.
11720         * vec.c (struct vec_usage): Remove operator< and compare
11721         function. Can be simply inherited.
11722
11723 2018-01-12  Martin Jambor  <mjambor@suse.cz>
11724
11725         PR target/81616
11726         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11727         * tree-ssa-math-opts.c: Include domwalk.h.
11728         (convert_mult_to_fma_1): New function.
11729         (fma_transformation_info): New type.
11730         (fma_deferring_state): Likewise.
11731         (cancel_fma_deferring): New function.
11732         (result_of_phi): Likewise.
11733         (last_fma_candidate_feeds_initial_phi): Likewise.
11734         (convert_mult_to_fma): Added deferring logic, split actual
11735         transformation to convert_mult_to_fma_1.
11736         (math_opts_dom_walker): New type.
11737         (math_opts_dom_walker::after_dom_children): New method, body moved
11738         here from pass_optimize_widening_mul::execute, added deferring logic
11739         bits.
11740         (pass_optimize_widening_mul::execute): Moved most of code to
11741         math_opts_dom_walker::after_dom_children.
11742         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11743         * config/i386/i386.c (ix86_option_override_internal): Added
11744         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11745
11746 2018-01-12  Richard Biener  <rguenther@suse.de>
11747
11748         PR debug/83157
11749         * dwarf2out.c (gen_variable_die): Do not reset old_die for
11750         inline instance vars.
11751
11752 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
11753
11754         PR target/81819
11755         * config/rx/rx.c (rx_is_restricted_memory_address):
11756         Handle SUBREG case.
11757
11758 2018-01-12  Richard Biener  <rguenther@suse.de>
11759
11760         PR tree-optimization/80846
11761         * target.def (split_reduction): New target hook.
11762         * targhooks.c (default_split_reduction): New function.
11763         * targhooks.h (default_split_reduction): Declare.
11764         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11765         target requests first reduce vectors by combining low and high
11766         parts.
11767         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11768         (get_vectype_for_scalar_type_and_size): Export.
11769         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11770         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11771         * doc/tm.texi: Regenerate.
11772         * config/i386/i386.c (ix86_split_reduction): Implement
11773         TARGET_VECTORIZE_SPLIT_REDUCTION.
11774
11775 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
11776
11777         PR target/83368
11778         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11779         in PIC mode except for TARGET_VXWORKS_RTP.
11780         * config/sparc/sparc.c: Include cfgrtl.h.
11781         (TARGET_INIT_PIC_REG): Define.
11782         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11783         (sparc_pic_register_p): New predicate.
11784         (sparc_legitimate_address_p): Use it.
11785         (sparc_legitimize_pic_address): Likewise.
11786         (sparc_delegitimize_address): Likewise.
11787         (sparc_mode_dependent_address_p): Likewise.
11788         (gen_load_pcrel_sym): Remove 4th parameter.
11789         (load_got_register): Adjust call to above.  Remove obsolete stuff.
11790         (sparc_expand_prologue): Do not call load_got_register here.
11791         (sparc_flat_expand_prologue): Likewise.
11792         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11793         (sparc_use_pseudo_pic_reg): New function.
11794         (sparc_init_pic_reg): Likewise.
11795         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11796         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11797
11798 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
11799
11800         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11801         Add item for branch_cost.
11802
11803 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
11804
11805         PR rtl-optimization/83565
11806         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11807         not extend the result to a larger mode for rotate operations.
11808         (num_sign_bit_copies1): Likewise.
11809
11810 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11811
11812         PR target/40411
11813         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11814         -symbolic.
11815         Use values-Xc.o for -pedantic.
11816         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11817
11818 2018-01-12  Martin Liska  <mliska@suse.cz>
11819
11820         PR ipa/83054
11821         * ipa-devirt.c (final_warning_record::grow_type_warnings):
11822         New function.
11823         (possible_polymorphic_call_targets): Use it.
11824         (ipa_devirt): Likewise.
11825
11826 2018-01-12  Martin Liska  <mliska@suse.cz>
11827
11828         * profile-count.h (enum profile_quality): Use 0 as invalid
11829         enum value of profile_quality.
11830
11831 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11832
11833         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11834         -mext-string options.
11835
11836 2018-01-12  Richard Biener  <rguenther@suse.de>
11837
11838         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11839         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11840         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11841         Likewise.
11842         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11843
11844 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
11845
11846         * configure.ac (--with-long-double-format): Add support for the
11847         configuration option to change the default long double format on
11848         PowerPC systems.
11849         * config.gcc (powerpc*-linux*-*): Likewise.
11850         * configure: Regenerate.
11851         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11852         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11853         used without modification.
11854
11855 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11856
11857         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11858         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11859         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11860         MISC_BUILTIN_SPEC_BARRIER.
11861         (rs6000_init_builtins): Likewise.
11862         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11863         enum value.
11864         (speculation_barrier): New define_insn.
11865         * doc/extend.texi: Document __builtin_speculation_barrier.
11866
11867 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
11868
11869         PR target/83203
11870         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11871         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11872         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11873         iterators.
11874         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
11875         integral modes instead of "ss" and "sd".
11876         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11877         vectors with 32-bit and 64-bit elements.
11878         (vecdupssescalarmodesuffix): New mode attribute.
11879         (vec_dup<mode>): Use it.
11880
11881 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
11882
11883         PR target/83330
11884         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11885         frame if argument is passed on stack.
11886
11887 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
11888
11889         PR target/82682
11890         * ree.c (combine_reaching_defs): Optimize also
11891         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11892         reg2=any_extend(exp); reg1=reg2;, formatting fix.
11893
11894 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
11895
11896         PR middle-end/83189
11897         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11898
11899 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
11900
11901         PR middle-end/83718
11902         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11903         after they are computed.
11904
11905 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
11906
11907         PR tree-optimization/83695
11908         * gimple-loop-linterchange.cc
11909         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11910         reset cached scev information after interchange.
11911         (pass_linterchange::execute): Remove call to scev_reset_htab.
11912
11913 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11914
11915         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11916         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11917         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11918         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11919         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11920         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11921         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11922         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11923         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11924         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11925         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11926         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11927         (V_lane_reg): Likewise.
11928         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11929         New define_expand.
11930         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11931         (vfmal_lane_low<mode>_intrinsic,
11932         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11933         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11934         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11935         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11936         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11937         vfmsl_lane_high<mode>_intrinsic): New define_insns.
11938
11939 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11940
11941         * config/arm/arm-cpus.in (fp16fml): New feature.
11942         (ALL_SIMD): Add fp16fml.
11943         (armv8.2-a): Add fp16fml as an option.
11944         (armv8.3-a): Likewise.
11945         (armv8.4-a): Add fp16fml as part of fp16.
11946         * config/arm/arm.h (TARGET_FP16FML): Define.
11947         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11948         when appropriate.
11949         * config/arm/arm-modes.def (V2HF): Define.
11950         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11951         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11952         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11953         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11954         vfmsl_low, vfmsl_high): New set of builtins.
11955         * config/arm/iterators.md (PLUSMINUS): New code iterator.
11956         (vfml_op): New code attribute.
11957         (VFMLHALVES): New int iterator.
11958         (VFML, VFMLSEL): New mode attributes.
11959         (V_reg): Define mapping for V2HF.
11960         (V_hi, V_lo): New mode attributes.
11961         (VF_constraint): Likewise.
11962         (vfml_half, vfml_half_selector): New int attributes.
11963         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11964         define_expand.
11965         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11966         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11967         New define_insn.
11968         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11969         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11970         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11971         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
11972         documentation.
11973         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11974         Document new effective target and option set.
11975
11976 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11977
11978         * config/arm/arm-cpus.in (armv8_4): New feature.
11979         (ARMv8_4a): New fgroup.
11980         (armv8.4-a): New arch.
11981         * config/arm/arm-tables.opt: Regenerate.
11982         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11983         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11984         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11985         Add matching rules for -march=armv8.4-a and extensions.
11986         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11987
11988 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
11989
11990         PR target/81821
11991         * config/rx/rx.md (BW): New mode attribute.
11992         (sync_lock_test_and_setsi): Add mode suffix to insn output.
11993
11994 2018-01-11  Richard Biener  <rguenther@suse.de>
11995
11996         PR tree-optimization/83435
11997         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11998         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11999         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12000
12001 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12002             Alan Hayward  <alan.hayward@arm.com>
12003             David Sherwood  <david.sherwood@arm.com>
12004
12005         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12006         field.
12007         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
12008         (aarch64_print_address_internal): Use it to check for a zero offset.
12009
12010 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12011             Alan Hayward  <alan.hayward@arm.com>
12012             David Sherwood  <david.sherwood@arm.com>
12013
12014         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12015         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12016         Return a poly_int64 rather than a HOST_WIDE_INT.
12017         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12018         rather than a HOST_WIDE_INT.
12019         * config/aarch64/aarch64.h (aarch64_frame): Protect with
12020         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
12021         hard_fp_offset, frame_size, initial_adjust, callee_offset and
12022         final_offset from HOST_WIDE_INT to poly_int64.
12023         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12024         to_constant when getting the number of units in an Advanced SIMD
12025         mode.
12026         (aarch64_builtin_vectorized_function): Check for a constant number
12027         of units.
12028         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12029         GET_MODE_SIZE.
12030         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12031         attribute instead of GET_MODE_NUNITS.
12032         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12033         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12034         GET_MODE_SIZE for fixed-size registers.
12035         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12036         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12037         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12038         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12039         (aarch64_print_operand, aarch64_print_address_internal)
12040         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12041         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12042         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12043         Handle polynomial GET_MODE_SIZE.
12044         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
12045         wider than SImode without modification.
12046         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12047         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12048         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12049         passing and returning SVE modes.
12050         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12051         rather than GEN_INT.
12052         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12053         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12054         (aarch64_allocate_and_probe_stack_space): Likewise.
12055         (aarch64_layout_frame): Cope with polynomial offsets.
12056         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12057         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
12058         polynomial offsets.
12059         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12060         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12061         poly_int64 rather than a HOST_WIDE_INT.
12062         (aarch64_get_separate_components, aarch64_process_components)
12063         (aarch64_expand_prologue, aarch64_expand_epilogue)
12064         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12065         (aarch64_anchor_offset): New function, split out from...
12066         (aarch64_legitimize_address): ...here.
12067         (aarch64_builtin_vectorization_cost): Handle polynomial
12068         TYPE_VECTOR_SUBPARTS.
12069         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12070         GET_MODE_NUNITS.
12071         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12072         number of elements from the PARALLEL rather than the mode.
12073         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12074         rather than GET_MODE_BITSIZE.
12075         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12076         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12077         (aarch64_expand_vec_perm_const_1): Handle polynomial
12078         d->perm.length () and d->perm elements.
12079         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
12080         Apply to_constant to d->perm elements.
12081         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12082         polynomial CONST_VECTOR_NUNITS.
12083         (aarch64_move_pointer): Take amount as a poly_int64 rather
12084         than an int.
12085         (aarch64_progress_pointer): Avoid temporary variable.
12086         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12087         the mode attribute instead of GET_MODE.
12088
12089 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12090             Alan Hayward  <alan.hayward@arm.com>
12091             David Sherwood  <david.sherwood@arm.com>
12092
12093         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12094         x exists before using it.
12095         (aarch64_add_constant_internal): Rename to...
12096         (aarch64_add_offset_1): ...this.  Replace regnum with separate
12097         src and dest rtxes.  Handle the case in which they're different,
12098         including when the offset is zero.  Replace scratchreg with an rtx.
12099         Use 2 additions if there is no spare register into which we can
12100         move a 16-bit constant.
12101         (aarch64_add_constant): Delete.
12102         (aarch64_add_offset): Replace reg with separate src and dest
12103         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
12104         Use aarch64_add_offset_1.
12105         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12106         an rtx rather than an int.  Take the delta as a poly_int64
12107         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
12108         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12109         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12110         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12111         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12112         and aarch64_add_sp.
12113         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12114         aarch64_add_constant.
12115
12116 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12117
12118         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12119         Use scalar_float_mode.
12120
12121 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12122
12123         * config/aarch64/aarch64-simd.md
12124         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12125         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12126         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12127         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12128         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12129         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12130         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12131         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12132         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12133         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12134
12135 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12136
12137         PR target/83514
12138         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12139         targ_options->x_arm_arch_string is non NULL.
12140
12141 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
12142
12143         * config/aarch64/aarch64.h
12144         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
12145
12146 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
12147
12148         PR target/82096
12149         * expmed.c (emit_store_flag_force): Swap if const op0
12150         and change VOIDmode to mode of op0.
12151
12152 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12153
12154         PR rtl-optimization/83761
12155         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12156         than bytes to mode_for_size.
12157
12158 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12159
12160         PR middle-end/83189
12161         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12162         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12163         profile.
12164
12165 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12166
12167         PR middle-end/83575
12168         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12169         when in layout mode.
12170         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12171         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12172         partition fixup.
12173
12174 2018-01-10  Michael Collison  <michael.collison@arm.com>
12175
12176         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12177         * config/aarch64/aarch64-option-extension.def: Add
12178         AARCH64_OPT_EXTENSION of 'fp16fml'.
12179         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12180         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12181         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12182         * config/aarch64/constraints.md (Ui7): New constraint.
12183         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12184         (VFMLA_SEL_W): Ditto.
12185         (f16quad): Ditto.
12186         (f16mac1): Ditto.
12187         (VFMLA16_LOW): New int iterator.
12188         (VFMLA16_HIGH): Ditto.
12189         (UNSPEC_FMLAL): New unspec.
12190         (UNSPEC_FMLSL): Ditto.
12191         (UNSPEC_FMLAL2): Ditto.
12192         (UNSPEC_FMLSL2): Ditto.
12193         (f16mac): New code attribute.
12194         * config/aarch64/aarch64-simd-builtins.def
12195         (aarch64_fmlal_lowv2sf): Ditto.
12196         (aarch64_fmlsl_lowv2sf): Ditto.
12197         (aarch64_fmlalq_lowv4sf): Ditto.
12198         (aarch64_fmlslq_lowv4sf): Ditto.
12199         (aarch64_fmlal_highv2sf): Ditto.
12200         (aarch64_fmlsl_highv2sf): Ditto.
12201         (aarch64_fmlalq_highv4sf): Ditto.
12202         (aarch64_fmlslq_highv4sf): Ditto.
12203         (aarch64_fmlal_lane_lowv2sf): Ditto.
12204         (aarch64_fmlsl_lane_lowv2sf): Ditto.
12205         (aarch64_fmlal_laneq_lowv2sf): Ditto.
12206         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12207         (aarch64_fmlalq_lane_lowv4sf): Ditto.
12208         (aarch64_fmlsl_lane_lowv4sf): Ditto.
12209         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12210         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12211         (aarch64_fmlal_lane_highv2sf): Ditto.
12212         (aarch64_fmlsl_lane_highv2sf): Ditto.
12213         (aarch64_fmlal_laneq_highv2sf): Ditto.
12214         (aarch64_fmlsl_laneq_highv2sf): Ditto.
12215         (aarch64_fmlalq_lane_highv4sf): Ditto.
12216         (aarch64_fmlsl_lane_highv4sf): Ditto.
12217         (aarch64_fmlalq_laneq_highv4sf): Ditto.
12218         (aarch64_fmlsl_laneq_highv4sf): Ditto.
12219         * config/aarch64/aarch64-simd.md:
12220         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12221         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12222         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12223         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12224         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12225         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12226         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12227         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12228         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12229         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12230         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12231         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12232         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12233         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12234         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12235         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12236         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12237         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12238         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12239         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12240         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12241         (vfmlsl_low_u32): Ditto.
12242         (vfmlalq_low_u32): Ditto.
12243         (vfmlslq_low_u32): Ditto.
12244         (vfmlal_high_u32): Ditto.
12245         (vfmlsl_high_u32): Ditto.
12246         (vfmlalq_high_u32): Ditto.
12247         (vfmlslq_high_u32): Ditto.
12248         (vfmlal_lane_low_u32): Ditto.
12249         (vfmlsl_lane_low_u32): Ditto.
12250         (vfmlal_laneq_low_u32): Ditto.
12251         (vfmlsl_laneq_low_u32): Ditto.
12252         (vfmlalq_lane_low_u32): Ditto.
12253         (vfmlslq_lane_low_u32): Ditto.
12254         (vfmlalq_laneq_low_u32): Ditto.
12255         (vfmlslq_laneq_low_u32): Ditto.
12256         (vfmlal_lane_high_u32): Ditto.
12257         (vfmlsl_lane_high_u32): Ditto.
12258         (vfmlal_laneq_high_u32): Ditto.
12259         (vfmlsl_laneq_high_u32): Ditto.
12260         (vfmlalq_lane_high_u32): Ditto.
12261         (vfmlslq_lane_high_u32): Ditto.
12262         (vfmlalq_laneq_high_u32): Ditto.
12263         (vfmlslq_laneq_high_u32): Ditto.
12264         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12265         (AARCH64_FL_FOR_ARCH8_4): New.
12266         (AARCH64_ISA_F16FML): New ISA flag.
12267         (TARGET_F16FML): New feature flag for fp16fml.
12268         (doc/invoke.texi): Document new fp16fml option.
12269
12270 2018-01-10  Michael Collison  <michael.collison@arm.com>
12271
12272         * config/aarch64/aarch64-builtins.c:
12273         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12274         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12275         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12276         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12277         (AARCH64_ISA_SHA3): New ISA flag.
12278         (TARGET_SHA3): New feature flag for sha3.
12279         * config/aarch64/iterators.md (sha512_op): New int attribute.
12280         (CRYPTO_SHA512): New int iterator.
12281         (UNSPEC_SHA512H): New unspec.
12282         (UNSPEC_SHA512H2): Ditto.
12283         (UNSPEC_SHA512SU0): Ditto.
12284         (UNSPEC_SHA512SU1): Ditto.
12285         * config/aarch64/aarch64-simd-builtins.def
12286         (aarch64_crypto_sha512hqv2di): New builtin.
12287         (aarch64_crypto_sha512h2qv2di): Ditto.
12288         (aarch64_crypto_sha512su0qv2di): Ditto.
12289         (aarch64_crypto_sha512su1qv2di): Ditto.
12290         (aarch64_eor3qv8hi): Ditto.
12291         (aarch64_rax1qv2di): Ditto.
12292         (aarch64_xarqv2di): Ditto.
12293         (aarch64_bcaxqv8hi): Ditto.
12294         * config/aarch64/aarch64-simd.md:
12295         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12296         (aarch64_crypto_sha512su0qv2di): Ditto.
12297         (aarch64_crypto_sha512su1qv2di): Ditto.
12298         (aarch64_eor3qv8hi): Ditto.
12299         (aarch64_rax1qv2di): Ditto.
12300         (aarch64_xarqv2di): Ditto.
12301         (aarch64_bcaxqv8hi): Ditto.
12302         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12303         (vsha512h2q_u64): Ditto.
12304         (vsha512su0q_u64): Ditto.
12305         (vsha512su1q_u64): Ditto.
12306         (veor3q_u16): Ditto.
12307         (vrax1q_u64): Ditto.
12308         (vxarq_u64): Ditto.
12309         (vbcaxq_u16): Ditto.
12310         * config/arm/types.md (crypto_sha512): New type attribute.
12311         (crypto_sha3): Ditto.
12312         (doc/invoke.texi): Document new sha3 option.
12313
12314 2018-01-10  Michael Collison  <michael.collison@arm.com>
12315
12316         * config/aarch64/aarch64-builtins.c:
12317         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12318         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12319         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12320         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12321         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12322         (AARCH64_ISA_SM4): New ISA flag.
12323         (TARGET_SM4): New feature flag for sm4.
12324         * config/aarch64/aarch64-simd-builtins.def
12325         (aarch64_sm3ss1qv4si): Ditto.
12326         (aarch64_sm3tt1aq4si): Ditto.
12327         (aarch64_sm3tt1bq4si): Ditto.
12328         (aarch64_sm3tt2aq4si): Ditto.
12329         (aarch64_sm3tt2bq4si): Ditto.
12330         (aarch64_sm3partw1qv4si): Ditto.
12331         (aarch64_sm3partw2qv4si): Ditto.
12332         (aarch64_sm4eqv4si): Ditto.
12333         (aarch64_sm4ekeyqv4si): Ditto.
12334         * config/aarch64/aarch64-simd.md:
12335         (aarch64_sm3ss1qv4si): Ditto.
12336         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12337         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12338         (aarch64_sm4eqv4si): Ditto.
12339         (aarch64_sm4ekeyqv4si): Ditto.
12340         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12341         (sm3part_op): Ditto.
12342         (CRYPTO_SM3TT): Ditto.
12343         (CRYPTO_SM3PART): Ditto.
12344         (UNSPEC_SM3SS1): New unspec.
12345         (UNSPEC_SM3TT1A): Ditto.
12346         (UNSPEC_SM3TT1B): Ditto.
12347         (UNSPEC_SM3TT2A): Ditto.
12348         (UNSPEC_SM3TT2B): Ditto.
12349         (UNSPEC_SM3PARTW1): Ditto.
12350         (UNSPEC_SM3PARTW2): Ditto.
12351         (UNSPEC_SM4E): Ditto.
12352         (UNSPEC_SM4EKEY): Ditto.
12353         * config/aarch64/constraints.md (Ui2): New constraint.
12354         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12355         * config/arm/types.md (crypto_sm3): New type attribute.
12356         (crypto_sm4): Ditto.
12357         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12358         (vsm3tt1aq_u32): Ditto.
12359         (vsm3tt1bq_u32): Ditto.
12360         (vsm3tt2aq_u32): Ditto.
12361         (vsm3tt2bq_u32): Ditto.
12362         (vsm3partw1q_u32): Ditto.
12363         (vsm3partw2q_u32): Ditto.
12364         (vsm4eq_u32): Ditto.
12365         (vsm4ekeyq_u32): Ditto.
12366         (doc/invoke.texi): Document new sm4 option.
12367
12368 2018-01-10  Michael Collison  <michael.collison@arm.com>
12369
12370         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12371         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12372         (AARCH64_FL_FOR_ARCH8_4): New.
12373         (AARCH64_FL_V8_4): New flag.
12374         (doc/invoke.texi): Document new armv8.4-a option.
12375
12376 2018-01-10  Michael Collison  <michael.collison@arm.com>
12377
12378         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12379         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12380         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12381         * config/aarch64/aarch64-option-extension.def: Add
12382         AARCH64_OPT_EXTENSION of 'sha2'.
12383         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12384         (crypto): Disable sha2 and aes if crypto disabled.
12385         (crypto): Enable aes and sha2 if enabled.
12386         (simd): Disable sha2 and aes if simd disabled.
12387         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12388         New flags.
12389         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12390         (TARGET_SHA2): New feature flag for sha2.
12391         (TARGET_AES): New feature flag for aes.
12392         * config/aarch64/aarch64-simd.md:
12393         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12394         conditional on TARGET_AES.
12395         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12396         (aarch64_crypto_sha1hsi): Make pattern conditional
12397         on TARGET_SHA2.
12398         (aarch64_crypto_sha1hv4si): Ditto.
12399         (aarch64_be_crypto_sha1hv4si): Ditto.
12400         (aarch64_crypto_sha1su1v4si): Ditto.
12401         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12402         (aarch64_crypto_sha1su0v4si): Ditto.
12403         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12404         (aarch64_crypto_sha256su0v4si): Ditto.
12405         (aarch64_crypto_sha256su1v4si): Ditto.
12406         (doc/invoke.texi): Document new aes and sha2 options.
12407
12408 2018-01-10  Martin Sebor  <msebor@redhat.com>
12409
12410         PR tree-optimization/83781
12411         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12412         as string arrays.
12413
12414 2018-01-11  Martin Sebor  <msebor@gmail.com>
12415             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12416
12417         PR tree-optimization/83501
12418         PR tree-optimization/81703
12419
12420         * tree-ssa-strlen.c (get_string_cst): Rename...
12421         (get_string_len): ...to this.  Handle global constants.
12422         (handle_char_store): Adjust.
12423
12424 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
12425             Jim Wilson  <jimw@sifive.com>
12426
12427         * config/riscv/riscv-protos.h (riscv_output_return): New.
12428         * config/riscv/riscv.c (struct machine_function): New naked_p field.
12429         (riscv_attribute_table, riscv_output_return),
12430         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12431         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12432         (riscv_compute_frame_info): Only compute frame->mask if not a naked
12433         function.
12434         (riscv_expand_prologue): Add early return for naked function.
12435         (riscv_expand_epilogue): Likewise.
12436         (riscv_function_ok_for_sibcall): Return false for naked function.
12437         (riscv_set_current_function): New.
12438         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12439         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12440         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12441         * doc/extend.texi (RISC-V Function Attributes): New.
12442
12443 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
12444
12445         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12446         check for 128-bit long double before checking TCmode.
12447         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12448         128-bit long doubles before checking TFmode or TCmode.
12449         (FLOAT128_IBM_P): Likewise.
12450
12451 2018-01-10  Martin Sebor  <msebor@redhat.com>
12452
12453         PR tree-optimization/83671
12454         * builtins.c (c_strlen): Unconditionally return zero for the empty
12455         string.
12456         Use -Warray-bounds for warnings.
12457         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12458         for non-constant array indices with COMPONENT_REF, arrays of
12459         arrays, and pointers to arrays.
12460         (gimple_fold_builtin_strlen): Determine and set length range for
12461         non-constant character arrays.
12462
12463 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
12464
12465         PR middle-end/81897
12466         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12467         empty blocks.
12468
12469 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
12470
12471         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12472
12473 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
12474
12475         PR target/83399
12476         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12477         VECTOR_MEM_ALTIVEC_OR_VSX_P.
12478         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12479         indexed_or_indirect_operand predicate.
12480         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12481         (*vsx_le_perm_load_v8hi): Likewise.
12482         (*vsx_le_perm_load_v16qi): Likewise.
12483         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12484         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12485         (*vsx_le_perm_store_v8hi): Likewise.
12486         (*vsx_le_perm_store_v16qi): Likewise.
12487         (eight unnamed splitters): Likewise.
12488
12489 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
12490
12491         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12492         * config/rs6000/emmintrin.h: Likewise.
12493         * config/rs6000/mmintrin.h: Likewise.
12494         * config/rs6000/xmmintrin.h: Likewise.
12495
12496 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
12497
12498         PR c++/43486
12499         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12500         "public_flag".
12501         * tree.c (tree_nop_conversion): Return true for location wrapper
12502         nodes.
12503         (maybe_wrap_with_location): New function.
12504         (selftest::check_strip_nops): New function.
12505         (selftest::test_location_wrappers): New function.
12506         (selftest::tree_c_tests): Call it.
12507         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12508         (maybe_wrap_with_location): New decl.
12509         (EXPR_LOCATION_WRAPPER_P): New macro.
12510         (location_wrapper_p): New inline function.
12511         (tree_strip_any_location_wrapper): New inline function.
12512
12513 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
12514
12515         PR target/83735
12516         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12517         stack_realign_offset for the largest alignment of stack slot
12518         actually used.
12519         (ix86_find_max_used_stack_alignment): New function.
12520         (ix86_finalize_stack_frame_flags): Use it.  Set
12521         max_used_stack_alignment if we don't realign stack.
12522         * config/i386/i386.h (machine_function): Add
12523         max_used_stack_alignment.
12524
12525 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
12526
12527         * config/arm/arm.opt (-mbranch-cost): New option.
12528         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12529         account.
12530
12531 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
12532
12533         PR target/83629
12534         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12535         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12536
12537 2018-01-10  Richard Biener  <rguenther@suse.de>
12538
12539         PR debug/83765
12540         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12541         early out so it also covers the case where we have a non-NULL
12542         origin.
12543
12544 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
12545
12546         PR tree-optimization/83753
12547         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12548         for non-strided grouped accesses if the number of elements is 1.
12549
12550 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12551
12552         PR target/81616
12553         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12554         * i386.h (TARGET_USE_GATHER): Define.
12555         * x86-tune.def (X86_TUNE_USE_GATHER): New.
12556
12557 2018-01-10  Martin Liska  <mliska@suse.cz>
12558
12559         PR bootstrap/82831
12560         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12561         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12562         partitioning.
12563         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12564         CLEANUP_NO_PARTITIONING is not set.
12565
12566 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
12567
12568         * doc/rtl.texi: Remove documentation of (const ...) wrappers
12569         for vectors, as a partial revert of r254296.
12570         * rtl.h (const_vec_p): Delete.
12571         (const_vec_duplicate_p): Don't test for vector CONSTs.
12572         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12573         * expmed.c (make_tree): Likewise.
12574
12575         Revert:
12576         * common.md (E, F): Use CONSTANT_P instead of checking for
12577         CONST_VECTOR.
12578         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12579         checking for CONST_VECTOR.
12580
12581 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
12582
12583         PR middle-end/83575
12584         * predict.c (force_edge_cold): Handle in more sane way edges
12585         with no prediction.
12586
12587 2018-01-09  Carl Love  <cel@us.ibm.com>
12588
12589         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12590         V4SI, V4SF types.
12591         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12592         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12593         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12594         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
12595         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12596         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
12597         * config/rs6000/rs6000-protos.h: Add extern defition for
12598         rs6000_generate_float2_double_code.
12599         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12600         function.
12601         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12602         (float2_v2df): Add define_expand.
12603
12604 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
12605
12606         PR target/83628
12607         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12608         op_mode in the force_to_mode call.
12609
12610 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
12611
12612         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12613         instead of checking each element individually.
12614         (aarch64_evpc_uzp): Likewise.
12615         (aarch64_evpc_zip): Likewise.
12616         (aarch64_evpc_ext): Likewise.
12617         (aarch64_evpc_rev): Likewise.
12618         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12619         instead of checking each element individually.  Return true without
12620         generating rtl if
12621         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12622         whether all selected elements come from the same input, instead of
12623         checking each element individually.  Remove calls to gen_rtx_REG,
12624         start_sequence and end_sequence and instead assert that no rtl is
12625         generated.
12626
12627 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
12628
12629         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12630         order of HIGH and CONST checks.
12631
12632 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
12633
12634         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12635         if the destination isn't an SSA_NAME.
12636
12637 2018-01-09  Richard Biener  <rguenther@suse.de>
12638
12639         PR tree-optimization/83668
12640         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12641         move prologue...
12642         (canonicalize_loop_form): ... here, renamed from ...
12643         (canonicalize_loop_closed_ssa_form): ... this and amended to
12644         swap successor edges for loop exit blocks to make us use
12645         the RPO order we need for initial schedule generation.
12646
12647 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
12648
12649         PR tree-optimization/64811
12650         * match.pd: When optimizing comparisons with Inf, avoid
12651         introducing or losing exceptions from comparisons with NaN.
12652
12653 2018-01-09  Martin Liska  <mliska@suse.cz>
12654
12655         PR sanitizer/82517
12656         * asan.c (shadow_mem_size): Add gcc_assert.
12657
12658 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
12659
12660         Don't save registers in main().
12661
12662         PR target/83738
12663         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12664         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12665         * config/avr/avr.c (avr_set_current_function): Don't error if
12666         naked, OS_task or OS_main are specified at the same time.
12667         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12668         OS_main.
12669         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12670         attribute.
12671         * common/config/avr/avr-common.c (avr_option_optimization_table):
12672         Switch on -mmain-is-OS_task for optimizing compilations.
12673
12674 2018-01-09  Richard Biener  <rguenther@suse.de>
12675
12676         PR tree-optimization/83572
12677         * graphite.c: Include cfganal.h.
12678         (graphite_transform_loops): Connect infinite loops to exit
12679         and remove fake edges at the end.
12680
12681 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
12682
12683         * ipa-inline.c (edge_badness): Revert accidental checkin.
12684
12685 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
12686
12687         PR ipa/80763
12688         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12689         symbols; not inline clones.
12690
12691 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
12692
12693         PR target/83507
12694         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12695         hard registers.  Formatting fixes.
12696
12697         PR preprocessor/83722
12698         * gcc.c (try_generate_repro): Pass
12699         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12700         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12701         do_report_bug.
12702
12703 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
12704             Kito Cheng  <kito.cheng@gmail.com>
12705
12706         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12707         (riscv_leaf_function_p): Delete.
12708         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12709
12710 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
12711
12712         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12713         function.
12714         (do_ifelse): New function.
12715         (do_isel): New function.
12716         (do_sub3): New function.
12717         (do_add3): New function.
12718         (do_load_mask_compare): New function.
12719         (do_overlap_load_compare): New function.
12720         (expand_compare_loop): New function.
12721         (expand_block_compare): Call expand_compare_loop() when appropriate.
12722         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12723         option description.
12724         (-mblock-compare-inline-loop-limit): New option.
12725
12726 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
12727
12728         PR target/83677
12729         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12730         Reverse order of second and third operands in first alternative.
12731         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12732         of first and second elements in UNSPEC_VPERMR vector.
12733         (altivec_expand_vec_perm_le): Likewise.
12734
12735 2018-01-08  Jeff Law  <law@redhat.com>
12736
12737         PR rtl-optimizatin/81308
12738         * tree-switch-conversion.c (cfg_altered): New file scoped static.
12739         (process_switch): If group_case_labels makes a change, then set
12740         cfg_altered.
12741         (pass_convert_switch::execute): If a switch is converted, then
12742         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
12743
12744         PR rtl-optimization/81308
12745         * recog.c (split_all_insns): Conditionally cleanup the CFG after
12746         splitting insns.
12747
12748 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
12749
12750         PR target/83663 - Revert r255946
12751         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12752         generation for cases where splatting a value is not useful.
12753         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12754         across a vec_duplicate and a paradoxical subreg forming a vector
12755         mode to a vec_concat.
12756
12757 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12758
12759         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12760         -march=armv8.3-a variants.
12761         * config/arm/t-multilib: Likewise.
12762         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
12763
12764 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
12765
12766         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12767         to generate rtl.
12768         (cceq_ior_compare_complement): Give it a name so I can use it, and
12769         change boolean_or_operator predicate to boolean_operator so it can
12770         be used to generate a crand.
12771         (eqne): New code iterator.
12772         (bd/bd_neg): New code_attrs.
12773         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12774         a single define_insn.
12775         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12776         decrement (bdnzt/bdnzf/bdzt/bdzf).
12777         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12778         with the new names of the branch decrement patterns, and added the
12779         names of the branch decrement conditional patterns.
12780
12781 2018-01-08  Richard Biener  <rguenther@suse.de>
12782
12783         PR tree-optimization/83563
12784         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12785         cache.
12786
12787 2018-01-08  Richard Biener  <rguenther@suse.de>
12788
12789         PR middle-end/83713
12790         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12791
12792 2018-01-08  Richard Biener  <rguenther@suse.de>
12793
12794         PR tree-optimization/83685
12795         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12796         references to abnormals.
12797
12798 2018-01-08  Richard Biener  <rguenther@suse.de>
12799
12800         PR lto/83719
12801         * dwarf2out.c (output_indirect_strings): Handle empty
12802         skeleton_debug_str_hash.
12803         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12804
12805 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
12806
12807         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12808         (emit_store_direct): Likewise.
12809         (arc_trampoline_adjust_address): Likewise.
12810         (arc_asm_trampoline_template): New function.
12811         (arc_initialize_trampoline): Use asm_trampoline_template.
12812         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12813         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12814         * config/arc/arc.md (flush_icache): Delete pattern.
12815
12816 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
12817
12818         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12819         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12820         munaligned-access.
12821
12822 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12823
12824         PR target/83681
12825         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12826         by not USED_FOR_TARGET.
12827         (make_pass_resolve_sw_modes): Likewise.
12828
12829 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12830
12831         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12832         USED_FOR_TARGET.
12833
12834 2018-01-08  Richard Biener  <rguenther@suse.de>
12835
12836         PR middle-end/83580
12837         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12838
12839 2018-01-08  Richard Biener  <rguenther@suse.de>
12840
12841         PR middle-end/83517
12842         * match.pd ((t * 2) / 2) -> t): Add missing :c.
12843
12844 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
12845
12846         PR middle-end/81897
12847         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12848         basic blocks with a small number of successors.
12849         (convert_control_dep_chain_into_preds): Improve handling of
12850         forwarder blocks.
12851         (dump_predicates): Split apart into...
12852         (dump_pred_chain): ...here...
12853         (dump_pred_info): ...and here.
12854         (can_one_predicate_be_invalidated_p): Add debugging printfs.
12855         (can_chain_union_be_invalidated_p): Improve check for invalidation
12856         of paths.
12857         (uninit_uses_cannot_happen): Avoid unnecessary if
12858         convert_control_dep_chain_into_preds yielded nothing.
12859
12860 2018-01-06  Martin Sebor  <msebor@redhat.com>
12861
12862         PR tree-optimization/83640
12863         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12864         subtracting negative offset from size.
12865         (builtin_access::overlap): Adjust offset bounds of the access to fall
12866         within the size of the object if possible.
12867
12868 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
12869
12870         PR rtl-optimization/83699
12871         * expmed.c (extract_bit_field_1): Restrict the vector usage of
12872         extract_bit_field_as_subreg to cases in which the extracted
12873         value is also a vector.
12874
12875         * lra-constraints.c (process_alt_operands): Test for the equivalence
12876         substitutions when detecting a possible reload cycle.
12877
12878 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
12879
12880         PR debug/83480
12881         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12882         by default if flag_selective_schedling{,2}.  Formatting fixes.
12883
12884         PR rtl-optimization/83682
12885         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12886         if it has non-VECTOR_MODE element mode.
12887         (vec_duplicate_p): Likewise.
12888
12889         PR middle-end/83694
12890         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12891         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12892
12893 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
12894
12895         PR target/83604
12896         * config/i386/i386-builtin.def
12897         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12898         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12899         Require also OPTION_MASK_ISA_AVX512F in addition to
12900         OPTION_MASK_ISA_GFNI.
12901         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12902         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12903         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12904         to OPTION_MASK_ISA_GFNI.
12905         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12906         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12907         OPTION_MASK_ISA_AVX512BW.
12908         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12909         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12910         addition to OPTION_MASK_ISA_GFNI.
12911         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12912         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12913         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12914         to OPTION_MASK_ISA_GFNI.
12915         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12916         a requirement for all ISAs rather than any of them with a few
12917         exceptions.
12918         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12919         processing.
12920         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12921         bitmasks to be enabled with 3 exceptions, instead of requiring any
12922         enabled ISA with lots of exceptions.
12923         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12924         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12925         Change avx512bw in isa attribute to avx512f.
12926         * config/i386/sgxintrin.h: Add license boilerplate.
12927         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
12928         to __AVX512F__ and __AVX512VL to __AVX512VL__.
12929         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12930         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12931         defined.
12932         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12933         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12934         temporarily sse2 rather than sse if not enabled already.
12935
12936         PR target/83604
12937         * config/i386/sse.md (VI248_VLBW): Rename to ...
12938         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
12939         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12940         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12941         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12942         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12943         mode iterator instead of VI248_VLBW.
12944
12945 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
12946
12947         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12948         (record_modified): Skip clobbers; add debug output.
12949         (param_change_prob): Use sreal frequencies.
12950
12951 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
12952
12953         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12954         punt for user-aligned variables.
12955
12956 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
12957
12958         * tree-chrec.c (chrec_contains_symbols): Return true for
12959         POLY_INT_CST.
12960
12961 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
12962
12963         PR target/82439
12964         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12965         of (x|y) == x for BICS pattern.
12966
12967 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
12968
12969         PR tree-optimization/83605
12970         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12971         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12972         can throw.
12973
12974 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12975
12976         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12977         * config/epiphany/rtems.h: New file.
12978
12979 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
12980             Uros Bizjak  <ubizjak@gmail.com>
12981
12982         PR target/83554
12983         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12984         QIreg_operand instead of register_operand predicate.
12985         * config/i386/i386.c (ix86_rop_should_change_byte_p,
12986         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12987         comments instead of -fmitigate[-_]rop.
12988
12989 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12990
12991         PR bootstrap/81926
12992         * cgraphunit.c (symbol_table::compile): Switch to text_section
12993         before calling assembly_start debug hook.
12994         * run-rtl-passes.c (run_rtl_passes): Likewise.
12995         Include output.h.
12996
12997 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
12998
12999         * tree-vrp.c (extract_range_from_binary_expr_1): Check
13000         range_int_cst_p rather than !symbolic_range_p before calling
13001         extract_range_from_multiplicative_op_1.
13002
13003 2018-01-04  Jeff Law  <law@redhat.com>
13004
13005         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13006         redundant test in assertion.
13007
13008 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13009
13010         * doc/rtl.texi: Document machine_mode wrapper classes.
13011
13012 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13013
13014         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13015         using tree_to_uhwi.
13016
13017 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13018
13019         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13020         the VEC_PERM_EXPR fold to fail.
13021
13022 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13023
13024         PR debug/83585
13025         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13026         to switched_sections.
13027
13028 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13029
13030         PR target/83680
13031         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13032         test for d.testing.
13033
13034 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
13035
13036         PR target/83387
13037         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13038         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13039
13040 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13041
13042         PR debug/83666
13043         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13044         is BLKmode and bitpos not zero or mode change is needed.
13045
13046 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13047
13048         PR target/83675
13049         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13050         TARGET_VIS2.
13051
13052 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
13053
13054         PR target/83628
13055         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13056         instead of MULT rtx.  Update all corresponding splitters.
13057         (*saddl_se): Ditto.
13058         (*ssub<modesuffix>): Ditto.
13059         (*ssubl_se): Ditto.
13060         (*cmp_sadd_di): Update split patterns.
13061         (*cmp_sadd_si): Ditto.
13062         (*cmp_sadd_sidi): Ditto.
13063         (*cmp_ssub_di): Ditto.
13064         (*cmp_ssub_si): Ditto.
13065         (*cmp_ssub_sidi): Ditto.
13066         * config/alpha/predicates.md (const23_operand): New predicate.
13067         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13068         Look for ASHIFT, not MULT inner operand.
13069         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13070
13071 2018-01-04  Martin Liska  <mliska@suse.cz>
13072
13073         PR gcov-profile/83669
13074         * gcov.c (output_intermediate_file): Add version to intermediate
13075         gcov file.
13076         * doc/gcov.texi: Document new field 'version' in intermediate
13077         file format. Fix location of '-k' option of gcov command.
13078
13079 2018-01-04  Martin Liska  <mliska@suse.cz>
13080
13081         PR ipa/82352
13082         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13083
13084 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13085
13086         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13087
13088 2018-01-03  Martin Sebor  <msebor@redhat.com>
13089
13090         PR tree-optimization/83655
13091         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13092         checking calls with invalid arguments.
13093
13094 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13095
13096         * tree-vect-stmts.c (vect_get_store_rhs): New function.
13097         (vectorizable_mask_load_store): Delete.
13098         (vectorizable_call): Return false for masked loads and stores.
13099         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
13100         instead of gimple_assign_rhs1.
13101         (vectorizable_load): Handle IFN_MASK_LOAD.
13102         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13103
13104 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13105
13106         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13107         split out from..,
13108         (vectorizable_mask_load_store): ...here.
13109         (vectorizable_load): ...and here.
13110
13111 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13112
13113         * tree-vect-stmts.c (vect_build_all_ones_mask)
13114         (vect_build_zero_merge_argument): New functions, split out from...
13115         (vectorizable_load): ...here.
13116
13117 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13118
13119         * tree-vect-stmts.c (vect_check_store_rhs): New function,
13120         split out from...
13121         (vectorizable_mask_load_store): ...here.
13122         (vectorizable_store): ...and here.
13123
13124 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13125
13126         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13127         split out from...
13128         (vectorizable_mask_load_store): ...here.
13129
13130 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13131
13132         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13133         (vect_model_store_cost): Take a vec_load_store_type instead of a
13134         vect_def_type.
13135         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13136         (vect_model_store_cost): Take a vec_load_store_type instead of a
13137         vect_def_type.
13138         (vectorizable_mask_load_store): Update accordingly.
13139         (vectorizable_store): Likewise.
13140         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13141
13142 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13143
13144         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13145         IFN_MASK_LOAD calls here rather than...
13146         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13147
13148 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13149             Alan Hayward  <alan.hayward@arm.com>
13150             David Sherwood  <david.sherwood@arm.com>
13151
13152         * expmed.c (extract_bit_field_1): For vector extracts,
13153         fall back to extract_bit_field_as_subreg if vec_extract
13154         isn't available.
13155
13156 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13157             Alan Hayward  <alan.hayward@arm.com>
13158             David Sherwood  <david.sherwood@arm.com>
13159
13160         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13161         they are variable or constant sized.
13162         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13163         slots for constant-sized data.
13164
13165 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13166             Alan Hayward  <alan.hayward@arm.com>
13167             David Sherwood  <david.sherwood@arm.com>
13168
13169         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13170         handling COND_EXPRs with boolean comparisons, try to find a better
13171         basis for the mask type than the boolean itself.
13172
13173 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13174
13175         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13176         is calculated and how it can be overridden.
13177         * genmodes.c (max_bitsize_mode_any_mode): New variable.
13178         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13179         if defined.
13180         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13181         if nonzero.
13182
13183 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13184             Alan Hayward  <alan.hayward@arm.com>
13185             David Sherwood  <david.sherwood@arm.com>
13186
13187         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13188         Remove the mode argument.
13189         (aarch64_simd_valid_immediate): Remove the mode and inverse
13190         arguments.
13191         * config/aarch64/iterators.md (bitsize): New iterator.
13192         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13193         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13194         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13195         aarch64_simd_valid_immediate.
13196         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13197         (aarch64_reg_or_bic_imm): Likewise.
13198         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13199         with an insn_type enum and msl with a modifier_type enum.
13200         Replace element_width with a scalar_mode.  Change the shift
13201         to unsigned int.  Add constructors for scalar_float_mode and
13202         scalar_int_mode elements.
13203         (aarch64_vect_float_const_representable_p): Delete.
13204         (aarch64_can_const_movi_rtx_p)
13205         (aarch64_simd_scalar_immediate_valid_for_move)
13206         (aarch64_simd_make_constant): Update call to
13207         aarch64_simd_valid_immediate.
13208         (aarch64_advsimd_valid_immediate_hs): New function.
13209         (aarch64_advsimd_valid_immediate): Likewise.
13210         (aarch64_simd_valid_immediate): Remove mode and inverse
13211         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
13212         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13213         and aarch64_float_const_representable_p on the result.
13214         (aarch64_output_simd_mov_immediate): Remove mode argument.
13215         Update call to aarch64_simd_valid_immediate and use of
13216         simd_immediate_info.
13217         (aarch64_output_scalar_simd_mov_immediate): Update call
13218         accordingly.
13219
13220 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13221             Alan Hayward  <alan.hayward@arm.com>
13222             David Sherwood  <david.sherwood@arm.com>
13223
13224         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13225         (mode_nunits): Likewise CONST_MODE_NUNITS.
13226         * machmode.def (ADJUST_NUNITS): Document.
13227         * genmodes.c (mode_data::need_nunits_adj): New field.
13228         (blank_mode): Update accordingly.
13229         (adj_nunits): New variable.
13230         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13231         parameter.
13232         (emit_mode_size_inline): Set need_bytesize_adj for all modes
13233         listed in adj_nunits.
13234         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13235         listed in adj_nunits.  Don't emit case statements for such modes.
13236         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13237         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
13238         nothing if adj_nunits is nonnull.
13239         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13240         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13241         (emit_mode_fbit): Update use of print_maybe_const_decl.
13242         (emit_move_size): Likewise.  Treat the array as non-const
13243         if adj_nunits.
13244         (emit_mode_adjustments): Handle adj_nunits.
13245
13246 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13247
13248         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13249         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13250         (VECTOR_MODES): Use it.
13251         (make_vector_modes): Take the prefix as an argument.
13252
13253 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13254             Alan Hayward  <alan.hayward@arm.com>
13255             David Sherwood  <david.sherwood@arm.com>
13256
13257         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13258         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13259         for MODE_VECTOR_BOOL.
13260         * machmode.def (VECTOR_BOOL_MODE): Document.
13261         * genmodes.c (VECTOR_BOOL_MODE): New macro.
13262         (make_vector_bool_mode): New function.
13263         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13264         MODE_VECTOR_BOOL.
13265         * lto-streamer-in.c (lto_input_mode_table): Likewise.
13266         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13267         Likewise.
13268         * stor-layout.c (int_mode_for_mode): Likewise.
13269         * tree.c (build_vector_type_for_mode): Likewise.
13270         * varasm.c (output_constant_pool_2): Likewise.
13271         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13272         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
13273         for MODE_VECTOR_BOOL.
13274         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13275         of mode class checks.
13276         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13277         instead of a list of mode class checks.
13278         (expand_vector_scalar_condition): Likewise.
13279         (type_for_widest_vector_mode): Handle BImode as an inner mode.
13280
13281 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13282             Alan Hayward  <alan.hayward@arm.com>
13283             David Sherwood  <david.sherwood@arm.com>
13284
13285         * machmode.h (mode_size): Change from unsigned short to
13286         poly_uint16_pod.
13287         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13288         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13289         or if measurement_type is not polynomial.
13290         (fixed_size_mode::includes_p): Check for constant-sized modes.
13291         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13292         return a poly_uint16 rather than an unsigned short.
13293         (emit_mode_size): Change the type of mode_size from unsigned short
13294         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
13295         (emit_mode_adjustments): Cope with polynomial vector sizes.
13296         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13297         for GET_MODE_SIZE.
13298         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13299         for GET_MODE_SIZE.
13300         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13301         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13302         * caller-save.c (setup_save_areas): Likewise.
13303         (replace_reg_with_saved_mem): Likewise.
13304         * calls.c (emit_library_call_value_1): Likewise.
13305         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13306         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13307         (gen_lowpart_for_combine): Likewise.
13308         * convert.c (convert_to_integer_1): Likewise.
13309         * cse.c (equiv_constant, cse_insn): Likewise.
13310         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13311         (cselib_subst_to_values): Likewise.
13312         * dce.c (word_dce_process_block): Likewise.
13313         * df-problems.c (df_word_lr_mark_ref): Likewise.
13314         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13315         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13316         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13317         (rtl_for_decl_location): Likewise.
13318         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13319         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13320         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13321         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13322         (expand_expr_real_1): Likewise.
13323         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13324         (pad_below): Likewise.
13325         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13326         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13327         * ira.c (get_subreg_tracking_sizes): Likewise.
13328         * ira-build.c (ira_create_allocno_objects): Likewise.
13329         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13330         (ira_sort_regnos_for_alter_reg): Likewise.
13331         * ira-costs.c (record_operand_costs): Likewise.
13332         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13333         (resolve_simple_move): Likewise.
13334         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13335         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13336         (lra_constraints): Likewise.
13337         (CONST_POOL_OK_P): Reject variable-sized modes.
13338         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13339         (add_pseudo_to_slot, lra_spill): Likewise.
13340         * omp-low.c (omp_clause_aligned_alignment): Likewise.
13341         * optabs-query.c (get_best_extraction_insn): Likewise.
13342         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13343         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13344         (expand_mult_highpart, valid_multiword_target_p): Likewise.
13345         * recog.c (offsettable_address_addr_space_p): Likewise.
13346         * regcprop.c (maybe_mode_change): Likewise.
13347         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13348         * regrename.c (build_def_use): Likewise.
13349         * regstat.c (dump_reg_info): Likewise.
13350         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13351         (find_reloads, find_reloads_subreg_address): Likewise.
13352         * reload1.c (eliminate_regs_1): Likewise.
13353         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13354         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13355         (simplify_binary_operation_1, simplify_subreg): Likewise.
13356         * targhooks.c (default_function_arg_padding): Likewise.
13357         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13358         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13359         (verify_gimple_assign_ternary): Likewise.
13360         * tree-inline.c (estimate_move_cost): Likewise.
13361         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13362         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13363         (get_address_cost_ainc): Likewise.
13364         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13365         (vect_supportable_dr_alignment): Likewise.
13366         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13367         (vectorizable_reduction): Likewise.
13368         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13369         (vectorizable_operation, vectorizable_load): Likewise.
13370         * tree.c (build_same_sized_truth_vector_type): Likewise.
13371         * valtrack.c (cleanup_auto_inc_dec): Likewise.
13372         * var-tracking.c (emit_note_insn_var_location): Likewise.
13373         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13374         (ADDR_VEC_ALIGN): Likewise.
13375
13376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13377             Alan Hayward  <alan.hayward@arm.com>
13378             David Sherwood  <david.sherwood@arm.com>
13379
13380         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13381         unsigned short.
13382         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13383         or if measurement_type is polynomial.
13384         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13385         * combine.c (make_extraction): Likewise.
13386         * dse.c (find_shift_sequence): Likewise.
13387         * dwarf2out.c (mem_loc_descriptor): Likewise.
13388         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13389         (extract_bit_field, extract_low_bits): Likewise.
13390         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13391         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13392         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13393         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13394         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13395         * reload.c (find_reloads): Likewise.
13396         * reload1.c (alter_reg): Likewise.
13397         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13398         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13399         * tree-if-conv.c (predicate_mem_writes): Likewise.
13400         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13401         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13402         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13403         * valtrack.c (dead_debug_insert_temp): Likewise.
13404         * varasm.c (mergeable_constant_section): Likewise.
13405         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13406
13407 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13408             Alan Hayward  <alan.hayward@arm.com>
13409             David Sherwood  <david.sherwood@arm.com>
13410
13411         * expr.c (expand_assignment): Cope with polynomial mode sizes
13412         when assigning to a CONCAT.
13413
13414 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13415             Alan Hayward  <alan.hayward@arm.com>
13416             David Sherwood  <david.sherwood@arm.com>
13417
13418         * machmode.h (mode_precision): Change from unsigned short to
13419         poly_uint16_pod.
13420         (mode_to_precision): Return a poly_uint16 rather than an unsigned
13421         short.
13422         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13423         or if measurement_type is not polynomial.
13424         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
13425         in which the mode is already known to be a scalar_int_mode.
13426         * genmodes.c (emit_mode_precision): Change the type of mode_precision
13427         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
13428         initializer.
13429         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13430         for GET_MODE_PRECISION.
13431         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13432         for GET_MODE_PRECISION.
13433         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13434         as polynomial.
13435         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13436         (expand_field_assignment, make_extraction): Likewise.
13437         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13438         (get_last_value): Likewise.
13439         * convert.c (convert_to_integer_1): Likewise.
13440         * cse.c (cse_insn): Likewise.
13441         * expr.c (expand_expr_real_1): Likewise.
13442         * lra-constraints.c (simplify_operand_subreg): Likewise.
13443         * optabs-query.c (can_atomic_load_p): Likewise.
13444         * optabs.c (expand_atomic_load): Likewise.
13445         (expand_atomic_store): Likewise.
13446         * ree.c (combine_reaching_defs): Likewise.
13447         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13448         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13449         * tree.h (type_has_mode_precision_p): Likewise.
13450         * ubsan.c (instrument_si_overflow): Likewise.
13451
13452 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13453             Alan Hayward  <alan.hayward@arm.com>
13454             David Sherwood  <david.sherwood@arm.com>
13455
13456         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13457         polynomial numbers of units.
13458         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13459         (valid_vector_subparts_p): New function.
13460         (build_vector_type): Remove temporary shim and take the number
13461         of units as a poly_uint64 rather than an int.
13462         (build_opaque_vector_type): Take the number of units as a
13463         poly_uint64 rather than an int.
13464         * tree.c (build_vector_from_ctor): Handle polynomial
13465         TYPE_VECTOR_SUBPARTS.
13466         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13467         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13468         (build_vector_from_val): If the number of units is variable,
13469         use build_vec_duplicate_cst for constant operands and
13470         VEC_DUPLICATE_EXPR otherwise.
13471         (make_vector_type): Remove temporary is_constant ().
13472         (build_vector_type, build_opaque_vector_type): Take the number of
13473         units as a poly_uint64 rather than an int.
13474         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13475         VECTOR_CST_NELTS.
13476         * cfgexpand.c (expand_debug_expr): Likewise.
13477         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13478         (store_constructor, expand_expr_real_1): Likewise.
13479         (const_scalar_mask_from_tree): Likewise.
13480         * fold-const-call.c (fold_const_reduction): Likewise.
13481         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13482         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13483         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13484         (fold_relational_const): Likewise.
13485         (native_interpret_vector): Likewise.  Change the size from an
13486         int to an unsigned int.
13487         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13488         TYPE_VECTOR_SUBPARTS.
13489         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13490         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13491         duplicating a non-constant operand into a variable-length vector.
13492         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13493         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13494         * ipa-icf.c (sem_variable::equals): Likewise.
13495         * match.pd: Likewise.
13496         * omp-simd-clone.c (simd_clone_subparts): Likewise.
13497         * print-tree.c (print_node): Likewise.
13498         * stor-layout.c (layout_type): Likewise.
13499         * targhooks.c (default_builtin_vectorization_cost): Likewise.
13500         * tree-cfg.c (verify_gimple_comparison): Likewise.
13501         (verify_gimple_assign_binary): Likewise.
13502         (verify_gimple_assign_ternary): Likewise.
13503         (verify_gimple_assign_single): Likewise.
13504         * tree-pretty-print.c (dump_generic_node): Likewise.
13505         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13506         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13507         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13508         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13509         (vect_shift_permute_load_chain): Likewise.
13510         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13511         (expand_vector_condition, optimize_vector_constructor): Likewise.
13512         (lower_vec_perm, get_compute_type): Likewise.
13513         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13514         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13515         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13516         (vect_recog_mask_conversion_pattern): Likewise.
13517         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13518         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13519         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13520         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13521         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13522         (vectorizable_shift, vectorizable_operation, vectorizable_store)
13523         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13524         (supportable_widening_operation): Likewise.
13525         (supportable_narrowing_operation): Likewise.
13526         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13527         Likewise.
13528         * varasm.c (output_constant): Likewise.
13529
13530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13531             Alan Hayward  <alan.hayward@arm.com>
13532             David Sherwood  <david.sherwood@arm.com>
13533
13534         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13535         so that both the length == 3 and length != 3 cases set up their
13536         own permute vectors.  Add comments explaining why we know the
13537         number of elements is constant.
13538         (vect_permute_load_chain): Likewise.
13539
13540 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13541             Alan Hayward  <alan.hayward@arm.com>
13542             David Sherwood  <david.sherwood@arm.com>
13543
13544         * machmode.h (mode_nunits): Change from unsigned char to
13545         poly_uint16_pod.
13546         (ONLY_FIXED_SIZE_MODES): New macro.
13547         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13548         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13549         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13550         New typedefs.
13551         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13552         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13553         or if measurement_type is not polynomial.
13554         * genmodes.c (ZERO_COEFFS): New macro.
13555         (emit_mode_nunits_inline): Make mode_nunits_inline return a
13556         poly_uint16.
13557         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13558         Use ZERO_COEFFS when emitting initializers.
13559         * data-streamer.h (bp_pack_poly_value): New function.
13560         (bp_unpack_poly_value): Likewise.
13561         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13562         for GET_MODE_NUNITS.
13563         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13564         for GET_MODE_NUNITS.
13565         * tree.c (make_vector_type): Remove temporary shim and make
13566         the real function take the number of units as a poly_uint64
13567         rather than an int.
13568         (build_vector_type_for_mode): Handle polynomial nunits.
13569         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13570         * emit-rtl.c (const_vec_series_p_1): Likewise.
13571         (gen_rtx_CONST_VECTOR): Likewise.
13572         * fold-const.c (test_vec_duplicate_folding): Likewise.
13573         * genrecog.c (validate_pattern): Likewise.
13574         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13575         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13576         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13577         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13578         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13579         * rtlanal.c (subreg_get_info): Likewise.
13580         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13581         (vect_grouped_load_supported): Likewise.
13582         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13583         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13584         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13585         (simplify_const_unary_operation, simplify_binary_operation_1)
13586         (simplify_const_binary_operation, simplify_ternary_operation)
13587         (test_vector_ops_duplicate, test_vector_ops): Likewise.
13588         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13589         instead of CONST_VECTOR_NUNITS.
13590         * varasm.c (output_constant_pool_2): Likewise.
13591         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13592         explicit-encoded elements in the XVEC for variable-length vectors.
13593
13594 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13595
13596         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13597
13598 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13599             Alan Hayward  <alan.hayward@arm.com>
13600             David Sherwood  <david.sherwood@arm.com>
13601
13602         * coretypes.h (fixed_size_mode): Declare.
13603         (fixed_size_mode_pod): New typedef.
13604         * builtins.h (target_builtins::x_apply_args_mode)
13605         (target_builtins::x_apply_result_mode): Change type to
13606         fixed_size_mode_pod.
13607         * builtins.c (apply_args_size, apply_result_size, result_vector)
13608         (expand_builtin_apply_args_1, expand_builtin_apply)
13609         (expand_builtin_return): Update accordingly.
13610
13611 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13612
13613         * cse.c (hash_rtx_cb): Hash only the encoded elements.
13614         * cselib.c (cselib_hash_rtx): Likewise.
13615         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13616         CONST_VECTOR encoding.
13617
13618 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
13619             Jeff Law  <law@redhat.com>
13620
13621         PR target/83641
13622         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13623         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13624         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13625         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13626
13627         PR target/83641
13628         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13629         explicitly probe *sp in a noreturn function if there were any callee
13630         register saves or frame pointer is needed.
13631
13632 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
13633
13634         PR debug/83621
13635         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13636         BLKmode for ternary, binary or unary expressions.
13637
13638         PR debug/83645
13639         * var-tracking.c (delete_vta_debug_insn): New inline function.
13640         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13641         insns from get_insns () to NULL instead of each bb separately.
13642         Use delete_vta_debug_insn.  No longer static.
13643         (vt_debug_insns_local, variable_tracking_main_1): Adjust
13644         delete_vta_debug_insns callers.
13645         * rtl.h (delete_vta_debug_insns): Declare.
13646         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13647         instead of variable_tracking_main.
13648
13649 2018-01-03  Martin Sebor  <msebor@redhat.com>
13650
13651         PR tree-optimization/83603
13652         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13653         arguments past the endof the argument list in functions declared
13654         without a prototype.
13655         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13656         Avoid checking when arguments are null.
13657
13658 2018-01-03  Martin Sebor  <msebor@redhat.com>
13659
13660         PR c/83559
13661         * doc/extend.texi (attribute const): Fix a typo.
13662         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13663         issuing -Wsuggest-attribute for void functions.
13664
13665 2018-01-03  Martin Sebor  <msebor@redhat.com>
13666
13667         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13668         offset_int::from instead of wide_int::to_shwi.
13669         (maybe_diag_overlap): Remove assertion.
13670         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13671         * gimple-ssa-sprintf.c (format_directive): Same.
13672         (parse_directive): Same.
13673         (sprintf_dom_walker::compute_format_length): Same.
13674         (try_substitute_return_value): Same.
13675
13676 2018-01-03  Jeff Law  <law@redhat.com>
13677
13678         PR middle-end/83654
13679         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13680         non-constant residual for zero at runtime and avoid probing in
13681         that case.  Reorganize code for trailing problem to mirror handling
13682         of the residual.
13683
13684 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13685
13686         PR tree-optimization/83501
13687         * tree-ssa-strlen.c (get_string_cst): New.
13688         (handle_char_store): Call get_string_cst.
13689
13690 2018-01-03  Martin Liska  <mliska@suse.cz>
13691
13692         PR tree-optimization/83593
13693         * tree-ssa-strlen.c: Include tree-cfg.h.
13694         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13695         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13696         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13697         to false.
13698         (strlen_dom_walker::before_dom_children): Call
13699         gimple_purge_dead_eh_edges. Dump tranformation with details
13700         dump flags.
13701         (strlen_dom_walker::before_dom_children): Update call by adding
13702         new argument cleanup_eh.
13703         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13704
13705 2018-01-03  Martin Liska  <mliska@suse.cz>
13706
13707         PR ipa/83549
13708         * cif-code.def (VARIADIC_THUNK): New enum value.
13709         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13710         thunks.
13711
13712 2018-01-03  Jan Beulich  <jbeulich@suse.com>
13713
13714         * sse.md (mov<mode>_internal): Tighten condition for when to use
13715         vmovdqu<ssescalarsize> for TI and OI modes.
13716
13717 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
13718
13719         Update copyright years.
13720
13721 2018-01-03  Martin Liska  <mliska@suse.cz>
13722
13723         PR ipa/83594
13724         * ipa-visibility.c (function_and_variable_visibility): Skip
13725         functions with noipa attribure.
13726
13727 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
13728
13729         * gcc.c (process_command): Update copyright notice dates.
13730         * gcov-dump.c (print_version): Ditto.
13731         * gcov.c (print_version): Ditto.
13732         * gcov-tool.c (print_version): Ditto.
13733         * gengtype.c (create_file): Ditto.
13734         * doc/cpp.texi: Bump @copying's copyright year.
13735         * doc/cppinternals.texi: Ditto.
13736         * doc/gcc.texi: Ditto.
13737         * doc/gccint.texi: Ditto.
13738         * doc/gcov.texi: Ditto.
13739         * doc/install.texi: Ditto.
13740         * doc/invoke.texi: Ditto.
13741
13742 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13743
13744         * vector-builder.h (vector_builder::m_full_nelts): Change from
13745         unsigned int to poly_uint64.
13746         (vector_builder::full_nelts): Update prototype accordingly.
13747         (vector_builder::new_vector): Likewise.
13748         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13749         (vector_builder::operator ==): Likewise.
13750         (vector_builder::finalize): Likewise.
13751         * int-vector-builder.h (int_vector_builder::int_vector_builder):
13752         Take the number of elements as a poly_uint64 rather than an
13753         unsigned int.
13754         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13755         from unsigned int to poly_uint64.
13756         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13757         (vec_perm_indices::new_vector): Likewise.
13758         (vec_perm_indices::length): Likewise.
13759         (vec_perm_indices::nelts_per_input): Likewise.
13760         (vec_perm_indices::input_nelts): Likewise.
13761         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13762         number of elements per input as a poly_uint64 rather than an
13763         unsigned int.  Use the original encoding for variable-length
13764         vectors, rather than clamping each individual element.
13765         For the second and subsequent elements in each pattern,
13766         clamp the step and base before clamping their sum.
13767         (vec_perm_indices::series_p): Handle polynomial element counts.
13768         (vec_perm_indices::all_in_range_p): Likewise.
13769         (vec_perm_indices_to_tree): Likewise.
13770         (vec_perm_indices_to_rtx): Likewise.
13771         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13772         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13773         (tree_vector_builder::new_binary_operation): Handle polynomial
13774         element counts.  Return false if we need to know the number
13775         of elements at compile time.
13776         * fold-const.c (fold_vec_perm): Punt if the number of elements
13777         isn't known at compile time.
13778
13779 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13780
13781         * vec-perm-indices.h (vec_perm_builder): Change element type
13782         from HOST_WIDE_INT to poly_int64.
13783         (vec_perm_indices::element_type): Update accordingly.
13784         (vec_perm_indices::clamp): Handle polynomial element_types.
13785         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13786         (vec_perm_indices::all_in_range_p): Likewise.
13787         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13788         than shwi trees.
13789         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13790         polynomial vec_perm_indices element types.
13791         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13792         * fold-const.c (fold_vec_perm): Likewise.
13793         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13794         * tree-vect-generic.c (lower_vec_perm): Likewise.
13795         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13796         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13797         element type to HOST_WIDE_INT.
13798
13799 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13800             Alan Hayward  <alan.hayward@arm.com>
13801             David Sherwood  <david.sherwood@arm.com>
13802
13803         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13804         rather than an int.  Use plus_constant.
13805         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13806         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13807
13808 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13809             Alan Hayward  <alan.hayward@arm.com>
13810             David Sherwood  <david.sherwood@arm.com>
13811
13812         * calls.c (emit_call_1, expand_call): Change struct_value_size from
13813         a HOST_WIDE_INT to a poly_int64.
13814
13815 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13816             Alan Hayward  <alan.hayward@arm.com>
13817             David Sherwood  <david.sherwood@arm.com>
13818
13819         * calls.c (load_register_parameters): Cope with polynomial
13820         mode sizes.  Require a constant size for BLKmode parameters
13821         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
13822         forces a parameter to be padded at the lsb end in order to
13823         fill a complete number of words, require the parameter size
13824         to be ordered wrt UNITS_PER_WORD.
13825
13826 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13827             Alan Hayward  <alan.hayward@arm.com>
13828             David Sherwood  <david.sherwood@arm.com>
13829
13830         * reload1.c (spill_stack_slot_width): Change element type
13831         from unsigned int to poly_uint64_pod.
13832         (alter_reg): Treat mode sizes as polynomial.
13833
13834 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13835             Alan Hayward  <alan.hayward@arm.com>
13836             David Sherwood  <david.sherwood@arm.com>
13837
13838         * reload.c (complex_word_subreg_p): New function.
13839         (reload_inner_reg_of_subreg, push_reload): Use it.
13840
13841 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13842             Alan Hayward  <alan.hayward@arm.com>
13843             David Sherwood  <david.sherwood@arm.com>
13844
13845         * lra-constraints.c (process_alt_operands): Reject matched
13846         operands whose sizes aren't ordered.
13847         (match_reload): Refer to this check here.
13848
13849 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13850             Alan Hayward  <alan.hayward@arm.com>
13851             David Sherwood  <david.sherwood@arm.com>
13852
13853         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13854         that the mode size is in the set {1, 2, 4, 8, 16}.
13855
13856 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13857             Alan Hayward  <alan.hayward@arm.com>
13858             David Sherwood  <david.sherwood@arm.com>
13859
13860         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13861         Use plus_constant instead of gen_rtx_PLUS.
13862
13863 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13864             Alan Hayward  <alan.hayward@arm.com>
13865             David Sherwood  <david.sherwood@arm.com>
13866
13867         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13868         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13869         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13870         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13871         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13872         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13873         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13874         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13875         * config/i386/i386.c (ix86_push_rounding): ...this new function.
13876         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13877         a poly_int64.
13878         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13879         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13880         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13881         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13882         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13883         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13884         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13885         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13886         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13887         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13888         function.
13889         * expr.c (emit_move_resolve_push): Treat the input and result
13890         of PUSH_ROUNDING as a poly_int64.
13891         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13892         (emit_push_insn): Likewise.
13893         * lra-eliminations.c (mark_not_eliminable): Likewise.
13894         * recog.c (push_operand): Likewise.
13895         * reload1.c (elimination_effects): Likewise.
13896         * rtlanal.c (nonzero_bits1): Likewise.
13897         * calls.c (store_one_arg): Likewise.  Require the padding to be
13898         known at compile time.
13899
13900 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13901             Alan Hayward  <alan.hayward@arm.com>
13902             David Sherwood  <david.sherwood@arm.com>
13903
13904         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13905         Use plus_constant instead of gen_rtx_PLUS.
13906
13907 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13908             Alan Hayward  <alan.hayward@arm.com>
13909             David Sherwood  <david.sherwood@arm.com>
13910
13911         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13912         rather than an int.
13913
13914 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13915             Alan Hayward  <alan.hayward@arm.com>
13916             David Sherwood  <david.sherwood@arm.com>
13917
13918         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13919         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13920         via stack temporaries.  Treat the mode size as polynomial too.
13921
13922 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13923             Alan Hayward  <alan.hayward@arm.com>
13924             David Sherwood  <david.sherwood@arm.com>
13925
13926         * expr.c (expand_expr_real_2): When handling conversions involving
13927         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13928         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
13929         as a poly_uint64 too.
13930
13931 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13932             Alan Hayward  <alan.hayward@arm.com>
13933             David Sherwood  <david.sherwood@arm.com>
13934
13935         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13936
13937 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13938             Alan Hayward  <alan.hayward@arm.com>
13939             David Sherwood  <david.sherwood@arm.com>
13940
13941         * combine.c (can_change_dest_mode): Handle polynomial
13942         REGMODE_NATURAL_SIZE.
13943         * expmed.c (store_bit_field_1): Likewise.
13944         * expr.c (store_constructor): Likewise.
13945         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13946         and polynomial REGMODE_NATURAL_SIZE.
13947         (gen_lowpart_common): Likewise.
13948         * reginfo.c (record_subregs_of_mode): Likewise.
13949         * rtlanal.c (read_modify_subreg_p): Likewise.
13950
13951 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13952             Alan Hayward  <alan.hayward@arm.com>
13953             David Sherwood  <david.sherwood@arm.com>
13954
13955         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13956         numbers of elements.
13957
13958 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13959             Alan Hayward  <alan.hayward@arm.com>
13960             David Sherwood  <david.sherwood@arm.com>
13961
13962         * match.pd: Cope with polynomial numbers of vector elements.
13963
13964 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13965             Alan Hayward  <alan.hayward@arm.com>
13966             David Sherwood  <david.sherwood@arm.com>
13967
13968         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13969         in a POINTER_PLUS_EXPR.
13970
13971 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13972             Alan Hayward  <alan.hayward@arm.com>
13973             David Sherwood  <david.sherwood@arm.com>
13974
13975         * omp-simd-clone.c (simd_clone_subparts): New function.
13976         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13977         (ipa_simd_modify_function_body): Likewise.
13978
13979 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13980             Alan Hayward  <alan.hayward@arm.com>
13981             David Sherwood  <david.sherwood@arm.com>
13982
13983         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13984         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13985         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13986         (expand_vector_condition, vector_element): Likewise.
13987         (subparts_gt): New function.
13988         (get_compute_type): Use subparts_gt.
13989         (count_type_subparts): Delete.
13990         (expand_vector_operations_1): Use subparts_gt instead of
13991         count_type_subparts.
13992
13993 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13994             Alan Hayward  <alan.hayward@arm.com>
13995             David Sherwood  <david.sherwood@arm.com>
13996
13997         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13998         (vect_compile_time_alias): ...this new function.  Do the calculation
13999         on poly_ints rather than trees.
14000         (vect_prune_runtime_alias_test_list): Update call accordingly.
14001
14002 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14003             Alan Hayward  <alan.hayward@arm.com>
14004             David Sherwood  <david.sherwood@arm.com>
14005
14006         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14007         numbers of units.
14008         (vect_schedule_slp_instance): Likewise.
14009
14010 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14011             Alan Hayward  <alan.hayward@arm.com>
14012             David Sherwood  <david.sherwood@arm.com>
14013
14014         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14015         constant and extern definitions for variable-length vectors.
14016         (vect_get_constant_vectors): Note that the number of units
14017         is known to be constant.
14018
14019 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14020             Alan Hayward  <alan.hayward@arm.com>
14021             David Sherwood  <david.sherwood@arm.com>
14022
14023         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14024         of units as polynomial.  Choose between WIDE and NARROW based
14025         on multiple_p.
14026
14027 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14028             Alan Hayward  <alan.hayward@arm.com>
14029             David Sherwood  <david.sherwood@arm.com>
14030
14031         * tree-vect-stmts.c (simd_clone_subparts): New function.
14032         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14033
14034 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14035             Alan Hayward  <alan.hayward@arm.com>
14036             David Sherwood  <david.sherwood@arm.com>
14037
14038         * tree-vect-stmts.c (vectorizable_call): Treat the number of
14039         vectors as polynomial.  Use build_index_vector for
14040         IFN_GOMP_SIMD_LANE.
14041
14042 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14043             Alan Hayward  <alan.hayward@arm.com>
14044             David Sherwood  <david.sherwood@arm.com>
14045
14046         * tree-vect-stmts.c (get_load_store_type): Treat the number of
14047         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14048         for variable-length vectors.
14049         (vectorizable_mask_load_store): Treat the number of units as
14050         polynomial, asserting that it is constant if the condition has
14051         already been enforced.
14052         (vectorizable_store, vectorizable_load): Likewise.
14053
14054 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14055             Alan Hayward  <alan.hayward@arm.com>
14056             David Sherwood  <david.sherwood@arm.com>
14057
14058         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14059         of units as polynomial.  Punt if we can't tell at compile time
14060         which vector contains the final result.
14061
14062 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14063             Alan Hayward  <alan.hayward@arm.com>
14064             David Sherwood  <david.sherwood@arm.com>
14065
14066         * tree-vect-loop.c (vectorizable_induction): Treat the number
14067         of units as polynomial.  Punt on SLP inductions.  Use an integer
14068         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
14069         cast of such a series for variable-length floating-point
14070         reductions.
14071
14072 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14073             Alan Hayward  <alan.hayward@arm.com>
14074             David Sherwood  <david.sherwood@arm.com>
14075
14076         * tree.h (build_index_vector): Declare.
14077         * tree.c (build_index_vector): New function.
14078         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14079         of units as polynomial, forcibly converting it to a constant if
14080         vectorizable_reduction has already enforced the condition.
14081         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
14082         to create a {1,2,3,...} vector.
14083         (vectorizable_reduction): Treat the number of units as polynomial.
14084         Choose vectype_in based on the largest scalar element size rather
14085         than the smallest number of units.  Enforce the restrictions
14086         relied on above.
14087
14088 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14089             Alan Hayward  <alan.hayward@arm.com>
14090             David Sherwood  <david.sherwood@arm.com>
14091
14092         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14093         number of units as polynomial.
14094
14095 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14096             Alan Hayward  <alan.hayward@arm.com>
14097             David Sherwood  <david.sherwood@arm.com>
14098
14099         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14100         * target.def (autovectorize_vector_sizes): Return the vector sizes
14101         by pointer, using vector_sizes rather than a bitmask.
14102         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14103         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14104         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14105         Likewise.
14106         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14107         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14108         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14109         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14110         * omp-general.c (omp_max_vf): Likewise.
14111         * omp-low.c (omp_clause_aligned_alignment): Likewise.
14112         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14113         * tree-vect-loop.c (vect_analyze_loop): Likewise.
14114         * tree-vect-slp.c (vect_slp_bb): Likewise.
14115         * doc/tm.texi: Regenerate.
14116         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14117         to a poly_uint64.
14118         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14119         the vector size as a poly_uint64 rather than an unsigned int.
14120         (current_vector_size): Change from an unsigned int to a poly_uint64.
14121         (get_vectype_for_scalar_type): Update accordingly.
14122         * tree.h (build_truth_vector_type): Take the size and number of
14123         units as a poly_uint64 rather than an unsigned int.
14124         (build_vector_type): Add a temporary overload that takes
14125         the number of units as a poly_uint64 rather than an unsigned int.
14126         * tree.c (make_vector_type): Likewise.
14127         (build_truth_vector_type): Take the number of units as a poly_uint64
14128         rather than an unsigned int.
14129
14130 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14131             Alan Hayward  <alan.hayward@arm.com>
14132             David Sherwood  <david.sherwood@arm.com>
14133
14134         * target.def (get_mask_mode): Take the number of units and length
14135         as poly_uint64s rather than unsigned ints.
14136         * targhooks.h (default_get_mask_mode): Update accordingly.
14137         * targhooks.c (default_get_mask_mode): Likewise.
14138         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14139         * doc/tm.texi: Regenerate.
14140
14141 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14142             Alan Hayward  <alan.hayward@arm.com>
14143             David Sherwood  <david.sherwood@arm.com>
14144
14145         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14146         * omp-general.c (omp_max_vf): Likewise.
14147         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14148         (expand_omp_simd): Handle polynomial safelen.
14149         * omp-low.c (omplow_simd_context): Add a default constructor.
14150         (omplow_simd_context::max_vf): Change from int to poly_uint64.
14151         (lower_rec_simd_input_clauses): Update accordingly.
14152         (lower_rec_input_clauses): Likewise.
14153
14154 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14155             Alan Hayward  <alan.hayward@arm.com>
14156             David Sherwood  <david.sherwood@arm.com>
14157
14158         * tree-vectorizer.h (vect_nunits_for_cost): New function.
14159         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14160         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14161         (vect_analyze_slp_cost): Likewise.
14162         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14163         (vect_model_load_cost): Likewise.
14164
14165 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14166             Alan Hayward  <alan.hayward@arm.com>
14167             David Sherwood  <david.sherwood@arm.com>
14168
14169         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14170         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14171         from an unsigned int * to a poly_uint64_pod *.
14172         (calculate_unrolling_factor): New function.
14173         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
14174
14175 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14176             Alan Hayward  <alan.hayward@arm.com>
14177             David Sherwood  <david.sherwood@arm.com>
14178
14179         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14180         from an unsigned int to a poly_uint64.
14181         (_loop_vec_info::slp_unrolling_factor): Likewise.
14182         (_loop_vec_info::vectorization_factor): Change from an int
14183         to a poly_uint64.
14184         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14185         (vect_get_num_vectors): New function.
14186         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14187         (vect_get_num_copies): Use vect_get_num_vectors.
14188         (vect_analyze_data_ref_dependences): Change max_vf from an int *
14189         to an unsigned int *.
14190         (vect_analyze_data_refs): Change min_vf from an int * to a
14191         poly_uint64 *.
14192         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14193         than an unsigned HOST_WIDE_INT.
14194         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14195         (vect_analyze_data_ref_dependence): Change max_vf from an int *
14196         to an unsigned int *.
14197         (vect_analyze_data_ref_dependences): Likewise.
14198         (vect_compute_data_ref_alignment): Handle polynomial vf.
14199         (vect_enhance_data_refs_alignment): Likewise.
14200         (vect_prune_runtime_alias_test_list): Likewise.
14201         (vect_shift_permute_load_chain): Likewise.
14202         (vect_supportable_dr_alignment): Likewise.
14203         (dependence_distance_ge_vf): Take the vectorization factor as a
14204         poly_uint64 rather than an unsigned HOST_WIDE_INT.
14205         (vect_analyze_data_refs): Change min_vf from an int * to a
14206         poly_uint64 *.
14207         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14208         vfm1 as a poly_uint64 rather than an int.  Make the same change
14209         for the returned bound_scalar.
14210         (vect_gen_vector_loop_niters): Handle polynomial vf.
14211         (vect_do_peeling): Likewise.  Update call to
14212         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14213         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14214         be constant.
14215         * tree-vect-loop.c (vect_determine_vectorization_factor)
14216         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14217         (vect_get_known_peeling_cost): Likewise.
14218         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14219         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14220         (vect_transform_loop): Likewise.  Use the lowest possible VF when
14221         updating the upper bounds of the loop.
14222         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
14223         rather than an int.
14224         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14225         polynomial unroll factors.
14226         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14227         (vect_make_slp_decision): Likewise.
14228         (vect_supported_load_permutation_p): Likewise, and polynomial
14229         vf too.
14230         (vect_analyze_slp_cost): Handle polynomial vf.
14231         (vect_slp_analyze_node_operations): Likewise.
14232         (vect_slp_analyze_bb_1): Likewise.
14233         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14234         than an unsigned HOST_WIDE_INT.
14235         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14236         (vectorizable_load): Handle polynomial vf.
14237         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14238         a poly_uint64.
14239         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14240
14241 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14242             Alan Hayward  <alan.hayward@arm.com>
14243             David Sherwood  <david.sherwood@arm.com>
14244
14245         * match.pd: Handle bit operations involving three constants
14246         and try to fold one pair.
14247
14248 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14249
14250         * tree-vect-loop-manip.c: Include gimple-fold.h.
14251         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14252         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
14253         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14254         Add a path that uses a step of VF instead of 1, but disable it
14255         for now.
14256         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14257         and niters_no_overflow parameters.  Update calls to
14258         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14259         Create a new SSA name if the latter choses to use a ste other
14260         than zero, and return it via niters_vector_mult_vf_var.
14261         * tree-vect-loop.c (vect_transform_loop): Update calls to
14262         vect_do_peeling, vect_gen_vector_loop_niters and
14263         slpeel_make_loop_iterate_ntimes.
14264         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14265         (vect_gen_vector_loop_niters): Update declarations after above changes.
14266
14267 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
14268
14269         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14270         128-bit round to integer instructions.
14271         (ceil<mode>2): Likewise.
14272         (btrunc<mode>2): Likewise.
14273         (round<mode>2): Likewise.
14274
14275 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14276
14277         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14278         unaligned VSX load/store on P8/P9.
14279         (expand_block_clear): Allow the use of unaligned VSX
14280         load/store on P8/P9.
14281
14282 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14283
14284         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14285         New function.
14286         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14287         swap associated with both a load and a store.
14288
14289 2018-01-02  Andrew Waterman  <andrew@sifive.com>
14290
14291         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14292         * config/riscv/riscv.md (clear_cache): Use it.
14293
14294 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
14295
14296         * web.c: Remove out-of-date comment.
14297
14298 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14299
14300         * expr.c (fixup_args_size_notes): Check that any existing
14301         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14302         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14303         (emit_single_push_insn): ...here.
14304
14305 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14306
14307         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14308         (const_vector_encoded_nelts): New function.
14309         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14310         (const_vector_int_elt, const_vector_elt): Declare.
14311         * emit-rtl.c (const_vector_int_elt_1): New function.
14312         (const_vector_elt): Likewise.
14313         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14314         of CONST_VECTOR_ELT.
14315
14316 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14317
14318         * expr.c: Include rtx-vector-builder.h.
14319         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14320         directly on the tree encoding.
14321         (const_vector_from_tree): Likewise.
14322         * optabs.c: Include rtx-vector-builder.h.
14323         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14324         sequence of "u" values.
14325         * vec-perm-indices.c: Include rtx-vector-builder.h.
14326         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14327         directly on the vec_perm_indices encoding.
14328
14329 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14330
14331         * doc/rtl.texi (const_vector): Describe new encoding scheme.
14332         * Makefile.in (OBJS): Add rtx-vector-builder.o.
14333         * rtx-vector-builder.h: New file.
14334         * rtx-vector-builder.c: Likewise.
14335         * rtl.h (rtx_def::u2): Add a const_vector field.
14336         (CONST_VECTOR_NPATTERNS): New macro.
14337         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14338         (CONST_VECTOR_DUPLICATE_P): Likewise.
14339         (CONST_VECTOR_STEPPED_P): Likewise.
14340         (CONST_VECTOR_ENCODED_ELT): Likewise.
14341         (const_vec_duplicate_p): Check for a duplicated vector encoding.
14342         (unwrap_const_vec_duplicate): Likewise.
14343         (const_vec_series_p): Check for a non-duplicated vector encoding.
14344         Say that the function only returns true for integer vectors.
14345         * emit-rtl.c: Include rtx-vector-builder.h.
14346         (gen_const_vec_duplicate_1): Delete.
14347         (gen_const_vector): Call gen_const_vec_duplicate instead of
14348         gen_const_vec_duplicate_1.
14349         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14350         (gen_const_vec_duplicate): Use rtx_vector_builder.
14351         (gen_const_vec_series): Likewise.
14352         (gen_rtx_CONST_VECTOR): Likewise.
14353         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14354         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14355         Build a new vector rather than modifying a CONST_VECTOR in-place.
14356         (handle_special_swappables): Update call accordingly.
14357         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14358         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14359         Build a new vector rather than modifying a CONST_VECTOR in-place.
14360         (handle_special_swappables): Update call accordingly.
14361
14362 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14363
14364         * simplify-rtx.c (simplify_const_binary_operation): Use
14365         CONST_VECTOR_ELT instead of XVECEXP.
14366
14367 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14368
14369         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14370         the selector elements to be different from the data elements
14371         if the selector is a VECTOR_CST.
14372         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14373         ssizetype for the selector.
14374
14375 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14376
14377         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14378         before testing each element individually.
14379         * tree-vect-generic.c (lower_vec_perm): Likewise.
14380
14381 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14382
14383         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14384         * selftest-run-tests.c (selftest::run_tests): Call it.
14385         * vector-builder.h (vector_builder::operator ==): New function.
14386         (vector_builder::operator !=): Likewise.
14387         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14388         (vec_perm_indices::all_from_input_p): New function.
14389         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14390         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14391         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14392         instead of reading the VECTOR_CST directly.  Detect whether both
14393         vector inputs are the same before constructing the vec_perm_indices,
14394         and update the number of inputs argument accordingly.  Use the
14395         utility functions added above.  Only construct sel2 if we need to.
14396
14397 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14398
14399         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14400         the broadcast of the low byte.
14401         (expand_mult_highpart): Use an explicit encoding for the permutes.
14402         * optabs-query.c (can_mult_highpart_p): Likewise.
14403         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14404         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14405         (vectorizable_bswap): Likewise.
14406         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14407         explicit encoding for the power-of-2 permutes.
14408         (vect_permute_store_chain): Likewise.
14409         (vect_grouped_load_supported): Likewise.
14410         (vect_permute_load_chain): Likewise.
14411
14412 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14413
14414         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14415         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14416         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14417         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14418         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14419         (vect_gen_perm_mask_any): Likewise.
14420
14421 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14422
14423         * int-vector-builder.h: New file.
14424         * vec-perm-indices.h: Include int-vector-builder.h.
14425         (vec_perm_indices): Redefine as an int_vector_builder.
14426         (auto_vec_perm_indices): Delete.
14427         (vec_perm_builder): Redefine as a stand-alone class.
14428         (vec_perm_indices::vec_perm_indices): New function.
14429         (vec_perm_indices::clamp): Likewise.
14430         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14431         (vec_perm_indices::new_vector): New function.
14432         (vec_perm_indices::new_expanded_vector): Update for new
14433         vec_perm_indices class.
14434         (vec_perm_indices::rotate_inputs): New function.
14435         (vec_perm_indices::all_in_range_p): Operate directly on the
14436         encoded form, without computing elided elements.
14437         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14438         encoding.  Update for new vec_perm_indices class.
14439         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14440         the given vec_perm_builder.
14441         (expand_vec_perm_var): Update vec_perm_builder constructor.
14442         (expand_mult_highpart): Use vec_perm_builder instead of
14443         auto_vec_perm_indices.
14444         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14445         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
14446         or double series encoding as appropriate.
14447         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14448         vec_perm_indices instead of auto_vec_perm_indices.
14449         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14450         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14451         (vect_permute_store_chain): Likewise.
14452         (vect_grouped_load_supported): Likewise.
14453         (vect_permute_load_chain): Likewise.
14454         (vect_shift_permute_load_chain): Likewise.
14455         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14456         (vect_transform_slp_perm_load): Likewise.
14457         (vect_schedule_slp_instance): Likewise.
14458         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14459         (vectorizable_mask_load_store): Likewise.
14460         (vectorizable_bswap): Likewise.
14461         (vectorizable_store): Likewise.
14462         (vectorizable_load): Likewise.
14463         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14464         vec_perm_indices instead of auto_vec_perm_indices.  Use
14465         tree_to_vec_perm_builder to read the vector from a tree.
14466         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14467         vec_perm_builder instead of a vec_perm_indices.
14468         (have_whole_vector_shift): Use vec_perm_builder and
14469         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
14470         truncation to calc_vec_perm_mask_for_shift.
14471         (vect_create_epilog_for_reduction): Likewise.
14472         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14473         from auto_vec_perm_indices to vec_perm_indices.
14474         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14475         instead of changing individual elements.
14476         (aarch64_vectorize_vec_perm_const): Use new_vector to install
14477         the vector in d.perm.
14478         * config/arm/arm.c (expand_vec_perm_d::perm): Change
14479         from auto_vec_perm_indices to vec_perm_indices.
14480         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14481         instead of changing individual elements.
14482         (arm_vectorize_vec_perm_const): Use new_vector to install
14483         the vector in d.perm.
14484         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14485         Update vec_perm_builder constructor.
14486         (rs6000_expand_interleave): Likewise.
14487         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14488         (rs6000_expand_interleave): Likewise.
14489
14490 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14491
14492         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14493         to qimode could truncate the indices.
14494         * optabs.c (expand_vec_perm_var): Likewise.
14495
14496 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14497
14498         * Makefile.in (OBJS): Add vec-perm-indices.o.
14499         * vec-perm-indices.h: New file.
14500         * vec-perm-indices.c: Likewise.
14501         * target.h (vec_perm_indices): Replace with a forward class
14502         declaration.
14503         (auto_vec_perm_indices): Move to vec-perm-indices.h.
14504         * optabs.h: Include vec-perm-indices.h.
14505         (expand_vec_perm): Delete.
14506         (selector_fits_mode_p, expand_vec_perm_var): Declare.
14507         (expand_vec_perm_const): Declare.
14508         * target.def (vec_perm_const_ok): Replace with...
14509         (vec_perm_const): ...this new hook.
14510         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14511         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14512         * doc/tm.texi: Regenerate.
14513         * optabs.def (vec_perm_const): Delete.
14514         * doc/md.texi (vec_perm_const): Likewise.
14515         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14516         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14517         expand_vec_perm for constant permutation vectors.  Assert that
14518         the mode of variable permutation vectors is the integer equivalent
14519         of the mode that is being permuted.
14520         * optabs-query.h (selector_fits_mode_p): Declare.
14521         * optabs-query.c: Include vec-perm-indices.h.
14522         (selector_fits_mode_p): New function.
14523         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14524         is defined, instead of checking whether the vec_perm_const_optab
14525         exists.  Use targetm.vectorize.vec_perm_const instead of
14526         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
14527         fit in the vector mode before using a variable permute.
14528         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14529         vec_perm_indices instead of an rtx.
14530         (expand_vec_perm): Replace with...
14531         (expand_vec_perm_const): ...this new function.  Take the selector
14532         as a vec_perm_indices rather than an rtx.  Also take the mode of
14533         the selector.  Update call to shift_amt_for_vec_perm_mask.
14534         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14535         Use vec_perm_indices::new_expanded_vector to expand the original
14536         selector into bytes.  Check whether the indices fit in the vector
14537         mode before using a variable permute.
14538         (expand_vec_perm_var): Make global.
14539         (expand_mult_highpart): Use expand_vec_perm_const.
14540         * fold-const.c: Includes vec-perm-indices.h.
14541         * tree-ssa-forwprop.c: Likewise.
14542         * tree-vect-data-refs.c: Likewise.
14543         * tree-vect-generic.c: Likewise.
14544         * tree-vect-loop.c: Likewise.
14545         * tree-vect-slp.c: Likewise.
14546         * tree-vect-stmts.c: Likewise.
14547         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14548         Delete.
14549         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14550         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14551         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14552         (aarch64_vectorize_vec_perm_const): ...this new function.
14553         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14554         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14555         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14556         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14557         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14558         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14559         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14560         into...
14561         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
14562         check for NEON modes.
14563         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14564         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14565         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14566         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14567         into...
14568         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
14569         the old VEC_PERM_CONST conditions.
14570         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14571         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14572         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14573         (ia64_vectorize_vec_perm_const_ok): Merge into...
14574         (ia64_vectorize_vec_perm_const): ...this new function.
14575         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14576         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14577         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14578         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14579         * config/mips/mips.c (mips_expand_vec_perm_const)
14580         (mips_vectorize_vec_perm_const_ok): Merge into...
14581         (mips_vectorize_vec_perm_const): ...this new function.
14582         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14583         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14584         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14585         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14586         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14587         (rs6000_expand_vec_perm_const): Delete.
14588         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14589         Delete.
14590         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14591         (altivec_expand_vec_perm_const_le): Take each operand individually.
14592         Operate on constant selectors rather than rtxes.
14593         (altivec_expand_vec_perm_const): Likewise.  Update call to
14594         altivec_expand_vec_perm_const_le.
14595         (rs6000_expand_vec_perm_const): Delete.
14596         (rs6000_vectorize_vec_perm_const_ok): Delete.
14597         (rs6000_vectorize_vec_perm_const): New function.
14598         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14599         an element count and rtx array.
14600         (rs6000_expand_extract_even): Update call accordingly.
14601         (rs6000_expand_interleave): Likewise.
14602         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14603         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14604         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14605         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14606         (rs6000_expand_vec_perm_const): Delete.
14607         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14608         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14609         (altivec_expand_vec_perm_const_le): Take each operand individually.
14610         Operate on constant selectors rather than rtxes.
14611         (altivec_expand_vec_perm_const): Likewise.  Update call to
14612         altivec_expand_vec_perm_const_le.
14613         (rs6000_expand_vec_perm_const): Delete.
14614         (rs6000_vectorize_vec_perm_const_ok): Delete.
14615         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
14616         reference to the SPE evmerge intructions.
14617         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14618         an element count and rtx array.
14619         (rs6000_expand_extract_even): Update call accordingly.
14620         (rs6000_expand_interleave): Likewise.
14621         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14622         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14623         new function.
14624         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14625
14626 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14627
14628         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14629         vector mode and that that mode matches the mode of the data
14630         being permuted.
14631         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14632         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
14633         directly using expand_vec_perm_1 when forcing selectors into
14634         registers.
14635         (expand_vec_perm_var): New function, split out from expand_vec_perm.
14636
14637 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14638
14639         * optabs-query.h (can_vec_perm_p): Delete.
14640         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14641         * optabs-query.c (can_vec_perm_p): Split into...
14642         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14643         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14644         particular selector is valid.
14645         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14646         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14647         (vect_grouped_load_supported): Likewise.
14648         (vect_shift_permute_load_chain): Likewise.
14649         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14650         (vect_transform_slp_perm_load): Likewise.
14651         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14652         (vectorizable_bswap): Likewise.
14653         (vect_gen_perm_mask_checked): Likewise.
14654         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
14655         implementations of variable permutation vectors into account
14656         when deciding which selector to use.
14657         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14658         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14659         with a false third argument.
14660         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14661         to test whether the constant selector is valid and can_vec_perm_var_p
14662         to test whether a variable selector is valid.
14663
14664 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14665
14666         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14667         * optabs-query.c (can_vec_perm_p): Likewise.
14668         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14669         instead of vec_perm_indices.
14670         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14671         (vect_gen_perm_mask_checked): Likewise,
14672         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14673         (vect_gen_perm_mask_checked): Likewise,
14674
14675 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14676
14677         * optabs-query.h (qimode_for_vec_perm): Declare.
14678         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14679         (qimode_for_vec_perm): ...this new function.
14680         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14681
14682 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14683
14684         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14685         does not have a conditional at the top.
14686
14687 2018-01-02  Richard Biener  <rguenther@suse.de>
14688
14689         * ipa-inline.c (big_speedup_p): Fix expression.
14690
14691 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
14692
14693         PR target/81616
14694         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14695         for generic 4->6.
14696
14697 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
14698
14699         PR target/81616
14700         Generic tuning.
14701         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14702         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14703         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14704         cond_taken_branch_cost 3->4.
14705
14706 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
14707
14708         PR tree-optimization/83581
14709         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14710         TODO_cleanup_cfg if any changes have been made.
14711
14712         PR middle-end/83608
14713         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14714         convert_modes if target mode has the right side, but different mode
14715         class.
14716
14717         PR middle-end/83609
14718         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14719         last argument when extracting from CONCAT.  If either from_real or
14720         from_imag is NULL, use expansion through memory.  If result is not
14721         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14722         the parts directly to inner mode, if even that fails, use expansion
14723         through memory.
14724
14725         PR middle-end/83623
14726         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14727         check for bswap in mode rather than HImode and use that in expand_unop
14728         too.
14729 \f
14730 Copyright (C) 2018 Free Software Foundation, Inc.
14731
14732 Copying and distribution of this file, with or without modification,
14733 are permitted in any medium without royalty provided the copyright
14734 notice and this notice are preserved.