gcc/ChangeLog

   1 2018-04-18  Martin Liska  <mliska@suse.cz>
   2
   3         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
   4         style.
   5
   6 2018-04-18  Martin Liska  <mliska@suse.cz>
   7
   8         Revert
   9         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
  10
  11         PR ipa/83983
  12         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
  13         arguments if they are comparable.
  14
  15 2018-04-18  Martin Liska  <mliska@suse.cz>
  16
  17         Revert
  18         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
  19
  20         PR lto/84805
  21         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
  22         incomplete types.
  23
  24 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
  25
  26         PR target/85388
  27         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
  28         ENDBR after calling __morestack.
  29
  30 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
  31
  32         PR jit/85384
  33         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
  34         by using gcc_base_ver to generate a gcc_driver_version, and use
  35         it when generating GCC_DRIVER_NAME.
  36         * configure: Regenerate.
  37
  38 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
  39
  40         PR target/81084
  41         * config.gcc: Obsolete powerpc*-*-*spe*.
  42
  43 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
  44
  45         PR debug/84637
  46         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
  47         (stabstr_D): Change type of unum from unsigned int to
  48         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
  49         type.
  50
  51 2018-04-17  Jim Wilson  <jimw@sifive.com>
  52
  53         PR 84856
  54         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
  55         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
  56         Set arg_pointer_offset after using pretend_args_size.
  57
  58 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
  59
  60         PR rtl-optimization/85431
  61         * dse.c (record_store): Ignore zero width stores.
  62
  63         PR sanitizer/85230
  64         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
  65         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
  66         __builtin_stack_restore rather than after it.
  67         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
  68         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
  69         argument instead of virtual_dynamic_stack_rtx.
  70
  71 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
  72
  73         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
  74         New prototype.
  75         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
  76         Add note to error message to explain internal mapping of overloaded
  77         built-in function name to non-overloaded built-in function name.
  78         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
  79         function.
  80
  81 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
  82
  83         PR target/85424
  84         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
  85         where the inputs overlap with the output.
  86
  87 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
  88
  89         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
  90         (=v, v) alternative and explicit "memory" attribute.
  91         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
  92         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
  93         attributes.
  94         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
  95         "sselog1" type instead of "sselog".
  96         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
  97         "sselog".  Remove explicit "memory" attribute.
  98         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
  99         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
 100         attributes.
 101         (vec_extract_hi_v32hi): Merge all alternatives into one, use
 102         "sselog1" type instead of "sselog".  Remove explicit "memory"
 103         attribute.
 104         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
 105         use "sselog1" type instead of "sselog".  Remove explicit "memory"
 106         attribute.
 107         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
 108         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
 109         attributes.
 110         (vec_extract_hi_v64qi): Merge all alternatives into one, use
 111         "sselog1" type instead of "sselog".  Remove explicit "memory"
 112         attribute.
 113         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
 114         use "sselog1" type instead of "sselog".  Remove explicit "memory"
 115         attribute.
 116
 117         PR target/85430
 118         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
 119
 120         PR middle-end/85414
 121         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
 122         on a SUBREG.
 123
 124 2018-04-17  Martin Jambor  <mjambor@suse.cz>
 125
 126         PR ipa/85421
 127         * ipa-cp.c (create_specialized_node): Call
 128         expand_all_artificial_thunks if necessary.
 129
 130 2018-04-17  Martin Liska  <mliska@suse.cz>
 131
 132         PR lto/85405
 133         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
 134         in message, remote space in between '_G' and '('.
 135
 136 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
 137
 138         PR target/85281
 139         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
 140         avx512f_vmcmp<mode>3<round_saeonly_name>,
 141         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
 142         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
 143         avx512f_rndscale<mode><round_saeonly_name>,
 144         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
 145         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
 146         Use %<iptr>2 instead of %2 for -masm=intel.
 147         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
 148         avx512f_vcvttss2usi<round_saeonly_name>,
 149         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
 150         -masm=intel.
 151         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
 152         avx512f_vcvttsd2usi<round_saeonly_name>,
 153         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
 154         Use %q1 instead of %1 for -masm=intel.
 155         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
 156         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
 157         of %3 for -masm=intel.
 158         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
 159         -masm=intel.
 160         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
 161         -masm=intel.
 162         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
 163         -masm=intel.
 164         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
 165         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
 166         %g1.
 167         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
 168         -masm=intel.
 169         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
 170         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
 171         %g1 and one with %0 and %1.
 172         (avx512er_vmrcp28<mode><round_saeonly_name>,
 173         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
 174         %1 for -masm=intel.
 175         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
 176         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
 177         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
 178         of %0 and %{%4%} for -masm=intel.
 179         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
 180         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
 181         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
 182         order of %0 and %{%5%}%{z%} for -masm=intel.
 183
 184 2018-04-17  Jan Hubicka  <jh@suse.cz>
 185
 186         PR lto/85405
 187         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
 188
 189 2018-04-17  Martin Liska  <mliska@suse.cz>
 190
 191         PR ipa/85329
 192         * multiple_target.c (create_dispatcher_calls): Set apostrophes
 193         for target_clone error message.  Make default implementation
 194         clone to be a local declaration.
 195         (separate_attrs): Add new argument and check for an empty
 196         string.
 197         (expand_target_clones): Handle it.
 198         (ipa_target_clone): Make redirection just for target_clones
 199         functions.
 200
 201 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
 202             Tom de Vries  <tom@codesourcery.com>
 203
 204         PR middle-end/84955
 205         * omp-expand.c (expand_oacc_for): Add dummy false branch for
 206         tiled basic blocks without omp continue statements.
 207
 208 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
 209
 210         PR target/83660
 211         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
 212         vec_extract expression as having side effects to make sure it gets
 213         a cleanup point.
 214
 215 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
 216
 217         PR target/85403
 218         * config/i386/i386.c (get_builtin_code_for_version): Check
 219         error_mark_node.
 220
 221 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
 222
 223         PR target/84331
 224         * gcc/config.gcc: Support "skylake".
 225         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
 226         PROCESSOR_SKYLAKE.
 227         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
 228         (processor_target_table): Add "skylake".
 229         (ix86_option_override_internal): Add "skylake".
 230         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
 231         PROCESSOR_CANNONLAKE.
 232         (get_builtin_code_for_version): Fix priority for
 233         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
 234         PROCESSOR_SKYLAKE-AVX512.
 235         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
 236         (processor_type): Add PROCESSOR_SKYLAKE.
 237
 238 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
 239             Jason Merrill  <jason@redhat.com>
 240
 241         PR c++/85112
 242         * convert.c (convert_to_integer_1): Use direct recursion for
 243         enumeral types and types with a precision less than the number
 244         of bits in their mode.
 245
 246 2018-04-16  Julia Koval  <julia.koval@intel.com>
 247
 248         PR target/84413
 249         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
 250         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
 251
 252 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
 253
 254         PR target/85293
 255         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
 256         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
 257         and -mno-direct-move.
 258
 259 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
 260
 261         PR target/83402
 262         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
 263         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
 264         Ensure negative shifts result in {0}.
 265
 266 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
 267
 268         PR rtl-optimization/79916
 269         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
 270         regs (if any) to define how to gnerate SD moves when LRA is in
 271         progress.
 272
 273 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
 274
 275         PR rtl-optimization/85393
 276         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
 277         * except.c (expand_dw2_landing_pad_for_region): Make static.
 278         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
 279         a label and unconditional jump to old_bb, rather than
 280         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
 281         basic block.
 282
 283         PR rtl-optimization/85376
 284         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
 285         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
 286         instead of a specific value.
 287
 288 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
 289             Bin Cheng  <bin.cheng@arm.com>
 290
 291         PR tree-optimization/82965
 292         PR tree-optimization/83991
 293         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
 294         by_profile_only parameter.
 295         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
 296         information if the loop was predicted to iterate too many times.
 297         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
 298
 299 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
 300
 301         PR lto/71991
 302         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
 303         always inline.
 304
 305 2018-04-13  Martin Liska  <mliska@suse.cz>
 306             Jakub Jelinek  <jakub@redhat.com>
 307
 308         PR middle-end/81657
 309         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
 310         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
 311         * builtins.c (expand_builtin_memory_copy_args): Use
 312         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
 313         handle dest_addr == pc_rtx.
 314
 315 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
 316
 317         PR target/85291
 318         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
 319         asked to not generate direct moves.
 320         (fix_trunc<mode>si2_stfiwx): Similar.
 321         (fix_trunc<mode>si2_internal): Similar.
 322
 323 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
 324
 325         PR debug/83157
 326         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
 327         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
 328         lookup if dest in some wider mode is known to be const0_rtx and
 329         if so, record permanent equivalence for it to be ZERO_EXTEND of
 330         the narrower mode destination.
 331
 332 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
 333
 334         * lto-streamer-out.c (output_function): Revert 259346.
 335         * omp-expand.c (expand_oacc_for): Likewise.
 336
 337 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
 338
 339         PR rtl-optimization/85354
 340         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
 341         * sel-sched.c (sel_global_init): ... here.
 342
 343 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
 344
 345         PR target/85238
 346         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
 347         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
 348         mode for PE-COFF targets.
 349         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
 350         (i386_pe_asm_lto_end): Likewise.
 351         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
 352         (TARGET_ASM_LTO_END): Likewise.
 353         * config/i386/winnt.c (saved_debug_info_level): New static variable.
 354         (i386_pe_asm_lto_start): New function.
 355         (i386_pe_asm_lto_end): Likewise.
 356
 357 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
 358             Richard Biener  <rguenther@suse.de>
 359
 360         PR middle-end/84955
 361         * lto-streamer-out.c (output_function): Fix CFG loop state before
 362         streaming out.
 363         * omp-expand.c (expand_oacc_for): Handle calls to internal
 364         functions like regular functions.
 365
 366 2018-04-12  Richard Biener  <rguenther@suse.de>
 367
 368         PR lto/85371
 369         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
 370         for the early LTO debug to properly generate references to it
 371         during DIE emission.  Do not re-use that for the skeleton for
 372         split-dwarf.
 373         (dwarf2out_early_finish): Likewise.
 374
 375 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
 376
 377         PR target/85328
 378         * config/i386/sse.md
 379         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
 380         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
 381         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
 382         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
 383         and output is a reg, avoid creating invalid lowpart subreg, but
 384         instead split into a 512-bit move.  Don't split if not AVX512VL,
 385         input is xmm16+ reg and output is a mem.
 386         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
 387         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
 388         xmm16+ reg and output is a mem.
 389
 390 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
 391
 392         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
 393         also for flag_dwarf2_cfi_asm.
 394
 395 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
 396
 397         PR rtl-optimization/85342
 398         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
 399         a bool scalar var inside of the loop instead.  Don't try to update
 400         recog_data.operand after failed apply_change_group.
 401
 402 2018-04-12  Tom de Vries  <tom@codesourcery.com>
 403
 404         PR target/85296
 405         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
 406         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
 407         array with flexible array member as array without given dimension.
 408         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
 409         argument for undefined param to true.
 410
 411 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
 412
 413         PR target/85321
 414         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
 415         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
 416         from PowerPC section.
 417         * config/rs6000/sysv4.opt (mcall-): Improve help text.
 418         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
 419         help text that is too long.
 420         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
 421         help text that is too long.
 422         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
 423         help text that is too long.
 424
 425 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
 426
 427         * config/alpha/alpha.md (stack_probe_internal): Rename
 428         from "probe_stack".  Update all callers.
 429
 430 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
 431
 432         PR rtl-optimization/84566
 433         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
 434         sched_macro_fuse_insns.
 435
 436 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
 437
 438         PR target/84301
 439         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
 440         (compute_block_dependences): ... from here.
 441
 442 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
 443
 444         PR tree-optimization/85331
 445         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
 446         from int to HOST_WIDE_INT.
 447
 448 2018-04-11  Martin Jambor  <mjambor@suse.cz>
 449
 450         PR ipa/84149
 451         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
 452         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
 453         not the same as the source val.
 454         (cgraph_edge_brings_value_p): New parameter.
 455         (gather_edges_for_value): Pass destination value to
 456         cgraph_edge_brings_value_p.
 457         (perhaps_add_new_callers): Likewise.
 458         (get_info_about_necessary_edges): Likewise and exclude values brought
 459         only by self-recursive edges.
 460         (create_specialized_node): Redirect only clones of self-calling edges.
 461         (+self_recursive_pass_through_p): New function.
 462         (find_more_scalar_values_for_callers_subset): Use it.
 463         (find_aggregate_values_for_callers_subset): Likewise.
 464         (known_aggs_to_agg_replacement_list): Removed.
 465         (decide_whether_version_node): Re-calculate known constants for all
 466         remaining context clones.
 467
 468 2018-04-11  Richard Biener  <rguenther@suse.de>
 469
 470         PR lto/85339
 471         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
 472         from early DWARF output.
 473         (dwarf2out_early_finish): Output line info unconditionally into
 474         early DWARF and add reference to it.
 475
 476 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
 477
 478         PR target/85281
 479         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
 480         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
 481         other than V2DFmode using iptr mode attribute.
 482         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
 483
 484 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
 485
 486         PR rtl-optimization/84659
 487         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
 488
 489 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
 490
 491         PR debug/85302
 492         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
 493         SIZEP is NULL.
 494         (output_loc_list): Pass address of a dummy size variable even in the
 495         locview handling loop.
 496         (index_location_lists): Add comment on why skip_loc_list_entry can't
 497         call size_of_locs.
 498
 499 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
 500
 501         PR target/85261
 502         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
 503         into register.
 504
 505 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
 506
 507         PR target/85321
 508         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
 509         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
 510         and -mstring-compare-inline-limit.
 511
 512 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
 513
 514         PR target/85287
 515         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
 516         for stack clash protection in a register whenever we need it to be in
 517         a register.
 518
 519 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
 520
 521         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
 522         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
 523
 524 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
 525
 526         PR target/85321
 527         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
 528         the help text.
 529         (mlong-double-): Ditto.
 530         * config/rs6000/sysv4.opt (msdata=): Ditto.
 531         (mtls-size=): Ditto.
 532
 533 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 534
 535         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
 536         erroneous entries for
 537         "vector int vec_ldl (int, long int *)", and
 538         "vector unsigned int vec_ldl (int, unsigned long int *)".
 539         Add comments and entries for
 540         "vector bool char vec_ldl (int, bool char *)",
 541         "vector bool short vec_ldl (int, bool short *)",
 542         "vector bool int vec_ldl (int, bool int *)",
 543         "vector bool long long vec_ldl (int, bool long long *)",
 544         "vector pixel vec_ldl (int, pixel *)",
 545         "vector long long vec_ldl (int, long long *)",
 546         "vector unsigned long long vec_ldl (int, unsigned long long *)".
 547         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
 548         type tree bool_long_long_type_node and correct definition of
 549         bool_V2DI_type_node to make reference to this new type tree.
 550         (rs6000_mangle_type): Replace erroneous reference to
 551         bool_long_type_node with bool_long_long_type_node.
 552         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
 553         comments to emphasize sign distinctions for char and int types and
 554         replace RS6000_BTI_bool_long constant with
 555         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
 556         use of RS6000_BTI_pixel.
 557         (bool_long_type_node): Remove this macro definition.
 558         (bool_long_long_type_node): New macro definition
 559
 560 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
 561
 562         PR rtl-optimization/85300
 563         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
 564         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
 565         simplify_unary_operation fails.
 566
 567 2018-04-10  Martin Liska  <mliska@suse.cz>
 568
 569         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
 570         cgraph_edge and ipa_ref.
 571
 572 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
 573
 574         PR target/85177
 575         PR target/85255
 576         * config/i386/sse.md
 577         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
 578         computation of the VEC_MERGE selector from mask.
 579         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
 580         Fix decoding of the VEC_MERGE selector into mask.
 581
 582 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
 583
 584         PR tree-optimization/85286
 585         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
 586
 587 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
 588
 589         * final.c (final_1): Set insn_last_address as well as
 590         insn_current_address.
 591
 592 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 593
 594         PR target/85173
 595         * explow.c (emit_stack_probe): Call validize_mem on memory location
 596         before passing it to gen_probe_stack.  Create address operand and
 597         legitimize it for the probe_stack_address case.
 598
 599 2018-04-09  Jan Hubicka  <jh@suse.cz>
 600
 601         PR lto/85078
 602         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
 603         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
 604         * tree.c (free_lang_data_in_type): Fix handling of binfos;
 605         walk basetypes.
 606         (free_lang_data): Rebuild type inheritance graph.
 607
 608 2018-04-09  Martin Sebor  <msebor@redhat.com>
 609
 610         * invoke.texi (-finline-small-functions): Mention other optimization
 611         options.
 612         (-findirect-inlining, -fpartial-inlining): Same.
 613         (-finline-functions-called-once): Same.
 614         (-freorder-blocks-and-partition): Same.
 615
 616 2018-04-09  Jan Hubicka  <jh@suse.cz>
 617
 618         PR rtl/84058
 619         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
 620         jumps; choose last target that matches the criteria (i.e.
 621         no partition changes for non-crossing jumps).
 622         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
 623         support for redirecting crossing jumps to non-crossing.
 624
 625 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
 626
 627         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
 628         also for naked functions.
 629
 630 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
 631
 632         * config/arc/arc.md (add_shift): New pattern.
 633         (add_shift2): Likewise.
 634         (sub_shift): Likewise.
 635         (sub_shift_cmp0_noout): Likewise.
 636         (compare_si_ashiftsi): Likewise.
 637         (xbfu_cmp0_noout): New combine pattern.
 638         (xbfu_cmp0"): Likewise.
 639         (movsi_set_cc_insn): Place the predicable variant first.
 640         (commutative_binary_cmp0_noout): Remove clobber.
 641         (commutative_binary_cmp0): New pattern.
 642         (noncommutative_binary_cmp0): Likewise.
 643         (noncommutative_binary_cmp0_noout): Likewise.
 644         (noncommutative_binary_comparison_result_used): Removed.
 645         (rsub_cmp0): New pattern.
 646         (rsub_cmp0_noout): Likewise.
 647         (extzvsi): Changed, keep only meaningful variants.
 648         (SQH, SEZ): New iterators.
 649         (SQH_postfix): New mode attribute.
 650         (SEZ_prefix): New code attribute.
 651         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
 652         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
 653         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
 654         of numerical value.
 655         (noncommutative_operator): Check the availability of barrel
 656         shifter option.
 657
 658 2018-04-09  Richard Biener  <rguenther@suse.de>
 659
 660         PR tree-optimization/85284
 661         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
 662         Only use the niter constraining form of simple_iv when the exit
 663         is always executed.
 664
 665 2018-04-09  Tom de Vries  <tom@codesourcery.com>
 666
 667         PR target/84041
 668         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
 669         (define_expand "*memory_barrier"): New define_expand.
 670         (define_insn "memory_barrier"): New insn.
 671
 672 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
 673
 674         PR rtl-optimization/80463
 675         PR rtl-optimization/83972
 676         PR rtl-optimization/83480
 677
 678         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
 679         correct producer for the insn.
 680         (tidy_control_flow): Fixup seqnos in case of debug insns.
 681
 682 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
 683
 684         PR rtl-optimization/83913
 685
 686         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
 687         different sched-times when merging exprs.
 688
 689 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
 690
 691         PR rtl-optimization/83962
 692
 693         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
 694         tidy_fallthru_edge and tidy_control_flow.
 695
 696 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
 697
 698         PR rtl-optimization/83530
 699
 700         * sel-sched.c (force_next_insn): New global variable.
 701         (remove_insn_for_debug): When force_next_insn is true, also leave only
 702         next insn in the ready list.
 703         (sel_sched_region): When the region wasn't scheduled, make another pass
 704         over it with force_next_insn set to 1.
 705
 706 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
 707
 708         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
 709         into tm_file.
 710         * config/nds32/constants.md (unspec_volatile_element): Add enum values
 711         for interrupt control.
 712         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
 713         functions for interrupt control.
 714         * config/nds32/nds32-intrinsic.md: Likewise.
 715         * config/nds32/nds32_intrinsic.h: Likewise.
 716         * config/nds32/nds32.h (nds32_builtins): Likewise.
 717
 718 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
 719
 720         * config/nds32/nds32.c (nds32_init_machine_status,
 721         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
 722         strict_aligned_p field.
 723         (nds32_expand_to_rtl_hook): New function.
 724         (TARGET_EXPAND_TO_RTL_HOOK): Define.
 725         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
 726
 727 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
 728             Chung-Ju Wu  <jasonwucj@gmail.com>
 729
 730         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
 731         * config/nds32/nds32-n7.md: New file.
 732         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
 733         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
 734         pipeline.
 735         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
 736         * config/nds32/nds32.md (pipeline_model): Add n7.
 737         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
 738         * config/nds32/pipelines.md: Include n7 settings.
 739
 740 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
 741             Chung-Ju Wu  <jasonwucj@gmail.com>
 742
 743         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
 744         * config/nds32/nds32-e8.md: New file.
 745         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
 746         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
 747         pipeline.
 748         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
 749         * config/nds32/nds32.md (pipeline_model): Add e8.
 750         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
 751         * config/nds32/pipelines.md: Include e8 settings.
 752
 753 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
 754             Chung-Ju Wu  <jasonwucj@gmail.com>
 755
 756         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
 757         * config/nds32/nds32-n8.md: New file.
 758         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
 759         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
 760         pipeline.
 761         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
 762         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
 763         * config/nds32/nds32.md (pipeline_model): Add n8.
 764         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
 765         * config/nds32/pipelines.md: Include n8 settings.
 766
 767 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
 768             Chung-Ju Wu  <jasonwucj@gmail.com>
 769
 770         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
 771         * config/nds32/nds32-n9-2r1w.md: New file.
 772         * config/nds32/nds32-n9-3r2w.md: New file.
 773         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
 774         nds32_register_ports): New or modify for cpu n9.
 775         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
 776         pipeline.
 777         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
 778         * config/nds32/nds32-utils.c: New file.
 779         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
 780         TARGET_MUL_SLOW): Define.
 781         * config/nds32/nds32.md (pipeline_model): New attribute.
 782         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
 783         New options that support cpu n9.
 784         * config/nds32/pipelines.md: Include n9 settings.
 785         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
 786
 787 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
 788
 789         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
 790         information if necessary.
 791         (output_cond_branch_compare_zero): Likewise.
 792         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
 793         (nds32_target_alignment): Refine for alignment.
 794         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
 795         (FUNCTION_BOUNDARY): Modify.
 796         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
 797         align case.
 798         * config/nds32/nds32.opt (malways-align, malign-functions): New.
 799
 800 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
 801
 802         * config/nds32/constants.md (unspec_volatile_element): Add values for
 803         TLB operation and data prefetch.
 804         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
 805         functions for TLB operation and data prefetch.
 806         * config/nds32/nds32-intrinsic.md: Likewise.
 807         * config/nds32/nds32_intrinsic.h: Likewise.
 808         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
 809         (nds32_print_operand): Likewise.
 810         * config/nds32/nds32.h (nds32_builtins): Likewise.
 811
 812 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
 813         Andrew Pinski <pinsika@gcc.gnu.org>
 814
 815         PR middle-end/82976
 816         * match.pd: Use constant_boolean_node of correct type instead of
 817         boolean_true_node or boolean_false_node for simplifying
 818         pointer comparisons to zero.
 819
 820 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
 821
 822         PR tree-optimization/80021
 823         * tree.c (verify_type_variant): Make error call in verify_variant_match
 824         translatable and remove final full stop.
 825
 826 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
 827
 828         * config/nds32/constants.md (unspec_volatile_element): Add
 829         UNSPEC_VOLATILE_EH_RETURN.
 830         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
 831         nds32_output_stack_pop): Support dwarf exception handling process.
 832         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
 833         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
 834         exception handling process.
 835         (nds32_compute_stack_frame): Likewise.
 836         (nds32_return_addr_rtx): Likewise.
 837         (nds32_initial_elimination_offset): Likewise.
 838         (nds32_expand_prologue): Likewise.
 839         (nds32_expand_epilogue): Likewise.
 840         (nds32_dynamic_chain_address): New function.
 841         * config/nds32/nds32.h (machine_function): Add fields for dwarf
 842         exception handling.
 843         (DYNAMIC_CHAIN_ADDRESS): Define.
 844         (EH_RETURN_DATA_REGNO): Define.
 845         (EH_RETURN_STACKADJ_RTX): Define.
 846         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
 847         patterns for dwarf exception handling.
 848
 849 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
 850
 851         * config/nds32/nds32.h: Clean up obsolete macros.
 852
 853 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
 854
 855         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
 856         Add enum values for particular instructions.
 857         * config/nds32/nds32-intrinsic.c: Implementation of expanding
 858         particular intrinsic functions.
 859         * config/nds32/nds32-intrinsic.md: Likewise.
 860         * config/nds32/nds32_intrinsic.h: Likewise.
 861         * config/nds32/nds32.h (nds32_builtins): Likewise.
 862         * config/nds32/nds32.md (type): Add pbsad and pbsada.
 863         (btst, ave): New patterns for particular instructions.
 864
 865 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
 866
 867         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
 868         Add enum values for atomic load/store and memory sync.
 869         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
 870         and memory sync.
 871         * config/nds32/nds32-intrinsic.md: Likewise.
 872         * config/nds32/nds32_intrinsic.h: Likewise.
 873         * config/nds32/nds32.h (nds32_builtins): Likewise.
 874
 875 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
 876
 877         PR tree-optimization/85257
 878         * fold-const.c (native_encode_vector): If not all elts could fit
 879         and off is -1, return 0 rather than offset.
 880         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
 881         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
 882         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
 883         adjust buffer in native_interpret_expr call.
 884
 885 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
 886
 887         * config/nds32/constants.md (unspec_volatile_element): Add cache
 888         control enum values.
 889         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
 890         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
 891         * config/nds32/nds32.c (nds32_cctl_names): New.
 892         (nds32_print_operand): Handle cache control register names.
 893         * config/nds32/nds32.h (nds32_builtins): New enum values.
 894         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
 895         macros.
 896         * config/nds32/nds32.md (type): Add mmu.
 897         * config/nds32/pipelines.md (simple_insn): Add mmu.
 898
 899 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
 900
 901         * config/nds32/nds32.md (type): Remove call.
 902         * config/nds32/pipelines.md (simple_insn): Likewise.
 903
 904 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
 905
 906         * config/nds32/constants.md (unspec_volatile_element): Add
 907         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
 908         UNSPEC_VOLATILE_FMFCFG.
 909         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
 910         description for fmfcfg and fmfcsr.
 911         (bdesc_1arg): Add fmtcsr.
 912         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
 913         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
 914         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
 915         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
 916         unspec_fmfcfg): New patterns.
 917         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
 918         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
 919         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
 920         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
 921         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
 922         __nds32__fmfcfg): Define.
 923
 924 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
 925
 926         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
 927         intrinsic register names.
 928         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
 929         intrinsic register enum values and macros.
 930
 931 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
 932
 933         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
 934         for load/store addressing form.
 935         (nds32_print_operand_address): Likewise.
 936
 937 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
 938
 939         PR target/85196
 940         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
 941         based on LABEL_REF.  Remove useless assertion.
 942         (pic_address_needs_scratch): Fix formatting.
 943         (sparc_legitimize_pic_address): Minor tweaks.
 944         (sparc_delegitimize_address): Adjust assertion accordingly.
 945         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
 946         into symbolic_operand.
 947         (movsi_high_pic_label_ref): Likewise.
 948         (movsi_lo_sum_pic_label_ref): Likewise.
 949         (movdi_pic_label_ref): Likewise.
 950         (movdi_high_pic_label_ref): Likewise.
 951         (movdi_lo_sum_pic_label_ref): Likewise.
 952
 953 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
 954
 955         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
 956         custom LIB_SPEC setup.
 957
 958 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
 959             Kito Cheng  <kito.cheng@gmail.com>
 960
 961         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
 962         * config/riscv/freebsd.h: New.
 963
 964 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
 965
 966         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
 967         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
 968         file.
 969
 970 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
 971             Kito Cheng  <kito.cheng@gmail.com>
 972
 973         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
 974         nds32_output_call, nds32_symbol_binds_local_p): New functions.
 975         * config/nds32/nds32-protos.h (nds32_output_call,
 976         nds32_output_return): Declare.
 977         * config/nds32/nds32.md: Refine all the call and return patterns.
 978
 979 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
 980
 981         PR debug/85252
 982         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
 983         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
 984
 985         PR rtl-optimization/84872
 986         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
 987         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
 988         EDGE_CROSSING edge.
 989
 990 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
 991
 992         * expr.c (copy_blkmode_to_reg): Revert 254862.
 993         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
 994
 995 2018-04-06  Richard Biener  <rguenther@suse.de>
 996
 997         PR middle-end/85244
 998         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
 999         after seeing a component reference with an adjacent field.  Treat
1000         refs to arrays at struct end of external decls similar to
1001         refs to unconstrained commons.
1002
1003 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
1004
1005         PR sanitizer/85213
1006         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1007         look through SAVE_EXPRs with non-side-effects argument.  Adjust
1008         recursive calls.
1009         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1010         save_p here.
1011
1012 2018-04-06  Richard Biener  <rguenther@suse.de>
1013
1014         PR middle-end/85180
1015         * alias.c (find_base_term): New wrapper around find_base_term
1016         unwinding CSELIB_VAL_PTR changes.
1017         (find_base_term): Do not restore CSELIB_VAL_PTR during the
1018         recursion.
1019
1020 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
1021
1022         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1023         instructions.
1024         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1025         constant definitions.
1026         ("nop"): lr 0,0 -> nopr r0
1027         ("nop_lr0", "nop_lr1"): New insn definitions.
1028
1029 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
1030
1031         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1032         NDS32_V3PUSH_AVAILABLE_P macro.
1033
1034 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
1035             Chung-Ju Wu  <jasonwucj@gmail.com>
1036
1037         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1038         (nds32*-*-*): Add float and fpu_config into supported_defaults.
1039         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1040         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1041         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1042         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1043         * config/nds32/constraints.md: New constraints and checking for hard
1044         float configuration.
1045         * config/nds32/iterators.md: New mode iterator and attribute for hard
1046         float configuration.
1047         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1048         patterns.
1049         * config/nds32/nds32-fpu.md: New file.
1050         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1051         deal with hard float code generation.
1052         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1053         ARCH_V3S.
1054         (abi_type, float_reg_number): New enum type.
1055         * config/nds32/nds32-predicates.c: New predicates for hard float.
1056         * config/nds32/nds32-protos.h: Declare functions for hard float.
1057         * config/nds32/nds32.c: Implementation for hard float configuration.
1058         * config/nds32/nds32.h: Definitions for hard float configuration.
1059         * config/nds32/nds32.md: Include hard float machine description and
1060         modify patterns for hard float configuration.
1061         * config/nds32/nds32.opt: New options for hard float configuration.
1062         * config/nds32/predicates.md: New predicates for hard float
1063         configuration.
1064
1065 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
1066
1067         * common/config/nds32/nds32-common.c
1068         (nds32_option_optimization_table): Enable -mreleax-hint by default.
1069
1070 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
1071
1072         PR middle-end/85195
1073         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1074         CONSTRUCTOR_ELT (ctor, ...)->value.
1075
1076 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
1077
1078         PR target/85193
1079         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1080
1081 2018-04-05  Tom de Vries  <tom@codesourcery.com>
1082
1083         PR target/85204
1084         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1085         cond jump.
1086
1087 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
1088             Kito Cheng  <kito.cheng@gmail.com>
1089
1090         * config/nds32/constraints.md (U33): Fine-tune checking condition.
1091         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1092         * config/nds32/nds32.h (nds32_16bit_address_type): Add
1093         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1094
1095 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
1096             Kito Cheng  <kito.cheng@gmail.com>
1097
1098         * config/nds32/constraints.md (Ufe): New memory constraint.
1099         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1100         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1101         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1102         operands.
1103         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1104         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1105
1106 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
1107
1108         * config/nds32/nds32.md: Use optimize_size in the condition for
1109         alu-shift instructions.
1110
1111 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
1112
1113         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1114
1115 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
1116
1117         * config/nds32/nds32.md (negsi2): Refine pattern.
1118
1119 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
1120             Chung-Ju Wu  <jasonwucj@gmail.com>
1121
1122         * config/nds32/iterators.md (shift_rotate): New code iterator.
1123         (shift): New code attribute.
1124         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1125         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1126         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1127         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1128         bit-wise operations.
1129         (andsi3, *andsi3): Ditto.
1130         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1131         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1132         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1133         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1134         nds32_ior_operand, nds32_xor_operand): New predicates.
1135
1136 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
1137
1138         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1139         (addsi3, subsi3): ... this.
1140
1141 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
1142
1143         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1144
1145 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
1146
1147         * config/nds32/nds32.md: Adjust indention.
1148
1149 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
1150
1151         * config/nds32/nds32.md (feature): New attribute.
1152
1153 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
1154
1155         * config/nds32/nds32.md (subtype): New attribute.
1156
1157 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1158
1159         PR target/85203
1160         * config/arm/arm-builtins.c (arm_expand_builtin): Change
1161         expansion to perform a bitwise AND of the argument followed by a
1162         boolean negation of the result.
1163
1164 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
1165
1166         PR rtl-optimization/84878
1167         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1168         the basic block.  Assert the use reference is not artificial and that
1169         it has an associated insn.
1170
1171 2018-04-04  Michael Matz  <matz@suse.de>
1172
1173         * builtins.c (compute_objsize): Pass correct operand
1174         to array_at_struct_end_p.
1175
1176 2018-04-04  Richard Biener  <rguenther@suse.de>
1177
1178         PR lto/85176
1179         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1180         from contexts for DINFO_LEVEL_TERSE and below.
1181
1182 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
1183
1184         * config/nds32/nds32-doubleword.md (move_<mode>): Require
1185         resiter_operand condition.
1186         * config/nds32/nds32.md (*move<mode>): Ditto.
1187
1188 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
1189             Monk Chiang  <sh.chiang04@gmail.com>
1190
1191         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1192
1193 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
1194
1195         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1196
1197 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
1198             Kito Cheng  <kito.cheng@gmail.com>
1199
1200         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1201         nds32_cond_code_str, output_cond_branch,
1202         output_cond_branch_compare_zero, nds32_expand_cbranch,
1203         nds32_expand_cstore, nds32_expand_movcc,
1204         nds32_output_cbranchsi4_equality_zero,
1205         nds32_output_cbranchsi4_equality_reg,
1206         nds32_output_cbranchsi4_equality_reg_or_const_int,
1207         nds32_output_cbranchsi4_greater_less_zero: New functions.
1208         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1209         nds32_expand_cstore, nds32_expand_movcc,
1210         nds32_output_cbranchsi4_equality_zero,
1211         nds32_output_cbranchsi4_equality_reg,
1212         nds32_output_cbranchsi4_equality_reg_or_const_int,
1213         nds32_output_cbranchsi4_greater_less_zero): Declare.
1214         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1215         nds32_rimm11s_operand): New predicates.
1216         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1217         * config/nds32/nds32.md: Rewrite all the branch and conditional move
1218         patterns.
1219
1220 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
1221
1222         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1223         * config/nds32/nds32.md: Ditto.
1224         * config/nds32/pipelines.md: Ditto.
1225
1226 2018-04-04  Richard Biener  <rguenther@suse.de>
1227
1228         PR tree-optimization/85168
1229         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1230         propagating abnormals.
1231
1232 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
1233
1234         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1235
1236 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
1237             Kito Cheng  <kito.cheng@gmail.com>
1238
1239         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1240         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1241         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1242         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1243         * config/nds32/nds32.md (sibcall_internal): New.
1244         (sibcall_register): Remove.
1245         (sibcall_immediate): Remove.
1246         (sibcall_value_internal): New.
1247         (sibcall_value_register): Remove.
1248         (sibcall_value_immediate): Remove.
1249         * config/nds32/predicates.md (nds32_general_register_operand): New.
1250         (nds32_call_address_operand): New.
1251
1252 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
1253
1254         PR rtl-optimization/85167
1255         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1256         bb_defs if *split_p, instead preinitialize it to NULL.
1257
1258         PR tree-optimization/85156
1259         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1260         evaluating the argument multiple times.
1261
1262 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
1263
1264         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1265         than vector.
1266         (_mm_cvtpd_ps): Likewise.
1267         (_mm_cvttpd_epi32): Likewise.
1268         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1269         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1270         vector, pixel, and bool following altivec.h include.
1271
1272 2018-04-03  Martin Sebor  <msebor@redhat.com>
1273
1274         * doc/extend.texi (Common Function Attributes): Clarify.
1275         (const attribute): Likewise.
1276         (pure attribute): Likewise.
1277
1278 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
1279
1280         PR target/85169
1281         * config/i386/i386.c (ix86_expand_vector_set): Use
1282         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
1283
1284 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
1285
1286         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1287         instructions when changing rounding bits to preserve precision bits
1288         in the x87 control word.
1289
1290 2018-04-03  Martin Liska  <mliska@suse.cz>
1291
1292         PR tree-optimization/82491
1293         * rtl.h (strip_offset_and_add): Replace += suboffset with
1294         poly_uint64 () + suboffset.
1295
1296 2018-03-29  Martin Liska  <mliska@suse.cz>
1297             Martin Jambor  <mjambor@suse.cz>
1298
1299         PR ipa/84947
1300         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1301         param_type is not an integral or pointer type.
1302
1303 2018-04-03  Richard Biener  <rguenther@suse.de>
1304
1305         * sese.h (recompute_all_dominators): Remove.
1306
1307 2018-04-02  Martin Sebor  <msebor@redhat.com>
1308
1309         * doc/invoke.texi (-Wrestrict): Fix typos.
1310
1311 2018-04-02  Jim Wilson  <jimw@sifive.com>
1312
1313         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1314         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1315         (<optab>di3, <optab>si3_extend): Likewise.
1316         (<optab>si3_mask, <optab>si3_mask_1): New.
1317         (<optab>di3_mask, <optab>di3_mask_1): New.
1318         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1319         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1320         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1321
1322 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
1323
1324         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1325         example.
1326
1327 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
1328
1329         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1330         (nds32_canonicalize_comparison): New function.
1331
1332 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
1333             Kito Cheng  <kito.cheng@gmail.com>
1334             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
1335
1336         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1337         * config/nds32/constants.md (unspec_volatile_element): Add
1338         UNSPEC_VOLATILE_RELAX_GROUP.
1339         * config/nds32/nds32-relax-opt.c: New file.
1340         * config/nds32/nds32-predicates.c
1341         (nds32_symbol_load_store_p): New function.
1342         * config/nds32/nds32-protos.h
1343         (nds32_symbol_load_store_p): Declare function.
1344         (make_pass_nds32_relax_opt): Declare new rtl pass function.
1345         * config/nds32/nds32.c
1346         (nds32_register_pass): New function to register pass.
1347         (nds32_register_passes): New function to register passes.
1348         * config/nds32/nds32.md (relax_group): New pattern.
1349         * config/nds32/nds32.opt (mrelax-hint): New option.
1350         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1351
1352 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
1353
1354         * config/nds32/t-nds32: Modify files dependency.
1355
1356 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
1357
1358         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1359         (PROFILE_HOOK): Define its implementation.
1360
1361 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
1362
1363         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1364         type and 32-bit size.
1365
1366 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
1367
1368         PR middle-end/85090
1369         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1370         (V_128_256): New mode iterator.
1371         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1372         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1373         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1374         of V.
1375         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1376         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1377
1378 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
1379
1380         PR target/83315
1381         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1382         NaN inputs correctly.
1383
1384 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
1385
1386         PR target/80546
1387         * config/rs6000/vsx.md (??r): New mode attribute.
1388         (*vsx_mov<mode>_64bit): Use it.
1389         (*vsx_mov<mode>_32bit): Likewise.
1390
1391 2018-03-30  Martin Sebor  <msebor@redhat.com>
1392
1393         PR tree-optimization/84818
1394         * builtins.c (check_access): Use warning_n.
1395
1396 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
1397
1398         PR target/83822
1399         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1400         condition.
1401         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1402         condition.
1403
1404 2018-03-30  Julia Koval  <julia.koval@intel.com>
1405
1406         PR target/84413
1407         * x86-tune.def (movx, partial_reg_dependency): Enable for
1408         m_SKYLAKE_AVX512.
1409
1410 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
1411
1412         PR inline-asm/84985
1413         * lra-constraints.c (process_alt_operands): Move setting
1414         this_alternative_matches below.
1415
1416 2018-03-29  Martin Liska  <mliska@suse.cz>
1417
1418         PR lto/84995.
1419         * doc/invoke.texi: Document how LTO works with debug info.
1420         Describe auto-load support of binutils.  Mention 'x86-64'
1421         as valid option value of -march option.
1422
1423 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
1424
1425         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1426
1427         PR c/85094
1428         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1429         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1430         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1431         checking.
1432
1433 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
1434
1435         PR target/84912
1436         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1437         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1438         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1439         for RS6000_BTM_POWERPC64.
1440         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1441         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1442         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1443         definition.
1444         (DIVDE): Use it.
1445         (DIVDEU): Likewise.
1446
1447 2018-03-28 Carl Love  <cel@us.ibm.com>
1448
1449         Revert
1450         2017-09-27  Carl Love  <cel@us.ibm.com>
1451
1452         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1453         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1454         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1455         fctiw instruction.
1456
1457 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1458
1459         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1460         instead of __vector bool.
1461         (_mm_max_pu8): Likewise.
1462         (_mm_min_pi16): Likewise.
1463
1464 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
1465
1466         PR target/84912
1467         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1468         (DIVWEUO): Likewise.
1469         (DIVDEO): Likewise.
1470         (DIVDEUO): Likewise.
1471         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1472         DIVWEUO and DIVDEUO.
1473         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1474         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1475         (div_extend): Likewise.
1476         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1477         builtin function.
1478         (__builtin_divweuo): Likewise.
1479         (__builtin_divdeo): Likewise.
1480         (__builtin_divdeuo): Likewise.
1481
1482 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
1483
1484         PR target/85095
1485         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1486         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1487
1488         PR tree-optimization/82004
1489         * gimple-match-head.c (optimize_pow_to_exp): New function.
1490         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1491         Don't fold to exp if optimize_pow_to_exp is false.
1492
1493 2018-03-28  Martin Liska  <mliska@suse.cz>
1494
1495         PR other/84819
1496         * calls.c (initialize_argument_information): Fix trailing space.
1497         * common.opt: Fix typo and provide better explanation for
1498         -fsanitize-coverage option.
1499         * config/i386/i386.opt: Fix typo.
1500
1501 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
1502             Martin Liska  <mliska@suse.cz>
1503
1504         PR sanitizer/85081
1505         * gimplify.c (asan_poison_variable): Don't do the check for
1506         gimplify_omp_ctxp here.
1507         (gimplify_decl_expr): Do it here.
1508         (gimplify_target_expr): Likewise.
1509
1510 2018-03-28  Martin Liska  <mliska@suse.cz>
1511
1512         PR target/84988
1513         * config/i386/i386.c (ix86_function_arg_advance): Do not call
1514         chkp_type_bounds_count if MPX is not enabled.
1515
1516 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
1517
1518         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1519
1520 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
1521
1522         PR target/84914
1523         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1524         function to create the function decl for complex long double
1525         multiply and divide for -mabi=ieeelongdouble.
1526         (init_float128_ieee): Call it.
1527
1528 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
1529
1530         PR target/85044
1531         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1532         -fcf-protection=branch -mibt.
1533         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1534
1535 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1536
1537         PR target/81863
1538         * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1539
1540 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
1541
1542         PR target/85056
1543         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1544         extern array declarations.
1545
1546 2018-03-27  Richard Biener  <rguenther@suse.de>
1547
1548         PR middle-end/84067
1549         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1550         explicit single_use checks.
1551
1552 2018-03-27  Richard Biener  <rguenther@suse.de>
1553
1554         PR tree-optimization/85082
1555         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1556         Valueize the VUSE.
1557
1558 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1559
1560         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1561         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1562         Turn on fasynchronous-unwind-tables and funwind-tables.
1563
1564 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
1565
1566         PR target/85073
1567         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1568         (*bmi_blsr_<mode>_ccz): Ditto.
1569
1570 2018-03-26  Tom de Vries  <tom@codesourcery.com>
1571
1572         PR tree-optimization/85063
1573         * omp-general.c (offloading_function_p): New function.  Factor out
1574         of ...
1575         * omp-offload.c (pass_omp_target_link::gate): ... here.
1576         * omp-general.h (offloading_function_p): Declare.
1577         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1578         with attribute omp declare target for offloading functions.
1579
1580 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
1581
1582         PR tree-optimization/84005
1583         * tree-data-ref.h (get_base_for_alignment): Declare.
1584         * tree-data-ref.c (get_base_for_alignment_1): New function.
1585         (get_base_for_alignment): Likewise.
1586         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1587         get_base_for_alignment to find a suitable base object, instead
1588         of always using drb->base_address.
1589
1590 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
1591
1592         PR inline-asm/85022
1593         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1594         known size by default.
1595
1596 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
1597
1598         PR inline-asm/85030
1599         * lra-constraints.c (process_alt_operands): Don't match BLKmode
1600         and non BLKmode operands.
1601
1602 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1603
1604         PR target/85026
1605         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1606         Clean up attributes.
1607
1608 2018-03-23  Richard Biener  <rguenther@suse.de>
1609
1610         PR debug/85020
1611         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1612         we are going to emit early debug for LTO.
1613
1614 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
1615
1616         PR inline-asm/85034
1617         * function.c (match_asm_constraints_1): Don't optimize if input
1618         doesn't satisfy general_operand predicate for output's mode.
1619
1620         PR inline-asm/85022
1621         * alias.c (write_dependence_p): Don't require for x_canonicalized
1622         non-VOIDmode if x has VOIDmode.
1623
1624         PR sanitizer/85029
1625         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1626         just don't try to optimize it rather than assert it never happens.
1627
1628 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1629
1630         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1631         macro expansions for definition of ST_INTERNAL_<mode> and
1632         LD_INTERNAL_<mode> builtins.
1633         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1634         Remove prototype.
1635         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1636         function.
1637         (altivec_expand_st_builtin): Likewise.
1638         (altivec_expand_builtin): Remove calls to deleted functions.
1639         (rs6000_address_for_altivec): Delete this function.
1640         * config/rs6000/vector.md: Remove expands for
1641         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1642
1643 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
1644
1645         PR target/84826
1646         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1647         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1648         re-computing once computed.
1649         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1650         (arm_init_machine_status): Initialize
1651         machine->static_chain_stack_bytes.
1652
1653 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1654
1655         PR target/84760
1656         * doc/extend.texi: Add four new prototypes for vec_ld.
1657         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1658         definitions for more logical presentation.
1659         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1660         entries for V1TI variants of __builtin_altivec_ld builtin.
1661         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1662         handling of V1TI variant of LVX icode pattern.
1663         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1664         (rs6000_gimple_fold_builtin): Likewise.
1665         (altivec_init_builtins): Add code to define
1666         __builtin_altivec_lvx_v1ti function.
1667
1668 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
1669
1670         PR inline-asm/84941
1671         * function.c (match_asm_constraints_1): Don't do the optimization
1672         if input isn't a REG, SUBREG, MEM or constant.
1673
1674 2018-03-22  Tom de Vries  <tom@codesourcery.com>
1675
1676         PR tree-optimization/84956
1677         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1678         bb_has_abnormal_pred.
1679
1680 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
1681
1682         PR sanitizer/85018
1683         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1684         DECL_INITIAL (decl) to decl at the end.
1685         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1686         adjust the comment.
1687
1688 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
1689
1690         * doc/extend.texi (__builtin_tgmath): Document when complex
1691         integer types are treated as _Complex _Float64.
1692
1693 2018-03-21  Tom de Vries  <tom@codesourcery.com>
1694
1695         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1696
1697 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
1698
1699         PR tree-optimization/84960
1700         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1701         if it is ENTRY block, move them into single succ of ENTRY in that case.
1702
1703 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
1704
1705         PR tree-optimization/84811
1706         * poly-int.h (poly_span_traits): Remove the T3 parameter and
1707         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1708         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1709         (known_subrange_p): Update accordingly.  Cast each value involved
1710         in the size comparison, rather than casting the result of the
1711         subtraction.
1712
1713 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
1714
1715         PR tree-optimization/84982
1716         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1717         by flipping the least significant bit rather than all bits from
1718         bitpos to bitpos + bitsize - 1.
1719
1720 2018-03-21  Nathan Sidwell  <nathan@acm.org>
1721
1722         * doc/extend.texi (Deprecated Features): Remove mention of
1723         long-deleted deprecations.
1724
1725 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1726
1727         PR jit/84288
1728         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1729         * configure: Regenerate.
1730
1731 2018-03-21  Tom de Vries  <tom@codesourcery.com>
1732
1733         PR tree-optimization/83126
1734         * tree-parloops.c (num_phis): New function.
1735         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1736
1737 2018-03-21  Nathan Sidwell  <nathan@acm.org>
1738
1739         * doc/extend.texi (Deprecated Features): Update deprecated flags,
1740         mention anon-struct/union members and trailing attributes.
1741
1742 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
1743
1744         PR tree-optimization/84969
1745         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1746         builtin memset partitions if they set different rhs values.
1747
1748 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
1749
1750         PR rtl-optimization/84989
1751         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1752         VEC_DUPLICATE with scalar result mode.
1753
1754 2018-03-21  Martin Liska  <mliska@suse.cz>
1755
1756         PR ipa/84963
1757         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1758         not intended return statement.
1759
1760 2018-03-21  Martin Liska  <mliska@suse.cz>
1761
1762         PR target/84988
1763         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1764         (chkp_find_bound_slots_1): Limit number of iterations.
1765
1766 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
1767
1768         PR target/84838
1769         * Minor grammar fixes for x86 options.
1770
1771 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
1772
1773         PR debug/84875
1774         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1775         holding REG_CFA_RESTORE notes, instead turn them into a USE.
1776
1777 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
1778
1779         PR target/83789
1780         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1781         (altivec_lvx_<mode>_1op): Likewise.
1782         (altivec_stvx_<mode>_2op): Likewise.
1783         (altivec_stvx_<mode>_1op): Likewise.
1784         (altivec_lvx_<VM2:mode>): New define_expand.
1785         (altivec_stvx_<VM2:mode>): Likewise.
1786         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1787         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1788         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1789         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1790         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1791         (rs6000_gen_lvx): Likewise.
1792         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1793         (altivec_expand_stv_builtin): Likewise.
1794         (altivec_expand_builtin): Likewise.
1795         * config/rs6000/vector.md: Likewise.
1796
1797 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1798
1799         PR target/82518
1800         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1801         BYTES_BIG_ENDIAN.
1802
1803 2018-03-20  Richard Biener  <rguenther@suse.de>
1804
1805         PR target/84986
1806         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1807         sign-conversions as zero, fall back to standard scalar_stmt
1808         cost for the rest.
1809
1810 2018-03-20  Martin Liska  <mliska@suse.cz>
1811
1812         PR ipa/84825
1813         * predict.c (rebuild_frequencies): Handle case when we have
1814         PROFILE_ABSENT, but flag_guess_branch_prob is false.
1815
1816 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
1817
1818         PR target/84990
1819         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1820         flag_section_anchors.
1821         * varasm.c (use_blocks_for_decl_p): Remove hack for
1822         dw2_force_const_mem.
1823
1824         PR target/84845
1825         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1826         to ...
1827         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
1828         be created, use lowpart_subreg of operands[0] rather than operands[0]
1829         itself.
1830         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1831         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1832         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1833         and n constraint instead of aarch64_shift_imm_di and Usd.
1834         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1835         (*aarch64_<optab>_reg_minus<mode>3): ... this.
1836
1837 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
1838
1839         PR target/82989
1840         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1841         to favor GPR over NEON registers.
1842         (<shift>di3_neon): Likewise.
1843
1844 2018-03-20  Tom de Vries  <tom@codesourcery.com>
1845
1846         PR target/84952
1847         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1848         (nvptx_process_pars): Emit bar.sync asap and alap.
1849
1850 2018-03-20  Tom de Vries  <tom@codesourcery.com>
1851
1852         PR target/84954
1853         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1854         seen_label if seen_label is already set.
1855
1856 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
1857
1858         PR target/84945
1859         * config/i386/i386.c (fold_builtin_cpu): For features above 31
1860         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1861         Use 1U instead of 1.  Formatting fixes.
1862
1863         PR c/84953
1864         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1865         instead of TREE_TYPE (s1) for the return value.
1866
1867 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
1868
1869         PR tree-optimization/84946
1870         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1871         bitsize + bitsize in poly_uint64 rather than poly_int64.
1872
1873         PR sanitizer/78651
1874         * dwarf2asm.c: Include fold-const.c.
1875         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1876         of decl rather than decl itself.
1877
1878         PR rtl-optimization/84643
1879         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1880
1881 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
1882
1883         PR sanitizer/78651
1884         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1885         calling assemble_variable.
1886
1887 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
1888
1889         PR target/81647
1890         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1891         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1892
1893 2018-03-19  Jim Wilson  <jimw@sifive.com>
1894
1895         PR bootstrap/84856
1896         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1897         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1898         (riscv_first_stack_step): Likewise.
1899         (riscv_option_override): Use STACK_BOUNDARY instead of
1900         MIN_STACK_BOUNDARY.
1901         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1902         MIN_STACK_BOUNDARY.
1903         (BIGGEST_ALIGNMENT): Set to 128.
1904         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1905         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1906         STACK_BOUNDARY.
1907
1908 2018-03-19  Richard Biener  <rguenther@suse.de>
1909
1910         PR tree-optimization/84933
1911         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1912         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1913
1914 2018-03-19  Richard Biener  <rguenther@suse.de>
1915
1916         PR tree-optimization/84859
1917         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1918         (cond_if_else_store_replacement): Perform sinking operation on
1919         single-store BBs regardless of MAX_STORES_TO_SINK setting.
1920         Generalize what a BB with a single eligible store is.
1921
1922 2018-03-19  Richard Biener  <rguenther@suse.de>
1923
1924         PR tree-optimization/84929
1925         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1926         chrec_is_positive against non-chrec arg.
1927
1928 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
1929
1930         PR target/84711
1931         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1932
1933 2018-03-18  Martin Liska  <mliska@suse.cz>
1934
1935         PR rtl-optimization/84635
1936         * regrename.c (build_def_use): Use matches_mode only when
1937         matches >= 0.
1938
1939 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
1940
1941         PR tree-optimization/84913
1942         * tree-vect-loop.c (vectorizable_reduction): Don't try to
1943         vectorize chains of COND_EXPRs.
1944
1945 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
1946
1947         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1948
1949 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
1950
1951         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1952
1953 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
1954
1955         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1956
1957 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
1958             Kito Cheng  <kito.cheng@gmail.com>
1959
1960         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1961         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1962         (nds32_adjust_reg_alloc_order): New function.
1963         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1964
1965 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
1966
1967         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1968         nds32_print_operand, nds32_print_operand_address): Use
1969         HOST_WIDE_INT_PRINT_DEC instead.
1970
1971 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
1972
1973         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1974
1975 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
1976
1977         PR target/84902
1978         * config/i386/i386.c (initial_ix86_tune_features,
1979         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1980         unsigned long long.
1981         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1982         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1983         rather than 1u << ix86_tune.  Formatting fix.
1984         (ix86_option_override_internal): Change ix86_arch_mask from
1985         unsigned int to unsigned HOST_WIDE_INT, initialize to
1986         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1987         (ix86_function_specific_restore): Likewise.
1988
1989 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
1990
1991         PR target/84899
1992         * postreload.c (reload_combine_recognize_pattern): Perform
1993         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1994         truncate_int_for_mode the result for the destination's mode.
1995
1996         PR c/84909
1997         * hsa-gen.c (mem_type_for_type): Fix comment typo.
1998         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1999         Likewise.
2000         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2001         Likewise.
2002
2003 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
2004
2005         PR target/84876
2006         * lra-assigns.c (lra_split_hard_reg_for): Don't use
2007         regno_allocno_class_array and sorted_pseudos.
2008         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2009         insns where regno is used.
2010
2011 2018-03-16  Martin Liska  <mliska@suse.cz>
2012
2013         PR ipa/84833
2014         * multiple_target.c (create_dispatcher_calls): Redirect
2015         reference in the symbol table.
2016
2017 2018-03-16  Martin Liska  <mliska@suse.cz>
2018
2019         PR ipa/84722
2020         * multiple_target.c (create_dispatcher_calls): Redirect also
2021         an alias.
2022
2023 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
2024
2025         PR c++/79937
2026         PR c++/82410
2027         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2028         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2029         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2030
2031 2018-03-16  Julia Koval  <julia.koval@intel.com>
2032
2033         * doc/invoke.texi (Skylake Server): Add CLWB.
2034         Cannonlake): Remove CLWB.
2035
2036 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
2037
2038         PR tree-optimization/84841
2039         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2040         1 << 3.
2041         (FLOAT_ONE_CONST_TYPE): Define.
2042         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2043         (sort_by_operand_rank): Put entries with higher constant_type last
2044         rather than first to match comments.
2045
2046 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
2047
2048         * config/nios2/nios2.md (movsi_internal): Fix thinko in
2049         split predicate.
2050
2051 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
2052
2053         PR c++/79085
2054         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2055         check and use address of target always.
2056
2057 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
2058
2059         PR target/84574
2060         * config/i386/i386.c (indirect_thunk_needed): Update comments.
2061         (indirect_thunk_bnd_needed): Likewise.
2062         (indirect_thunks_used): Likewise.
2063         (indirect_thunks_bnd_used): Likewise.
2064         (indirect_return_needed): New.
2065         (indirect_return_bnd_needed): Likewise.
2066         (output_indirect_thunk_function): Add a bool argument for
2067         function return.
2068         (output_indirect_thunk_function): Don't generate alias for
2069         function return thunk.
2070         (ix86_code_end): Call output_indirect_thunk_function to generate
2071         function return thunks.
2072         (ix86_output_function_return): Set indirect_return_bnd_needed
2073         and indirect_return_needed instead of indirect_thunk_bnd_needed
2074         and indirect_thunk_needed.
2075
2076 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
2077
2078         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2079         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2080         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2081
2082 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
2083             Paul Hua <paul.hua.gm@gmail.com>
2084
2085         PR c/84852
2086         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2087
2088 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
2089
2090         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2091         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2092         resp. SFmode cases.
2093
2094 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
2095
2096         PR target/84711
2097         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2098         instead of GET_MODE_SIZE when comparing Units.
2099
2100 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
2101
2102         PR target/68256
2103         * varasm.c (hash_section): Return an unchangeble hash value
2104         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2105         Return !aarch64_can_use_per_function_literal_pools_p ().
2106
2107 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
2108
2109         PR target/84860
2110         * optabs.c (emit_conditional_move): Pass address of cmode's copy
2111         rather than address of cmode as last argument to prepare_cmp_insn.
2112
2113 2018-03-15  Julia Koval  <julia.koval@intel.com>
2114
2115         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2116         F_AVX512VNNI, F_AVX512BITALG): New.
2117
2118 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
2119
2120         PR target/83451
2121         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
2122         insn for floating-point loads and stores.
2123
2124 2018-03-14  Carl Love  <cel@us.ibm.com>
2125
2126         * config/rs6000/rs6000-c.c: Add macro definitions for
2127         ALTIVEC_BUILTIN_VEC_PERMXOR.
2128         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2129         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2130         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2131         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2132         UNSPEC_VPERMXOR.
2133         * config/doc/extend.texi: Add prototypes for vec_permxor.
2134
2135 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
2136
2137         PR c/84852
2138         * diagnostic-show-locus.c (class layout_point): Convert m_line
2139         from int to linenum_type.
2140         (line_span::comparator): Use linenum "compare" function when
2141         comparing line numbers.
2142         (test_line_span): New function.
2143         (layout_range::contains_point): Convert param "row" from int to
2144         linenum_type.
2145         (layout_range::intersects_line_p): Likewise.
2146         (layout::will_show_line_p): Likewise.
2147         (layout::print_source_line): Likewise.
2148         (layout::should_print_annotation_line_p): Likewise.
2149         (layout::print_annotation_line): Likewise.
2150         (layout::print_leading_fixits): Likewise.
2151         (layout::annotation_line_showed_range_p): Likewise.
2152         (struct line_corrections): Likewise for field m_row.
2153         (line_corrections::line_corrections): Likewise for param "row".
2154         (layout::print_trailing_fixits): Likewise.
2155         (layout::get_state_at_point): Likewise.
2156         (layout::get_x_bound_for_row): Likewise.
2157         (layout::print_line): Likewise.
2158         (diagnostic_show_locus): Likewise for locals "last_line" and
2159         "row".
2160         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2161         * input.c (selftest::test_linenum_comparisons): New function.
2162         (selftest::input_c_tests): Call it.
2163         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2164         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2165         * selftest.h (ASSERT_GT): New macro.
2166         (ASSERT_GT_AT): New macro.
2167         (ASSERT_LT): New macro.
2168         (ASSERT_LT_AT): New macro.
2169
2170 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
2171
2172         PR rtl-optimization/84780
2173         * combine.c (distribute_links): Don't make a link based on pc_rtx.
2174
2175 2018-03-14  Martin Liska  <mliska@suse.cz>
2176
2177         * tree.c (record_node_allocation_statistics): Use
2178         get_stats_node_kind.
2179         (get_stats_node_kind): New function extracted from
2180         record_node_allocation_statistics.
2181         (free_node): Use get_stats_node_kind.
2182
2183 2018-03-14  Richard Biener  <rguenther@suse.de>
2184
2185         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2186         that the value-set of ANTIC_IN doesn't grow.
2187
2188         Revert
2189         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2190         member.
2191         (BB_VISITED_WITH_VISITED_SUCCS): New define.
2192         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2193
2194 2018-03-14  Julia Koval  <julia.koval@intel.com>
2195
2196         * config.gcc (icelake-client, icelake-server): New.
2197         (icelake): Remove.
2198         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2199         (initial_ix86_arch_features): Ditto.
2200         (PTA_SKYLAKE): Add SGX.
2201         (PTA_ICELAKE): Remove.
2202         (PTA_ICELAKE_CLIENT): New.
2203         (PTA_ICELAKE_SERVER): New.
2204         (ix86_option_override_internal): Split up icelake on icelake client and
2205         icelake server.
2206         (get_builtin_code_for_version): Ditto.
2207         (fold_builtin_cpu): Ditto.
2208         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2209         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2210         * config/i386/i386.h (processor_type): Ditto.
2211         * doc/invoke.texi: Ditto.
2212
2213 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
2214
2215         PR sanitizer/83392
2216         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2217         INTEGER_CST offset, add it together with bitpos / 8 and
2218         sign extend based on POINTER_SIZE.
2219
2220         PR target/84844
2221         Revert
2222         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
2223
2224         PR target/78090
2225         * config/i386/constraints.md (Yc): New register constraint.
2226         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2227         Use Yc constraint for alternative 2 of operand 0.  Remove
2228         preferred_for_speed attribute.
2229
2230 2018-03-14  Richard Biener  <rguenther@suse.de>
2231
2232         PR tree-optimization/84830
2233         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2234         with the old one to avoid oscillations.
2235
2236 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
2237
2238         PR target/83712
2239         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2240         pseudos.
2241         (assign_by_spills): Return a flag of reload assignment failure.
2242         Do not process the reload assignment failures.  Do not spill other
2243         reload pseudos if they has the same reg class.  Update n if
2244         necessary.
2245         (lra_assign): Add a return arg.  Set up from the result of
2246         assign_by_spills call.
2247         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2248         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
2249         usage_insns if it is not NULL.
2250         (spill_hard_reg_in_range): New function.
2251         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2252         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2253         function prototypes.
2254         (lra_assign): Change prototype.
2255         * lra.c (lra): Add code to deal with fails by splitting hard reg
2256         live ranges.
2257
2258 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
2259
2260         * config/riscv/riscv.opt (mrelax): New option.
2261         * config/riscv/riscv.c (riscv_file_start): Emit ".option
2262         "norelax" when riscv_mrelax is disabled.
2263         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2264
2265 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
2266
2267         PR target/84743
2268         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2269         reassociation for int modes.
2270
2271 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
2272
2273         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2274         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2275         for big-endian.
2276         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2277         * config/aarch64/aarch64-sve.md
2278         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2279         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2280         (*extend<mode><Vwide>2): Rename to...
2281         (aarch64_sve_extend<mode><Vwide>2): ...this.
2282         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2283         renaming the old pattern to...
2284         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
2285         unsigned packs.
2286         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2287         define_expand, renaming the old pattern to...
2288         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2289         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2290         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2291         account when deciding which SVE instruction the optab should use.
2292         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2293
2294 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
2295
2296         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2297         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2298         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2299         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2300         (tlsdesc_small_<mode>): Turn a define_expand and use
2301         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
2302         (tlsdesc_small_advsimd_<mode>): ...this.
2303         (tlsdesc_small_sve_<mode>): New pattern.
2304
2305 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
2306
2307         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2308         (UNSPEC_UMUL_HIGHPART): New constants.
2309         (MUL_HIGHPART): New int iteraor.
2310         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2311         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2312         define_expand.
2313         (*<su>mul<mode>3_highpart): New define_insn.
2314
2315 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
2316
2317         PR lto/84805
2318         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2319         incomplete types.
2320
2321 2018-03-13  Martin Liska  <mliska@suse.cz>
2322
2323         PR ipa/84658.
2324         * (sem_item_optimizer::sem_item_optimizer): Initialize new
2325         vector.
2326         (sem_item_optimizer::~sem_item_optimizer): Release it.
2327         (sem_item_optimizer::merge_classes): Register variable aliases.
2328         (sem_item_optimizer::fixup_pt_set): New function.
2329         (sem_item_optimizer::fixup_points_to_sets): Likewise.
2330         * ipa-icf.h: Declare new variables and functions.
2331
2332 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
2333
2334         PR middle-end/84834
2335         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2336         integer_pow2p@2 and test integer_pow2p in condition.
2337         (A < 0 ? C : 0): Similarly for @1.
2338
2339         PR middle-end/84831
2340         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2341         characters starting at p contain '\0' character, don't look beyond
2342         that.
2343
2344         PR target/84827
2345         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2346         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2347
2348         PR target/84828
2349         * reg-stack.c (change_stack): Change update_end var from int to
2350         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2351         also call set_block_for_insn on the newly added insns and rescan.
2352
2353         PR target/84786
2354         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2355         on the last operand.
2356
2357         PR c++/84704
2358         * tree.c (stabilize_reference_1): Return save_expr (e) for
2359         STATEMENT_LIST even if it doesn't have side-effects.
2360
2361 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
2362
2363         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2364
2365 2018-03-12  Renlin Li  <renlin.li@arm.com>
2366
2367         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2368         aarch64_output_scalar_simd_mov_immediate.
2369
2370 2018-03-12  Martin Sebor  <msebor@redhat.com>
2371
2372         PR tree-optimization/83456
2373         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2374         for perfectly overlapping calls to memcpy.
2375         (gimple_fold_builtin_memory_chk): Same.
2376         (gimple_fold_builtin_strcpy): Handle no-warning.
2377         (gimple_fold_builtin_stxcpy_chk): Same.
2378         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2379
2380 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
2381
2382         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2383         parameter.  Use it for SFmode.
2384         (rs6000_function_arg_advance_1): Adjust.
2385         (rs6000_function_arg): Adjust.
2386         (rs6000_gimplify_va_arg): Pass false for that new parameter.
2387
2388 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
2389
2390         PR rtl-optimization/84169
2391         PR rtl-optimization/84780
2392         * combine.c (can_combine_p): Check for a 2-insn combination whether
2393         the destination register is used between the two insns, too.
2394
2395 2018-03-12  Richard Biener  <rguenther@suse.de>
2396
2397         PR tree-optimization/84803
2398         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2399         for refs DR analysis didn't process.
2400
2401 2018-03-12  Richard Biener  <rguenther@suse.de>
2402
2403         PR tree-optimization/84777
2404         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2405         force-vectorize loops ignore whether we are optimizing for size.
2406
2407 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
2408
2409         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2410         (TARGET_MD_ASM_ADJUST): Define.
2411
2412 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
2413             Kito Cheng  <kito.cheng@gmail.com>
2414             Chung-Ju Wu  <jasonwucj@gmail.com>
2415
2416         * config/nds32/nds32.c (nds32_compute_stack_frame,
2417         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2418         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2419         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2420         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2421         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2422         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2423         * config/nds32/nds32.md (prologue, epilogue): Use macro
2424         NDS32_V3PUSH_AVAILABLE_P to do checking.
2425
2426 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
2427
2428         PR debug/58150
2429         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2430         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2431         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
2432         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
2433         addition of most attributes on !orig_type_die or the attribute not
2434         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2435
2436 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
2437             Chung-Ju Wu  <jasonwucj@gmail.com>
2438
2439         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2440         __NDS32_VH__ macro.
2441         * config/nds32/nds32.opt (mvh): New option.
2442
2443 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
2444             Chung-Ju Wu  <jasonwucj@gmail.com>
2445
2446         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2447         function.
2448         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2449         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2450         definition.
2451
2452 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
2453             Chung-Ju Wu  <jasonwucj@gmail.com>
2454
2455         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2456         function.
2457         * config/nds32/nds32-multiple.md (strlensi): New pattern.
2458         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2459
2460 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
2461             Kito Cheng  <kito.cheng@gmail.com>
2462             Chung-Ju Wu  <jasonwucj@gmail.com>
2463
2464         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2465         UNSPEC_FFMISM and UNSPEC_FLMISM.
2466         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2467         for ffb, ffmism and flmism.
2468         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2469         (unspec_ffmism): Ditto.
2470         (unspec_flmism): Ditto.
2471         (nds32_expand_builtin_impl): Check if string extension is available.
2472         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2473         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2474
2475 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
2476
2477         Reverting patch:
2478         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
2479
2480         PR target/83712
2481         * lra-assigns.c (assign_by_spills): Return a flag of reload
2482         assignment failure.  Do not process the reload assignment
2483         failures.  Do not spill other reload pseudos if they has the same
2484         reg class.
2485         (lra_assign): Add a return arg.  Set up from the result of
2486         assign_by_spills call.
2487         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2488         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
2489         usage_insns if it is not NULL.
2490         (spill_hard_reg_in_range): New function.
2491         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2492         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2493         function prototypes.
2494         (lra_assign): Change prototype.
2495         * lra.c (lra): Add code to deal with fails by splitting hard reg
2496         live ranges.
2497
2498 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
2499
2500         PR target/84807
2501         * config/i386/i386.opt: Replace Enforcment with Enforcement.
2502
2503 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
2504
2505         PR debug/84620
2506         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2507         (dw_val_node): Add val_symbolic_view.
2508         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2509         (symview_upper_bound): New.
2510         (new_line_info_table): Initialize symviews_since_reset.
2511         (dwarf2out_source_line): Count symviews_since_reset and set
2512         symview_upper_bound.
2513         (dw_val_equal_p): Handle symview.
2514         (add_AT_symview): New.
2515         (print_dw_val): Handle symview.
2516         (attr_checksum, attr_checksum_ordered): Likewise.
2517         (same_dw_val_p, size_of_die): Likewise.
2518         (value_format, output_die): Likewise.
2519         (add_high_low_attributes): Use add_AT_symview for entry_view.
2520         (dwarf2out_finish): Reset symview_upper_bound, clear
2521         zero_view_p.
2522
2523 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
2524
2525         PR target/83969
2526         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2527         Add strict argument and use it.
2528         (rs6000_split_multireg_move): Update for new strict argument.
2529         (mem_operand_gpr): Disallow all non-offsettable addresses.
2530         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2531
2532 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
2533
2534         PR target/84772
2535         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2536         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2537         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2538
2539         PR c++/84767
2540         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2541         decl, use remap_type if we want to use the type.
2542
2543 2018-03-09  Martin Sebor  <msebor@redhat.com>
2544
2545         PR tree-optimization/84526
2546         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2547         Remove dead code.
2548         (builtin_access::generic_overlap): Be prepared to handle non-array
2549         base objects.
2550
2551 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
2552
2553         PR rtl-optimization/84682
2554         * lra-constraints.c (process_address_1): Check is_address flag
2555         for address constraints.
2556         (process_alt_operands): Likewise.
2557         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2558         preprocess_constraints.
2559         * recog.h (preprocess_constraints): Add oploc parameter.
2560         Adjust callers.
2561         * recog.c (preprocess_constraints): Test address_operand for
2562         CT_ADDRESS constraints.
2563
2564 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
2565
2566         PR target/83712
2567         * lra-assigns.c (assign_by_spills): Return a flag of reload
2568         assignment failure.  Do not process the reload assignment
2569         failures.  Do not spill other reload pseudos if they has the same
2570         reg class.
2571         (lra_assign): Add a return arg.  Set up from the result of
2572         assign_by_spills call.
2573         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2574         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
2575         usage_insns if it is not NULL.
2576         (spill_hard_reg_in_range): New function.
2577         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2578         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2579         function prototypes.
2580         (lra_assign): Change prototype.
2581         * lra.c (lra): Add code to deal with fails by splitting hard reg
2582         live ranges.
2583
2584 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2585
2586         PR target/83193
2587         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2588         Accept complain bool parameter.  Only emit errors if it is true.
2589         (arm_parse_cpu_option_name): Likewise.
2590         (arm_target_thumb_only): Adjust callers of the above.
2591         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2592         prototype to take a default true bool parameter.
2593         (arm_parse_arch_option_name): Likewise.
2594
2595 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
2596             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
2597
2598         PR jit/64089
2599         PR jit/84288
2600         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2601         * configure: Regenerate.
2602         * configure.ac ("linker --version-script option"): New.
2603         ("linker soname option"): New.
2604
2605 2018-03-09  Richard Biener  <rguenther@suse.de>
2606
2607         PR tree-optimization/84775
2608         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2609         immediate uses of predicate stmts and mark them modified.
2610
2611         Revert
2612         PR tree-optimization/84178
2613         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2614         to caller.
2615         (version_loop_for_if_conversion): Delay update_ssa call.
2616         (tree_if_conversion): Delay update_ssa until after predicate
2617         insertion.
2618
2619 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
2620
2621         PR target/84763
2622         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2623         when the function accesses prior frames.
2624
2625 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
2626
2627         PR debug/84456
2628         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2629         gen_llsym, otherwise call maybe_gen_llsym.
2630
2631         PR inline-asm/84742
2632         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2633         has ',' character inside of it.
2634
2635 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2636
2637         PR target/84748
2638         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2639         as clobbering CC_REGNUM.
2640
2641 2018-03-08  Richard Biener  <rguenther@suse.de>
2642
2643         PR middle-end/84552
2644         * tree-scalar-evolution.c: Include tree-into-ssa.h.
2645         (follow_copies_to_constant): Do not follow SSA names registered
2646         for update.
2647
2648 2018-03-08  Richard Biener  <rguenther@suse.de>
2649
2650         PR tree-optimization/84178
2651         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2652         to caller.
2653         (version_loop_for_if_conversion): Delay update_ssa call.
2654         (tree_if_conversion): Delay update_ssa until after predicate
2655         insertion.
2656
2657 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
2658
2659         PR tree-optimization/84178
2660         * tree-if-conv.c (release_bb_predicate): Remove the
2661         the assertion that the stmts have NULL use_ops.
2662         Discard the statements, asserting that they haven't
2663         yet been added to a BB.
2664
2665 2018-03-08  Richard Biener  <rguenther@suse.de>
2666
2667         PR tree-optimization/84746
2668         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2669         (phi_translate): Pass in destination ANTIC_OUT set.
2670         (phi_translate_1): Likewise.  For a simplified result lookup
2671         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2672         (phi_translate_set): Adjust.
2673         (do_pre_regular_insertion): Likewise.
2674         (do_pre_partial_partial_insertion): Likewise.
2675
2676 2018-03-08  Martin Liska  <mliska@suse.cz>
2677
2678         PR gcov-profile/84735
2679         * doc/gcov.texi: Document usage of profile files.
2680         * gcov-io.h: Document changes in the format.
2681
2682 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
2683
2684         PR debug/84404
2685         PR debug/84408
2686         * dwarf2out.c (struct dw_line_info_table): Update comments for
2687         view == -1.
2688         (FORCE_RESET_NEXT_VIEW): New.
2689         (FORCE_RESETTING_VIEW_P): New.
2690         (RESETTING_VIEW_P): Check for -1 too.
2691         (ZERO_VIEW_P): Likewise.
2692         (new_line_info_table): Force-reset next view.
2693         (dwarf2out_begin_function): Likewise.
2694         (dwarf2out_source_line): Simplify zero_view_p initialization.
2695         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2696         view directly.  Omit view when omitting .loc at line 0.
2697
2698 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
2699
2700         PR tree-optimization/84740
2701         * tree-switch-conversion.c (process_switch): Call build_constructors
2702         only if info.phi_count is non-zero.
2703
2704         PR tree-optimization/84739
2705         * tree-tailcall.c (find_tail_calls): Check call arguments against
2706         DECL_ARGUMENTS (current_function_decl) rather than
2707         DECL_ARGUMENTS (func) when checking for tail recursion.
2708
2709 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
2710
2711         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2712         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2713         Volker Reichelt's entry and add entries for people that perform
2714         GCC fuzzy testing and report numerous bugs.
2715
2716 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
2717
2718         PR target/82411
2719         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2720         readonly data in sdata, if that is disabled.
2721         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2722         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2723         -mreadonly-in-sdata option.
2724
2725 2018-03-07  Martin Sebor  <msebor@redhat.com>
2726
2727         PR tree-optimization/84468
2728         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2729         basic block when looking for nul assignment.
2730
2731 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
2732
2733         PR target/84277
2734         * except.h (output_function_exception_table): Adjust prototype.
2735         * except.c (output_function_exception_table): Remove FNNAME parameter
2736         and add SECTION parameter.  Ouput one part of the table at a time.
2737         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2738         the first part of the exception table and emit unwind directives.
2739         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2740         (i386_pe_seh_cold_init): Likewise.
2741         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2742         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2743         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2744         (ix86_output_call_insn): Emit a nop in one more case for SEH.
2745         * config/i386/winnt.c: Include except.h.
2746         (struct seh_frame_state): Add reg_offset, after_prologue and
2747         in_cold_section fields.
2748         (i386_pe_seh_end_prologue): Set seh->after_prologue.
2749         (i386_pe_seh_cold_init): New function.
2750         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2751         to seh->in_cold_section.
2752         (seh_emit_push): Record the offset of the push.
2753         (seh_emit_save): Record the offet of the save.
2754         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2755         Test seh->after_prologue to disregard the epilogue.
2756         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2757         (i386_pe_end_cold_function): New function.
2758
2759 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
2760
2761         PR fortran/84565
2762         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2763         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2764
2765         PR c++/84704
2766         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2767         on tmp_var.
2768         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2769         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2770
2771         PR middle-end/84723
2772         * multiple_target.c: Include tree-inline.h and intl.h.
2773         (expand_target_clones): Diagnose and fail if node->definition and
2774         !tree_versionable_function_p (node->decl).
2775
2776 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
2777
2778         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2779         sprint_ul.
2780         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2781         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2782         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2783
2784 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
2785
2786         PR target/84710
2787         * combine.c (try_combine): Use reg_or_subregno instead of handling
2788         just paradoxical SUBREGs and REGs.
2789
2790 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
2791
2792          * config/arc/arc.c (arc_finalize_pic): Remove function.
2793          (arc_must_save_register): We use single base PIC register, remove
2794          checks to save/restore the PIC register.
2795          (arc_expand_prologue): Likewise.
2796          * config/arc/arc-protos.h (arc_set_default_type_attributes):
2797          Remove.
2798          (arc_verify_short): Likewise.
2799          (arc_attr_type): Likewise.
2800          * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2801          (walk_stores): Likewise.
2802          (arc_address_cost): Make it static.
2803          (arc_verify_short): Likewise.
2804          (branch_dest): Likewise.
2805          (arc_attr_type): Likewise.
2806          * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2807          (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2808          (arc_final_prescan_insn): Remove inserting the nops due to
2809          hardware hazards.  It is done in reorg step.
2810          (insn_length_variant_t): Remove.
2811          (insn_length_parameters_t): Likewise.
2812          (arc_insn_length_parameters): Likewise.
2813          (arc_get_insn_variants): Likewise.
2814          * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2815
2816 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
2817
2818         PR inline-asm/84683
2819         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2820         assertion failure.
2821
2822         PR tree-optimization/84687
2823         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2824         on new_node->decl.
2825         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2826
2827 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2828
2829         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2830         Rename to ppc_speculation_barrier.
2831         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2832         __builtin_ppc_speculation_barrier.
2833
2834 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
2835
2836         PR target/84700
2837         * combine.c (combine_simplify_rtx): Don't try to simplify if
2838         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2839         are equal to x.
2840
2841 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
2842
2843         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2844         to 32 bytes when compiling for POWER9.
2845
2846 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
2847
2848         PR target/84564
2849         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2850         regparm >= 3 with no arg reg available also for calls with
2851         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
2852
2853         PR target/84524
2854         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2855         orig,vex.
2856         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
2857
2858 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
2859
2860         PR target/84264
2861         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2862
2863 2018-03-05  Richard Biener  <rguenther@suse.de>
2864
2865         PR tree-optimization/84486
2866         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2867         When inserting a __builtin_assume_aligned call set the LHS
2868         SSA name alignment info accordingly.
2869
2870 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
2871
2872         PR tree-optimization/84114
2873         * config/aarch64/aarch64.c (aarch64_reassociation_width)
2874         Avoid reassociation of FLOAT_MODE addition.
2875
2876 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
2877
2878         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2879         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2880         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2881         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2882         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2883         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2884         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2885         and -mwbnoinvd.
2886         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2887         __builtin_ia32_wbinvd): New builtins.
2888         (SPECIAL_ARGS2): New.
2889         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2890         (SPECIAL_ARGS2): New.
2891         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2892         (ix86_valid_target_attribute_inner_p): Ditto.
2893         (ix86_init_mmx_sse_builtins): Add special_args2.
2894         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2895         TARGET_WBNOINVD_P): New.
2896         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2897         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2898         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2899         * config/i386/immintrin.h (_wbinvd): New intrinsic.
2900         * config/i386/pconfigintrin.h: New file.
2901         * config/i386/wbnoinvdintrin.h: Ditto.
2902         * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2903         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2904
2905 2018-03-05  Richard Biener  <rguenther@suse.de>
2906
2907         PR tree-optimization/84670
2908         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2909         member.
2910         (BB_VISITED_WITH_VISITED_SUCCS): New define.
2911         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2912         (compute_antic_aux): Only assert the number of values in ANTIC_IN
2913         doesn't grow if all successors (recursively) were visited at least
2914         once.
2915
2916 2018-03-05  Richard Biener  <rguenther@suse.de>
2917
2918         PR tree-optimization/84650
2919         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2920         if executed in the loop pipeline.
2921
2922 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
2923
2924         * doc/configfiles.texi (Configuration Files): Move info about
2925         conditionalizing $target-protos.h to...
2926         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
2927         differs from $target-protos.h.
2928
2929 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
2930             Chung-Ju Wu  <jasonwucj@gmail.com>
2931
2932         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2933         * config/nds32/nds32-multiple.md (setmemsi): Define.
2934         * config/nds32/nds32-memory-manipulation.c
2935         (nds32_gen_dup_4_byte_to_word_value): New.
2936         (emit_setmem_word_loop): New.
2937         (emit_setmem_byte_loop): New.
2938         (nds32_expand_setmem_loop): New.
2939         (nds32_expand_setmem_loop_v3m): New.
2940         (nds32_expand_setmem_unroll): New.
2941         (nds32_expand_setmem): New.
2942
2943 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
2944             Chung-Ju Wu  <jasonwucj@gmail.com>
2945
2946         * config/nds32/nds32-memory-manipulation.c
2947         (nds32_emit_load_store): New.
2948         (nds32_emit_post_inc_load_store): New.
2949         (nds32_emit_mem_move): New.
2950         (nds32_emit_mem_move_block): New.
2951         (nds32_expand_movmemsi_loop_unknown_size): New.
2952         (nds32_expand_movmemsi_loop_known_size): New.
2953         (nds32_expand_movmemsi_loop): New.
2954         (nds32_expand_movmemsi_unroll): New.
2955         (nds32_expand_movmemqi): Rename ...
2956         (nds32_expand_movmemsi): ... to this.
2957         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2958         (movmemsi): ... to this.
2959         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2960         (nds32_expand_movmemsi): ... to this.
2961
2962 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
2963             Monk Chiang  <sh.chiang04@gmail.com>
2964             Chung-Ju Wu  <jasonwucj@gmail.com>
2965
2966         * config/nds32/nds32-protos.h
2967         (nds32_expand_load_multiple): New arguments.
2968         (nds32_expand_store_multiple): Ditto.
2969         (nds32_valid_multiple_load_store): Rename ...
2970         (nds32_valid_multiple_load_store_p): ... to this.
2971         * config/nds32/nds32-memory-manipulation.c
2972         (nds32_expand_load_multiple): Refine implementation.
2973         (nds32_expand_store_multiple): Ditto.
2974         * config/nds32/nds32-multiple.md
2975         (load_multiple): Update nds32_expand_load_multiple interface.
2976         (store_multiple): Update nds32_expand_store_multiple interface.
2977         * config/nds32/nds32-predicates.c
2978         (nds32_valid_multiple_load_store): Rename ...
2979         (nds32_valid_multiple_load_store_p): ... to this and refine
2980         implementation.
2981         * config/nds32/predicates.md
2982         (nds32_load_multiple_and_update_address_operation): New predicate.
2983         (nds32_store_multiple_and_update_address_operation): New predicate.
2984
2985 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
2986             Chung-Ju Wu  <jasonwucj@gmail.com>
2987
2988         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2989         (combo): New attribute.
2990         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2991
2992 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
2993
2994         * config/nds32/nds32.opt: Change -mcmodel= default value.
2995
2996 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
2997             Monk Chiang  <sh.chiang04@gmail.com>
2998             Chung-Ju Wu  <jasonwucj@gmail.com>
2999
3000         * config/nds32/constants.md (unspec_element): New enum.
3001         * config/nds32/constraints.md (Umw): New constraint.
3002         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3003         * config/nds32/nds32-intrinsic.md: Likewise.
3004         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3005         (nds32_valid_smw_lwm_base_p): New.
3006         (nds32_output_smw_single_word): New.
3007         (nds32_output_lmw_single_word): New.
3008         (nds32_expand_unaligned_load): New.
3009         (nds32_expand_unaligned_store): New.
3010         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3011         (nds32_output_smw_single_word): Declare.
3012         (nds32_output_lmw_single_word): Declare.
3013         (nds32_expand_unaligned_load): Declare.
3014         (nds32_expand_unaligned_store): Declare.
3015         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3016         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3017         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3018         NDS32_BUILTIN_UASTORE_DW.
3019         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3020         predicate.
3021
3022 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
3023             Kito Cheng  <kito.cheng@gmail.com>
3024             Chung-Ju Wu  <jasonwucj@gmail.com>
3025
3026         * config/nds32/nds32-intrinsic.c
3027         (nds32_expand_builtin_null_ftype_reg): Delete.
3028         (nds32_expand_builtin_reg_ftype_imm): Ditto.
3029         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3030         (nds32_read_argument): New.
3031         (nds32_legitimize_target): Ditto.
3032         (nds32_legitimize_argument): Ditto.
3033         (nds32_check_constant_argument): Ditto.
3034         (nds32_expand_unop_builtin): Ditto.
3035         (nds32_expand_unopimm_builtin): Ditto.
3036         (nds32_expand_binop_builtin): Ditto.
3037         (nds32_builtin_decl_impl): Ditto.
3038         (builtin_description): Ditto.
3039         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3040         (nds32_init_builtins_impl): Ditto.
3041         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3042         (nds32_builtin_decl): New.
3043         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3044         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3045
3046 2018-03-02  Jeff Law  <law@redhat.com>
3047
3048         * reorg.c (stop_search_p): Handle DEBUG_INSN.
3049         (redundant_insn, fill_simple_delay_slots): Likewise.
3050         (fill_slots_from_thread): Likewise.
3051         * resource.c (mark_referenced_resources): Likewise.
3052         (mark_set_resources, find_dead_or_set_registers): Likewise.
3053
3054 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
3055
3056         * substring-locations.h (format_warning_va): Formatting fix for
3057         ATTRIBUTE_GCC_DIAG.
3058         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3059         argument.
3060         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3061         * substring-locations.c: Include intl.h.
3062         (format_warning_va): Turned into small wrapper around
3063         format_warning_n_va, renamed to ...
3064         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3065         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3066         use ngettext.
3067         (format_warning_at_substring_n): New function.
3068         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3069         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
3070         format_warning_at_substring with just a shorter name instead of
3071         const function pointer.
3072         (fmtwarn_n): New function.
3073         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3074         appropriate, get rid of all the fmtstr temporaries, move conditionals
3075         with G_() wrapped string literals directly into fmtwarn arguments,
3076         cast dir.len to (int), formatting fixes.
3077
3078 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
3079
3080         * doc/invoke.texi: Remove "Cilk Plus" references.
3081
3082 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
3083             Richard Biener  <rguenther@suse.de>
3084
3085         PR ipa/84628
3086         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3087         for error or warning attributes if CALL_FROM_THUNK_P is set.
3088         Formatting fixes.
3089
3090 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
3091
3092         PR target/56540
3093         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3094         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3095
3096         PR target/56540
3097         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3098         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3099
3100         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3101         instead of -1U in last predictors element's probability member.
3102
3103 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
3104
3105         PR ipa/83983
3106         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3107         arguments if they are comparable.
3108
3109 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
3110
3111         PR tree-optimization/84634
3112         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3113         masks and masked_loop_p with a single loop_masks, making sure it's
3114         null for bb vectorization.
3115
3116 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
3117
3118         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3119         (vect_analyze_data_ref_access): Use loop->safe_len rather than
3120         loop->force_vectorize to check whether there is no alias.
3121
3122 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
3123
3124         PR target/84614
3125         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3126         prototypes.
3127         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3128         comments.
3129         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3130         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3131         instead of a loop around prev_real_insn.
3132         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3133         prev_real_insn.
3134
3135         PR inline-asm/84625
3136         * config/i386/i386.c (ix86_print_operand): Use conditional
3137         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3138         zero vector.
3139
3140 2018-03-02  Richard Biener  <rguenther@suse.de>
3141
3142         PR tree-optimization/84427
3143         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3144         (bitmap_set_subtract_values): Rewrite to handle multiple
3145         exprs per value.
3146         (clean): Likewise.
3147         (prune_clobbered_mems): Likewise.
3148         (phi_translate): Take edge instead of pred/phiblock.
3149         (phi_translate_1): Likewise.
3150         (phi_translate_set): Likewise.  Insert all translated
3151         exprs for a value into the set, keeping possibly multiple
3152         expressions per value.
3153         (compute_antic_aux): Adjust for phi_translate changes.
3154         When intersecting union the expressions and prune those
3155         not in the final value set, keeping possibly multiple
3156         expressions per value.  Do not use value-insertion
3157         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3158         all expressions.  Add verification that the value-sets
3159         only shrink during iteration.
3160         (compute_partial_antic_aux): Adjust for the phi_translate changes.
3161         (do_pre_regular_insertion): Likewise.
3162         (do_pre_partial_partial_insertion): Likewise.
3163
3164 2018-03-02  Richard Biener  <rguenther@suse.de>
3165
3166         PR target/82005
3167         * config/darwin.c (saved_debug_info_level): New static global.
3168         (darwin_asm_lto_start): Disable debug info generation for LTO out.
3169         (darwin_asm_lto_end): Restore debug info generation settings.
3170
3171 2018-03-01  Martin Liska  <mliska@suse.cz>
3172
3173         PR sanitizer/82484
3174         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3175         volatile arguments.
3176
3177 2018-03-01  Richard Biener  <rguenther@suse.de>
3178
3179         PR debug/84645
3180         * dwarf2out.c (gen_variable_die): Properly handle late VLA
3181         type annotation with LTO when debug was disabled at compile-time.
3182
3183 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
3184
3185         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3186         XINT with INTVAL.
3187         (mips_final_postscan_insn): Likewise.
3188
3189 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
3190
3191         PR rtl-optimization/84528
3192         * alias.c (init_alias_target): Add commentary.
3193         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3194         a unique base value if the frame pointer is not eliminated
3195         to the stack pointer.
3196
3197 2018-03-01  Tom de Vries  <tom@codesourcery.com>
3198
3199         PR rtl-optimization/83327
3200         * lra-int.h (hard_regs_spilled_into): Declare.
3201         * lra.c (hard_regs_spilled_into): Define.
3202         (init_reg_info): Init hard_regs_spilled_into.
3203         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3204         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3205         (process_bb_lives): Handle hard_regs_spilled_into.
3206         (lra_create_live_ranges_1): Before doing liveness propagation, clear
3207         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3208
3209 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
3210
3211         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3212         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3213         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3214         * config/rs6000/aix72.h: New file.
3215
3216 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
3217
3218         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3219         instead of warning_at with conditional singular and plural messages
3220         where possible.
3221
3222         PR target/52991
3223         * stor-layout.c (update_alignment_for_field): For
3224         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3225         && !DECL_PACKED (field), do the alignment update, just use
3226         only desired_align instead of MAX (type_align, desired_align)
3227         as the alignment.
3228         (place_field): Don't do known_align < desired_align handling
3229         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3230         is non-NULL, instead do it after rli->prev_field handling and
3231         only if not within a bitfield word.  For DECL_PACKED (field)
3232         use type_align of BITS_PER_UNIT.
3233
3234 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
3235
3236         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3237         superfluous parentheses and trailing spaces.
3238
3239 2018-02-28  Richard Biener  <rguenther@suse.de>
3240
3241         PR tree-optimization/84584
3242         * graphite-scop-detection.c (scop_detection::add_scop): Discard
3243         SCoPs with fake exit edge.
3244
3245 2018-02-28  Martin Liska  <mliska@suse.cz>
3246
3247         PR testsuite/84597
3248         * timevar.c (timer::print): Fix format to properly print 100%
3249         values.
3250
3251 2018-02-28  Richard Biener  <rguenther@suse.de>
3252
3253         PR middle-end/84607
3254         * genmatch.c (capture_info::walk_match): Do not mark
3255         captured expressions without operands as expr_p given
3256         they act more like predicates and should be subject to
3257         "lost tail" side-effect preserving.
3258
3259 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
3260
3261         PR rtl-optimization/81611
3262         * auto-inc-dec.c (attempt_change): Move dead note from
3263         mem_insn if it's the next use of regno
3264         (find_address): Take address use of reg holding
3265         non-incremented value.  Add parm to limit search to the named
3266         reg only.
3267         (merge_in_block): Attempt to use a mem insn that is the next
3268         use of the original regno.
3269
3270 2018-02-27  Martin Sebor  <msebor@redhat.com>
3271
3272         PR c++/83871
3273         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3274         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3275
3276 2018-02-27  Martin Sebor  <msebor@redhat.com>
3277
3278         PR translation/84207
3279         * diagnostic-core.h (warning_n, error_n, inform_n): Change
3280         n argument to unsigned HOST_WIDE_INT.
3281         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3282         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
3283         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3284         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3285
3286 2018-02-27  Richard Biener  <rguenther@suse.de>
3287
3288         PR tree-optimization/84512
3289         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3290         Do not use the estimate returned from record_stmt_cost for
3291         the scalar iteration cost but sum properly using add_stmt_cost.
3292
3293 2018-02-27  Richard Biener  <rguenther@suse.de>
3294
3295         PR tree-optimization/84466
3296         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3297         Adjust last change to less strictly validate use operands.
3298
3299 2018-02-27  Martin Liska  <mliska@suse.cz>
3300
3301         PR gcov-profile/84548
3302         * gcov.c (process_file): Allow partial overlap and consider it
3303         also as group functions.
3304         (output_lines): Properly calculate range of lines for a group.
3305
3306 2018-02-27  Martin Liska  <mliska@suse.cz>
3307
3308         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3309         'ggc' suffixes.  Change first column width.
3310         (timer::print): Fix formatting of the column.
3311
3312 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
3313
3314         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3315         preserve inline entry blocks for the sake of debug inline
3316         entry point markers alone.
3317         (remove_unused_locals): Suggest in comments a better place to
3318         force the preservation of inline entry blocks that are
3319         otherwise unused, but do not preserve them.
3320
3321 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
3322
3323         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3324
3325 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
3326
3327         PR target/84039
3328         * config/i386/constraints.md (Bs): Replace
3329         ix86_indirect_branch_register with
3330         TARGET_INDIRECT_BRANCH_REGISTER.
3331         (Bw): Likewise.
3332         * config/i386/i386.md (indirect_jump): Likewise.
3333         (tablejump): Likewise.
3334         (*sibcall_memory): Likewise.
3335         (*sibcall_value_memory): Likewise.
3336         Peepholes of indirect call and jump via memory: Likewise.
3337         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3338         (*sibcall_value_GOT_32): Likewise.
3339         * config/i386/predicates.md (indirect_branch_operand): Likewise.
3340         (GOT_memory_operand): Likewise.
3341         (call_insn_operand): Likewise.
3342         (sibcall_insn_operand): Likewise.
3343         (GOT32_symbol_operand): Likewise.
3344         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3345
3346 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
3347
3348         PR rtl-optimization/83496
3349         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3350         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
3351         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3352         redundant insn, if any.
3353         (relax_delay_slots): Likewise.
3354         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3355
3356 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
3357
3358         PR tree-optimization/83965
3359         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3360         that grouped statements are part of a reduction chain.  Return
3361         true if the statement is not marked as a reduction itself but
3362         is part of a group.
3363         (vect_recog_dot_prod_pattern): Don't check whether the statement
3364         is part of a group here.
3365         (vect_recog_sad_pattern): Likewise.
3366         (vect_recog_widen_sum_pattern): Likewise.
3367
3368 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
3369
3370         PR debug/84545
3371         * final.c (rest_of_clean_state): Also look for calls inside sequences.
3372
3373 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
3374
3375         PR target/84530
3376         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3377         the bool argument.
3378         (ix86_output_indirect_function_return): New prototype.
3379         (ix86_split_simple_return_pop_internal): Likewise.
3380         * config/i386/i386.c (indirect_return_via_cx): New.
3381         (indirect_return_via_cx_bnd): Likewise.
3382         (indirect_thunk_name): Handle return va CX_REG.
3383         (output_indirect_thunk_function): Create alias for
3384         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3385         (ix86_output_indirect_jmp): Remove the bool argument.
3386         (ix86_output_indirect_function_return): New function.
3387         (ix86_split_simple_return_pop_internal): Likewise.
3388         * config/i386/i386.md (*indirect_jump): Don't pass false
3389         to ix86_output_indirect_jmp.
3390         (*tablejump_1): Likewise.
3391         (simple_return_pop_internal): Change it to define_insn_and_split.
3392         Call ix86_split_simple_return_pop_internal to split it for
3393         -mfunction-return=.
3394         (simple_return_indirect_internal): Call
3395         ix86_output_indirect_function_return instead of
3396         ix86_output_indirect_jmp.
3397
3398 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
3399
3400         PR bootstrap/84405
3401         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3402         memset and value initialization afterwards.
3403
3404 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
3405
3406         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3407
3408 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
3409
3410         PR target/84521
3411         * common/config/aarch64/aarch64-common.c
3412         (aarch_option_optimization_table[]): Switch
3413         off fomit-frame-pointer
3414
3415 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
3416             Chung-Ju Wu  <jasonwucj@gmail.com>
3417
3418         * config/nds32/nds32-multiple.md (load_multiple): Disallow
3419         volatile memory.
3420         (store_multiple): Ditto.
3421
3422 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
3423
3424         * config.gcc: Add --with-cpu support for nds32 target.
3425         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3426         * config/nds32/nds32.opt: Add -mcpu= option.
3427
3428 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
3429
3430         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3431         isel=yes): Warn for these deprecated options.
3432
3433 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
3434
3435         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3436         ISA_2_5_MASKS_EMBEDDED.
3437
3438 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
3439
3440         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3441         p->max as pointers rather than using iterative_hash_expr.
3442
3443 2018-02-23  Carl Love  <cel@us.ibm.com>
3444
3445         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3446         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3447         BU_P8V_OVERLOAD_2.
3448         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3449         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3450         P8V_BUILTIN_VEC_VUNSIGNED2.
3451
3452 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
3453
3454         PR target/81572
3455         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3456         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3457         LRA_UNKNOWN_ALT.
3458         * lra-constraints.c (curr_insn_transform): Set up
3459         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
3460         LRA_UNKNOWN_ALT.
3461         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3462         * lra-eliminations.c (spill_pseudos): Ditto.
3463         (process_insn_for_elimination): Ditto.
3464         * lra-lives.c (reg_early_clobber_p): Use the new macros.
3465         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3466         LRA_NON_CLOBBERED_ALT.
3467
3468 2018-02-22  Martin Sebor  <msebor@redhat.com>
3469
3470         PR tree-optimization/84480
3471         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3472         to maybe_diag_stxncpy_trunc.  Call it.
3473         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3474         from gimple_fold_builtin_strcpy.  Print inlining stack.
3475         (handle_builtin_stxncpy): Print inlining stack.
3476         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3477
3478 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
3479
3480         PR target/84176
3481         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3482         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3483         and -fcheck-pointer-bounds are used together.
3484         (indirect_thunk_prefix): New enum.
3485         (indirect_thunk_need_prefix): New function.
3486         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
3487         "_nt" instead of "_bnd" for NOTRACK prefix.
3488         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3489         (output_indirect_thunk_function): Likewise.
3490         (): Likewise.
3491         (ix86_code_end): Update output_indirect_thunk_function calls.
3492         (ix86_output_indirect_branch_via_reg): Replace
3493         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3494         (ix86_output_indirect_branch_via_push): Likewise.
3495         (ix86_output_function_return): Likewise.
3496         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3497         incompatible with -fcf-protection=branch and
3498         -fcheck-pointer-bounds.
3499
3500 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
3501
3502         PR target/83335
3503         * config/aarch64/aarch64.c (aarch64_print_address_internal):
3504         Change gcc_assert call to output_operand_lossage.
3505
3506 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
3507
3508         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3509
3510 2018-02-22  DJ Delorie  <dj@redhat.com>
3511             Sebastian Perta  <sebastian.perta@renesas.com>
3512             Oleg Endo  <olegendo@gcc.gnu.org>
3513
3514         * config/rx/rx.c (rx_rtx_costs): New function.
3515         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3516
3517 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
3518
3519         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3520
3521 2018-02-22  Martin Liska  <mliska@suse.cz>
3522
3523         PR driver/83193
3524         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3525         Add "native" as a possible value.
3526
3527 2018-02-22  Martin Liska  <mliska@suse.cz>
3528
3529         PR driver/83193
3530         * config/i386/i386.c (ix86_option_override_internal):
3531         Add "native" as a possible value for -march and -mtune.
3532
3533 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
3534
3535         PR target/84502
3536         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3537         to all type variants.
3538
3539         PR tree-optimization/84503
3540         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3541         width as info->bitpos + info->bitsize - start.
3542         (merged_store_group::merge_overlapping): Simplify width computation.
3543         (check_no_overlap): New function.
3544         (imm_store_chain_info::try_coalesce_bswap): Compute expected
3545         start + width and last_order of the group, fail if check_no_overlap
3546         fails.
3547         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3548         to group if check_no_overlap fails.
3549
3550 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
3551
3552         * config/rs6000/altivec.md: Delete contraint arguments to
3553         define_expand, define_split, and define_peephole2, and in
3554         define_insn_and_split if always unused.
3555         * config/rs6000/darwin.md: Ditto.
3556         * config/rs6000/dfp.md: Ditto.
3557         * config/rs6000/rs6000.md: Ditto.
3558         * config/rs6000/sync.md: Ditto.
3559         * config/rs6000/vector.md: Ditto.
3560         * config/rs6000/vsx.md: Ditto.
3561
3562 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
3563
3564         * config/rs6000/altivec.md: Write output control strings as braced
3565         blocks instead of double-quoted strings.
3566         * config/rs6000/darwin.md: Ditto.
3567         * config/rs6000/rs6000.md: Ditto.
3568         * config/rs6000/vector.md: Ditto.
3569         * config/rs6000/vsx.md: Ditto.
3570
3571 2018-02-21  Jason Merrill  <jason@redhat.com>
3572
3573         PR c++/84314 - ICE with templates and fastcall attribute.
3574         * attribs.c (build_type_attribute_qual_variant): Remove assert.
3575
3576 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
3577
3578         * ipa-cp.c (determine_versionability): Fix comment typos.
3579
3580 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
3581
3582         PR c/84229
3583         * ipa-cp.c (determine_versionability): Do not version functions caling
3584         va_arg_pack.
3585
3586 2018-02-21  Martin Liska  <mliska@suse.cz>
3587
3588         PR driver/83193
3589         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3590         Add "native" as a possible value.
3591         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
3592         the macro when native cpu detection is available.
3593
3594 2018-02-21  Martin Liska  <mliska@suse.cz>
3595
3596         PR driver/83193
3597         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3598         Add "native" as a possible value.
3599         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3600         when native cpu detection is available.
3601
3602 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
3603             Martin Sebor  <msebor@redhat.com>
3604
3605         PR tree-optimization/84478
3606         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3607         false.
3608         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3609         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3610         support which is conservatively correct, for 2 only stay conservative
3611         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
3612         argument to the 2 argument get_range_strlen, adjust 6 arg
3613         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3614         false.
3615         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3616         (gimple_fold_builtin_strlen): Pass true as last argument to
3617         get_range_strlen.
3618
3619 2018-02-20  Martin Sebor  <msebor@redhat.com>
3620
3621         PR middle-end/84095
3622         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3623         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
3624         (builtin_memref::builtin_memref): Factor out parts into
3625         set_base_and_offset and call it.
3626
3627 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
3628
3629         PR middle-end/84406
3630         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3631         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3632         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
3633         search at the associated MODE_INT.
3634
3635 2018-02-20  Jeff Law  <law@redhat.com>
3636
3637         PR middle-end/82123
3638         PR tree-optimization/81592
3639         PR middle-end/79257
3640         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3641         for range data rather than using global data.
3642         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3643         range data rather than using global data.
3644         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3645         pass it to children as needed.
3646         (struct directive::fmtresult): Similarly.
3647         (struct directive::set_width): Similarly.
3648         (struct directive::set_precision): Similarly.
3649         (format_integer, format_directive, parse_directive): Similarly.
3650         (format_none): Accept unnamed vr_values parameter.
3651         (format_percent, format_floating, format_character): Similarly.
3652         (format_string, format_plain): Similarly.
3653         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3654         the EVRP range analyzer for range data rather than using global data.
3655         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3656         gimple-ssa-evrp-analyze.h
3657         (class sprintf_dom_walker): Add after_dom_children member function.
3658         Add evrp_range_analyzer member.
3659         (sprintf_dom_walker::before_dom_children): Call into the EVRP
3660         range analyzer as needed.
3661         (sprintf_dom_walker::after_dom_children): New member function.
3662         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3663         if not optimizing.
3664         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3665         (evrp_range_analyzer::pop_to_marker): Likewise.
3666
3667 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
3668
3669         PR tree-optimization/84419
3670         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3671         with the required type if its current type is compatible but
3672         different.
3673
3674 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
3675
3676         PR middle-end/82004
3677         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3678         after vectorization.
3679
3680 2018-02-20  Martin Liska  <mliska@suse.cz>
3681
3682         PR driver/83193
3683         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3684         possible values if we don't have a hint.
3685
3686 2018-02-20  Martin Liska  <mliska@suse.cz>
3687
3688         PR c/84310
3689         PR target/79747
3690         * final.c (shorten_branches): Build align_tab array with one
3691         more element.
3692         * opts.c (finish_options): Add alignment option limit check.
3693         (MAX_CODE_ALIGN): Likewise.
3694         (MAX_CODE_ALIGN_VALUE): Likewise.
3695         * doc/invoke.texi: Document maximum allowed option value for
3696         all -falign-* options.
3697
3698 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
3699
3700         PR target/84146
3701         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3702         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3703         * var-tracking.c (emit_note_insn_var_location): Remove all references
3704         to NOTE_INSN_CALL_ARG_LOCATION.
3705         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3706         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3707         Use copy_rtx_if_shared.
3708         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3709         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3710         (dwarf2out_var_location): Remove handling of
3711         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3712         on call_insn.
3713         * final.c (final_scan_insn): Remove all references to
3714         NOTE_INSN_CALL_ARG_LOCATION.
3715         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
3716         before dumping final insns.
3717         * except.c (emit_note_eh_region_end): Remove all references to
3718         NOTE_INSN_CALL_ARG_LOCATION.
3719         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3720         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3721         * config/arc/arc.c (hwloop_optimize): Likewise.
3722         * config/arm/arm.c (create_fix_barrier): Likewise.
3723         * config/s390/s390.c (s390_chunkify_start): Likewise.
3724         * config/sh/sh.c (find_barrier): Likewise.
3725         * config/i386/i386.c (rest_of_insert_endbranch,
3726         ix86_seh_fixup_eh_fallthru): Likewise.
3727         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3728         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3729         * config/frv/frv.c (frv_function_prologue): Likewise.
3730         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
3731         reg note.
3732         (note_outside_basic_block_p): Remove all references to
3733         NOTE_INSN_CALL_ARG_LOCATION.
3734         * gengtype.c (adjust_field_rtx_def): Likewise.
3735         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3736         Likewise.
3737         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3738         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3739
3740         PR c++/84444
3741         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3742         is ADDR_EXPR.
3743
3744         PR tree-optimization/84452
3745         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3746         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3747         is NULL.
3748
3749 2018-02-19  Martin Liska  <mliska@suse.cz>
3750
3751         PR sanitizer/82183
3752         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3753
3754 2018-02-19  Martin Liska  <mliska@suse.cz>
3755             Richard Sandiford  <richard.sandiford@linaro.org>
3756
3757         PR tree-optimization/82491
3758         * gimple-fold.c (get_base_constructor): Make earlier bail out
3759         to prevent ubsan.
3760
3761 2018-02-19  Carl Love  <cel@us.ibm.com>
3762
3763         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3764         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3765         BU_P8V_OVERLOAD_1.
3766         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3767         P8V_BUILTIN_VEC_NEG.
3768
3769 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
3770
3771         * config/rl78/rl78.md (movdf): New define expand.
3772
3773 2018-02-19  Martin Liska  <mliska@suse.cz>
3774
3775         PR other/80589
3776         * doc/invoke.texi: Fix typo.
3777         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3778
3779 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
3780
3781         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3782         handle rs6000_single_float and rs6000_double_float specially for
3783         e500 family CPUs.
3784
3785 2018-02-16  Jeff Law  <law@redhat.com>
3786
3787         * config/rx/rx.c (add_pop_cfi_notes): New function.;
3788         (pop_regs): Use it.
3789
3790 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
3791
3792         PR ipa/84425
3793         * ipa-inline.c (inline_small_functions): Fix a typo.
3794
3795 2018-02-16  Nathan Sidwell  <nathan@acm.org>
3796
3797         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3798
3799 2018-02-16  Carl Love  <cel@us.ibm.com>
3800
3801         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3802         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3803         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3804         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3805         expansion to P8V_BUILTIN_VEC_FLOAT2.
3806
3807 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
3808
3809         PR rtl-optimization/70023
3810         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3811         src_regno into account.
3812
3813 2018-02-16  Carl Love  <cel@us.ibm.com>
3814
3815         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3816         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3817         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3818         * config/rs6000/rs6000.c: Remove case statements for
3819         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3820         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3821         and P9V_BUILTIN_VEC_VINSERT4B.
3822         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3823         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3824         * config/rs6000/vsx.md:
3825         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3826         vec_insert4b.
3827
3828 2018-02-16  Carl Love  <cel@us.ibm.com>
3829
3830         * config/rs6000/altivec.h: Add builtin names vec_extract4b
3831         vec_insert4b.
3832         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3833         definitions.
3834         * config/rs6000/rs6000-c.c: Add the definitions for
3835         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3836         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3837         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3838         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
3839         definition for insert4b and define insn *insert3b_internal.
3840         * doc/extend.texi: Add documentation for vec_extract4b.
3841
3842 2018-02-16  Nathan Sidwell  <nathan@acm.org>
3843
3844         * doc/extend.texi (Backwards Compatibility): Mention friend
3845         injection.  Note for-scope is deprecated.
3846         * doc/invoke.texi (-ffriend-injection): Deprecate.
3847
3848 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
3849
3850         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3851         that moved to I2, also allow destinations that are a paradoxical
3852         subreg (instead of a normal reg).
3853
3854 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
3855
3856         PR target/83831
3857         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3858         to QImode.
3859
3860 2018-02-16  Richard Biener  <rguenther@suse.de>
3861
3862         PR tree-optimization/84037
3863         PR tree-optimization/84016
3864         PR target/82862
3865         * config/i386/i386.c (ix86_builtin_vectorization_cost):
3866         Adjust vec_construct for the fact we need additional higher latency
3867         128bit inserts for AVX256 and AVX512 vector builds.
3868         (ix86_add_stmt_cost): Scale vector construction cost for
3869         elementwise loads.
3870
3871 2018-02-16  Richard Biener  <rguenther@suse.de>
3872
3873         PR tree-optimization/84417
3874         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3875         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3876         (non_rewritable_lvalue_p): Likewise, use poly-ints.
3877
3878 2018-02-16  Martin Liska  <mliska@suse.cz>
3879
3880         PR sanitizer/84307
3881         * internal-fn.def (ASAN_CHECK): Set proper flags.
3882         (ASAN_MARK): Likewise.
3883
3884 2018-02-16  Julia Koval  <julia.koval@intel.com>
3885
3886         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3887         from PTA_CANNONLAKE.
3888
3889 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
3890
3891         PR target/84272
3892         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3893         Use ++iter rather than iter++ for std::list iterators.
3894         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
3895         defer deleting them until all nodes in the forest are processed.  Do
3896         free even leaf nodes.  Change to_process into auto_vec.
3897
3898         PR bootstrap/84405
3899         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3900         * vec.h (vec_default_construct): Use memset instead of placement new
3901         if BROKEN_VALUE_INITIALIZATION is defined.
3902         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3903         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3904         is defined.
3905
3906         PR rtl-optimization/83723
3907         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3908         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
3909         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
3910         recursive calls.
3911         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3912         callers.
3913         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3914
3915 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
3916
3917         PR rtl-optimization/81443
3918         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3919         from inner REGs to paradoxical SUBREGs.
3920
3921 2018-02-16  Richard Biener  <rguenther@suse.de>
3922
3923         PR tree-optimization/84399
3924         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3925         For operands we can analyze at their definition make sure we can
3926         analyze them at each use as well.
3927
3928 2018-02-16  Richard Biener  <rguenther@suse.de>
3929
3930         PR tree-optimization/84190
3931         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3932         volatile accesses if the decl isn't volatile.
3933
3934 2018-02-15  Jason Merrill  <jason@redhat.com>
3935
3936         PR c++/84314 - ICE with templates and fastcall attribute.
3937         * attribs.c (build_type_attribute_qual_variant): Don't clobber
3938         TYPE_CANONICAL on an existing type.
3939
3940 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
3941
3942         PR tree-optimization/84383
3943         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3944         dstoff nor call operand_equal_p if dstbase is NULL.
3945
3946         PR tree-optimization/84334
3947         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3948         also a CONSTANT_CLASS_P, punt.
3949
3950 2018-02-14  Jim Wilson  <jimw@sifive.com>
3951
3952         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3953         first SMALL_OPERAND check.  New local min_second_step.  Move assert
3954         to where locals are set.  Add TARGET_RVC support.
3955         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3956
3957 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
3958
3959         * doc/invoke.texi: Correct -Wformat-overflow code sample.
3960
3961 2018-02-14  Martin Sebor  <msebor@redhat.com>
3962
3963         PR tree-optimization/83698
3964         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3965         arrays constrain the offset range to their bounds.
3966         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3967         (builtin_access::overlap): Avoid setting the size of overlap if it's
3968         already been set.
3969         (maybe_diag_overlap): Also consider arrays when deciding what values
3970         of offsets to include in diagnostics.
3971
3972 2018-02-14  Martin Sebor  <msebor@redhat.com>
3973
3974         PR c/84108
3975         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3976         that correspond to the kind of a declaration.
3977
3978 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
3979
3980         PR target/83984
3981         * config/pa/pa.md: Load address of PIC label using the linkage table
3982         if the label is nonlocal.
3983
3984 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3985
3986         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3987         warning message if user requests -maltivec=be.
3988         * doc/invoke.texi: Document deprecation of -maltivec=be.
3989
3990 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
3991
3992         PR target/84220
3993         * config/rs6000/rs6000-c.c: Update definitions for
3994         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3995         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3996
3997 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
3998
3999         PR target/84239
4000         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4001         add _get_ssp intrinsics. Remove argument from
4002         __builtin_ia32_rdssp[d|q].
4003         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4004         * config/i386/i386-builtin.def: Remove argument from
4005         __builtin_ia32_rdssp[d|q].
4006         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4007         ix86_expand_special_args_builtin for _rdssp[d|q].
4008         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4009         Clear register before usage.
4010         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4011         Add documentation for new _get_ssp and _inc_ssp intrinsics.
4012
4013 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
4014
4015         PR tree-optimization/84357
4016         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4017         operand 1 of an ARRAY_REF too.
4018
4019 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
4020
4021         PR target/83831
4022         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4023         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4024         declarations.
4025         (set_of_reg): New struct.
4026         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4027         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4028         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4029         functions.
4030         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4031         Split into bitclr, bitset, bitinvert patterns if appropriate.
4032         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4033         use rx_fuse_in_memory_bitop.
4034         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4035         to named insn, correct maximum insn length.
4036
4037 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
4038
4039         PR target/79242
4040         * machmode.def: Define a complex mode for PARTIAL_INT.
4041         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4042         MODE_PARTIAL_INT.
4043         * doc/rtl.texi: Document CSPImode.
4044         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4045         handling.
4046         (msp430_hard_regno_nregs_with_padding): Likewise.
4047
4048 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
4049
4050         PR target/84279
4051         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4052
4053 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
4054
4055         PR rtl-optimization/84169
4056         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
4057         we generated a parallel as new i3 and we split that to new i2 and i3
4058         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
4059         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
4060         those to i2, not i1.  Partially rewrite this scan code.
4061
4062 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
4063
4064         PR c/82210
4065         * stor-layout.c (place_field): For variable length fields, adjust
4066         offset_align afterwards not just based on the field's alignment,
4067         but also on the size.
4068
4069         PR middle-end/84309
4070         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4071         of exps and logs in the use_exp2 case.
4072
4073 2018-02-13  Jeff Law  <law@redhat.com>
4074
4075         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4076         entry for "vector".
4077
4078         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4079         ARGS as unused.
4080
4081 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
4082
4083         PR debug/84342
4084         PR debug/84319
4085         * common.opt (gas-loc-support, gas-locview-support): New.
4086         (ginline-points, ginternal-reset-location-views): New.
4087         * doc/invoke.texi: Document them.  Use @itemx where intended.
4088         (gvariable-location-views): Adjust.
4089         * target.def (reset_location_view): New.
4090         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4091         (TARGET_RESET_LOCATION_VIEW): New.
4092         * doc/tm.texi: Rebuilt.
4093         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4094         (dwarf2out_default_as_locview_support): New.
4095         (output_asm_line_debug_info): Use option variables.
4096         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4097         (output_loc_list): Likewise.
4098         (add_high_low_attributes): Check option variables.
4099         Don't output entry view attribute in strict mode.
4100         (gen_inlined_subroutine_die): Check option variables.
4101         (dwarf2out_inline_entry): Likewise.
4102         (init_sections_and_labels): Likewise.
4103         (dwarf2out_early_finish): Likewise.
4104         (maybe_reset_location_view): New, from...
4105         (dwarf2out_var_location): ... here.  Call it.
4106         * debug.h (dwarf2out_default_as_loc_support): Declare.
4107         (dwarf2out_default_as_locview_support): Declare.
4108         * hooks.c (hook_int_rtx_insn_0): New.
4109         * hooks.h (hook_int_rtx_insn_0): Declare.
4110         * toplev.c (process_options): Take -gas-loc-support and
4111         -gas-locview-support from dwarf2out.  Enable
4112         -gvariable-location-views by default only with locview
4113         assembler support.  Enable -ginternal-reset-location-views by
4114         default only if the target defines the corresponding hook.
4115         Enable -ginline-points by default if location views are
4116         enabled; force it disabled if statement frontiers are
4117         disabled.
4118         * tree-inline.c (expand_call_inline): Check option variables.
4119         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4120
4121 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
4122
4123         PR tree-optimization/84321
4124         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4125         handling.  Also check whether the anti-range contains any values
4126         that satisfy the mask; switch to a VR_RANGE if not.
4127
4128 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
4129
4130         PR sanitizer/84340
4131         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4132
4133 2018-02-13  Martin Jambor  <mjambor@suse.cz>
4134
4135         PR c++/83990
4136         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4137         of call statements, also set location of a load to a temporary.
4138
4139 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
4140
4141         * config/rl78/rl78.c (add_vector_labels): New function.
4142         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4143         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4144         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4145         which checks that no arguments are passed.
4146         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4147         * doc/extend.texi: Documentation for the new attribute.
4148
4149 2018-02-13  Andreas Schwab  <schwab@suse.de>
4150
4151         * config/riscv/linux.h (CPP_SPEC): Define.
4152
4153 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
4154
4155         PR target/84335
4156         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4157         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4158         OPTION_MASK_ISA_AES as first argument to def_builtin_const
4159         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4160         instead of OPTION_MASK_ISA_PCLMUL as first argument to
4161         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4162         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4163         temporarily for AES and PCLMUL builtins.
4164
4165         PR tree-optimization/84339
4166         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4167         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4168         Formatting fixes.
4169
4170         PR middle-end/84309
4171         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4172         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4173         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4174         inline function.
4175         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4176         inline function.
4177         * omp-simd-clone.h: New file.
4178         * omp-simd-clone.c: Include omp-simd-clone.h.
4179         (expand_simd_clones): No longer static.
4180         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4181         cgraph.h and omp-simd-clone.h.
4182         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4183         (vect_recog_widen_shift_pattern): Formatting fix.
4184         (vect_pattern_recog_1): Don't check optab for calls.
4185
4186         PR target/84336
4187         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4188         operands[2] into a REG before using gen_lowpart on it.
4189
4190 2018-02-12  Jeff Law  <law@redhat.com>
4191
4192         PR target/83760
4193         * config/sh/sh.c (find_barrier): Consider a sibling call
4194         a barrier as well.
4195
4196         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4197         successfully back substituting a reg.
4198
4199 2018-02-12  Richard Biener  <rguenther@suse.de>
4200
4201         PR tree-optimization/84037
4202         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4203         parameter, move visited init to caller.
4204         (vect_slp_analyze_operations): Separate cost from validity
4205         check, initialize visited once for all instances.
4206         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4207         for all instances.
4208         * tree-vect-stmts.c (vect_model_simple_cost): Make early
4209         out an assert.
4210         (vect_model_promotion_demotion_cost): Likewise.
4211         (vectorizable_bswap): Guard cost modeling with !slp_node
4212         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4213         SLP stmts.
4214         (vectorizable_call): Likewise.
4215         (vectorizable_conversion): Likewise.
4216         (vectorizable_assignment): Likewise.
4217         (vectorizable_shift): Likewise.
4218         (vectorizable_operation): Likewise.
4219         (vectorizable_store): Likewise.
4220         (vectorizable_load): Likewise.
4221         (vectorizable_condition): Likewise.
4222         (vectorizable_comparison): Likewise.
4223
4224 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
4225
4226         PR sanitizer/84307
4227         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4228         (ASAN_MARK): Fix fnspec to account for return value, change pointer
4229         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4230
4231 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
4232
4233         PR middle-end/83665
4234         * params.def (inline-min-speedup): Increase from 8 to 15.
4235         (max-inline-insns-auto): Decrease from 40 to 30.
4236         * ipa-split.c (consider_split): Add some buffer for function to
4237         be considered inlining candidate.
4238         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4239         default values.
4240
4241 2018-02-12  Richard Biener  <rguenther@suse.de>
4242
4243         PR tree-optimization/84037
4244         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4245         matched stmts if we cannot swap the non-matched ones.
4246
4247 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
4248
4249         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4250         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4251         _mm_maskz_scalef_round_ss): New intrinsics.
4252         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4253         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4254         __builtin_ia32_scalefss_round): Remove.
4255         (__builtin_ia32_scalefsd_mask_round,
4256         __builtin_ia32_scalefss_mask_round): New intrinsics.
4257         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4258         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4259         ((match_operand:VF_128 2 "<round_nimm_predicate>"
4260         "<round_constraint>")): Changed to ...
4261         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4262         "<round_scalar_constraint>")): ... this.
4263         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4264         %0, %1, %2<round_op3>}"): Changed to ...
4265         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4266         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4267         %2<round_scalar_mask_op3>}"): ... this.
4268         * config/i386/subst.md (round_scalar_nimm_predicate): New.
4269
4270 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
4271
4272         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4273         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4274         (_mm_maskz_sqrt_round_ss): New intrinsics.
4275         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4276         (__builtin_ia32_sqrtsd_mask_round)
4277         (__builtin_ia32_sqrtss_mask_round): New builtins.
4278         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4279         (__builtin_ia32_sqrtss_round): Remove.
4280         (__builtin_ia32_sqrtsd_mask_round)
4281         (__builtin_ia32_sqrtss_mask_round): New builtins.
4282         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4283         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4284         ((match_operand:VF_128 1 "vector_operand"
4285         "xBm,<round_constraint>")): Changed to ...
4286         ((match_operand:VF_128 1 "vector_operand"
4287         "xBm,<round_scalar_constraint>")): ... this.
4288         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4289         %0, %2, %<iptr>1<round_op3>}): Changed to ...
4290         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4291         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4292         %<iptr>1<round_scalar_mask_op3>}): ... this.
4293         ((set_attr "prefix" "<round_prefix>")): Changed to ...
4294         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4295
4296 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
4297
4298         PR target/84266
4299         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4300         Cast vec_cmpeq result to correct type.
4301         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4302         Cast vec_cmpgt result to correct type.
4303
4304 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
4305
4306         * final.c (final_scan_insn_1): Renamed from...
4307         (final_scan_insn): ... this.  New wrapper, to recover
4308         seen from the outermost call in recursive ones.
4309         * config/sparc/sparc.c (output_return): Drop seen from call.
4310         (output_sibcall): Likewise.
4311         * config/visium/visium.c (output_branch): Likewise.
4312
4313 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
4314
4315         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4316         function label.
4317
4318 2018-02-10  Alan Modra  <amodra@gmail.com>
4319
4320         PR target/84300
4321         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4322         Specify LR as an input.
4323
4324 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
4325
4326         PR sanitizer/83987
4327         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4328         remove_member_access_dummy_vars): New functions.
4329         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4330         lower_omp_1, execute_lower_omp): Use them.
4331
4332         PR rtl-optimization/84308
4333         * shrink-wrap.c (spread_components): Release todo vector.
4334
4335 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
4336
4337         PR rtl-optimization/57193
4338         * ira-color.c (struct allocno_color_data): Add member
4339         conflict_allocno_hard_prefs.
4340         (update_conflict_allocno_hard_prefs): New.
4341         (bucket_allocno_compare_func): Add a preference based on
4342         conflict_allocno_hard_prefs.
4343         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4344         (color_allocnos): Remove a dead code.  Initiate
4345         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
4346
4347 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
4348
4349         PR target/84226
4350         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4351         constraint from =wa to wa.  Avoid a subreg on the output operand,
4352         instead use a pseudo and subreg it in a move.
4353         (p9_xxbrd_<mode>): Changed to ...
4354         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4355         (p9_xxbrd_v2df): New expander.
4356         (p9_xxbrw_<mode>): Changed to ...
4357         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4358         (p9_xxbrw_v4sf): New expander.
4359
4360 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
4361
4362         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4363
4364 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
4365
4366         PR target/83926
4367         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4368         multiply in 32-bit mode.
4369         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4370         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4371         mode.
4372
4373 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
4374
4375         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4376         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4377         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4378         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4379
4380 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
4381
4382         PR lto/84213
4383         * dwarf2out.c (is_trivial_indirect_ref): New function.
4384         (dwarf2out_late_global_decl): Do not generate a location
4385         attribute for variables that have a non-trivial DECL_VALUE_EXPR
4386         and that are not defined in the current unit.
4387
4388 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
4389
4390         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4391         instead of a libcall for UNORDERED.
4392
4393 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
4394
4395         PR target/82641
4396         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4397         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4398
4399 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4400
4401         PR target/PR84295
4402         * config/s390/s390.c (s390_set_current_function): Invoke
4403         s390_indirect_branch_settings also if fndecl didn't change.
4404
4405 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
4406
4407         * config/rs6000/rs6000.md (blockage): Set length to zero.
4408
4409 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
4410
4411         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4412
4413 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
4414
4415         PR sanitizer/84285
4416         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4417         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4418         -static-lib*san.
4419
4420         PR debug/84252
4421         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4422         PARALLEL incoming that failed vt_get_decl_and_offset check.
4423
4424         PR middle-end/84237
4425         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4426         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4427         TREE_READONLY bit.
4428         (get_variable_section): For decls in named .bss* sections pass true as
4429         second argument to bss_initializer_p.
4430
4431 2018-02-09  Marek Polacek  <polacek@redhat.com>
4432             Jakub Jelinek  <jakub@redhat.com>
4433
4434         PR c++/83659
4435         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4436         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
4437         Sync some changes from cxx_fold_indirect_ref.
4438
4439 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
4440
4441         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4442         markers.
4443         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4444         (BLOCK_INLINE_ENTRY_LABEL): New.
4445         (dwarf2out_var_location): Disregard inline entry markers.
4446         (inline_entry_data): New struct.
4447         (inline_entry_data_hasher): New hashtable type.
4448         (inline_entry_data_hasher::hash): New.
4449         (inline_entry_data_hasher::equal): New.
4450         (inline_entry_data_table): New variable.
4451         (add_high_low_attributes): Add DW_AT_entry_pc and
4452         DW_AT_GNU_entry_view attributes if a pending entry is found
4453         in inline_entry_data_table.  Add old entry_pc attribute only
4454         if debug nonbinding markers are disabled.
4455         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4456         markers are enabled.
4457         (block_within_block_p, dwarf2out_inline_entry): New.
4458         (dwarf2out_finish): Check that no entries remained in
4459         inline_entry_data_table.
4460         * final.c (reemit_insn_block_notes): Handle inline entry notes.
4461         (final_scan_insn, notice_source_line): Likewise.
4462         (rest_of_clean_state): Skip inline entry markers.
4463         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4464         markers.
4465         * gimple.c (gimple_build_debug_inline_entry): New.
4466         * gimple.h (enum gimple_debug_subcode): Add
4467         GIMPLE_DEBUG_INLINE_ENTRY.
4468         (gimple_build_debug_inline_entry): Declare.
4469         (gimple_debug_inline_entry_p): New.
4470         (gimple_debug_nonbind_marker_p): Adjust.
4471         * insn-notes.def (INLINE_ENTRY): New.
4472         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4473         inline entry marker notes.
4474         (print_insn): Likewise.
4475         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4476         (INSN_DEBUG_MARKER_KIND): Likewise.
4477         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4478         * tree-inline.c (expand_call_inline): Build and insert
4479         debug_inline_entry stmt.
4480         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4481         inline entry blocks early, if nonbind markers are enabled.
4482         (dump_scope_block): Dump fragment info.
4483         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4484         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4485         (gimple_build_debug_inline_entry): New.
4486         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4487         Enable/disable inline entry points too.
4488         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4489         (DEBUG_INSN): Describe inline entry markers.
4490
4491         * common.opt (gvariable-location-views): New.
4492         (gvariable-location-views=incompat5): New.
4493         * config.in: Rebuilt.
4494         * configure: Rebuilt.
4495         * configure.ac: Test assembler for view support.
4496         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4497         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4498         * dwarf2out.c (var_loc_view): New typedef.
4499         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4500         (dwarf2out_locviews_in_attribute): New.
4501         (dwarf2out_locviews_in_loclist): New.
4502         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4503         (enum dw_line_info_opcode): Add LI_adv_address.
4504         (struct dw_line_info_table): Add view.
4505         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4506         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4507         (zero_view_p): New variable.
4508         (ZERO_VIEW_P): New macro.
4509         (output_asm_line_debug_info): New.
4510         (struct var_loc_node): Add view.
4511         (add_AT_view_list, AT_loc_list): New.
4512         (add_var_loc_to_decl): Add view param.  Test it against last.
4513         (new_loc_list): Add view params.  Record them.
4514         (AT_loc_list_ptr): Handle loc and view lists.
4515         (view_list_to_loc_list_val_node): New.
4516         (print_dw_val): Handle dw_val_class_view_list.
4517         (size_of_die): Likewise.
4518         (value_format): Likewise.
4519         (loc_list_has_views): New.
4520         (gen_llsym): Set vl_symbol too.
4521         (maybe_gen_llsym, skip_loc_list_entry): New.
4522         (dwarf2out_maybe_output_loclist_view_pair): New.
4523         (output_loc_list): Output view list or entries too.
4524         (output_view_list_offset): New.
4525         (output_die): Handle dw_val_class_view_list.
4526         (output_dwarf_version): New.
4527         (output_compilation_unit_header): Use it.
4528         (output_skeleton_debug_sections): Likewise.
4529         (output_rnglists, output_line_info): Likewise.
4530         (output_pubnames, output_aranges): Update version comments.
4531         (output_one_line_info_table): Output view numbers in asm comments.
4532         (dw_loc_list): Determine current endview, pass it to new_loc_list.
4533         Call maybe_gen_llsym.
4534         (loc_list_from_tree_1): Adjust.
4535         (add_AT_location_description): Create view list attribute if
4536         needed, check it's absent otherwise.
4537         (convert_cfa_to_fb_loc_list): Adjust.
4538         (maybe_emit_file): Call output_asm_line_debug_info for test.
4539         (dwarf2out_var_location): Reset views as needed.  Precompute
4540         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
4541         attribute.  Set view.
4542         (new_line_info_table): Reset next view.
4543         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4544         (dwarf2out_source_line): Likewise.  Output view resets and labels to
4545         the assembler, or select appropriate line info opcodes.
4546         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4547         (optimize_string_length): Catch it.  Adjust.
4548         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
4549         dw_val_class_view_list, and remove it if no longer needed.
4550         (hash_loc_list): Hash view numbers.
4551         (loc_list_hasher::equal): Compare them.
4552         (optimize_location_lists): Check whether a view list symbol is
4553         needed, and whether the locview attribute is present, and
4554         whether they match.  Remove the locview attribute if no longer
4555         needed.
4556         (index_location_lists): Call skip_loc_list_entry for test.
4557         (dwarf2out_finish): Call output_asm_line_debug_info for test.
4558         Use output_dwarf_version.
4559         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4560         (struct dw_val_node): Add val_view_list.
4561         * final.c (SEEN_NEXT_VIEW): New.
4562         (set_next_view_needed): New.
4563         (clear_next_view_needed): New.
4564         (maybe_output_next_view): New.
4565         (final_start_function): Rename to...
4566         (final_start_function_1): ... this.  Take pointer to FIRST,
4567         add SEEN parameter.  Emit param bindings in the initial view.
4568         (final_start_function): Reintroduce SEEN-less interface.
4569         (final): Rename to...
4570         (final_1): ... this.  Take SEEN parameter.  Output final pending
4571         next view at the end.
4572         (final): Reintroduce seen-less interface.
4573         (final_scan_insn): Output pending next view before switching
4574         sections or ending a block.  Mark the next view as needed when
4575         outputting variable locations.  Notify debug backend of section
4576         changes, and of location view changes.
4577         (rest_of_handle_final): Adjust.
4578         * toplev.c (process_options): Autodetect value for debug variable
4579         location views option.  Warn on incompat5 without -gdwarf-5.
4580         * doc/invoke.texi (gvariable-location-views): New.
4581         (gvariable-location-views=incompat5): New.
4582         (gno-variable-location-views): New.
4583
4584 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
4585
4586         PR tree-optimization/84136
4587         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4588         that the result of find_edge is non-NULL.
4589
4590 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
4591
4592         PR target/83008
4593         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4594         storing integer register in SImode.  Fix cost of 256 and 512
4595         byte aligned SSE register store.
4596
4597 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
4598
4599         * config/i386/i386.c (ix86_multiplication_cost): Fix
4600         multiplication cost for TARGET_AVX512DQ.
4601
4602 2018-02-08  Marek Polacek  <polacek@redhat.com>
4603
4604         PR tree-optimization/84238
4605         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4606         get_range_strlen.
4607
4608 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
4609
4610         PR tree-optimization/84265
4611         * tree-vect-stmts.c (vectorizable_store): Don't treat
4612         VMAT_CONTIGUOUS accesses as grouped.
4613         (vectorizable_load): Likewise.
4614
4615 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
4616
4617         PR tree-optimization/81635
4618         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4619         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4620         (test_round_for_mask): New functions.
4621         (wide_int_cc_tests): Call test_round_for_mask.
4622         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4623         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4624         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4625         range returned by get_range_info.
4626
4627 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
4628
4629         PR ipa/81360
4630         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4631         * symtab.c: Include builtins.h
4632         (symtab_node::output_to_lto_symbol_table_p): Move here
4633         from lto-streamer-out.c:output_symbol_p.
4634         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4635         (output_symbol_p): Move all logic to symtab.c
4636         (produce_symtab): Update.
4637
4638 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4639
4640         * config/s390/s390-opts.h (enum indirect_branch): Define.
4641         * config/s390/s390-protos.h (s390_return_addr_from_memory)
4642         (s390_indirect_branch_via_thunk)
4643         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4644         (enum s390_indirect_branch_type): Define.
4645         * config/s390/s390.c (struct s390_frame_layout, struct
4646         machine_function): Remove.
4647         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4648         (indirect_branch_table_label_no, indirect_branch_table_name):
4649         Define variables.
4650         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4651         (enum s390_indirect_branch_option): Define.
4652         (s390_return_addr_from_memory): New function.
4653         (s390_handle_string_attribute): New function.
4654         (s390_attribute_table): Add new attribute handler.
4655         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4656         (s390_indirect_branch_via_thunk): New function.
4657         (s390_indirect_branch_via_inline_thunk): New function.
4658         (s390_function_ok_for_sibcall): When jumping via thunk disallow
4659         sibling call optimization for non z10 compiles.
4660         (s390_emit_call): Force indirect branch target to be a single
4661         register.  Add r1 clobber for non-z10 compiles.
4662         (s390_emit_epilogue): Emit return jump via return_use expander.
4663         (s390_reorg): Handle JUMP_INSNs as execute targets.
4664         (s390_option_override_internal): Perform validity checks for the
4665         new command line options.
4666         (s390_indirect_branch_attrvalue): New function.
4667         (s390_indirect_branch_settings): New function.
4668         (s390_set_current_function): Invoke s390_indirect_branch_settings.
4669         (s390_output_indirect_thunk_function):  New function.
4670         (s390_code_end): Implement target hook.
4671         (s390_case_values_threshold): Implement target hook.
4672         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4673         macros.
4674         * config/s390/s390.h (struct s390_frame_layout)
4675         (struct machine_function): Move here from s390.c.
4676         (TARGET_INDIRECT_BRANCH_NOBP_RET)
4677         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4678         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4679         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4680         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4681         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4682         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4683         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4684         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4685         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4686         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4687         (mnemonic attribute): Add values which aren't recognized
4688         automatically.
4689         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4690         pattern for branch conversion.  Fix mnemonic attribute.
4691         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4692         indirect branch via thunk if requested.
4693         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4694         ("*indirect_jump"): Disable for branch conversion using out of
4695         line thunks.
4696         ("indirect_jump_via_thunk<mode>_z10")
4697         ("indirect_jump_via_thunk<mode>")
4698         ("indirect_jump_via_inlinethunk<mode>_z10")
4699         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4700         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4701         ("casesi_jump_via_inlinethunk<mode>_z10")
4702         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4703         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4704         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4705         ("*indirect2_jump"): Disable for branch conversion.
4706         ("casesi_jump"): Turn into expander and expand patterns for branch
4707         conversion.
4708         ("return_use"): New expander.
4709         ("*return"): Emit return via thunk and rename it to ...
4710         ("*return<mode>"): ... this one.
4711         * config/s390/s390.opt: Add new options and and enum for the
4712         option values.
4713
4714 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
4715
4716         * lra-constraints.c (match_reload): Unconditionally use
4717         gen_lowpart_SUBREG, rather than selecting between that
4718         and equivalent gen_rtx_SUBREG code.
4719
4720 2018-02-08  Richard Biener  <rguenther@suse.de>
4721
4722         PR tree-optimization/84233
4723         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4724         changed flag instead of boguously re-using phi_inserted.
4725
4726 2018-02-08  Martin Jambor  <mjambor@suse.cz>
4727
4728         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4729         static local variables.
4730
4731 2018-02-08  Richard Biener  <rguenther@suse.de>
4732
4733         PR tree-optimization/84278
4734         * tree-vect-stmts.c (vectorizable_store): When looking for
4735         smaller vector types to perform grouped strided loads/stores
4736         make sure the mode is supported by the target.
4737         (vectorizable_load): Likewise.
4738
4739 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
4740
4741         * config/aarch64/aarch64.c (aarch64_components_for_bb):
4742         Increase LDP/STP opportunities by adding adjacent callee-saves.
4743
4744 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
4745
4746         PR rtl-optimization/84068
4747         PR rtl-optimization/83459
4748         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4749
4750 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
4751
4752         PR tree-optimization/84224
4753         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4754         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4755         non-zero arguments.
4756
4757 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
4758
4759         PR target/84113
4760         * config/rs6000/altivec.md (*restore_world): Remove LR use.
4761         * config/rs6000/predicates.md (restore_world_operation): Adjust op
4762         count, remove one USE.
4763
4764 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
4765
4766         * doc/install.texi (Configuration): Document the
4767         --with-long-double-format={ibm,ieee} PowerPC configuration
4768         options.
4769
4770         PR target/84154
4771         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4772         Convert from define_expand to be define_insn_and_split.  Rework
4773         float/double/_Float128 conversions to QI/HI/SImode to work with
4774         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
4775         conversions to QI/HImode types did a store and then a load to
4776         truncate the value.  For conversions to VSX registers, don't split
4777         the insn, instead emit the code directly.  Use the code iterator
4778         any_fix to combine signed and unsigned conversions.
4779         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4780         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4781         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4782         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4783         (fix_<mode>di2_hw): Likewise.
4784         (fixuns_<mode>di2_hw): Likewise.
4785         (fix_<mode>si2_hw): Likewise.
4786         (fixuns_<mode>si2_hw): Likewise.
4787         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4788         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4789         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4790         fix<uns>_trunc<SFDF:mode>si2_p8.
4791         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4792         used.
4793         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4794         (fix<uns>_<mode>_mem): Likewise.
4795         (fctiw<u>z_<mode>_mem): Likewise.
4796         (fix<uns>_<mode>_mem): Likewise.
4797         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4798         the register allocator from doing a direct move to the GPRs to do
4799         a store, and instead use the ISA 3.0 store byte/half-word from
4800         vector register instruction.  For IEEE 128-bit floating point,
4801         also optimize stores of 32-bit ints.
4802         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4803
4804 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
4805
4806         * genextract.c (push_pathstr_operand): New function to support
4807         [a-zA-Z].
4808         (walk_rtx): Call push_pathstr_operand.
4809         (print_path): Support [a-zA-Z].
4810
4811 2018-02-07  Richard Biener  <rguenther@suse.de>
4812
4813         PR tree-optimization/84037
4814         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4815         (cse_and_gimplify_to_preheader): Declare.
4816         (vect_get_place_in_interleaving_chain): Likewise.
4817         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4818         ivexpr_map.
4819         (_loop_vec_info::~_loop_vec_info): Delete it.
4820         (cse_and_gimplify_to_preheader): New function.
4821         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4822         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4823         (vectorizable_load): Likewise.  For grouped stores always base
4824         the IV on the first element.
4825         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4826         condition before gimplifying.
4827
4828 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
4829
4830         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4831         *DIV_EXPR and *MOD_EXPR.
4832
4833 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
4834
4835         PR target/84248
4836         * config/i386/i386.c (ix86_option_override_internal): Mask out
4837         the CF_SET bit when checking -fcf-protection.
4838
4839 2018-02-07  Tom de Vries  <tom@codesourcery.com>
4840
4841         PR libgomp/84217
4842         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4843         enough.
4844
4845 2018-02-07  Richard Biener  <rguenther@suse.de>
4846
4847         PR tree-optimization/84204
4848         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4849         this place.
4850
4851         PR tree-optimization/84205
4852         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4853         special-case isl_ast_op_zdiv_r.
4854
4855         PR tree-optimization/84223
4856         * graphite-scop-detection.c (gather_bbs::before_dom_children):
4857         Only add conditions from within the region.
4858         (gather_bbs::after_dom_children): Adjust.
4859
4860 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
4861
4862         PR target/84209
4863         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4864         * config/avr/avr.md: Only post-reload split REG-REG moves if
4865         either register is GENERAL_REG_P.
4866
4867 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
4868
4869         PR tree-optimization/84235
4870         * tree-ssa-scopedtables.c
4871         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4872         if the subtraction is performed in floating point type where NaNs are
4873         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4874         build 1.  Formatting fix.
4875
4876 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
4877
4878         PR target/84146
4879         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4880         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4881         and skip it regardless of bb boundaries.  Use CALL_P macro,
4882         don't test INSN_P (insn) together with CALL_P or JUMP_P check
4883         unnecessarily, formatting fix.
4884
4885 2018-02-06  Michael Collison  <michael.collison@arm.com>
4886
4887         * config/arm/thumb2.md:
4888         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4889         (*thumb_mov_notscc): Ditto.
4890
4891 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
4892
4893         PR target/84154
4894         * config/rs6000/rs6000.md (su code attribute): Use "u" for
4895         unsigned_fix, not "s".
4896
4897 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4898
4899         * configure.ac (gcc_fn_eh_frame_ro): New function.
4900         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4901         correct .eh_frame permissions.
4902         * configure: Regenerate.
4903
4904 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
4905
4906         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4907         irrelevant options.
4908
4909 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4910
4911         * config/rs6000/rs6000.c (rs6000_option_override_internal):
4912         Display warning message for -mno-speculate-indirect-jumps.
4913
4914 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
4915
4916         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4917         Undocumented.
4918         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4919
4920 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
4921
4922         PR tree-optimization/84225
4923         * tree-eh.c (find_trapping_overflow): Only call
4924         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4925
4926 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
4927
4928         PR target/84145
4929         * config/i386/i386.c: Reimplement the check of possible options
4930         -mibt/-mshstk conbination. Change error messages.
4931         * doc/invoke.texi: Fix a typo: remove extra '='.
4932
4933 2018-02-06  Marek Polacek  <polacek@redhat.com>
4934
4935         PR tree-optimization/84228
4936         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4937
4938 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
4939
4940         PR target/82641
4941         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4942         emitted arch directives.
4943         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4944         __ARM_FEATURE_COPROC before changing architectures.
4945
4946 2018-02-06  Richard Biener  <rguenther@suse.de>
4947
4948         * config/i386/i386.c (print_reg): Fix typo.
4949         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4950
4951 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
4952
4953         * configure: Regenerate.
4954
4955 2018-02-05  Martin Sebor  <msebor@redhat.com>
4956
4957         PR tree-optimization/83369
4958         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4959         inlining context.
4960
4961 2018-02-05  Martin Liska  <mliska@suse.cz>
4962
4963         * doc/invoke.texi: Cherry-pick upstream r323995.
4964
4965 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
4966
4967         * ira.c (ira_init_register_move_cost): Adjust comment.
4968
4969 2018-02-05  Martin Liska  <mliska@suse.cz>
4970
4971         PR gcov-profile/84137
4972         * doc/gcov.texi: Fix typo in documentation.
4973
4974 2018-02-05  Martin Liska  <mliska@suse.cz>
4975
4976         PR gcov-profile/83879
4977         * doc/gcov.texi: Document necessity of --dynamic-list-data when
4978         using dlopen functionality.
4979
4980 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
4981
4982         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4983         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4984         _mm_maskz_range_ss, _mm_mask_range_round_ss,
4985         _mm_maskz_range_round_ss): New intrinsics.
4986         (__builtin_ia32_rangesd128_round)
4987         (__builtin_ia32_rangess128_round): Remove.
4988         (__builtin_ia32_rangesd128_mask_round,
4989         __builtin_ia32_rangess128_mask_round): New builtins.
4990         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4991         __builtin_ia32_rangess128_round): Remove.
4992         (__builtin_ia32_rangesd128_mask_round,
4993         __builtin_ia32_rangess128_mask_round): New builtins.
4994         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4995         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4996         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4997         "<round_saeonly_constraint>")): Changed to ...
4998         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4999         "<round_saeonly_scalar_constraint>")): ... this.
5000         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5001         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5002         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5003         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5004         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5005
5006 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
5007
5008         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5009         options.
5010         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5011         Remove all values except native, 8540 and 8548.
5012
5013 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
5014
5015         * config/i386/i386.c (ix86_output_function_return): Pass
5016         INVALID_REGNUM, instead of -1, as invalid register number to
5017         indirect_thunk_name and output_indirect_thunk.
5018
5019 2018-02-02  Julia Koval  <julia.koval@intel.com>
5020
5021         * config.gcc: Add -march=icelake.
5022         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5023         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5024         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5025         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5026         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5027         (processor_target_table): Add icelake.
5028         (ix86_option_override_internal): Handle new PTAs.
5029         (get_builtin_code_for_version): Handle icelake.
5030         (M_INTEL_COREI7_ICELAKE): New.
5031         (fold_builtin_cpu): Handle icelake.
5032         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5033         * doc/invoke.texi: Add -march=icelake.
5034
5035 2018-02-02  Julia Koval  <julia.koval@intel.com>
5036
5037         * config/i386/i386.c (ix86_option_override_internal): Change flags type
5038         to wide_int_bitmask.
5039         * wide-int-bitmask.h: New.
5040
5041 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5042
5043         PR target/84066
5044         * config/i386/i386.md: Replace Pmode with word_mode in
5045         builtin_setjmp_setup and builtin_longjmp to support x32.
5046
5047 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
5048
5049         PR target/56010
5050         PR target/83743
5051         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5052         #include "opts.h".
5053         (rs6000_supported_cpu_names): New static variable.
5054         (linux_cpu_translation_table): Likewise.
5055         (elf_platform) <cpu>: Define new static variable and use it.
5056         Translate kernel AT_PLATFORM name to canonical name if needed.
5057         Error if platform name is unknown.
5058
5059 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
5060
5061         PR target/84089
5062         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5063
5064 2018-02-01  Jeff Law  <law@redhat.com>
5065
5066         PR target/84128
5067         * config/i386/i386.c (release_scratch_register_on_entry): Add new
5068         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
5069         the scratch if RELEASE_VIA_POP is false.
5070         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5071         If we have to save a temporary register, decrement SIZE appropriately.
5072         Pass new arguments to release_scratch_register_on_entry.
5073         (ix86_adjust_stack_and_probe): Likewise.
5074         (ix86_emit_probe_stack_range): Pass new arguments to
5075         release_scratch_register_on_entry.
5076
5077 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
5078
5079         PR rtl-optimization/84157
5080         * combine.c (change_zero_ext): Use REG_P predicate in
5081         front of HARD_REGISTER_P predicate.
5082
5083 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
5084
5085         * config/avr/avr.c (avr_option_override): Move disabling of
5086         -fdelete-null-pointer-checks to...
5087         * common/config/avr/avr-common.c (avr_option_optimization_table):
5088         ...here.
5089
5090 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
5091
5092         PR tree-optimization/81635
5093         * tree-data-ref.c (split_constant_offset_1): For types that
5094         wrap on overflow, try to use range info to prove that wrapping
5095         cannot occur.
5096
5097 2018-02-01  Renlin Li  <renlin.li@arm.com>
5098
5099         PR target/83370
5100         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5101         TAILCALL_ADDR_REGS.
5102         (aarch64_register_move_cost): Likewise.
5103         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5104         TAILCALL_ADDR_REGS.
5105         (REG_CLASS_NAMES): Likewise.
5106         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5107         TAILCALL_ADDR_REGS. Remove IP registers.
5108         * config/aarch64/aarch64.md (Ucs): Update register constraint.
5109
5110 2018-02-01  Richard Biener  <rguenther@suse.de>
5111
5112         * domwalk.h (dom_walker::dom_walker): Add additional constructor
5113         for specifying RPO order and allow NULL for that.
5114         * domwalk.c (dom_walker::dom_walker): Likewise.
5115         (dom_walker::walk): Handle NULL RPO order.
5116         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5117         in RPO order.
5118         (rewrite_update_dom_walker): Likewise.
5119         (mark_def_dom_walker): Likewise.
5120
5121 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
5122
5123         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5124         (aarch64_maybe_expand_sve_subreg_move): Declare.
5125         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5126         * config/aarch64/predicates.md (aarch64_any_register_operand): New
5127         predicate.
5128         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5129         that are semantically a reverse operation.
5130         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5131         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5132         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5133         functions.
5134         (aarch64_can_change_mode_class): For big-endian, forbid changes
5135         between two SVE modes if they have different element sizes.
5136
5137 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
5138
5139         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5140         the TImode handling for big-endian targets.
5141
5142 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
5143
5144         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5145         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
5146         not just bytes.
5147         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5148         Remove BSWAP handing for big-endian targets and use the form of
5149         LD1RQ appropariate for the mode.
5150
5151 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
5152
5153         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5154         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5155         duplicated element.
5156
5157 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
5158
5159         PR tearget/83845
5160         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5161         check for operands that need to go through aarch64_sve_reload_be.
5162
5163 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
5164
5165         PR tree-optimization/81661
5166         PR tree-optimization/84117
5167         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5168         * tree-eh.c: Include gimplify.h.
5169         (find_trapping_overflow, replace_trapping_overflow,
5170         rewrite_to_non_trapping_overflow): New functions.
5171         * tree-vect-loop.c: Include tree-eh.h.
5172         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5173         * tree-data-ref.c: Include tree-eh.h.
5174         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5175
5176 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
5177
5178         PR rtl-optimization/84123
5179         * combine.c (change_zero_ext): Check if hard register satisfies
5180         can_change_dest_mode before calling gen_lowpart_SUBREG.
5181
5182 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
5183
5184         PR target/82444
5185         * ira.c (ira_init_register_move_cost): Remove assert.
5186
5187 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
5188
5189         PR rtl-optimization/84071
5190         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5191         * doc/tm.texi: Regenerate.
5192
5193 2018-01-31  Richard Biener  <rguenther@suse.de>
5194
5195         PR tree-optimization/84132
5196         * tree-data-ref.c (analyze_miv_subscript): Properly
5197         check whether evolution_function_is_affine_multivariate_p
5198         before calling gcd_of_steps_may_divide_p.
5199
5200 2018-01-31  Julia Koval  <julia.koval@intel.com>
5201
5202         PR target/83618
5203         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5204         * config/i386/i386.md (rdpid_rex64) New.
5205         (rdpid): Make 32bit only.
5206
5207 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
5208
5209         PR lto/84105
5210         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5211         an IDENTIFIER_NODE for FUNCTION_TYPE's.
5212
5213 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
5214
5215         Revert
5216         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
5217
5218         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5219
5220 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
5221
5222         PR rtl-optimization/84071
5223         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5224         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5225
5226 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
5227
5228         * config/arc/arc.c (arc_handle_aux_attribute): New function.
5229         (arc_attribute_table): Add 'aux' attribute.
5230         (arc_in_small_data_p): Consider aux like variables.
5231         (arc_is_aux_reg_p): New function.
5232         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5233         (arc_get_aux_arg): New function.
5234         (prepare_move_operands): Handle aux-register access.
5235         (arc_handle_aux_attribute): New function.
5236         * doc/extend.texi (ARC Variable attributes): Add subsection.
5237
5238 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
5239
5240         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5241         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5242         (arc_attribute_table): Add 'uncached' attribute.
5243         (arc_print_operand): Print '.di' flag for uncached memory
5244         accesses.
5245         (arc_in_small_data_p): Do not consider for small data the uncached
5246         types.
5247         (arc_is_uncached_mem_p): New function.
5248         * config/arc/predicates.md (compact_store_memory_operand): Check
5249         for uncached memory accesses.
5250         (nonvol_nonimm_operand): Likewise.
5251         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5252
5253 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
5254
5255         PR c/84100
5256         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5257         falign-loops=): Add Optimization flag.
5258
5259 2018-01-30  Jeff Law  <law@redhat.com>
5260
5261         PR target/84064
5262         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5263         INT_REGISTERS_SAVED.  Check it prior to calling
5264         get_scratch_register_on_entry.
5265         (ix86_adjust_stack_and_probe): Similarly.
5266         (ix86_emit_probe_stack_range): Similarly.
5267         (ix86_expand_prologue): Corresponding changes.
5268
5269 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5270
5271         PR target/40411
5272         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5273         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5274
5275 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
5276
5277         PR target/84112
5278         * lra-constraints.c (curr_insn_transform): Process AND in the
5279         address.
5280
5281 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
5282
5283         PR rtl-optimization/83986
5284         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5285         dependence against last_pending_memory_flush in addition to
5286         pending_jump_insns.
5287
5288 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
5289
5290         PR tree-optimization/81611
5291         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5292         copies.
5293
5294 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
5295
5296         PR target/83758
5297         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5298         a reg rtx.
5299
5300 2018-01-30  Richard Biener  <rguenther@suse.de>
5301             Jakub Jelinek  <jakub@redhat.com>
5302
5303         PR tree-optimization/84111
5304         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5305         inner loops added during recursion, as they don't have up-to-date
5306         SSA form.
5307
5308 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
5309
5310         PR ipa/81360
5311         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5312         (can_inline_edge_by_limits_p): ... here.
5313         (can_early_inline_edge_p, check_callers,
5314         update_caller_keys, update_callee_keys, recursive_inlining,
5315         add_new_edges_to_heap, speculation_useful_p,
5316         inline_small_functions,
5317         inline_small_functions, flatten_function,
5318         inline_to_all_callers_1): Update.
5319
5320 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
5321
5322         * profile-count.c (profile_count::combine_with_ipa_count): Handle
5323         zeros correctly.
5324
5325 2018-01-30  Richard Biener  <rguenther@suse.de>
5326
5327         PR tree-optimization/83008
5328         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5329         invariant and constant vector uses in stmts when they need
5330         more than one stmt.
5331
5332 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5333
5334         PR bootstrap/84017
5335         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5336         * configure: Regenerate.
5337
5338 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
5339
5340         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5341         pattern.
5342         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5343         Use gen_rtx_REG rather than gen_lowpart.
5344
5345 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
5346
5347         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5348         rather than 0 when creating partial subregs.
5349
5350 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
5351
5352         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5353         of usage.
5354
5355 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
5356
5357         PR target/81550
5358         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5359         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5360         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5361         flags.  This restores the settings used before the 2017-07-24.
5362         Turning off pre increment/decrement/modify allows IVOPTS to
5363         optimize DF/SF loops where the index is an int.
5364
5365 2018-01-29  Richard Biener  <rguenther@suse.de>
5366             Kelvin Nilsen  <kelvin@gcc.gnu.org>
5367
5368         PR bootstrap/80867
5369         * tree-vect-stmts.c (vectorizable_call): Don't call
5370         targetm.vectorize_builtin_md_vectorized_function if callee is
5371         NULL.
5372
5373 2018-01-22  Carl Love  <cel@us.ibm.com>
5374
5375         * doc/extend.tex: Fix typo in second arg in
5376         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5377
5378 2018-01-29  Richard Biener  <rguenther@suse.de>
5379
5380         PR tree-optimization/84086
5381         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5382         (flush_ssaname_freelist): When SSA names were released reset
5383         the SCEV hash table.
5384
5385 2018-01-29  Richard Biener  <rguenther@suse.de>
5386
5387         PR tree-optimization/84057
5388         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5389         removed paths when removing edges.
5390
5391 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
5392
5393         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5394         -mfunction-return=@var{choice}.
5395
5396 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5397
5398         PR diagnostic/84034
5399         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5400         Handle CR like TAB.
5401         (layout::print_source_line): Likewise.
5402         (test_get_line_width_without_trailing_whitespace): Add test cases.
5403
5404 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
5405
5406         PR middle-end/84040
5407         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5408         debug insns.
5409
5410 2018-01-26  Jim Wilson  <jimw@sifive.com>
5411
5412         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5413
5414         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5415         specified.
5416
5417 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5418
5419         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5420         and CMP + SUB-immediate -> SUBS.
5421
5422 2018-01-26  Martin Sebor  <msebor@redhat.com>
5423
5424         PR tree-optimization/83896
5425         * tree-ssa-strlen.c (get_string_len): Rename...
5426         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
5427         Avoid assuming length is constant.
5428         (handle_char_store): Use HOST_WIDE_INT for string length.
5429
5430 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
5431
5432         PR target/81763
5433         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5434         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5435
5436 2018-01-26  Richard Biener  <rguenther@suse.de>
5437
5438         PR rtl-optimization/84003
5439         * dse.c (record_store): Only record redundant stores when
5440         the earlier store aliases at least all accesses the later one does.
5441
5442 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
5443
5444         PR rtl-optimization/83985
5445         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5446         REG_CFA_RESTORE insns.
5447         (delete_unmarked_insns): Don't ignore separate shrink wrapping
5448         REG_CFA_RESTORE insns here.
5449
5450         PR c/83989
5451         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5452         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5453
5454 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
5455
5456         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5457         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5458         (arc_init): Likewise.
5459         (arc_override_options): Likewise.
5460         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5461         value.
5462         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5463         support.
5464         * config/arc/arc.h (TARGET_DBNZ): Define.
5465         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5466         properly set the tune attribute.
5467         (dbnz): Use TARGET_DBNZ guard.
5468         * config/arc/arc.opt (mtune): Add core3 option.
5469
5470 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
5471
5472         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5473         recognize new pic like addresses.
5474         (arc_delegitimize_address): Clean up.
5475
5476 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
5477
5478         * config/arc/arc-arches.def: Option mrf16 valid for all
5479         architectures.
5480         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5481         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5482         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5483         * config/arc/arc-tables.opt: Regenerate.
5484         * config/arc/arc.c (arc_conditional_register_usage): Handle
5485         reduced register file case.
5486         (arc_file_start): Set must have build attributes.
5487         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5488         mrf16 option value.
5489         * config/arc/arc.opt (mrf16): Add new option.
5490         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5491         * config/arc/genmultilib.awk: Handle new mrf16 option.
5492         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5493         * config/arc/t-multilib: Regenerate.
5494         * doc/invoke.texi (ARC Options): Document mrf16 option.
5495
5496 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
5497
5498         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5499         * config/arc/arc.c (arc_handle_secure_attribute): New function.
5500         (arc_attribute_table): Add 'secure_call' attribute.
5501         (arc_print_operand): Print secure call operand.
5502         (arc_function_ok_for_sibcall): Don't optimize tail calls when
5503         secure.
5504         (arc_is_secure_call_p): New function.  * config/arc/arc.md
5505         (call_i): Add support for sjli instruction.
5506         (call_value_i): Likewise.
5507         * config/arc/constraints.md (Csc): New constraint.
5508
5509 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
5510             John Eric Martin  <John.Martin@emmicro-us.com>
5511
5512         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5513         * config/arc/arc.c (_arc_jli_section): New struct.
5514         (arc_jli_section): New type.
5515         (rc_jli_sections): New static variable.
5516         (arc_handle_jli_attribute): New function.
5517         (arc_attribute_table): Add jli_always and jli_fixed attribute.
5518         (arc_file_end): New function.
5519         (TARGET_ASM_FILE_END): Define.
5520         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5521         (arc_add_jli_section): New function.
5522         (jli_call_scan): Likewise.
5523         (arc_reorg): Call jli_call_scan.
5524         (arc_output_addsi): Remove 'S' from printing asm operand.
5525         (arc_is_jli_call_p): New function.
5526         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5527         operand.
5528         (movhi_insn): Likewise.
5529         (movsi_insn): Likewise.
5530         (movsi_set_cc_insn): Likewise.
5531         (loadqi_update): Likewise.
5532         (load_zeroextendqisi_update): Likewise.
5533         (load_signextendqisi_update): Likewise.
5534         (loadhi_update): Likewise.
5535         (load_zeroextendhisi_update): Likewise.
5536         (load_signextendhisi_update): Likewise.
5537         (loadsi_update): Likewise.
5538         (loadsf_update): Likewise.
5539         (movsicc_insn): Likewise.
5540         (bset_insn): Likewise.
5541         (bxor_insn): Likewise.
5542         (bclr_insn): Likewise.
5543         (bmsk_insn): Likewise.
5544         (bicsi3_insn): Likewise.
5545         (cmpsi_cc_c_insn): Likewise.
5546         (movsi_ne): Likewise.
5547         (movsi_cond_exec): Likewise.
5548         (clrsbsi2): Likewise.
5549         (norm_f): Likewise.
5550         (normw): Likewise.
5551         (swap): Likewise.
5552         (divaw): Likewise.
5553         (flag): Likewise.
5554         (sr): Likewise.
5555         (kflag): Likewise.
5556         (ffs): Likewise.
5557         (ffs_f): Likewise.
5558         (fls): Likewise.
5559         (call_i): Remove 'S' asm letter, add jli instruction.
5560         (call_value_i): Likewise.
5561         * config/arc/arc.op (mjli-always): New option.
5562         * config/arc/constraints.md (Cji): New constraint.
5563         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5564         operand.
5565         (subsf3_fpx): Likewise.
5566         (mulsf3_fpx): Likewise.
5567         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5568         asm operand.
5569         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5570         function attrbutes.
5571         * doc/invoke.texi (ARC): Document mjli-always option.
5572
5573 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
5574
5575         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5576         avoid addition with 0 and use incw and decw where possible.
5577
5578 2018-01-26  Richard Biener  <rguenther@suse.de>
5579
5580         PR tree-optimization/81082
5581         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5582         association if it requires casting to unsigned.
5583         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5584         from fold_plusminus_mult_expr to catch important cases late when
5585         range info is available.
5586
5587 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5588
5589         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5590         * configure.ac (hidden_linkonce): New test.
5591         * configure: Regenerate.
5592         * config.in: Regenerate.
5593
5594 2018-01-26  Julia Koval  <julia.koval@intel.com>
5595
5596         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5597         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5598         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5599         _mm_mask_bitshuffle_epi64_mask): Fix type.
5600         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5601         USI_FTYPE_V4DI_V4DI_USI): Remove.
5602         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5603         __builtin_ia32_vpshufbitqmb256_mask,
5604         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5605         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5606         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5607
5608 2018-01-26  Alan Modra  <amodra@gmail.com>
5609
5610         PR target/84033
5611         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5612         UNSPEC_VBPERMQ.  Sort other unspecs.
5613
5614 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
5615
5616         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5617
5618 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
5619
5620         PR middle-end/83055
5621         * predict.c (drop_profile): Do not push/pop cfun; update also
5622         node->count.
5623         (handle_missing_profiles): Fix logic looking for zero profiles.
5624
5625 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
5626
5627         PR middle-end/83977
5628         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5629         on functions with #pragma omp declare simd or functions with simd
5630         attribute.
5631         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5632         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5633         Remove trailing \n from warning_at calls.
5634
5635 2018-01-25  Tom de Vries  <tom@codesourcery.com>
5636
5637         PR target/84028
5638         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5639         for neutered workers.
5640
5641 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
5642
5643         PR target/68467
5644         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5645         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5646
5647 2018-01-24  Jeff Law  <law@redhat.com>
5648
5649         PR target/83994
5650         * i386.c (get_probe_interval): Move to earlier point.
5651         (ix86_compute_frame_layout): If -fstack-clash-protection and
5652         the frame is larger than the probe interval, then use pushes
5653         to save registers rather than reg->mem moves.
5654         (ix86_expand_prologue): Remove conditional for int_registers_saved
5655         assertion.
5656
5657 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
5658
5659         PR target/84014
5660         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5661         min/max for never referenced object.
5662
5663 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
5664
5665         PR middle-end/83977
5666         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5667         here.
5668         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5669         attributes from DECL_ATTRIBUTES (decl) without affecting
5670         DECL_ATTRIBUTES (current_function_decl).
5671         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5672         functions with non-NULL DECL_ABSTRACT_ORIGIN.
5673
5674 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
5675
5676         PR tree-optimization/83979
5677         * fold-const.c (fold_comparison): Use constant_boolean_node
5678         instead of boolean_{true,false}_node.
5679
5680 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
5681
5682         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5683         with zero counts.
5684
5685 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5686
5687         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5688         Simplify the clause that sets the length attribute.
5689         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5690         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5691         clause that sets the length attribute.
5692         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5693
5694 2018-01-24  Tom de Vries  <tom@codesourcery.com>
5695
5696         PR target/83589
5697         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5698         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5699         Add strict parameter.
5700         (prevent_branch_around_nothing): Insert dummy insn between branch to
5701         label and label with no ptx insn inbetween.
5702         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5703
5704 2018-01-24  Tom de Vries  <tom@codesourcery.com>
5705
5706         PR target/81352
5707         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5708         for neutered threads in warp.
5709         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5710
5711 2018-01-24  Richard Biener  <rguenther@suse.de>
5712
5713         PR tree-optimization/83176
5714         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5715         operands.
5716
5717 2018-01-24  Richard Biener  <rguenther@suse.de>
5718
5719         PR tree-optimization/82819
5720         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5721         code generating pluses that are no-ops in the target precision.
5722
5723 2018-01-24  Richard Biener  <rguenther@suse.de>
5724
5725         PR middle-end/84000
5726         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5727
5728 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
5729
5730         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5731         to merge probabilities.
5732         * predict.c (probably_never_executed): Also mark as cold functions
5733         with global 0 profile and guessed local profile.
5734         * profile-count.c (profile_probability::combine_with_count): New
5735         member function.
5736         * profile-count.h (profile_probability::operator*,
5737         profile_probability::operator*=, profile_probability::operator/,
5738         profile_probability::operator/=): Reduce precision to adjusted
5739         and set value to guessed on contradictory divisions.
5740         (profile_probability::combine_with_freq): Remove.
5741         (profile_probability::combine_wiht_count): Declare.
5742         (profile_count::force_nonzero):: Set to adjusted.
5743         (profile_count::probability_in):: Set quality to adjusted.
5744         * tree-ssa-tail-merge.c (replace_block_by): Use
5745         combine_with_count.
5746
5747 2018-01-23  Andrew Waterman  <andrew@sifive.com>
5748             Jim Wilson  <jimw@sifive.com>
5749
5750         * config/riscv/riscv.c (riscv_stack_boundary): New.
5751         (riscv_option_override): Set riscv_stack_boundary.  Handle
5752         riscv_preferred_stack_boundary_arg.
5753         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5754         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5755         (STACK_BOUNDARY): Set to riscv_stack_boundary.
5756         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5757         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5758         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5759
5760 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
5761
5762         PR target/83905
5763         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5764         of struct ix86_frame.
5765         (ix86_expand_epilogue): Likewise.  Add a local variable for
5766         the reg_save_offset field in struct ix86_frame.
5767
5768 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
5769
5770         PR tree-optimization/82604
5771         * tree-loop-distribution.c (enum partition_kind): New enum item
5772         PKIND_PARTIAL_MEMSET.
5773         (partition_builtin_p): Support above new enum item.
5774         (generate_code_for_partition): Ditto.
5775         (compute_access_range): Differentiate cases that equality can be
5776         proven at all loops, the innermost loops or no loops.
5777         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5778         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
5779         (finalize_partitions, distribute_loop): Don't fuse partition of
5780         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5781         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5782         parloop is enabled.
5783
5784 2018-01-23  Martin Liska  <mliska@suse.cz>
5785
5786         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5787         order to ignore the predictor.
5788         (PRED_POLYMORPHIC_CALL): Likewise.
5789         (PRED_RECURSIVE_CALL): Likewise.
5790
5791 2018-01-23  Martin Liska  <mliska@suse.cz>
5792
5793         * tree-profile.c (tree_profiling): Print function header to
5794         aware reader which function we are working on.
5795         * value-prof.c (gimple_find_values_to_profile): Do not print
5796         not interesting value histograms.
5797
5798 2018-01-23  Martin Liska  <mliska@suse.cz>
5799
5800         * profile-count.h (enum profile_quality): Add
5801         profile_uninitialized as the first value. Do not number values
5802         as they are zero based.
5803         (profile_count::verify): Update sanity check.
5804         (profile_probability::verify): Likewise.
5805
5806 2018-01-23  Nathan Sidwell  <nathan@acm.org>
5807
5808         * doc/invoke.texi (ffor-scope): Deprecate.
5809
5810 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
5811
5812         PR tree-optimization/83510
5813         * domwalk.c (set_all_edges_as_executable): New function.
5814         (dom_walker::dom_walker): Convert bool param
5815         "skip_unreachable_blocks" to enum reachability.  Move setup of
5816         edge flags to set_all_edges_as_executable and only do it when
5817         reachability is REACHABLE_BLOCKS.
5818         * domwalk.h (enum dom_walker::reachability): New enum.
5819         (dom_walker::dom_walker): Convert bool param
5820         "skip_unreachable_blocks" to enum reachability.
5821         (set_all_edges_as_executable): New decl.
5822         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
5823         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5824         "reachability".
5825         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5826         but converting true to REACHABLE_BLOCKS.
5827         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5828         * tree-vrp.c
5829         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5830         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5831         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5832         REACHABLE_BLOCKS.
5833         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5834         if check_all_array_refs will be called.
5835
5836 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
5837
5838         * tree.c (selftest::test_location_wrappers): Add more test
5839         coverage.
5840
5841 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
5842
5843         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5844         (selftest::test_bit_in_range): Likewise.
5845
5846 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
5847
5848         PR testsuite/83888
5849         * doc/sourcebuild.texi (vect_float): Say that the selector
5850         only describes the situation when -funsafe-math-optimizations is on.
5851         (vect_float_strict): Document.
5852
5853 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
5854
5855         PR tree-optimization/83965
5856         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5857         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5858         instead of checking only for a reduction.
5859         (vect_recog_widen_sum_pattern): Likewise.
5860
5861 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
5862
5863         * predict.c (probably_never_executed): Only use precise profile info.
5864         (compute_function_frequency): Skip after inlining hack since we now
5865         have quality checking.
5866
5867 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
5868
5869         * profile-count.h (profile_probability::very_unlikely,
5870         profile_probability::unlikely, profile_probability::even): Set
5871         precision to guessed.
5872
5873 2018-01-23  Richard Biener  <rguenther@suse.de>
5874
5875         PR tree-optimization/83963
5876         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5877         Properly terminate dominator walk when crossing the exit edge not
5878         when visiting its source block.
5879
5880 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
5881
5882         PR c++/83918
5883         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5884         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5885
5886 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
5887
5888         PR tree-optimization/83957
5889         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
5890         semicolon after for body surrounded by braces.
5891
5892         PR tree-optimization/83081
5893         * profile-count.h (profile_probability::split): New method.
5894         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5895         Use profile_probability::split.
5896         (do_compare_rtx_and_jump): Fix adjustment of probabilities
5897         when splitting a single conditional jump into 2.
5898
5899 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
5900
5901         PR tree-optimization/69452
5902         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5903         decl.
5904
5905 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
5906
5907         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5908         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5909         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5910
5911 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
5912
5913         * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5914         * config/rl78/rl78.md (movdi): New define_expand.
5915         * config/rl78/rl78.c (rl78_split_movdi): New function.
5916
5917 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
5918
5919         PR target/83862
5920         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5921         no longer used.
5922         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5923         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5924         128-bit to produce an UNSPEC move to get the double word with the
5925         signbit and then a shift directly to do signbit.
5926         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5927         implementation with a new version that just does either a direct
5928         move or a regular move.  Move memory interface to separate insns.
5929         Move insns so they are next to the expander.
5930         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5931         with signbit move.  Split big and little endian case.
5932         (signbit<mode>2_dm_mem_le): Likewise.
5933         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5934         (signbit<mode>2_dm2): Likewise.
5935
5936 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
5937
5938         * config/rl78/rl78.md (anddi3): New define_expand.
5939
5940 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
5941
5942         * config/rl78/rl78.md (umindi3): New define_expand.
5943
5944 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
5945
5946         * config/rl78/rl78.md (smindi3): New define_expand.
5947
5948 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
5949
5950         * config/rl78/rl78.md (smaxdi3): New define_expand.
5951
5952 2018-01-22  Carl Love  <cel@us.ibm.com>
5953
5954         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5955         LVX_V1TI): Add macro expansion.
5956         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5957         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5958         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5959         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5960         Change check to determine if the instruction is a byte reversing
5961         entry.  Fix typo in comment.
5962         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5963         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5964         Add def_builtin calls for new builtins.
5965         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5966         Add define_insn expansion.
5967
5968 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
5969
5970         * config/rl78/rl78.md (umaxdi3): New define_expand.
5971
5972 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
5973
5974         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5975         for non-QImode registers.
5976
5977 2018-01-22  Richard Biener  <rguenther@suse.de>
5978
5979         PR tree-optimization/83963
5980         * graphite-scop-detection.c (scop_detection::get_sese): Delay
5981         including the loop exit block.
5982         (scop_detection::merge_sese): Likewise.
5983         (scop_detection::add_scop): Do it here instead.
5984
5985 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5986
5987         * doc/sourcebuild.texi (arm_softfloat): Document.
5988
5989 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
5990
5991         PR gcc/77734
5992         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5993         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5994         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5995
5996 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5997             David Edelsohn  <dje.gcc@gmail.com>
5998
5999         PR target/83946
6000         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6001         Change "crset eq" to "crset 2".
6002         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6003         (*call_indirect_aix<mode>_nospec): Likewise.
6004         (*call_value_indirect_aix<mode>_nospec): Likewise.
6005         (*call_indirect_elfv2<mode>_nospec): Likewise.
6006         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6007         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6008         change assembly output from . to $.
6009         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6010         (indirect_jump<mode>_nospec): Change assembly output from . to $.
6011         (*tablejump<mode>_internal1_nospec): Likewise.
6012
6013 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
6014
6015         PR target/80870
6016         * config/sh/sh_optimize_sett_clrt.cc:
6017         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6018
6019 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
6020
6021         PR tree-optimization/83940
6022         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6023         offset_dt to vect_constant_def rather than vect_unknown_def_type.
6024         (vect_check_load_store_mask): Add a mask_dt_out parameter and
6025         use it to pass back the definition type.
6026         (vect_check_store_rhs): Likewise rhs_dt_out.
6027         (vect_build_gather_load_calls): Add a mask_dt argument and use
6028         it instead of a call to vect_is_simple_use.
6029         (vectorizable_store): Update calls to vect_check_load_store_mask
6030         and vect_check_store_rhs.  Use the dt returned by the latter instead
6031         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
6032         instead of calls to vect_is_simple_use.  Pass the scalar rather
6033         than the vector operand to vect_is_simple_use when handling
6034         second and subsequent copies of an rhs value.
6035         (vectorizable_load): Update calls to vect_check_load_store_mask
6036         and vect_build_gather_load_calls.  Use the cached mask_dt and
6037         gs_info.offset_dt instead of calls to vect_is_simple_use.
6038
6039 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
6040
6041         PR middle-end/83945
6042         * tree-emutls.c: Include gimplify.h.
6043         (lower_emutls_2): New function.
6044         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6045         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6046         it before further processing.
6047
6048         PR target/83930
6049         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6050         UINTVAL (trueop1) instead of INTVAL (op1).
6051
6052 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
6053
6054         PR debug/81570
6055         PR debug/83728
6056         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6057         INCOMING_FRAME_SP_OFFSET if not defined.
6058         (scan_trace): Add ENTRY argument.  If true and
6059         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6060         emit a note to adjust the CFA offset.
6061         (create_cfi_notes): Adjust scan_trace callers.
6062         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6063         INCOMING_FRAME_SP_OFFSET in the CIE.
6064         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6065         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6066         Likewise.
6067         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6068         * doc/tm.texi: Regenerated.
6069
6070 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
6071
6072         PR rtl-optimization/83147
6073         * lra-constraints.c (remove_inheritance_pseudos): Use
6074         lra_substitute_pseudo_within_insn.
6075
6076 2018-01-19  Tom de Vries  <tom@codesourcery.com>
6077             Cesar Philippidis  <cesar@codesourcery.com>
6078
6079         PR target/83920
6080         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6081
6082 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
6083
6084         PR target/83790
6085         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6086         spaces for function labels.
6087
6088 2018-01-19  Martin Liska  <mliska@suse.cz>
6089
6090         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6091         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6092         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6093         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6094         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6095         (PRED_CONST_RETURN): Change from 69 to 65.
6096         (PRED_NULL_RETURN): Change from 91 to 71.
6097         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6098         (PRED_LOOP_GUARD): Change from 66 to 73.
6099
6100 2018-01-19  Martin Liska  <mliska@suse.cz>
6101
6102         * predict.c (predict_insn_def): Add new assert.
6103         (struct branch_predictor): Change type to signed integer.
6104         (test_prediction_value_range): Amend test to cover
6105         PROB_UNINITIALIZED.
6106         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6107         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6108         (PRED_LOOP_ITERATIONS_MAX): Likewise.
6109         (PRED_LOOP_IV_COMPARE): Likewise.
6110         * predict.h (PROB_UNINITIALIZED): Define new constant.
6111
6112 2018-01-19  Martin Liska  <mliska@suse.cz>
6113
6114         * predict.c (dump_prediction): Add new format for
6115         analyze_brprob.py script which is enabled with -details
6116         suboption.
6117         * profile-count.h (precise_p): New function.
6118
6119 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
6120
6121         PR tree-optimization/83922
6122         * tree-vect-loop.c (vect_verify_full_masking): Return false if
6123         there are no statements that need masking.
6124         (vect_active_double_reduction_p): New function.
6125         (vect_analyze_loop_operations): Use it when handling phis that
6126         are not in the loop header.
6127
6128 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
6129
6130         PR tree-optimization/83914
6131         * tree-vect-loop.c (vectorizable_induction): Don't convert
6132         init_expr or apply the peeling adjustment for inductions
6133         that are nested within the vectorized loop.
6134
6135 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6136
6137         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6138         instead of NEG.
6139
6140 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
6141
6142         PR sanitizer/81715
6143         PR testsuite/83882
6144         * function.h (gimplify_parameters): Add gimple_seq * argument.
6145         * function.c: Include gimple.h and options.h.
6146         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6147         for the added local temporaries if needed.
6148         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6149         if there are any parameter cleanups, wrap whole body into a
6150         try/finally with the cleanups.
6151
6152 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
6153
6154         PR target/82964
6155         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6156         Use GET_MODE_CLASS for scalar floating point.
6157
6158 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
6159
6160         PR ipa/82256
6161         patch by PaX Team
6162         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6163         Fix call of call_cgraph_insertion_hooks.
6164
6165 2018-01-18  Martin Sebor  <msebor@redhat.com>
6166
6167         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6168
6169 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
6170
6171         PR ipa/83619
6172         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6173         frequencies.
6174
6175 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
6176
6177         PR other/70268
6178         * common.opt: (-ffile-prefix-map): New option.
6179         * opts.c (common_handle_option): Defer it.
6180         * opts-global.c (handle_common_deferred_options): Handle it.
6181         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6182         * file-prefix-map.h: New file.
6183         (remap_debug_filename, add_debug_prefix_map): ...here.
6184         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6185         * final.c (debug_prefix_map, add_debug_prefix_map
6186         remap_debug_filename): Move to...
6187         * file-prefix-map.c: New file.
6188         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6189         generalize, get rid of alloca(), use strrchr() instead of strchr().
6190         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6191         Implement in terms of add_prefix_map().
6192         (remap_macro_filename, remap_debug_filename): Implement in term of
6193         remap_filename().
6194         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6195         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6196         * dbxout.c: Include file-prefix-map.h.
6197         * varasm.c: Likewise.
6198         * vmsdbgout.c: Likewise.
6199         * xcoffout.c: Likewise.
6200         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6201         * doc/cppopts.texi (-fmacro-prefix-map): Document.
6202         * doc/invoke.texi (-ffile-prefix-map): Document.
6203         (-fdebug-prefix-map): Update description.
6204
6205 2018-01-18  Martin Liska  <mliska@suse.cz>
6206
6207         * config/i386/i386.c (indirect_thunk_name): Document that also
6208         lfence is emitted.
6209         (output_indirect_thunk): Document why both instructions
6210         (pause and lfence) are generated.
6211
6212 2018-01-18  Richard Biener  <rguenther@suse.de>
6213
6214         PR tree-optimization/83887
6215         * graphite-scop-detection.c
6216         (scop_detection::get_nearest_dom_with_single_entry): Remove.
6217         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6218         (scop_detection::merge_sese): Re-implement with a flood-fill
6219         algorithm that properly finds a SESE region if it exists.
6220
6221 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
6222
6223         PR c/61240
6224         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6225         pointer_diff optimizations use view_convert instead of convert.
6226
6227 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6228
6229         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6230         Generate different code for -mno-speculate-indirect-jumps.
6231         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6232         (*call_indirect_aix<mode>): Disable for
6233         -mno-speculate-indirect-jumps.
6234         (*call_indirect_aix<mode>_nospec): New define_insn.
6235         (*call_value_indirect_aix<mode>): Disable for
6236         -mno-speculate-indirect-jumps.
6237         (*call_value_indirect_aix<mode>_nospec): New define_insn.
6238         (*sibcall_nonlocal_sysv<mode>): Generate different code for
6239         -mno-speculate-indirect-jumps.
6240         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6241
6242 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
6243
6244         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6245         long double type, set the flags for noting the default long double
6246         type, even if we don't pass or return a long double type.
6247
6248 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
6249
6250         PR ipa/83051
6251         * ipa-inline.c (flatten_function): Do not overwrite final inlining
6252         failure.
6253
6254 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
6255
6256         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6257         support for merge[hl].
6258         (fold_mergehl_helper): New helper function.
6259         (tree-vector-builder.h): New #include for tree_vector_builder usage.
6260         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6261         (altivec_vmrglw_direct): Add xxmrglw insn.
6262
6263 2018-01-17  Andrew Waterman  <andrew@sifive.com>
6264
6265         * config/riscv/riscv.c (riscv_conditional_register_usage): If
6266         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6267
6268 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
6269
6270         PR lto/83121
6271         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6272         call the lto_location_cache before reading the
6273         DECL_SOURCE_LOCATION of the types.
6274
6275 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
6276             Richard Sandiford  <richard.sandiford@linaro.org>
6277
6278         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6279         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6280         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6281         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6282         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6283         Add declaration.
6284         * config/aarch64/constraints.md (aarch64_movti_operand):
6285         Limit immediates.
6286         * config/aarch64/predicates.md (Uti): Add new constraint.
6287
6288 2018-01-17 Carl Love  <cel@us.ibm.com>
6289
6290         * config/rs6000/vsx.md (define_expand xl_len_r,
6291         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6292         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6293         lxvll.
6294         (define_expand, define_insn): Move the shift left from  the
6295         define_insn to the define_expand for lxvl and stxvl instructions.
6296         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6297         and XL_LEN_R definitions to PURE.
6298
6299 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
6300
6301         * config/i386/i386.c (indirect_thunk_name): Declare regno
6302         as unsigned int.  Compare regno with INVALID_REGNUM.
6303         (output_indirect_thunk): Ditto.
6304         (output_indirect_thunk_function): Ditto.
6305         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
6306         in the call to output_indirect_thunk_function.
6307
6308 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
6309
6310         PR middle-end/83884
6311         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6312         rather than the size of inner_type to determine the stack slot size
6313         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6314
6315 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
6316
6317         PR target/83546
6318         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6319         to PTA_SILVERMONT.
6320
6321 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
6322
6323         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6324         endian Linux systems to optionally enable multilibs for selecting
6325         the long double type if the user configured an explicit type.
6326         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6327         have no long double multilibs if not defined.
6328         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6329         warn if the user used -mabi={ieee,ibm}longdouble and we built
6330         multilibs for long double.
6331         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6332         appropriate multilib option.
6333         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6334         multilib options.
6335         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6336         for building long double multilibs.
6337         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6338
6339 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
6340
6341         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6342         copies.
6343
6344         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6345         64 bits.
6346         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6347         128 bits.
6348
6349         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6350         variables.
6351
6352         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6353         return value.
6354
6355 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
6356
6357         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6358         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6359
6360 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6361
6362         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6363         different rtl trees depending on TARGET_64BIT.
6364         (rs6000_gen_lvx): Likewise.
6365
6366 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
6367
6368         * config/visium/visium.md (nop): Tweak comment.
6369         (hazard_nop): Likewise.
6370
6371 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6372
6373         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6374         -mspeculate-indirect-jumps.
6375         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6376         for -mno-speculate-indirect-jumps.
6377         (*call_indirect_elfv2<mode>_nospec): New define_insn.
6378         (*call_value_indirect_elfv2<mode>): Disable for
6379         -mno-speculate-indirect-jumps.
6380         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6381         (indirect_jump): Emit different RTL for
6382         -mno-speculate-indirect-jumps.
6383         (*indirect_jump<mode>): Disable for
6384         -mno-speculate-indirect-jumps.
6385         (*indirect_jump<mode>_nospec): New define_insn.
6386         (tablejump): Emit different RTL for
6387         -mno-speculate-indirect-jumps.
6388         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6389         (tablejumpsi_nospec): New define_expand.
6390         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6391         (tablejumpdi_nospec): New define_expand.
6392         (*tablejump<mode>_internal1): Disable for
6393         -mno-speculate-indirect-jumps.
6394         (*tablejump<mode>_internal1_nospec): New define_insn.
6395         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6396         option.
6397
6398 2018-01-16  Artyom Skrobov tyomitch@gmail.com
6399
6400         * caller-save.c (insert_save): Drop unnecessary parameter.  All
6401         callers updated.
6402
6403 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
6404             Richard Biener  <rguenth@suse.de>
6405
6406         PR libgomp/83590
6407         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6408         return early, inline manually is_gimple_sizepos.  Make sure if we
6409         call gimplify_expr we don't end up with a gimple constant.
6410         * tree.c (variably_modified_type_p): Don't return true for
6411         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
6412         * gimplify.h (is_gimple_sizepos): Remove.
6413
6414 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
6415
6416         PR tree-optimization/83857
6417         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6418         vectorizable_live_operation for pure SLP statements.
6419         (vectorizable_live_operation): Handle PHIs.
6420
6421 2018-01-16  Richard Biener  <rguenther@suse.de>
6422
6423         PR tree-optimization/83867
6424         * tree-vect-stmts.c (vect_transform_stmt): Precompute
6425         nested_in_vect_loop_p since the scalar stmt may get invalidated.
6426
6427 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
6428
6429         PR c/83844
6430         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6431         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6432         If off is not INTEGER_CST, issue a may not be aligned warning
6433         rather than isn't aligned.  Use isn%'t rather than isn't.
6434         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6435         into MULT_EXPR.
6436         <case MULT_EXPR>: Improve the case when bottom and one of the
6437         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6438         operand, in that case check if the other operand is multiple of
6439         bottom divided by the INTEGER_CST operand.
6440
6441 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
6442
6443         PR target/83858
6444         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6445         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6446         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6447         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6448         * config/pa/pa.c (pa_function_arg_advance): Likewise.
6449         (pa_function_arg, pa_arg_partial_bytes): Likewise.
6450         (pa_function_arg_size): New function.
6451
6452 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
6453
6454         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6455         in a separate statement.
6456
6457 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
6458
6459         PR tree-optimization/83847
6460         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6461         group gathers and scatters.
6462
6463 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
6464
6465         PR rtl-optimization/86620
6466         * params.def (max-sched-ready-insns): Bump minimum value to 1.
6467
6468         PR rtl-optimization/83213
6469         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6470         to last if both are JUMP_INSNs.
6471
6472         PR tree-optimization/83843
6473         * gimple-ssa-store-merging.c
6474         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6475         store_immediate_info for bswap/nop orig_stores.
6476
6477 2018-01-15  Andrew Waterman  <andrew@sifive.com>
6478
6479         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6480         !TARGET_MUL.
6481         <UDIV>: Increase cost if !TARGET_DIV.
6482
6483 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
6484
6485         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6486         (define_attr "cr_logical_3op"): New.
6487         (cceq_ior_compare): Adjust.
6488         (cceq_ior_compare_complement): Adjust.
6489         (*cceq_rev_compare): Adjust.
6490         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6491         (is_cracked_insn): Adjust.
6492         (insn_must_be_first_in_group): Adjust.
6493         * config/rs6000/40x.md: Adjust.
6494         * config/rs6000/440.md: Adjust.
6495         * config/rs6000/476.md: Adjust.
6496         * config/rs6000/601.md: Adjust.
6497         * config/rs6000/603.md: Adjust.
6498         * config/rs6000/6xx.md: Adjust.
6499         * config/rs6000/7450.md: Adjust.
6500         * config/rs6000/7xx.md: Adjust.
6501         * config/rs6000/8540.md: Adjust.
6502         * config/rs6000/cell.md: Adjust.
6503         * config/rs6000/e300c2c3.md: Adjust.
6504         * config/rs6000/e500mc.md: Adjust.
6505         * config/rs6000/e500mc64.md: Adjust.
6506         * config/rs6000/e5500.md: Adjust.
6507         * config/rs6000/e6500.md: Adjust.
6508         * config/rs6000/mpc.md: Adjust.
6509         * config/rs6000/power4.md: Adjust.
6510         * config/rs6000/power5.md: Adjust.
6511         * config/rs6000/power6.md: Adjust.
6512         * config/rs6000/power7.md: Adjust.
6513         * config/rs6000/power8.md: Adjust.
6514         * config/rs6000/power9.md: Adjust.
6515         * config/rs6000/rs64.md: Adjust.
6516         * config/rs6000/titan.md: Adjust.
6517
6518 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
6519
6520         * config/i386/predicates.md (indirect_branch_operand): Rewrite
6521         ix86_indirect_branch_register logic.
6522
6523 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
6524
6525         * config/i386/constraints.md (Bs): Update
6526         ix86_indirect_branch_register check.  Don't check
6527         ix86_indirect_branch_register with GOT_memory_operand.
6528         (Bw): Likewise.
6529         * config/i386/predicates.md (GOT_memory_operand): Don't check
6530         ix86_indirect_branch_register here.
6531         (GOT32_symbol_operand): Likewise.
6532
6533 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
6534
6535         * config/i386/predicates.md (constant_call_address_operand):
6536         Rewrite ix86_indirect_branch_register logic.
6537         (sibcall_insn_operand): Likewise.
6538
6539 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
6540
6541         * config/i386/constraints.md (Bs): Replace
6542         ix86_indirect_branch_thunk_register with
6543         ix86_indirect_branch_register.
6544         (Bw): Likewise.
6545         * config/i386/i386.md (indirect_jump): Likewise.
6546         (tablejump): Likewise.
6547         (*sibcall_memory): Likewise.
6548         (*sibcall_value_memory): Likewise.
6549         Peepholes of indirect call and jump via memory: Likewise.
6550         * config/i386/i386.opt: Likewise.
6551         * config/i386/predicates.md (indirect_branch_operand): Likewise.
6552         (GOT_memory_operand): Likewise.
6553         (call_insn_operand): Likewise.
6554         (sibcall_insn_operand): Likewise.
6555         (GOT32_symbol_operand): Likewise.
6556
6557 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
6558
6559         PR middle-end/83837
6560         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6561         type rather than type addr's type points to.
6562         (expand_omp_atomic_mutex): Likewise.
6563         (expand_omp_atomic): Likewise.
6564
6565 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
6566
6567         PR target/83839
6568         * config/i386/i386.c (output_indirect_thunk_function): Use
6569         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6570         for  __x86_return_thunk.
6571
6572 2018-01-15  Richard Biener  <rguenther@suse.de>
6573
6574         PR middle-end/83850
6575         * expmed.c (extract_bit_field_1): Fix typo.
6576
6577 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6578
6579         PR target/83687
6580         * config/arm/iterators.md (VF): New mode iterator.
6581         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6582         Remove integer-related logic from pattern.
6583         (neon_vabd<mode>_3): Likewise.
6584
6585 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
6586
6587         PR middle-end/82694
6588         * common.opt (fstrict-overflow): No longer an alias.
6589         (fwrapv-pointer): New option.
6590         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6591         also for pointer types based on flag_wrapv_pointer.
6592         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6593         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6594         opts->x_flag_wrapv got set.
6595         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6596         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6597         POINTER_TYPE_OVERFLOW_UNDEFINED.
6598         * match.pd: Likewise in address comparison pattern.
6599         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6600
6601 2018-01-15  Richard Biener  <rguenther@suse.de>
6602
6603         PR lto/83804
6604         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6605         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
6606         Reset type names to their identifier if their TYPE_DECL doesn't
6607         have linkage (and thus is used for ODR and devirt).
6608         (save_debug_info_for_decl): Remove.
6609         (save_debug_info_for_type): Likewise.
6610         (add_tree_to_fld_list): Adjust.
6611         * tree-pretty-print.c (dump_generic_node): Make dumping of
6612         type names more robust.
6613
6614 2018-01-15  Richard Biener  <rguenther@suse.de>
6615
6616         * BASE-VER: Bump to 8.0.1.
6617
6618 2018-01-14  Martin Sebor  <msebor@redhat.com>
6619
6620         PR other/83508
6621         * builtins.c (check_access): Avoid warning when the no-warning bit
6622         is set.
6623
6624 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
6625
6626         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6627         * ira-color (allocno_hard_regs_compare): Likewise.
6628
6629 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
6630
6631         PR target/83013
6632         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6633         Use .pushsection/.popsection.
6634
6635 2018-01-14  Martin Sebor  <msebor@redhat.com>
6636
6637         PR c++/81327
6638         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6639
6640 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
6641
6642         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6643         entry from extra_headers.
6644         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6645         extra_headers, make the list bitwise identical to the i?86-*-* one.
6646
6647 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
6648
6649         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6650         -mcmodel=large with -mindirect-branch=thunk,
6651         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6652         -mfunction-return=thunk-extern.
6653         * doc/invoke.texi: Document -mcmodel=large is incompatible with
6654         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6655         -mfunction-return=thunk and -mfunction-return=thunk-extern.
6656
6657 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
6658
6659         * config/i386/i386.c (print_reg): Print the name of the full
6660         integer register without '%'.
6661         (ix86_print_operand): Handle 'V'.
6662          * doc/extend.texi: Document 'V' modifier.
6663
6664 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
6665
6666         * config/i386/constraints.md (Bs): Disallow memory operand for
6667         -mindirect-branch-register.
6668         (Bw): Likewise.
6669         * config/i386/predicates.md (indirect_branch_operand): Likewise.
6670         (GOT_memory_operand): Likewise.
6671         (call_insn_operand): Likewise.
6672         (sibcall_insn_operand): Likewise.
6673         (GOT32_symbol_operand): Likewise.
6674         * config/i386/i386.md (indirect_jump): Call convert_memory_address
6675         for -mindirect-branch-register.
6676         (tablejump): Likewise.
6677         (*sibcall_memory): Likewise.
6678         (*sibcall_value_memory): Likewise.
6679         Disallow peepholes of indirect call and jump via memory for
6680         -mindirect-branch-register.
6681         (*call_pop): Replace m with Bw.
6682         (*call_value_pop): Likewise.
6683         (*sibcall_pop_memory): Replace m with Bs.
6684         * config/i386/i386.opt (mindirect-branch-register): New option.
6685         * doc/invoke.texi: Document -mindirect-branch-register option.
6686
6687 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
6688
6689         * config/i386/i386-protos.h (ix86_output_function_return): New.
6690         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6691         set function_return_type.
6692         (indirect_thunk_name): Add ret_p to indicate thunk for function
6693         return.
6694         (output_indirect_thunk_function): Pass false to
6695         indirect_thunk_name.
6696         (ix86_output_indirect_branch_via_reg): Likewise.
6697         (ix86_output_indirect_branch_via_push): Likewise.
6698         (output_indirect_thunk_function): Create alias for function
6699         return thunk if regno < 0.
6700         (ix86_output_function_return): New function.
6701         (ix86_handle_fndecl_attribute): Handle function_return.
6702         (ix86_attribute_table): Add function_return.
6703         * config/i386/i386.h (machine_function): Add
6704         function_return_type.
6705         * config/i386/i386.md (simple_return_internal): Use
6706         ix86_output_function_return.
6707         (simple_return_internal_long): Likewise.
6708         * config/i386/i386.opt (mfunction-return=): New option.
6709         (indirect_branch): Mention -mfunction-return=.
6710         * doc/extend.texi: Document function_return function attribute.
6711         * doc/invoke.texi: Document -mfunction-return= option.
6712
6713 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
6714
6715         * config/i386/i386-opts.h (indirect_branch): New.
6716         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6717         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6718         with local indirect jump when converting indirect call and jump.
6719         (ix86_set_indirect_branch_type): New.
6720         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6721         (indirectlabelno): New.
6722         (indirect_thunk_needed): Likewise.
6723         (indirect_thunk_bnd_needed): Likewise.
6724         (indirect_thunks_used): Likewise.
6725         (indirect_thunks_bnd_used): Likewise.
6726         (INDIRECT_LABEL): Likewise.
6727         (indirect_thunk_name): Likewise.
6728         (output_indirect_thunk): Likewise.
6729         (output_indirect_thunk_function): Likewise.
6730         (ix86_output_indirect_branch_via_reg): Likewise.
6731         (ix86_output_indirect_branch_via_push): Likewise.
6732         (ix86_output_indirect_branch): Likewise.
6733         (ix86_output_indirect_jmp): Likewise.
6734         (ix86_code_end): Call output_indirect_thunk_function if needed.
6735         (ix86_output_call_insn): Call ix86_output_indirect_branch if
6736         needed.
6737         (ix86_handle_fndecl_attribute): Handle indirect_branch.
6738         (ix86_attribute_table): Add indirect_branch.
6739         * config/i386/i386.h (machine_function): Add indirect_branch_type
6740         and has_local_indirect_jump.
6741         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6742         to true.
6743         (tablejump): Likewise.
6744         (*indirect_jump): Use ix86_output_indirect_jmp.
6745         (*tablejump_1): Likewise.
6746         (simple_return_indirect_internal): Likewise.
6747         * config/i386/i386.opt (mindirect-branch=): New option.
6748         (indirect_branch): New.
6749         (keep): Likewise.
6750         (thunk): Likewise.
6751         (thunk-inline): Likewise.
6752         (thunk-extern): Likewise.
6753         * doc/extend.texi: Document indirect_branch function attribute.
6754         * doc/invoke.texi: Document -mindirect-branch= option.
6755
6756 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
6757
6758         PR ipa/83051
6759         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6760
6761 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
6762
6763         * ipa-inline.c (want_inline_small_function_p): Return false if
6764         inlining has already failed with CIF_FINAL_ERROR.
6765         (update_caller_keys): Call want_inline_small_function_p before
6766         can_inline_edge_p.
6767         (update_callee_keys): Likewise.
6768
6769 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6770
6771         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6772         New function.
6773         (rs6000_quadword_masked_address_p): Likewise.
6774         (quad_aligned_load_p): Likewise.
6775         (quad_aligned_store_p): Likewise.
6776         (const_load_sequence_p): Add comment to describe the outer-most loop.
6777         (mimic_memory_attributes_and_flags): New function.
6778         (rs6000_gen_stvx): Likewise.
6779         (replace_swapped_aligned_store): Likewise.
6780         (rs6000_gen_lvx): Likewise.
6781         (replace_swapped_aligned_load): Likewise.
6782         (replace_swapped_load_constant): Capitalize argument name in
6783         comment describing this function.
6784         (rs6000_analyze_swaps): Add a third pass to search for vector loads
6785         and stores that access quad-word aligned addresses and replace
6786         with stvx or lvx instructions when appropriate.
6787         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6788         New function prototype.
6789         (rs6000_quadword_masked_address_p): Likewise.
6790         (rs6000_gen_lvx): Likewise.
6791         (rs6000_gen_stvx): Likewise.
6792         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6793         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6794         when memory address is aligned.
6795         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6796         this split to select lvx instruction when memory address is aligned.
6797         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6798         instruction when memory address is aligned.
6799         (*vsx_le_perm_load_v16qi): Likewise.
6800         (four unnamed splitters): Modify to select the stvx instruction
6801         when memory is aligned.
6802
6803 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
6804
6805         * predict.c (determine_unlikely_bbs): Handle correctly BBs
6806         which appears in the queue multiple times.
6807
6808 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
6809             Alan Hayward  <alan.hayward@arm.com>
6810             David Sherwood  <david.sherwood@arm.com>
6811
6812         * tree-vectorizer.h (vec_lower_bound): New structure.
6813         (_loop_vec_info): Add check_nonzero and lower_bounds.
6814         (LOOP_VINFO_CHECK_NONZERO): New macro.
6815         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6816         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6817         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6818         fields.  Make seg_len the distance travelled, not including the
6819         access size.
6820         (dr_direction_indicator): Declare.
6821         (dr_zero_step_indicator): Likewise.
6822         (dr_known_forward_stride_p): Likewise.
6823         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6824         tree-ssanames.h.
6825         (runtime_alias_check_p): Allow runtime alias checks with
6826         variable strides.
6827         (operator ==): Compare access_size and align.
6828         (prune_runtime_alias_test_list): Rework for new distinction between
6829         the access_size and seg_len.
6830         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
6831         segment lengths.
6832         (get_segment_min_max): New function.
6833         (create_intersect_range_checks): Use it.
6834         (dr_step_indicator): New function.
6835         (dr_direction_indicator): Likewise.
6836         (dr_zero_step_indicator): Likewise.
6837         (dr_known_forward_stride_p): Likewise.
6838         * tree-loop-distribution.c (data_ref_segment_size): Return
6839         DR_STEP * (niters - 1).
6840         (compute_alias_check_pairs): Update call to the dr_with_seg_len
6841         constructor.
6842         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6843         (vect_preserves_scalar_order_p): New function, split out from...
6844         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
6845         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6846         (vect_vfa_access_size): New function.
6847         (vect_vfa_align): Likewise.
6848         (vect_compile_time_alias): Take access_size_a and access_b arguments.
6849         (dump_lower_bound): New function.
6850         (vect_check_lower_bound): Likewise.
6851         (vect_small_gap_p): Likewise.
6852         (vectorizable_with_step_bound_p): Likewise.
6853         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6854         depencies if the vectorization factor is 1.  Convert the checks
6855         for nonzero steps into checks on the bounds of DR_STEP.  Try using
6856         a bunds check for variable steps if the minimum required step is
6857         relatively small. Update calls to the dr_with_seg_len
6858         constructor and to vect_compile_time_alias.
6859         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6860         function.
6861         (vect_loop_versioning): Call it.
6862         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6863         when retrying.
6864         (vect_estimate_min_profitable_iters): Account for any bounds checks.
6865
6866 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
6867             Alan Hayward  <alan.hayward@arm.com>
6868             David Sherwood  <david.sherwood@arm.com>
6869
6870         * doc/sourcebuild.texi (vect_scatter_store): Document.
6871         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6872         optabs.
6873         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6874         Document.
6875         * genopinit.c (main): Add supports_vec_scatter_store and
6876         supports_vec_scatter_store_cached to target_optabs.
6877         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6878         IFN_MASK_SCATTER_STORE.
6879         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6880         functions.
6881         * internal-fn.h (internal_store_fn_p): Declare.
6882         (internal_fn_stored_value_index): Likewise.
6883         * internal-fn.c (scatter_store_direct): New macro.
6884         (expand_scatter_store_optab_fn): New function.
6885         (direct_scatter_store_optab_supported_p): New macro.
6886         (internal_store_fn_p): New function.
6887         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6888         IFN_MASK_SCATTER_STORE.
6889         (internal_fn_mask_index): Likewise.
6890         (internal_fn_stored_value_index): New function.
6891         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6892         for scatter stores.
6893         * optabs-query.h (supports_vec_scatter_store_p): Declare.
6894         * optabs-query.c (supports_vec_scatter_store_p): New function.
6895         * tree-vectorizer.h (vect_get_store_rhs): Declare.
6896         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6897         true for scatter stores.
6898         (vect_gather_scatter_fn_p): Handle scatter stores too.
6899         (vect_check_gather_scatter): Consider using scatter stores if
6900         supports_vec_scatter_store_p.
6901         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6902         scatter stores too.
6903         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6904         internal_fn_stored_value_index.
6905         (check_load_store_masking): Handle scatter stores too.
6906         (vect_get_store_rhs): Make public.
6907         (vectorizable_call): Use internal_store_fn_p.
6908         (vectorizable_store): Handle scatter store internal functions.
6909         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6910         when deciding whether the end of the group has been reached.
6911         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6912         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6913         (mask_scatter_store<mode>): New insns.
6914
6915 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
6916             Alan Hayward  <alan.hayward@arm.com>
6917             David Sherwood  <david.sherwood@arm.com>
6918
6919         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6920         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6921         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6922         function.
6923         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6924         Use vect_truncate_gather_scatter_offset if we can't treat the
6925         operation as a normal gather load or scatter store.
6926         (get_group_load_store_type): Take the gather_scatter_info
6927         as argument.  Try using a gather load or scatter store for
6928         single-element groups.
6929         (get_load_store_type): Update calls to get_group_load_store_type
6930         and vect_use_strided_gather_scatters_p.
6931
6932 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
6933             Alan Hayward  <alan.hayward@arm.com>
6934             David Sherwood  <david.sherwood@arm.com>
6935
6936         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6937         optional tree argument.
6938         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6939         null target hooks.
6940         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6941         but continue to use the current value as a fallback.
6942         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6943         to compare the updates.
6944         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6945         (get_load_store_type): Use it when handling a strided access.
6946         (vect_get_strided_load_store_ops): New function.
6947         (vect_get_data_ptr_increment): Likewise.
6948         (vectorizable_load): Handle strided gather loads.  Always pass
6949         a step to vect_create_data_ref_ptr and bump_vector_ptr.
6950
6951 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
6952             Alan Hayward  <alan.hayward@arm.com>
6953             David Sherwood  <david.sherwood@arm.com>
6954
6955         * doc/md.texi (gather_load@var{m}): Document.
6956         (mask_gather_load@var{m}): Likewise.
6957         * genopinit.c (main): Add supports_vec_gather_load and
6958         supports_vec_gather_load_cached to target_optabs.
6959         * optabs-tree.c (init_tree_optimization_optabs): Use
6960         ggc_cleared_alloc to allocate target_optabs.
6961         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6962         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6963         functions.
6964         * internal-fn.h (internal_load_fn_p): Declare.
6965         (internal_gather_scatter_fn_p): Likewise.
6966         (internal_fn_mask_index): Likewise.
6967         (internal_gather_scatter_fn_supported_p): Likewise.
6968         * internal-fn.c (gather_load_direct): New macro.
6969         (expand_gather_load_optab_fn): New function.
6970         (direct_gather_load_optab_supported_p): New macro.
6971         (direct_internal_fn_optab): New function.
6972         (internal_load_fn_p): Likewise.
6973         (internal_gather_scatter_fn_p): Likewise.
6974         (internal_fn_mask_index): Likewise.
6975         (internal_gather_scatter_fn_supported_p): Likewise.
6976         * optabs-query.c (supports_at_least_one_mode_p): New function.
6977         (supports_vec_gather_load_p): Likewise.
6978         * optabs-query.h (supports_vec_gather_load_p): Declare.
6979         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6980         and memory_type field.
6981         (NUM_PATTERNS): Bump to 15.
6982         * tree-vect-data-refs.c: Include internal-fn.h.
6983         (vect_gather_scatter_fn_p): New function.
6984         (vect_describe_gather_scatter_call): Likewise.
6985         (vect_check_gather_scatter): Try using internal functions for
6986         gather loads.  Recognize existing calls to a gather load function.
6987         (vect_analyze_data_refs): Consider using gather loads if
6988         supports_vec_gather_load_p.
6989         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6990         (vect_get_gather_scatter_offset_type): Likewise.
6991         (vect_convert_mask_for_vectype): Likewise.
6992         (vect_add_conversion_to_patterm): Likewise.
6993         (vect_try_gather_scatter_pattern): Likewise.
6994         (vect_recog_gather_scatter_pattern): New pattern recognizer.
6995         (vect_vect_recog_func_ptrs): Add it.
6996         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6997         internal_fn_mask_index and internal_gather_scatter_fn_p.
6998         (check_load_store_masking): Take the gather_scatter_info as an
6999         argument and handle gather loads.
7000         (vect_get_gather_scatter_ops): New function.
7001         (vectorizable_call): Check internal_load_fn_p.
7002         (vectorizable_load): Likewise.  Handle gather load internal
7003         functions.
7004         (vectorizable_store): Update call to check_load_store_masking.
7005         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7006         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7007         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7008         (aarch64_gather_scale_operand_d): New predicates.
7009         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7010         (mask_gather_load<mode>): New insns.
7011
7012 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7013             Alan Hayward  <alan.hayward@arm.com>
7014             David Sherwood  <david.sherwood@arm.com>
7015
7016         * optabs.def (fold_left_plus_optab): New optab.
7017         * doc/md.texi (fold_left_plus_@var{m}): Document.
7018         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7019         * internal-fn.c (fold_left_direct): Define.
7020         (expand_fold_left_optab_fn): Likewise.
7021         (direct_fold_left_optab_supported_p): Likewise.
7022         * fold-const-call.c (fold_const_fold_left): New function.
7023         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7024         * tree-parloops.c (valid_reduction_p): New function.
7025         (gather_scalar_reductions): Use it.
7026         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7027         (vect_finish_replace_stmt): Declare.
7028         * tree-vect-loop.c (fold_left_reduction_fn): New function.
7029         (needs_fold_left_reduction_p): New function, split out from...
7030         (vect_is_simple_reduction): ...here.  Accept reductions that
7031         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7032         (vect_force_simple_reduction): Also store the reduction type in
7033         the assignment's STMT_VINFO_REDUC_TYPE.
7034         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7035         (merge_with_identity): New function.
7036         (vect_expand_fold_left): Likewise.
7037         (vectorize_fold_left_reduction): Likewise.
7038         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
7039         scalar phi in place for it.  Check for target support and reject
7040         cases that would reassociate the operation.  Defer the transform
7041         phase to vectorize_fold_left_reduction.
7042         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7043         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7044         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7045
7046 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7047
7048         * tree-if-conv.c (predicate_mem_writes): Remove redundant
7049         call to ifc_temp_var.
7050
7051 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7052             Alan Hayward  <alan.hayward@arm.com>
7053             David Sherwood  <david.sherwood@arm.com>
7054
7055         * target.def (legitimize_address_displacement): Take the original
7056         offset as a poly_int.
7057         * targhooks.h (default_legitimize_address_displacement): Update
7058         accordingly.
7059         * targhooks.c (default_legitimize_address_displacement): Likewise.
7060         * doc/tm.texi: Regenerate.
7061         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7062         as an argument, moving assert of ad->disp == ad->disp_term to...
7063         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
7064         Try calling targetm.legitimize_address_displacement before expanding
7065         the address rather than afterwards, and adjust for the new interface.
7066         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7067         Match the new hook interface.  Handle SVE addresses.
7068         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7069         new hook interface.
7070
7071 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7072
7073         * Makefile.in (OBJS): Add early-remat.o.
7074         * target.def (select_early_remat_modes): New hook.
7075         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7076         * doc/tm.texi: Regenerate.
7077         * targhooks.h (default_select_early_remat_modes): Declare.
7078         * targhooks.c (default_select_early_remat_modes): New function.
7079         * timevar.def (TV_EARLY_REMAT): New timevar.
7080         * passes.def (pass_early_remat): New pass.
7081         * tree-pass.h (make_pass_early_remat): Declare.
7082         * early-remat.c: New file.
7083         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7084         function.
7085         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7086
7087 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7088             Alan Hayward  <alan.hayward@arm.com>
7089             David Sherwood  <david.sherwood@arm.com>
7090
7091         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7092         vfm1 with a bound_epilog parameter.
7093         (vect_do_peeling): Update calls accordingly, and move the prologue
7094         call earlier in the function.  Treat the base bound_epilog as 0 for
7095         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
7096         this base when peeling for gaps.
7097         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7098         with fully-masked loops.
7099         (vect_estimate_min_profitable_iters): Handle the single peeled
7100         iteration in that case.
7101
7102 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7103             Alan Hayward  <alan.hayward@arm.com>
7104             David Sherwood  <david.sherwood@arm.com>
7105
7106         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7107         single-element interleaving even if the size is not a power of 2.
7108         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7109         accesses for single-element interleaving if the group size is
7110         not a power of 2.
7111
7112 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7113             Alan Hayward  <alan.hayward@arm.com>
7114             David Sherwood  <david.sherwood@arm.com>
7115
7116         * doc/md.texi (fold_extract_last_@var{m}): Document.
7117         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7118         * optabs.def (fold_extract_last_optab): New optab.
7119         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7120         * internal-fn.c (fold_extract_direct): New macro.
7121         (expand_fold_extract_optab_fn): Likewise.
7122         (direct_fold_extract_optab_supported_p): Likewise.
7123         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7124         * tree-vect-loop.c (vect_model_reduction_cost): Handle
7125         EXTRACT_LAST_REDUCTION.
7126         (get_initial_def_for_reduction): Do not create an initial vector
7127         for EXTRACT_LAST_REDUCTION reductions.
7128         (vectorizable_reduction): Leave the scalar phi in place for
7129         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
7130         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
7131         epilogue code for EXTRACT_LAST_REDUCTION and defer the
7132         transform phase to vectorizable_condition.
7133         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7134         split out from...
7135         (vect_finish_stmt_generation): ...here.
7136         (vect_finish_replace_stmt): New function.
7137         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7138         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7139         pattern.
7140         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7141
7142 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7143             Alan Hayward  <alan.hayward@arm.com>
7144             David Sherwood  <david.sherwood@arm.com>
7145
7146         * doc/md.texi (extract_last_@var{m}): Document.
7147         * optabs.def (extract_last_optab): New optab.
7148         * internal-fn.def (EXTRACT_LAST): New internal function.
7149         * internal-fn.c (cond_unary_direct): New macro.
7150         (expand_cond_unary_optab_fn): Likewise.
7151         (direct_cond_unary_optab_supported_p): Likewise.
7152         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7153         loops using EXTRACT_LAST.
7154         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7155         (extract_last_<mode>): ...this optab.
7156         (vec_extract<mode><Vel>): Update accordingly.
7157
7158 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7159             Alan Hayward  <alan.hayward@arm.com>
7160             David Sherwood  <david.sherwood@arm.com>
7161
7162         * target.def (empty_mask_is_expensive): New hook.
7163         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7164         * doc/tm.texi: Regenerate.
7165         * targhooks.h (default_empty_mask_is_expensive): Declare.
7166         * targhooks.c (default_empty_mask_is_expensive): New function.
7167         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7168         if the target says that empty masks are expensive.
7169         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7170         New function.
7171         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7172
7173 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7174             Alan Hayward  <alan.hayward@arm.com>
7175             David Sherwood  <david.sherwood@arm.com>
7176
7177         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7178         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7179         (vect_use_loop_mask_for_alignment_p): New function.
7180         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7181         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7182         niters_skip argument.  Make sure that the first niters_skip elements
7183         of the first iteration are inactive.
7184         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7185         Update call to vect_set_loop_masks_directly.
7186         (get_misalign_in_elems): New function, split out from...
7187         (vect_gen_prolog_loop_niters): ...here.
7188         (vect_update_init_of_dr): Take a code argument that specifies whether
7189         the adjustment should be added or subtracted.
7190         (vect_update_init_of_drs): Likewise.
7191         (vect_prepare_for_masked_peels): New function.
7192         (vect_do_peeling): Skip prologue peeling if we're using a mask
7193         instead.  Update call to vect_update_inits_of_drs.
7194         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7195         mask_skip_niters.
7196         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7197         alignment.  Do not include the number of peeled iterations in
7198         the minimum threshold in that case.
7199         (vectorizable_induction): Adjust the start value down by
7200         LOOP_VINFO_MASK_SKIP_NITERS iterations.
7201         (vect_transform_loop): Call vect_prepare_for_masked_peels.
7202         Take the number of skipped iterations into account when calculating
7203         the loop bounds.
7204         * tree-vect-stmts.c (vect_gen_while_not): New function.
7205
7206 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7207             Alan Hayward  <alan.hayward@arm.com>
7208             David Sherwood  <david.sherwood@arm.com>
7209
7210         * doc/sourcebuild.texi (vect_fully_masked): Document.
7211         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7212         default value to 0.
7213         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7214         split out from...
7215         (vect_analyze_loop_2): ...here. Don't check the vectorization
7216         factor against the number of loop iterations if the loop is
7217         fully-masked.
7218
7219 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7220             Alan Hayward  <alan.hayward@arm.com>
7221             David Sherwood  <david.sherwood@arm.com>
7222
7223         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7224         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7225         (dump_groups): Update accordingly.
7226         (iv_use::mem_type): New member variable.
7227         (address_p): New function.
7228         (record_use): Add a mem_type argument and initialize the new
7229         mem_type field.
7230         (record_group_use): Add a mem_type argument.  Use address_p.
7231         Remove obsolete null checks of base_object.  Update call to record_use.
7232         (find_interesting_uses_op): Update call to record_group_use.
7233         (find_interesting_uses_cond): Likewise.
7234         (find_interesting_uses_address): Likewise.
7235         (get_mem_type_for_internal_fn): New function.
7236         (find_address_like_use): Likewise.
7237         (find_interesting_uses_stmt): Try find_address_like_use before
7238         calling find_interesting_uses_op.
7239         (addr_offset_valid_p): Use the iv mem_type field as the type
7240         of the addressed memory.
7241         (add_autoinc_candidates): Likewise.
7242         (get_address_cost): Likewise.
7243         (split_small_address_groups_p): Use address_p.
7244         (split_address_groups): Likewise.
7245         (add_iv_candidate_for_use): Likewise.
7246         (autoinc_possible_for_pair): Likewise.
7247         (rewrite_groups): Likewise.
7248         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7249         (determine_group_iv_cost): Update after split of USE_ADDRESS.
7250         (get_alias_ptr_type_for_ptr_address): New function.
7251         (rewrite_use_address): Rewrite address uses in calls that were
7252         identified by find_address_like_use.
7253
7254 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7255             Alan Hayward  <alan.hayward@arm.com>
7256             David Sherwood  <david.sherwood@arm.com>
7257
7258         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7259         TARGET_MEM_REFs.
7260         * gimple-expr.h (is_gimple_addressable: Likewise.
7261         * gimple-expr.c (is_gimple_address): Likewise.
7262         * internal-fn.c (expand_call_mem_ref): New function.
7263         (expand_mask_load_optab_fn): Use it.
7264         (expand_mask_store_optab_fn): Likewise.
7265
7266 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7267             Alan Hayward  <alan.hayward@arm.com>
7268             David Sherwood  <david.sherwood@arm.com>
7269
7270         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7271         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7272         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7273         (cond_umax@var{mode}): Document.
7274         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7275         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7276         (cond_umin_optab, cond_umax_optab): New optabs.
7277         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7278         (COND_IOR, COND_XOR): New internal functions.
7279         * internal-fn.h (get_conditional_internal_fn): Declare.
7280         * internal-fn.c (cond_binary_direct): New macro.
7281         (expand_cond_binary_optab_fn): Likewise.
7282         (direct_cond_binary_optab_supported_p): Likewise.
7283         (get_conditional_internal_fn): New function.
7284         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7285         Cope with reduction statements that are vectorized as calls rather
7286         than assignments.
7287         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7288         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7289         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7290         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7291         (UNSPEC_COND_EOR): New unspecs.
7292         (optab): Add mappings for them.
7293         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7294         (sve_int_op, sve_fp_op): New int attributes.
7295
7296 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7297             Alan Hayward  <alan.hayward@arm.com>
7298             David Sherwood  <david.sherwood@arm.com>
7299
7300         * optabs.def (while_ult_optab): New optab.
7301         * doc/md.texi (while_ult@var{m}@var{n}): Document.
7302         * internal-fn.def (WHILE_ULT): New internal function.
7303         * internal-fn.h (direct_internal_fn_supported_p): New override
7304         that takes two types as argument.
7305         * internal-fn.c (while_direct): New macro.
7306         (expand_while_optab_fn): New function.
7307         (convert_optab_supported_p): Likewise.
7308         (direct_while_optab_supported_p): New macro.
7309         * wide-int.h (wi::udiv_ceil): New function.
7310         * tree-vectorizer.h (rgroup_masks): New structure.
7311         (vec_loop_masks): New typedef.
7312         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7313         and fully_masked_p.
7314         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7315         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7316         (vect_max_vf): New function.
7317         (slpeel_make_loop_iterate_ntimes): Delete.
7318         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7319         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7320         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7321         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7322         internal-fn.h, stor-layout.h and optabs-query.h.
7323         (vect_set_loop_mask): New function.
7324         (add_preheader_seq): Likewise.
7325         (add_header_seq): Likewise.
7326         (interleave_supported_p): Likewise.
7327         (vect_maybe_permute_loop_masks): Likewise.
7328         (vect_set_loop_masks_directly): Likewise.
7329         (vect_set_loop_condition_masked): Likewise.
7330         (vect_set_loop_condition_unmasked): New function, split out from
7331         slpeel_make_loop_iterate_ntimes.
7332         (slpeel_make_loop_iterate_ntimes): Rename to..
7333         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
7334         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7335         (vect_do_peeling): Update call accordingly.
7336         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7337         loops.
7338         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7339         mask_compare_type, can_fully_mask_p and fully_masked_p.
7340         (release_vec_loop_masks): New function.
7341         (_loop_vec_info): Use it to free the loop masks.
7342         (can_produce_all_loop_masks_p): New function.
7343         (vect_get_max_nscalars_per_iter): Likewise.
7344         (vect_verify_full_masking): Likewise.
7345         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7346         retries, and free the mask rgroups before retrying.  Check loop-wide
7347         reasons for disallowing fully-masked loops.  Make the final decision
7348         about whether use a fully-masked loop or not.
7349         (vect_estimate_min_profitable_iters): Do not assume that peeling
7350         for the number of iterations will be needed for fully-masked loops.
7351         (vectorizable_reduction): Disable fully-masked loops.
7352         (vectorizable_live_operation): Likewise.
7353         (vect_halve_mask_nunits): New function.
7354         (vect_double_mask_nunits): Likewise.
7355         (vect_record_loop_mask): Likewise.
7356         (vect_get_loop_mask): Likewise.
7357         (vect_transform_loop): Handle the case in which the final loop
7358         iteration might handle a partial vector.  Call vect_set_loop_condition
7359         instead of slpeel_make_loop_iterate_ntimes.
7360         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7361         (check_load_store_masking): New function.
7362         (prepare_load_store_mask): Likewise.
7363         (vectorizable_store): Handle fully-masked loops.
7364         (vectorizable_load): Likewise.
7365         (supportable_widening_operation): Use vect_halve_mask_nunits for
7366         booleans.
7367         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7368         (vect_gen_while): New function.
7369         * config/aarch64/aarch64.md (umax<mode>3): New expander.
7370         (aarch64_uqdec<mode>): New insn.
7371
7372 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7373             Alan Hayward  <alan.hayward@arm.com>
7374             David Sherwood  <david.sherwood@arm.com>
7375
7376         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7377         (reduc_xor_scal_optab): New optabs.
7378         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7379         (reduc_xor_scal_@var{m}): Document.
7380         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7381         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7382         internal functions.
7383         * fold-const-call.c (fold_const_call): Handle them.
7384         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7385         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7386         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7387         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7388         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7389         (UNSPEC_XORV): New unspecs.
7390         (optab): Add entries for them.
7391         (BITWISEV): New int iterator.
7392         (bit_reduc_op): New int attributes.
7393
7394 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7395             Alan Hayward  <alan.hayward@arm.com>
7396             David Sherwood  <david.sherwood@arm.com>
7397
7398         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7399         * internal-fn.def (VEC_SHL_INSERT): New internal function.
7400         * optabs.def (vec_shl_insert_optab): New optab.
7401         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7402         (duplicate_and_interleave): Likewise.
7403         * tree-vect-loop.c: Include internal-fn.h.
7404         (neutral_op_for_slp_reduction): New function, split out from
7405         get_initial_defs_for_reduction.
7406         (get_initial_def_for_reduction): Handle option 2 for variable-length
7407         vectors by loading the neutral value into a vector and then shifting
7408         the initial value into element 0.
7409         (get_initial_defs_for_reduction): Replace the code argument with
7410         the neutral value calculated by neutral_op_for_slp_reduction.
7411         Use gimple_build_vector for constant-length vectors.
7412         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7413         but the first group_size elements have a neutral value.
7414         Use duplicate_and_interleave otherwise.
7415         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7416         Update call to get_initial_defs_for_reduction.  Handle SLP
7417         reductions for variable-length vectors by creating one vector
7418         result for each scalar result, with the elements associated
7419         with other scalar results stubbed out with the neutral value.
7420         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7421         Require IFN_VEC_SHL_INSERT for double reductions on
7422         variable-length vectors, or SLP reductions that have
7423         a neutral value.  Require can_duplicate_and_interleave_p
7424         support for variable-length unchained SLP reductions if there
7425         is no neutral value, such as for MIN/MAX reductions.  Also require
7426         the number of vector elements to be a multiple of the number of
7427         SLP statements when doing variable-length unchained SLP reductions.
7428         Update call to vect_create_epilog_for_reduction.
7429         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7430         and remove initial values.
7431         (duplicate_and_interleave): Make public.
7432         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7433         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7434
7435 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7436             Alan Hayward  <alan.hayward@arm.com>
7437             David Sherwood  <david.sherwood@arm.com>
7438
7439         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7440         (can_duplicate_and_interleave_p): New function.
7441         (vect_get_and_check_slp_defs): Take the vector of statements
7442         rather than just the current one.  Remove excess parentheses.
7443         Restriction rejectinon of vect_constant_def and vect_external_def
7444         for variable-length vectors to boolean types, or types for which
7445         can_duplicate_and_interleave_p is false.
7446         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7447         (duplicate_and_interleave): New function.
7448         (vect_get_constant_vectors): Use gimple_build_vector for
7449         constant-length vectors and suitable variable-length constant
7450         vectors.  Use duplicate_and_interleave for other variable-length
7451         vectors.  Don't defer the update when inserting new statements.
7452
7453 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7454             Alan Hayward  <alan.hayward@arm.com>
7455             David Sherwood  <david.sherwood@arm.com>
7456
7457         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7458         min_profitable_iters doesn't go negative.
7459
7460 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7461             Alan Hayward  <alan.hayward@arm.com>
7462             David Sherwood  <david.sherwood@arm.com>
7463
7464         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7465         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7466         * optabs.def (vec_mask_load_lanes_optab): New optab.
7467         (vec_mask_store_lanes_optab): Likewise.
7468         * internal-fn.def (MASK_LOAD_LANES): New internal function.
7469         (MASK_STORE_LANES): Likewise.
7470         * internal-fn.c (mask_load_lanes_direct): New macro.
7471         (mask_store_lanes_direct): Likewise.
7472         (expand_mask_load_optab_fn): Handle masked operations.
7473         (expand_mask_load_lanes_optab_fn): New macro.
7474         (expand_mask_store_optab_fn): Handle masked operations.
7475         (expand_mask_store_lanes_optab_fn): New macro.
7476         (direct_mask_load_lanes_optab_supported_p): Likewise.
7477         (direct_mask_store_lanes_optab_supported_p): Likewise.
7478         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7479         parameter.
7480         (vect_load_lanes_supported): Likewise.
7481         * tree-vect-data-refs.c (strip_conversion): New function.
7482         (can_group_stmts_p): Likewise.
7483         (vect_analyze_data_ref_accesses): Use it instead of checking
7484         for a pair of assignments.
7485         (vect_store_lanes_supported): Take a masked_p parameter.
7486         (vect_load_lanes_supported): Likewise.
7487         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7488         vect_store_lanes_supported and vect_load_lanes_supported.
7489         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7490         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7491         parameter.  Don't allow gaps for masked accesses.
7492         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
7493         and vect_load_lanes_supported.
7494         (get_load_store_type): Take a masked_p parameter and update
7495         call to get_group_load_store_type.
7496         (vectorizable_store): Update call to get_load_store_type.
7497         Handle IFN_MASK_STORE_LANES.
7498         (vectorizable_load): Update call to get_load_store_type.
7499         Handle IFN_MASK_LOAD_LANES.
7500
7501 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7502             Alan Hayward  <alan.hayward@arm.com>
7503             David Sherwood  <david.sherwood@arm.com>
7504
7505         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7506         modes for SVE.
7507         * config/aarch64/aarch64-protos.h
7508         (aarch64_sve_struct_memory_operand_p): Declare.
7509         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7510         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7511         (VPRED, vpred): Handle SVE structure modes.
7512         * config/aarch64/constraints.md (Utx): New constraint.
7513         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7514         (aarch64_sve_struct_nonimmediate_operand): New predicates.
7515         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7516         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7517         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7518         structure modes.  Split into pieces after RA.
7519         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7520         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7521         New patterns.
7522         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7523         SVE structure modes.
7524         (aarch64_classify_address): Likewise.
7525         (sizetochar): Move earlier in file.
7526         (aarch64_print_operand): Handle SVE register lists.
7527         (aarch64_array_mode): New function.
7528         (aarch64_sve_struct_memory_operand_p): Likewise.
7529         (TARGET_ARRAY_MODE): Redefine.
7530
7531 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7532             Alan Hayward  <alan.hayward@arm.com>
7533             David Sherwood  <david.sherwood@arm.com>
7534
7535         * target.def (array_mode): New target hook.
7536         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7537         * doc/tm.texi: Regenerate.
7538         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7539         * hooks.c (hook_optmode_mode_uhwi_none): New function.
7540         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7541         targetm.array_mode.
7542         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
7543         type sizes.
7544
7545 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7546             Alan Hayward  <alan.hayward@arm.com>
7547             David Sherwood  <david.sherwood@arm.com>
7548
7549         * fold-const.c (fold_binary_loc): Check the argument types
7550         rather than the result type when testing for a vector operation.
7551
7552 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7553
7554         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7555         * doc/tm.texi: Regenerate.
7556
7557 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7558             Alan Hayward  <alan.hayward@arm.com>
7559             David Sherwood  <david.sherwood@arm.com>
7560
7561         * doc/invoke.texi (-msve-vector-bits=): Document new option.
7562         (sve): Document new AArch64 extension.
7563         * doc/md.texi (w): Extend the description of the AArch64
7564         constraint to include SVE vectors.
7565         (Upl, Upa): Document new AArch64 predicate constraints.
7566         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7567         enum.
7568         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7569         (msve-vector-bits=): New option.
7570         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7571         SVE when these are disabled.
7572         (sve): New extension.
7573         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7574         modes.  Adjust their number of units based on aarch64_sve_vg.
7575         (MAX_BITSIZE_MODE_ANY_MODE): Define.
7576         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7577         aarch64_addr_query_type.
7578         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7579         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7580         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7581         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7582         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7583         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7584         (aarch64_simd_imm_zero_p): Delete.
7585         (aarch64_check_zero_based_sve_index_immediate): Declare.
7586         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7587         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7588         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7589         (aarch64_sve_float_mul_immediate_p): Likewise.
7590         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7591         rather than an rtx.
7592         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7593         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7594         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7595         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7596         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7597         (aarch64_regmode_natural_size): Likewise.
7598         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7599         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7600         left one place.
7601         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7602         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7603         for VG and the SVE predicate registers.
7604         (V_ALIASES): Add a "z"-prefixed alias.
7605         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7606         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7607         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7608         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7609         (REG_CLASS_NAMES): Add entries for them.
7610         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
7611         and the predicate registers.
7612         (aarch64_sve_vg): Declare.
7613         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7614         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7615         (REGMODE_NATURAL_SIZE): Define.
7616         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7617         SVE macros.
7618         * config/aarch64/aarch64.c: Include cfgrtl.h.
7619         (simd_immediate_info): Add a constructor for series vectors,
7620         and an associated step field.
7621         (aarch64_sve_vg): New variable.
7622         (aarch64_dbx_register_number): Handle VG and the predicate registers.
7623         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7624         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7625         (VEC_ANY_DATA, VEC_STRUCT): New constants.
7626         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7627         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7628         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7629         (aarch64_get_mask_mode): New functions.
7630         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7631         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7632         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
7633         predicate modes and predicate registers.  Explicitly restrict
7634         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
7635         to store a vector mode if it is recognized by
7636         aarch64_classify_vector_mode.
7637         (aarch64_regmode_natural_size): New function.
7638         (aarch64_hard_regno_caller_save_mode): Return the original mode
7639         for predicates.
7640         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7641         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7642         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7643         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7644         functions.
7645         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
7646         does not overlap dest if the function is frame-related.  Handle
7647         SVE constants.
7648         (aarch64_split_add_offset): New function.
7649         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7650         them aarch64_add_offset.
7651         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7652         and update call to aarch64_sub_sp.
7653         (aarch64_add_cfa_expression): New function.
7654         (aarch64_expand_prologue): Pass extra temporary registers to the
7655         functions above.  Handle the case in which we need to emit new
7656         DW_CFA_expressions for registers that were originally saved
7657         relative to the stack pointer, but now have to be expressed
7658         relative to the frame pointer.
7659         (aarch64_output_mi_thunk): Pass extra temporary registers to the
7660         functions above.
7661         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
7662         IP0 and IP1 values for SVE frames.
7663         (aarch64_expand_vec_series): New function.
7664         (aarch64_expand_sve_widened_duplicate): Likewise.
7665         (aarch64_expand_sve_const_vector): Likewise.
7666         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7667         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
7668         into the register, rather than emitting a SET directly.
7669         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7670         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7671         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7672         (offset_9bit_signed_scaled_p): New functions.
7673         (aarch64_replicate_bitmask_imm): New function.
7674         (aarch64_bitmask_imm): Use it.
7675         (aarch64_cannot_force_const_mem): Reject expressions involving
7676         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
7677         (aarch64_classify_index): Handle SVE indices, by requiring
7678         a plain register index with a scale that matches the element size.
7679         (aarch64_classify_address): Handle SVE addresses.  Assert that
7680         the mode of the address is VOIDmode or an integer mode.
7681         Update call to aarch64_classify_symbol.
7682         (aarch64_classify_symbolic_expression): Update call to
7683         aarch64_classify_symbol.
7684         (aarch64_const_vec_all_in_range_p): New function.
7685         (aarch64_print_vector_float_operand): Likewise.
7686         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
7687         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
7688         and the FP immediates 1.0 and 0.5.
7689         (aarch64_print_address_internal): Handle SVE addresses.
7690         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7691         (aarch64_regno_regclass): Handle predicate registers.
7692         (aarch64_secondary_reload): Handle big-endian reloads of SVE
7693         data modes.
7694         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7695         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7696         (aarch64_convert_sve_vector_bits): New function.
7697         (aarch64_override_options): Use it to handle -msve-vector-bits=.
7698         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7699         rather than an rtx.
7700         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7701         Handle SVE vector and predicate modes.  Accept VL-based constants
7702         that need only one temporary register, and VL offsets that require
7703         no temporary registers.
7704         (aarch64_conditional_register_usage): Mark the predicate registers
7705         as fixed if SVE isn't available.
7706         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7707         Return true for SVE vector and predicate modes.
7708         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7709         rather than an unsigned int.  Handle SVE modes.
7710         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
7711         SVE modes.
7712         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7713         if SVE is enabled.
7714         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7715         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7716         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7717         (aarch64_sve_float_mul_immediate_p): New functions.
7718         (aarch64_sve_valid_immediate): New function.
7719         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7720         Explicitly reject structure modes.  Check for INDEX constants.
7721         Handle PTRUE and PFALSE constants.
7722         (aarch64_check_zero_based_sve_index_immediate): New function.
7723         (aarch64_simd_imm_zero_p): Delete.
7724         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7725         vector modes.  Accept constants in the range of CNT[BHWD].
7726         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7727         ask for an Advanced SIMD mode.
7728         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7729         (aarch64_simd_vector_alignment): Handle SVE predicates.
7730         (aarch64_vectorize_preferred_vector_alignment): New function.
7731         (aarch64_simd_vector_alignment_reachable): Use it instead of
7732         the vector size.
7733         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7734         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7735         functions.
7736         (MAX_VECT_LEN): Delete.
7737         (expand_vec_perm_d): Add a vec_flags field.
7738         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7739         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7740         (aarch64_evpc_ext): Don't apply a big-endian lane correction
7741         for SVE modes.
7742         (aarch64_evpc_rev): Rename to...
7743         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
7744         (aarch64_evpc_rev_global): New function.
7745         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7746         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7747         MAX_VECT_LEN.
7748         (aarch64_evpc_sve_tbl): New function.
7749         (aarch64_expand_vec_perm_const_1): Update after rename of
7750         aarch64_evpc_rev.  Handle SVE permutes too, trying
7751         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7752         than aarch64_evpc_tbl.
7753         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7754         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7755         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7756         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7757         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7758         (aarch64_expand_sve_vcond): New functions.
7759         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7760         of aarch64_vector_mode_p.
7761         (aarch64_dwarf_poly_indeterminate_value): New function.
7762         (aarch64_compute_pressure_classes): Likewise.
7763         (aarch64_can_change_mode_class): Likewise.
7764         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7765         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7766         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7767         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7768         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7769         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7770         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7771         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7772         constraints.
7773         (Dn, Dl, Dr): Accept const as well as const_vector.
7774         (Dz): Likewise.  Compare against CONST0_RTX.
7775         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7776         of "vector" where appropriate.
7777         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7778         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7779         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7780         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7781         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7782         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7783         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7784         (v_int_equiv): Extend to SVE modes.
7785         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7786         mode attributes.
7787         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7788         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7789         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7790         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7791         (SVE_COND_FP_CMP): New int iterators.
7792         (perm_hilo): Handle the new unpack unspecs.
7793         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7794         attributes.
7795         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7796         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7797         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7798         (aarch64_equality_operator, aarch64_constant_vector_operand)
7799         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7800         (aarch64_sve_nonimmediate_operand): Likewise.
7801         (aarch64_sve_general_operand): Likewise.
7802         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7803         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7804         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7805         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7806         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7807         (aarch64_sve_float_arith_immediate): Likewise.
7808         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7809         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7810         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7811         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7812         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7813         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7814         (aarch64_sve_float_arith_operand): Likewise.
7815         (aarch64_sve_float_arith_with_sub_operand): Likewise.
7816         (aarch64_sve_float_mul_operand): Likewise.
7817         (aarch64_sve_vec_perm_operand): Likewise.
7818         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7819         (aarch64_mov_operand): Accept const_poly_int and const_vector.
7820         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7821         as well as const_vector.
7822         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7823         in file.  Use CONST0_RTX and CONSTM1_RTX.
7824         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
7825         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7826         Use aarch64_simd_imm_zero.
7827         * config/aarch64/aarch64-sve.md: New file.
7828         * config/aarch64/aarch64.md: Include it.
7829         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7830         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7831         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7832         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7833         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7834         (sve): New attribute.
7835         (enabled): Disable instructions with the sve attribute unless
7836         TARGET_SVE.
7837         (movqi, movhi): Pass CONST_POLY_INT operaneds through
7838         aarch64_expand_mov_immediate.
7839         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7840         CNT[BHSD] immediates.
7841         (movti): Split CONST_POLY_INT moves into two halves.
7842         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7843         Split additions that need a temporary here if the destination
7844         is the stack pointer.
7845         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7846         (*add<mode>3_poly_1): New instruction.
7847         (set_clobber_cc): New expander.
7848
7849 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7850
7851         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7852         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
7853         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7854         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7855         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
7856         Change innermode from fixed_mode_size to machine_mode.
7857         (simplify_subreg): Update call accordingly.  Handle a constant-sized
7858         subreg of a variable-length CONST_VECTOR.
7859
7860 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
7861             Alan Hayward  <alan.hayward@arm.com>
7862             David Sherwood  <david.sherwood@arm.com>
7863
7864         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7865         (add_offset_to_base): New function, split out from...
7866         (create_mem_ref): ...here.  When handling a scale other than 1,
7867         check first whether the address is valid without the offset.
7868         Add it into the base if so, leaving the index and scale as-is.
7869
7870 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
7871
7872         PR c++/83778
7873         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7874         fold_for_warn before checking if arg2 is INTEGER_CST.
7875
7876 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
7877
7878         * config/rs6000/predicates.md (load_multiple_operation): Delete.
7879         (store_multiple_operation): Delete.
7880         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7881         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7882         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7883         guarded by TARGET_STRING.
7884         (rs6000_output_load_multiple): Delete.
7885         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7886         OPTION_MASK_STRING / TARGET_STRING handling.
7887         (print_operand) <'N', 'O'>: Add comment that these are unused now.
7888         (const rs6000_opt_masks) <"string">: Change mask to 0.
7889         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7890         (MASK_STRING): Delete.
7891         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7892         parts.  Simplify.
7893         (load_multiple): Delete.
7894         (*ldmsi8): Delete.
7895         (*ldmsi7): Delete.
7896         (*ldmsi6): Delete.
7897         (*ldmsi5): Delete.
7898         (*ldmsi4): Delete.
7899         (*ldmsi3): Delete.
7900         (store_multiple): Delete.
7901         (*stmsi8): Delete.
7902         (*stmsi7): Delete.
7903         (*stmsi6): Delete.
7904         (*stmsi5): Delete.
7905         (*stmsi4): Delete.
7906         (*stmsi3): Delete.
7907         (movmemsi_8reg): Delete.
7908         (corresponding unnamed define_insn): Delete.
7909         (movmemsi_6reg): Delete.
7910         (corresponding unnamed define_insn): Delete.
7911         (movmemsi_4reg): Delete.
7912         (corresponding unnamed define_insn): Delete.
7913         (movmemsi_2reg): Delete.
7914         (corresponding unnamed define_insn): Delete.
7915         (movmemsi_1reg): Delete.
7916         (corresponding unnamed define_insn): Delete.
7917         * config/rs6000/rs6000.opt (mno-string): New.
7918         (mstring): Replace by deprecation warning stub.
7919         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7920
7921 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
7922
7923         * regrename.c (regrename_do_replace): If replacing the same
7924         reg multiple times, try to reuse last created gen_raw_REG.
7925
7926         PR debug/81155
7927         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7928         main to workaround a bug in GDB.
7929
7930 2018-01-12  Tom de Vries  <tom@codesourcery.com>
7931
7932         PR target/83737
7933         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7934
7935 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
7936
7937         PR rtl-optimization/80481
7938         * ira-color.c (get_cap_member): New function.
7939         (allocnos_conflict_by_live_ranges_p): Use it.
7940         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7941         (setup_slot_coalesced_allocno_live_ranges): Ditto.
7942
7943 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
7944
7945         PR target/83628
7946         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7947         (*saddl_se_1): Ditto.
7948         (*ssubsi_1): Ditto.
7949         (*ssubl_se_1): Ditto.
7950
7951 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
7952
7953         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7954         rather than wi::to_widest for DR_INITs.
7955         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7956         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7957         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7958         INTEGER_CSTs.
7959         (vect_analyze_group_access_1): Note that here.
7960
7961 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
7962
7963         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7964         polynomial type sizes.
7965
7966 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
7967
7968         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7969         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7970         (gimple_add_tmp_var): Likewise.
7971
7972 2018-01-12  Martin Liska  <mliska@suse.cz>
7973
7974         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7975         (gimple_alloc_sizes): Likewise.
7976         (dump_gimple_statistics): Use PRIu64 in printf format.
7977         * gimple.h: Change uint64_t to int.
7978
7979 2018-01-12  Martin Liska  <mliska@suse.cz>
7980
7981         * tree-core.h: Use uint64_t instead of int.
7982         * tree.c (tree_node_counts): Likewise.
7983         (tree_node_sizes): Likewise.
7984         (dump_tree_statistics): Use PRIu64 in printf format.
7985
7986 2018-01-12  Martin Liska  <mliska@suse.cz>
7987
7988         * Makefile.in: As qsort_chk is implemented in vec.c, add
7989         vec.o to linkage of gencfn-macros.
7990         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7991         passing the info to record_node_allocation_statistics.
7992         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7993         and pass the info.
7994         * ggc-common.c (struct ggc_usage): Add operator== and use
7995         it in operator< and compare function.
7996         * mem-stats.h (struct mem_usage): Likewise.
7997         * vec.c (struct vec_usage): Remove operator< and compare
7998         function. Can be simply inherited.
7999
8000 2018-01-12  Martin Jambor  <mjambor@suse.cz>
8001
8002         PR target/81616
8003         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8004         * tree-ssa-math-opts.c: Include domwalk.h.
8005         (convert_mult_to_fma_1): New function.
8006         (fma_transformation_info): New type.
8007         (fma_deferring_state): Likewise.
8008         (cancel_fma_deferring): New function.
8009         (result_of_phi): Likewise.
8010         (last_fma_candidate_feeds_initial_phi): Likewise.
8011         (convert_mult_to_fma): Added deferring logic, split actual
8012         transformation to convert_mult_to_fma_1.
8013         (math_opts_dom_walker): New type.
8014         (math_opts_dom_walker::after_dom_children): New method, body moved
8015         here from pass_optimize_widening_mul::execute, added deferring logic
8016         bits.
8017         (pass_optimize_widening_mul::execute): Moved most of code to
8018         math_opts_dom_walker::after_dom_children.
8019         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8020         * config/i386/i386.c (ix86_option_override_internal): Added
8021         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8022
8023 2018-01-12  Richard Biener  <rguenther@suse.de>
8024
8025         PR debug/83157
8026         * dwarf2out.c (gen_variable_die): Do not reset old_die for
8027         inline instance vars.
8028
8029 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
8030
8031         PR target/81819
8032         * config/rx/rx.c (rx_is_restricted_memory_address):
8033         Handle SUBREG case.
8034
8035 2018-01-12  Richard Biener  <rguenther@suse.de>
8036
8037         PR tree-optimization/80846
8038         * target.def (split_reduction): New target hook.
8039         * targhooks.c (default_split_reduction): New function.
8040         * targhooks.h (default_split_reduction): Declare.
8041         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8042         target requests first reduce vectors by combining low and high
8043         parts.
8044         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8045         (get_vectype_for_scalar_type_and_size): Export.
8046         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8047         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8048         * doc/tm.texi: Regenerate.
8049         * config/i386/i386.c (ix86_split_reduction): Implement
8050         TARGET_VECTORIZE_SPLIT_REDUCTION.
8051
8052 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
8053
8054         PR target/83368
8055         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8056         in PIC mode except for TARGET_VXWORKS_RTP.
8057         * config/sparc/sparc.c: Include cfgrtl.h.
8058         (TARGET_INIT_PIC_REG): Define.
8059         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8060         (sparc_pic_register_p): New predicate.
8061         (sparc_legitimate_address_p): Use it.
8062         (sparc_legitimize_pic_address): Likewise.
8063         (sparc_delegitimize_address): Likewise.
8064         (sparc_mode_dependent_address_p): Likewise.
8065         (gen_load_pcrel_sym): Remove 4th parameter.
8066         (load_got_register): Adjust call to above.  Remove obsolete stuff.
8067         (sparc_expand_prologue): Do not call load_got_register here.
8068         (sparc_flat_expand_prologue): Likewise.
8069         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8070         (sparc_use_pseudo_pic_reg): New function.
8071         (sparc_init_pic_reg): Likewise.
8072         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8073         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8074
8075 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
8076
8077         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8078         Add item for branch_cost.
8079
8080 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
8081
8082         PR rtl-optimization/83565
8083         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8084         not extend the result to a larger mode for rotate operations.
8085         (num_sign_bit_copies1): Likewise.
8086
8087 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8088
8089         PR target/40411
8090         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8091         -symbolic.
8092         Use values-Xc.o for -pedantic.
8093         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8094
8095 2018-01-12  Martin Liska  <mliska@suse.cz>
8096
8097         PR ipa/83054
8098         * ipa-devirt.c (final_warning_record::grow_type_warnings):
8099         New function.
8100         (possible_polymorphic_call_targets): Use it.
8101         (ipa_devirt): Likewise.
8102
8103 2018-01-12  Martin Liska  <mliska@suse.cz>
8104
8105         * profile-count.h (enum profile_quality): Use 0 as invalid
8106         enum value of profile_quality.
8107
8108 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8109
8110         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8111         -mext-string options.
8112
8113 2018-01-12  Richard Biener  <rguenther@suse.de>
8114
8115         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8116         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8117         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8118         Likewise.
8119         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8120
8121 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
8122
8123         * configure.ac (--with-long-double-format): Add support for the
8124         configuration option to change the default long double format on
8125         PowerPC systems.
8126         * config.gcc (powerpc*-linux*-*): Likewise.
8127         * configure: Regenerate.
8128         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8129         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8130         used without modification.
8131
8132 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8133
8134         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8135         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8136         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8137         MISC_BUILTIN_SPEC_BARRIER.
8138         (rs6000_init_builtins): Likewise.
8139         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8140         enum value.
8141         (speculation_barrier): New define_insn.
8142         * doc/extend.texi: Document __builtin_speculation_barrier.
8143
8144 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
8145
8146         PR target/83203
8147         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8148         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8149         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8150         iterators.
8151         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
8152         integral modes instead of "ss" and "sd".
8153         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8154         vectors with 32-bit and 64-bit elements.
8155         (vecdupssescalarmodesuffix): New mode attribute.
8156         (vec_dup<mode>): Use it.
8157
8158 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
8159
8160         PR target/83330
8161         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8162         frame if argument is passed on stack.
8163
8164 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
8165
8166         PR target/82682
8167         * ree.c (combine_reaching_defs): Optimize also
8168         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8169         reg2=any_extend(exp); reg1=reg2;, formatting fix.
8170
8171 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
8172
8173         PR middle-end/83189
8174         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8175
8176 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
8177
8178         PR middle-end/83718
8179         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8180         after they are computed.
8181
8182 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
8183
8184         PR tree-optimization/83695
8185         * gimple-loop-linterchange.cc
8186         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8187         reset cached scev information after interchange.
8188         (pass_linterchange::execute): Remove call to scev_reset_htab.
8189
8190 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8191
8192         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8193         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8194         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8195         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8196         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8197         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8198         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8199         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8200         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8201         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8202         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8203         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8204         (V_lane_reg): Likewise.
8205         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8206         New define_expand.
8207         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8208         (vfmal_lane_low<mode>_intrinsic,
8209         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8210         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8211         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8212         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8213         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8214         vfmsl_lane_high<mode>_intrinsic): New define_insns.
8215
8216 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8217
8218         * config/arm/arm-cpus.in (fp16fml): New feature.
8219         (ALL_SIMD): Add fp16fml.
8220         (armv8.2-a): Add fp16fml as an option.
8221         (armv8.3-a): Likewise.
8222         (armv8.4-a): Add fp16fml as part of fp16.
8223         * config/arm/arm.h (TARGET_FP16FML): Define.
8224         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8225         when appropriate.
8226         * config/arm/arm-modes.def (V2HF): Define.
8227         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8228         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8229         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8230         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8231         vfmsl_low, vfmsl_high): New set of builtins.
8232         * config/arm/iterators.md (PLUSMINUS): New code iterator.
8233         (vfml_op): New code attribute.
8234         (VFMLHALVES): New int iterator.
8235         (VFML, VFMLSEL): New mode attributes.
8236         (V_reg): Define mapping for V2HF.
8237         (V_hi, V_lo): New mode attributes.
8238         (VF_constraint): Likewise.
8239         (vfml_half, vfml_half_selector): New int attributes.
8240         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8241         define_expand.
8242         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8243         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8244         New define_insn.
8245         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8246         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8247         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8248         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
8249         documentation.
8250         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8251         Document new effective target and option set.
8252
8253 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8254
8255         * config/arm/arm-cpus.in (armv8_4): New feature.
8256         (ARMv8_4a): New fgroup.
8257         (armv8.4-a): New arch.
8258         * config/arm/arm-tables.opt: Regenerate.
8259         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8260         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8261         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8262         Add matching rules for -march=armv8.4-a and extensions.
8263         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8264
8265 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
8266
8267         PR target/81821
8268         * config/rx/rx.md (BW): New mode attribute.
8269         (sync_lock_test_and_setsi): Add mode suffix to insn output.
8270
8271 2018-01-11  Richard Biener  <rguenther@suse.de>
8272
8273         PR tree-optimization/83435
8274         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8275         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8276         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8277
8278 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
8279             Alan Hayward  <alan.hayward@arm.com>
8280             David Sherwood  <david.sherwood@arm.com>
8281
8282         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8283         field.
8284         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
8285         (aarch64_print_address_internal): Use it to check for a zero offset.
8286
8287 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
8288             Alan Hayward  <alan.hayward@arm.com>
8289             David Sherwood  <david.sherwood@arm.com>
8290
8291         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8292         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8293         Return a poly_int64 rather than a HOST_WIDE_INT.
8294         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8295         rather than a HOST_WIDE_INT.
8296         * config/aarch64/aarch64.h (aarch64_frame): Protect with
8297         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
8298         hard_fp_offset, frame_size, initial_adjust, callee_offset and
8299         final_offset from HOST_WIDE_INT to poly_int64.
8300         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8301         to_constant when getting the number of units in an Advanced SIMD
8302         mode.
8303         (aarch64_builtin_vectorized_function): Check for a constant number
8304         of units.
8305         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8306         GET_MODE_SIZE.
8307         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8308         attribute instead of GET_MODE_NUNITS.
8309         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8310         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8311         GET_MODE_SIZE for fixed-size registers.
8312         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8313         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8314         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8315         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8316         (aarch64_print_operand, aarch64_print_address_internal)
8317         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8318         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8319         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8320         Handle polynomial GET_MODE_SIZE.
8321         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
8322         wider than SImode without modification.
8323         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8324         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8325         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8326         passing and returning SVE modes.
8327         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8328         rather than GEN_INT.
8329         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8330         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8331         (aarch64_allocate_and_probe_stack_space): Likewise.
8332         (aarch64_layout_frame): Cope with polynomial offsets.
8333         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8334         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
8335         polynomial offsets.
8336         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8337         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8338         poly_int64 rather than a HOST_WIDE_INT.
8339         (aarch64_get_separate_components, aarch64_process_components)
8340         (aarch64_expand_prologue, aarch64_expand_epilogue)
8341         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8342         (aarch64_anchor_offset): New function, split out from...
8343         (aarch64_legitimize_address): ...here.
8344         (aarch64_builtin_vectorization_cost): Handle polynomial
8345         TYPE_VECTOR_SUBPARTS.
8346         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8347         GET_MODE_NUNITS.
8348         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8349         number of elements from the PARALLEL rather than the mode.
8350         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8351         rather than GET_MODE_BITSIZE.
8352         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8353         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8354         (aarch64_expand_vec_perm_const_1): Handle polynomial
8355         d->perm.length () and d->perm elements.
8356         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
8357         Apply to_constant to d->perm elements.
8358         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8359         polynomial CONST_VECTOR_NUNITS.
8360         (aarch64_move_pointer): Take amount as a poly_int64 rather
8361         than an int.
8362         (aarch64_progress_pointer): Avoid temporary variable.
8363         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8364         the mode attribute instead of GET_MODE.
8365
8366 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
8367             Alan Hayward  <alan.hayward@arm.com>
8368             David Sherwood  <david.sherwood@arm.com>
8369
8370         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8371         x exists before using it.
8372         (aarch64_add_constant_internal): Rename to...
8373         (aarch64_add_offset_1): ...this.  Replace regnum with separate
8374         src and dest rtxes.  Handle the case in which they're different,
8375         including when the offset is zero.  Replace scratchreg with an rtx.
8376         Use 2 additions if there is no spare register into which we can
8377         move a 16-bit constant.
8378         (aarch64_add_constant): Delete.
8379         (aarch64_add_offset): Replace reg with separate src and dest
8380         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
8381         Use aarch64_add_offset_1.
8382         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8383         an rtx rather than an int.  Take the delta as a poly_int64
8384         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
8385         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8386         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8387         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8388         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8389         and aarch64_add_sp.
8390         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8391         aarch64_add_constant.
8392
8393 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
8394
8395         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8396         Use scalar_float_mode.
8397
8398 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
8399
8400         * config/aarch64/aarch64-simd.md
8401         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8402         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8403         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8404         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8405         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8406         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8407         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8408         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8409         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8410         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8411
8412 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8413
8414         PR target/83514
8415         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8416         targ_options->x_arm_arch_string is non NULL.
8417
8418 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
8419
8420         * config/aarch64/aarch64.h
8421         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
8422
8423 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
8424
8425         PR target/82096
8426         * expmed.c (emit_store_flag_force): Swap if const op0
8427         and change VOIDmode to mode of op0.
8428
8429 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
8430
8431         PR rtl-optimization/83761
8432         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8433         than bytes to mode_for_size.
8434
8435 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
8436
8437         PR middle-end/83189
8438         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8439         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8440         profile.
8441
8442 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
8443
8444         PR middle-end/83575
8445         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8446         when in layout mode.
8447         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8448         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8449         partition fixup.
8450
8451 2018-01-10  Michael Collison  <michael.collison@arm.com>
8452
8453         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8454         * config/aarch64/aarch64-option-extension.def: Add
8455         AARCH64_OPT_EXTENSION of 'fp16fml'.
8456         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8457         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8458         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8459         * config/aarch64/constraints.md (Ui7): New constraint.
8460         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8461         (VFMLA_SEL_W): Ditto.
8462         (f16quad): Ditto.
8463         (f16mac1): Ditto.
8464         (VFMLA16_LOW): New int iterator.
8465         (VFMLA16_HIGH): Ditto.
8466         (UNSPEC_FMLAL): New unspec.
8467         (UNSPEC_FMLSL): Ditto.
8468         (UNSPEC_FMLAL2): Ditto.
8469         (UNSPEC_FMLSL2): Ditto.
8470         (f16mac): New code attribute.
8471         * config/aarch64/aarch64-simd-builtins.def
8472         (aarch64_fmlal_lowv2sf): Ditto.
8473         (aarch64_fmlsl_lowv2sf): Ditto.
8474         (aarch64_fmlalq_lowv4sf): Ditto.
8475         (aarch64_fmlslq_lowv4sf): Ditto.
8476         (aarch64_fmlal_highv2sf): Ditto.
8477         (aarch64_fmlsl_highv2sf): Ditto.
8478         (aarch64_fmlalq_highv4sf): Ditto.
8479         (aarch64_fmlslq_highv4sf): Ditto.
8480         (aarch64_fmlal_lane_lowv2sf): Ditto.
8481         (aarch64_fmlsl_lane_lowv2sf): Ditto.
8482         (aarch64_fmlal_laneq_lowv2sf): Ditto.
8483         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8484         (aarch64_fmlalq_lane_lowv4sf): Ditto.
8485         (aarch64_fmlsl_lane_lowv4sf): Ditto.
8486         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8487         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8488         (aarch64_fmlal_lane_highv2sf): Ditto.
8489         (aarch64_fmlsl_lane_highv2sf): Ditto.
8490         (aarch64_fmlal_laneq_highv2sf): Ditto.
8491         (aarch64_fmlsl_laneq_highv2sf): Ditto.
8492         (aarch64_fmlalq_lane_highv4sf): Ditto.
8493         (aarch64_fmlsl_lane_highv4sf): Ditto.
8494         (aarch64_fmlalq_laneq_highv4sf): Ditto.
8495         (aarch64_fmlsl_laneq_highv4sf): Ditto.
8496         * config/aarch64/aarch64-simd.md:
8497         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8498         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8499         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8500         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8501         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8502         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8503         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8504         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8505         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8506         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8507         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8508         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8509         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8510         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8511         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8512         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8513         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8514         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8515         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8516         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8517         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8518         (vfmlsl_low_u32): Ditto.
8519         (vfmlalq_low_u32): Ditto.
8520         (vfmlslq_low_u32): Ditto.
8521         (vfmlal_high_u32): Ditto.
8522         (vfmlsl_high_u32): Ditto.
8523         (vfmlalq_high_u32): Ditto.
8524         (vfmlslq_high_u32): Ditto.
8525         (vfmlal_lane_low_u32): Ditto.
8526         (vfmlsl_lane_low_u32): Ditto.
8527         (vfmlal_laneq_low_u32): Ditto.
8528         (vfmlsl_laneq_low_u32): Ditto.
8529         (vfmlalq_lane_low_u32): Ditto.
8530         (vfmlslq_lane_low_u32): Ditto.
8531         (vfmlalq_laneq_low_u32): Ditto.
8532         (vfmlslq_laneq_low_u32): Ditto.
8533         (vfmlal_lane_high_u32): Ditto.
8534         (vfmlsl_lane_high_u32): Ditto.
8535         (vfmlal_laneq_high_u32): Ditto.
8536         (vfmlsl_laneq_high_u32): Ditto.
8537         (vfmlalq_lane_high_u32): Ditto.
8538         (vfmlslq_lane_high_u32): Ditto.
8539         (vfmlalq_laneq_high_u32): Ditto.
8540         (vfmlslq_laneq_high_u32): Ditto.
8541         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8542         (AARCH64_FL_FOR_ARCH8_4): New.
8543         (AARCH64_ISA_F16FML): New ISA flag.
8544         (TARGET_F16FML): New feature flag for fp16fml.
8545         (doc/invoke.texi): Document new fp16fml option.
8546
8547 2018-01-10  Michael Collison  <michael.collison@arm.com>
8548
8549         * config/aarch64/aarch64-builtins.c:
8550         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8551         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8552         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8553         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8554         (AARCH64_ISA_SHA3): New ISA flag.
8555         (TARGET_SHA3): New feature flag for sha3.
8556         * config/aarch64/iterators.md (sha512_op): New int attribute.
8557         (CRYPTO_SHA512): New int iterator.
8558         (UNSPEC_SHA512H): New unspec.
8559         (UNSPEC_SHA512H2): Ditto.
8560         (UNSPEC_SHA512SU0): Ditto.
8561         (UNSPEC_SHA512SU1): Ditto.
8562         * config/aarch64/aarch64-simd-builtins.def
8563         (aarch64_crypto_sha512hqv2di): New builtin.
8564         (aarch64_crypto_sha512h2qv2di): Ditto.
8565         (aarch64_crypto_sha512su0qv2di): Ditto.
8566         (aarch64_crypto_sha512su1qv2di): Ditto.
8567         (aarch64_eor3qv8hi): Ditto.
8568         (aarch64_rax1qv2di): Ditto.
8569         (aarch64_xarqv2di): Ditto.
8570         (aarch64_bcaxqv8hi): Ditto.
8571         * config/aarch64/aarch64-simd.md:
8572         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8573         (aarch64_crypto_sha512su0qv2di): Ditto.
8574         (aarch64_crypto_sha512su1qv2di): Ditto.
8575         (aarch64_eor3qv8hi): Ditto.
8576         (aarch64_rax1qv2di): Ditto.
8577         (aarch64_xarqv2di): Ditto.
8578         (aarch64_bcaxqv8hi): Ditto.
8579         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8580         (vsha512h2q_u64): Ditto.
8581         (vsha512su0q_u64): Ditto.
8582         (vsha512su1q_u64): Ditto.
8583         (veor3q_u16): Ditto.
8584         (vrax1q_u64): Ditto.
8585         (vxarq_u64): Ditto.
8586         (vbcaxq_u16): Ditto.
8587         * config/arm/types.md (crypto_sha512): New type attribute.
8588         (crypto_sha3): Ditto.
8589         (doc/invoke.texi): Document new sha3 option.
8590
8591 2018-01-10  Michael Collison  <michael.collison@arm.com>
8592
8593         * config/aarch64/aarch64-builtins.c:
8594         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8595         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8596         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8597         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8598         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8599         (AARCH64_ISA_SM4): New ISA flag.
8600         (TARGET_SM4): New feature flag for sm4.
8601         * config/aarch64/aarch64-simd-builtins.def
8602         (aarch64_sm3ss1qv4si): Ditto.
8603         (aarch64_sm3tt1aq4si): Ditto.
8604         (aarch64_sm3tt1bq4si): Ditto.
8605         (aarch64_sm3tt2aq4si): Ditto.
8606         (aarch64_sm3tt2bq4si): Ditto.
8607         (aarch64_sm3partw1qv4si): Ditto.
8608         (aarch64_sm3partw2qv4si): Ditto.
8609         (aarch64_sm4eqv4si): Ditto.
8610         (aarch64_sm4ekeyqv4si): Ditto.
8611         * config/aarch64/aarch64-simd.md:
8612         (aarch64_sm3ss1qv4si): Ditto.
8613         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8614         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8615         (aarch64_sm4eqv4si): Ditto.
8616         (aarch64_sm4ekeyqv4si): Ditto.
8617         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8618         (sm3part_op): Ditto.
8619         (CRYPTO_SM3TT): Ditto.
8620         (CRYPTO_SM3PART): Ditto.
8621         (UNSPEC_SM3SS1): New unspec.
8622         (UNSPEC_SM3TT1A): Ditto.
8623         (UNSPEC_SM3TT1B): Ditto.
8624         (UNSPEC_SM3TT2A): Ditto.
8625         (UNSPEC_SM3TT2B): Ditto.
8626         (UNSPEC_SM3PARTW1): Ditto.
8627         (UNSPEC_SM3PARTW2): Ditto.
8628         (UNSPEC_SM4E): Ditto.
8629         (UNSPEC_SM4EKEY): Ditto.
8630         * config/aarch64/constraints.md (Ui2): New constraint.
8631         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8632         * config/arm/types.md (crypto_sm3): New type attribute.
8633         (crypto_sm4): Ditto.
8634         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8635         (vsm3tt1aq_u32): Ditto.
8636         (vsm3tt1bq_u32): Ditto.
8637         (vsm3tt2aq_u32): Ditto.
8638         (vsm3tt2bq_u32): Ditto.
8639         (vsm3partw1q_u32): Ditto.
8640         (vsm3partw2q_u32): Ditto.
8641         (vsm4eq_u32): Ditto.
8642         (vsm4ekeyq_u32): Ditto.
8643         (doc/invoke.texi): Document new sm4 option.
8644
8645 2018-01-10  Michael Collison  <michael.collison@arm.com>
8646
8647         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8648         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8649         (AARCH64_FL_FOR_ARCH8_4): New.
8650         (AARCH64_FL_V8_4): New flag.
8651         (doc/invoke.texi): Document new armv8.4-a option.
8652
8653 2018-01-10  Michael Collison  <michael.collison@arm.com>
8654
8655         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8656         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8657         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8658         * config/aarch64/aarch64-option-extension.def: Add
8659         AARCH64_OPT_EXTENSION of 'sha2'.
8660         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8661         (crypto): Disable sha2 and aes if crypto disabled.
8662         (crypto): Enable aes and sha2 if enabled.
8663         (simd): Disable sha2 and aes if simd disabled.
8664         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8665         New flags.
8666         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8667         (TARGET_SHA2): New feature flag for sha2.
8668         (TARGET_AES): New feature flag for aes.
8669         * config/aarch64/aarch64-simd.md:
8670         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8671         conditional on TARGET_AES.
8672         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8673         (aarch64_crypto_sha1hsi): Make pattern conditional
8674         on TARGET_SHA2.
8675         (aarch64_crypto_sha1hv4si): Ditto.
8676         (aarch64_be_crypto_sha1hv4si): Ditto.
8677         (aarch64_crypto_sha1su1v4si): Ditto.
8678         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8679         (aarch64_crypto_sha1su0v4si): Ditto.
8680         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8681         (aarch64_crypto_sha256su0v4si): Ditto.
8682         (aarch64_crypto_sha256su1v4si): Ditto.
8683         (doc/invoke.texi): Document new aes and sha2 options.
8684
8685 2018-01-10  Martin Sebor  <msebor@redhat.com>
8686
8687         PR tree-optimization/83781
8688         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8689         as string arrays.
8690
8691 2018-01-11  Martin Sebor  <msebor@gmail.com>
8692             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8693
8694         PR tree-optimization/83501
8695         PR tree-optimization/81703
8696
8697         * tree-ssa-strlen.c (get_string_cst): Rename...
8698         (get_string_len): ...to this.  Handle global constants.
8699         (handle_char_store): Adjust.
8700
8701 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
8702             Jim Wilson  <jimw@sifive.com>
8703
8704         * config/riscv/riscv-protos.h (riscv_output_return): New.
8705         * config/riscv/riscv.c (struct machine_function): New naked_p field.
8706         (riscv_attribute_table, riscv_output_return),
8707         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8708         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8709         (riscv_compute_frame_info): Only compute frame->mask if not a naked
8710         function.
8711         (riscv_expand_prologue): Add early return for naked function.
8712         (riscv_expand_epilogue): Likewise.
8713         (riscv_function_ok_for_sibcall): Return false for naked function.
8714         (riscv_set_current_function): New.
8715         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8716         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8717         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8718         * doc/extend.texi (RISC-V Function Attributes): New.
8719
8720 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
8721
8722         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8723         check for 128-bit long double before checking TCmode.
8724         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8725         128-bit long doubles before checking TFmode or TCmode.
8726         (FLOAT128_IBM_P): Likewise.
8727
8728 2018-01-10  Martin Sebor  <msebor@redhat.com>
8729
8730         PR tree-optimization/83671
8731         * builtins.c (c_strlen): Unconditionally return zero for the empty
8732         string.
8733         Use -Warray-bounds for warnings.
8734         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8735         for non-constant array indices with COMPONENT_REF, arrays of
8736         arrays, and pointers to arrays.
8737         (gimple_fold_builtin_strlen): Determine and set length range for
8738         non-constant character arrays.
8739
8740 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
8741
8742         PR middle-end/81897
8743         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8744         empty blocks.
8745
8746 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
8747
8748         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8749
8750 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
8751
8752         PR target/83399
8753         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8754         VECTOR_MEM_ALTIVEC_OR_VSX_P.
8755         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8756         indexed_or_indirect_operand predicate.
8757         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8758         (*vsx_le_perm_load_v8hi): Likewise.
8759         (*vsx_le_perm_load_v16qi): Likewise.
8760         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8761         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8762         (*vsx_le_perm_store_v8hi): Likewise.
8763         (*vsx_le_perm_store_v16qi): Likewise.
8764         (eight unnamed splitters): Likewise.
8765
8766 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
8767
8768         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8769         * config/rs6000/emmintrin.h: Likewise.
8770         * config/rs6000/mmintrin.h: Likewise.
8771         * config/rs6000/xmmintrin.h: Likewise.
8772
8773 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
8774
8775         PR c++/43486
8776         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8777         "public_flag".
8778         * tree.c (tree_nop_conversion): Return true for location wrapper
8779         nodes.
8780         (maybe_wrap_with_location): New function.
8781         (selftest::check_strip_nops): New function.
8782         (selftest::test_location_wrappers): New function.
8783         (selftest::tree_c_tests): Call it.
8784         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8785         (maybe_wrap_with_location): New decl.
8786         (EXPR_LOCATION_WRAPPER_P): New macro.
8787         (location_wrapper_p): New inline function.
8788         (tree_strip_any_location_wrapper): New inline function.
8789
8790 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
8791
8792         PR target/83735
8793         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8794         stack_realign_offset for the largest alignment of stack slot
8795         actually used.
8796         (ix86_find_max_used_stack_alignment): New function.
8797         (ix86_finalize_stack_frame_flags): Use it.  Set
8798         max_used_stack_alignment if we don't realign stack.
8799         * config/i386/i386.h (machine_function): Add
8800         max_used_stack_alignment.
8801
8802 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
8803
8804         * config/arm/arm.opt (-mbranch-cost): New option.
8805         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8806         account.
8807
8808 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
8809
8810         PR target/83629
8811         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8812         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8813
8814 2018-01-10  Richard Biener  <rguenther@suse.de>
8815
8816         PR debug/83765
8817         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8818         early out so it also covers the case where we have a non-NULL
8819         origin.
8820
8821 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
8822
8823         PR tree-optimization/83753
8824         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8825         for non-strided grouped accesses if the number of elements is 1.
8826
8827 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
8828
8829         PR target/81616
8830         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8831         * i386.h (TARGET_USE_GATHER): Define.
8832         * x86-tune.def (X86_TUNE_USE_GATHER): New.
8833
8834 2018-01-10  Martin Liska  <mliska@suse.cz>
8835
8836         PR bootstrap/82831
8837         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8838         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8839         partitioning.
8840         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8841         CLEANUP_NO_PARTITIONING is not set.
8842
8843 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
8844
8845         * doc/rtl.texi: Remove documentation of (const ...) wrappers
8846         for vectors, as a partial revert of r254296.
8847         * rtl.h (const_vec_p): Delete.
8848         (const_vec_duplicate_p): Don't test for vector CONSTs.
8849         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8850         * expmed.c (make_tree): Likewise.
8851
8852         Revert:
8853         * common.md (E, F): Use CONSTANT_P instead of checking for
8854         CONST_VECTOR.
8855         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8856         checking for CONST_VECTOR.
8857
8858 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
8859
8860         PR middle-end/83575
8861         * predict.c (force_edge_cold): Handle in more sane way edges
8862         with no prediction.
8863
8864 2018-01-09  Carl Love  <cel@us.ibm.com>
8865
8866         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8867         V4SI, V4SF types.
8868         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8869         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8870         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8871         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
8872         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8873         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
8874         * config/rs6000/rs6000-protos.h: Add extern defition for
8875         rs6000_generate_float2_double_code.
8876         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8877         function.
8878         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8879         (float2_v2df): Add define_expand.
8880
8881 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
8882
8883         PR target/83628
8884         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8885         op_mode in the force_to_mode call.
8886
8887 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
8888
8889         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8890         instead of checking each element individually.
8891         (aarch64_evpc_uzp): Likewise.
8892         (aarch64_evpc_zip): Likewise.
8893         (aarch64_evpc_ext): Likewise.
8894         (aarch64_evpc_rev): Likewise.
8895         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8896         instead of checking each element individually.  Return true without
8897         generating rtl if
8898         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8899         whether all selected elements come from the same input, instead of
8900         checking each element individually.  Remove calls to gen_rtx_REG,
8901         start_sequence and end_sequence and instead assert that no rtl is
8902         generated.
8903
8904 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
8905
8906         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8907         order of HIGH and CONST checks.
8908
8909 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
8910
8911         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8912         if the destination isn't an SSA_NAME.
8913
8914 2018-01-09  Richard Biener  <rguenther@suse.de>
8915
8916         PR tree-optimization/83668
8917         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8918         move prologue...
8919         (canonicalize_loop_form): ... here, renamed from ...
8920         (canonicalize_loop_closed_ssa_form): ... this and amended to
8921         swap successor edges for loop exit blocks to make us use
8922         the RPO order we need for initial schedule generation.
8923
8924 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
8925
8926         PR tree-optimization/64811
8927         * match.pd: When optimizing comparisons with Inf, avoid
8928         introducing or losing exceptions from comparisons with NaN.
8929
8930 2018-01-09  Martin Liska  <mliska@suse.cz>
8931
8932         PR sanitizer/82517
8933         * asan.c (shadow_mem_size): Add gcc_assert.
8934
8935 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
8936
8937         Don't save registers in main().
8938
8939         PR target/83738
8940         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8941         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8942         * config/avr/avr.c (avr_set_current_function): Don't error if
8943         naked, OS_task or OS_main are specified at the same time.
8944         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8945         OS_main.
8946         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8947         attribute.
8948         * common/config/avr/avr-common.c (avr_option_optimization_table):
8949         Switch on -mmain-is-OS_task for optimizing compilations.
8950
8951 2018-01-09  Richard Biener  <rguenther@suse.de>
8952
8953         PR tree-optimization/83572
8954         * graphite.c: Include cfganal.h.
8955         (graphite_transform_loops): Connect infinite loops to exit
8956         and remove fake edges at the end.
8957
8958 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
8959
8960         * ipa-inline.c (edge_badness): Revert accidental checkin.
8961
8962 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
8963
8964         PR ipa/80763
8965         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8966         symbols; not inline clones.
8967
8968 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
8969
8970         PR target/83507
8971         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8972         hard registers.  Formatting fixes.
8973
8974         PR preprocessor/83722
8975         * gcc.c (try_generate_repro): Pass
8976         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8977         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8978         do_report_bug.
8979
8980 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
8981             Kito Cheng  <kito.cheng@gmail.com>
8982
8983         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8984         (riscv_leaf_function_p): Delete.
8985         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8986
8987 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
8988
8989         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8990         function.
8991         (do_ifelse): New function.
8992         (do_isel): New function.
8993         (do_sub3): New function.
8994         (do_add3): New function.
8995         (do_load_mask_compare): New function.
8996         (do_overlap_load_compare): New function.
8997         (expand_compare_loop): New function.
8998         (expand_block_compare): Call expand_compare_loop() when appropriate.
8999         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9000         option description.
9001         (-mblock-compare-inline-loop-limit): New option.
9002
9003 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9004
9005         PR target/83677
9006         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9007         Reverse order of second and third operands in first alternative.
9008         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9009         of first and second elements in UNSPEC_VPERMR vector.
9010         (altivec_expand_vec_perm_le): Likewise.
9011
9012 2018-01-08  Jeff Law  <law@redhat.com>
9013
9014         PR rtl-optimizatin/81308
9015         * tree-switch-conversion.c (cfg_altered): New file scoped static.
9016         (process_switch): If group_case_labels makes a change, then set
9017         cfg_altered.
9018         (pass_convert_switch::execute): If a switch is converted, then
9019         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
9020
9021         PR rtl-optimization/81308
9022         * recog.c (split_all_insns): Conditionally cleanup the CFG after
9023         splitting insns.
9024
9025 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
9026
9027         PR target/83663 - Revert r255946
9028         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9029         generation for cases where splatting a value is not useful.
9030         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9031         across a vec_duplicate and a paradoxical subreg forming a vector
9032         mode to a vec_concat.
9033
9034 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9035
9036         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9037         -march=armv8.3-a variants.
9038         * config/arm/t-multilib: Likewise.
9039         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
9040
9041 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
9042
9043         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9044         to generate rtl.
9045         (cceq_ior_compare_complement): Give it a name so I can use it, and
9046         change boolean_or_operator predicate to boolean_operator so it can
9047         be used to generate a crand.
9048         (eqne): New code iterator.
9049         (bd/bd_neg): New code_attrs.
9050         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9051         a single define_insn.
9052         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9053         decrement (bdnzt/bdnzf/bdzt/bdzf).
9054         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9055         with the new names of the branch decrement patterns, and added the
9056         names of the branch decrement conditional patterns.
9057
9058 2018-01-08  Richard Biener  <rguenther@suse.de>
9059
9060         PR tree-optimization/83563
9061         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9062         cache.
9063
9064 2018-01-08  Richard Biener  <rguenther@suse.de>
9065
9066         PR middle-end/83713
9067         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9068
9069 2018-01-08  Richard Biener  <rguenther@suse.de>
9070
9071         PR tree-optimization/83685
9072         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9073         references to abnormals.
9074
9075 2018-01-08  Richard Biener  <rguenther@suse.de>
9076
9077         PR lto/83719
9078         * dwarf2out.c (output_indirect_strings): Handle empty
9079         skeleton_debug_str_hash.
9080         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9081
9082 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
9083
9084         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9085         (emit_store_direct): Likewise.
9086         (arc_trampoline_adjust_address): Likewise.
9087         (arc_asm_trampoline_template): New function.
9088         (arc_initialize_trampoline): Use asm_trampoline_template.
9089         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9090         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9091         * config/arc/arc.md (flush_icache): Delete pattern.
9092
9093 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
9094
9095         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9096         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9097         munaligned-access.
9098
9099 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9100
9101         PR target/83681
9102         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9103         by not USED_FOR_TARGET.
9104         (make_pass_resolve_sw_modes): Likewise.
9105
9106 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9107
9108         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9109         USED_FOR_TARGET.
9110
9111 2018-01-08  Richard Biener  <rguenther@suse.de>
9112
9113         PR middle-end/83580
9114         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9115
9116 2018-01-08  Richard Biener  <rguenther@suse.de>
9117
9118         PR middle-end/83517
9119         * match.pd ((t * 2) / 2) -> t): Add missing :c.
9120
9121 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
9122
9123         PR middle-end/81897
9124         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9125         basic blocks with a small number of successors.
9126         (convert_control_dep_chain_into_preds): Improve handling of
9127         forwarder blocks.
9128         (dump_predicates): Split apart into...
9129         (dump_pred_chain): ...here...
9130         (dump_pred_info): ...and here.
9131         (can_one_predicate_be_invalidated_p): Add debugging printfs.
9132         (can_chain_union_be_invalidated_p): Improve check for invalidation
9133         of paths.
9134         (uninit_uses_cannot_happen): Avoid unnecessary if
9135         convert_control_dep_chain_into_preds yielded nothing.
9136
9137 2018-01-06  Martin Sebor  <msebor@redhat.com>
9138
9139         PR tree-optimization/83640
9140         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9141         subtracting negative offset from size.
9142         (builtin_access::overlap): Adjust offset bounds of the access to fall
9143         within the size of the object if possible.
9144
9145 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
9146
9147         PR rtl-optimization/83699
9148         * expmed.c (extract_bit_field_1): Restrict the vector usage of
9149         extract_bit_field_as_subreg to cases in which the extracted
9150         value is also a vector.
9151
9152         * lra-constraints.c (process_alt_operands): Test for the equivalence
9153         substitutions when detecting a possible reload cycle.
9154
9155 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
9156
9157         PR debug/83480
9158         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9159         by default if flag_selective_schedling{,2}.  Formatting fixes.
9160
9161         PR rtl-optimization/83682
9162         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9163         if it has non-VECTOR_MODE element mode.
9164         (vec_duplicate_p): Likewise.
9165
9166         PR middle-end/83694
9167         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9168         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9169
9170 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
9171
9172         PR target/83604
9173         * config/i386/i386-builtin.def
9174         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9175         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9176         Require also OPTION_MASK_ISA_AVX512F in addition to
9177         OPTION_MASK_ISA_GFNI.
9178         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9179         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9180         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9181         to OPTION_MASK_ISA_GFNI.
9182         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9183         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9184         OPTION_MASK_ISA_AVX512BW.
9185         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9186         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9187         addition to OPTION_MASK_ISA_GFNI.
9188         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9189         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9190         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9191         to OPTION_MASK_ISA_GFNI.
9192         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9193         a requirement for all ISAs rather than any of them with a few
9194         exceptions.
9195         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9196         processing.
9197         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9198         bitmasks to be enabled with 3 exceptions, instead of requiring any
9199         enabled ISA with lots of exceptions.
9200         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9201         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9202         Change avx512bw in isa attribute to avx512f.
9203         * config/i386/sgxintrin.h: Add license boilerplate.
9204         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
9205         to __AVX512F__ and __AVX512VL to __AVX512VL__.
9206         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9207         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9208         defined.
9209         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9210         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9211         temporarily sse2 rather than sse if not enabled already.
9212
9213         PR target/83604
9214         * config/i386/sse.md (VI248_VLBW): Rename to ...
9215         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
9216         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9217         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9218         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9219         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9220         mode iterator instead of VI248_VLBW.
9221
9222 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
9223
9224         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9225         (record_modified): Skip clobbers; add debug output.
9226         (param_change_prob): Use sreal frequencies.
9227
9228 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
9229
9230         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9231         punt for user-aligned variables.
9232
9233 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
9234
9235         * tree-chrec.c (chrec_contains_symbols): Return true for
9236         POLY_INT_CST.
9237
9238 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
9239
9240         PR target/82439
9241         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9242         of (x|y) == x for BICS pattern.
9243
9244 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
9245
9246         PR tree-optimization/83605
9247         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9248         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9249         can throw.
9250
9251 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9252
9253         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9254         * config/epiphany/rtems.h: New file.
9255
9256 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
9257             Uros Bizjak  <ubizjak@gmail.com>
9258
9259         PR target/83554
9260         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9261         QIreg_operand instead of register_operand predicate.
9262         * config/i386/i386.c (ix86_rop_should_change_byte_p,
9263         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9264         comments instead of -fmitigate[-_]rop.
9265
9266 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9267
9268         PR bootstrap/81926
9269         * cgraphunit.c (symbol_table::compile): Switch to text_section
9270         before calling assembly_start debug hook.
9271         * run-rtl-passes.c (run_rtl_passes): Likewise.
9272         Include output.h.
9273
9274 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
9275
9276         * tree-vrp.c (extract_range_from_binary_expr_1): Check
9277         range_int_cst_p rather than !symbolic_range_p before calling
9278         extract_range_from_multiplicative_op_1.
9279
9280 2018-01-04  Jeff Law  <law@redhat.com>
9281
9282         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9283         redundant test in assertion.
9284
9285 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
9286
9287         * doc/rtl.texi: Document machine_mode wrapper classes.
9288
9289 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
9290
9291         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9292         using tree_to_uhwi.
9293
9294 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
9295
9296         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9297         the VEC_PERM_EXPR fold to fail.
9298
9299 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
9300
9301         PR debug/83585
9302         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9303         to switched_sections.
9304
9305 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
9306
9307         PR target/83680
9308         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9309         test for d.testing.
9310
9311 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
9312
9313         PR target/83387
9314         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9315         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9316
9317 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
9318
9319         PR debug/83666
9320         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9321         is BLKmode and bitpos not zero or mode change is needed.
9322
9323 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
9324
9325         PR target/83675
9326         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9327         TARGET_VIS2.
9328
9329 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
9330
9331         PR target/83628
9332         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9333         instead of MULT rtx.  Update all corresponding splitters.
9334         (*saddl_se): Ditto.
9335         (*ssub<modesuffix>): Ditto.
9336         (*ssubl_se): Ditto.
9337         (*cmp_sadd_di): Update split patterns.
9338         (*cmp_sadd_si): Ditto.
9339         (*cmp_sadd_sidi): Ditto.
9340         (*cmp_ssub_di): Ditto.
9341         (*cmp_ssub_si): Ditto.
9342         (*cmp_ssub_sidi): Ditto.
9343         * config/alpha/predicates.md (const23_operand): New predicate.
9344         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9345         Look for ASHIFT, not MULT inner operand.
9346         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9347
9348 2018-01-04  Martin Liska  <mliska@suse.cz>
9349
9350         PR gcov-profile/83669
9351         * gcov.c (output_intermediate_file): Add version to intermediate
9352         gcov file.
9353         * doc/gcov.texi: Document new field 'version' in intermediate
9354         file format. Fix location of '-k' option of gcov command.
9355
9356 2018-01-04  Martin Liska  <mliska@suse.cz>
9357
9358         PR ipa/82352
9359         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9360
9361 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
9362
9363         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9364
9365 2018-01-03  Martin Sebor  <msebor@redhat.com>
9366
9367         PR tree-optimization/83655
9368         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9369         checking calls with invalid arguments.
9370
9371 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9372
9373         * tree-vect-stmts.c (vect_get_store_rhs): New function.
9374         (vectorizable_mask_load_store): Delete.
9375         (vectorizable_call): Return false for masked loads and stores.
9376         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
9377         instead of gimple_assign_rhs1.
9378         (vectorizable_load): Handle IFN_MASK_LOAD.
9379         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9380
9381 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9382
9383         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9384         split out from..,
9385         (vectorizable_mask_load_store): ...here.
9386         (vectorizable_load): ...and here.
9387
9388 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9389
9390         * tree-vect-stmts.c (vect_build_all_ones_mask)
9391         (vect_build_zero_merge_argument): New functions, split out from...
9392         (vectorizable_load): ...here.
9393
9394 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9395
9396         * tree-vect-stmts.c (vect_check_store_rhs): New function,
9397         split out from...
9398         (vectorizable_mask_load_store): ...here.
9399         (vectorizable_store): ...and here.
9400
9401 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9402
9403         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9404         split out from...
9405         (vectorizable_mask_load_store): ...here.
9406
9407 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9408
9409         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9410         (vect_model_store_cost): Take a vec_load_store_type instead of a
9411         vect_def_type.
9412         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9413         (vect_model_store_cost): Take a vec_load_store_type instead of a
9414         vect_def_type.
9415         (vectorizable_mask_load_store): Update accordingly.
9416         (vectorizable_store): Likewise.
9417         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9418
9419 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9420
9421         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9422         IFN_MASK_LOAD calls here rather than...
9423         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9424
9425 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9426             Alan Hayward  <alan.hayward@arm.com>
9427             David Sherwood  <david.sherwood@arm.com>
9428
9429         * expmed.c (extract_bit_field_1): For vector extracts,
9430         fall back to extract_bit_field_as_subreg if vec_extract
9431         isn't available.
9432
9433 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9434             Alan Hayward  <alan.hayward@arm.com>
9435             David Sherwood  <david.sherwood@arm.com>
9436
9437         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9438         they are variable or constant sized.
9439         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9440         slots for constant-sized data.
9441
9442 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9443             Alan Hayward  <alan.hayward@arm.com>
9444             David Sherwood  <david.sherwood@arm.com>
9445
9446         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9447         handling COND_EXPRs with boolean comparisons, try to find a better
9448         basis for the mask type than the boolean itself.
9449
9450 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9451
9452         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9453         is calculated and how it can be overridden.
9454         * genmodes.c (max_bitsize_mode_any_mode): New variable.
9455         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9456         if defined.
9457         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9458         if nonzero.
9459
9460 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9461             Alan Hayward  <alan.hayward@arm.com>
9462             David Sherwood  <david.sherwood@arm.com>
9463
9464         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9465         Remove the mode argument.
9466         (aarch64_simd_valid_immediate): Remove the mode and inverse
9467         arguments.
9468         * config/aarch64/iterators.md (bitsize): New iterator.
9469         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9470         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9471         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9472         aarch64_simd_valid_immediate.
9473         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9474         (aarch64_reg_or_bic_imm): Likewise.
9475         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9476         with an insn_type enum and msl with a modifier_type enum.
9477         Replace element_width with a scalar_mode.  Change the shift
9478         to unsigned int.  Add constructors for scalar_float_mode and
9479         scalar_int_mode elements.
9480         (aarch64_vect_float_const_representable_p): Delete.
9481         (aarch64_can_const_movi_rtx_p)
9482         (aarch64_simd_scalar_immediate_valid_for_move)
9483         (aarch64_simd_make_constant): Update call to
9484         aarch64_simd_valid_immediate.
9485         (aarch64_advsimd_valid_immediate_hs): New function.
9486         (aarch64_advsimd_valid_immediate): Likewise.
9487         (aarch64_simd_valid_immediate): Remove mode and inverse
9488         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
9489         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9490         and aarch64_float_const_representable_p on the result.
9491         (aarch64_output_simd_mov_immediate): Remove mode argument.
9492         Update call to aarch64_simd_valid_immediate and use of
9493         simd_immediate_info.
9494         (aarch64_output_scalar_simd_mov_immediate): Update call
9495         accordingly.
9496
9497 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9498             Alan Hayward  <alan.hayward@arm.com>
9499             David Sherwood  <david.sherwood@arm.com>
9500
9501         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9502         (mode_nunits): Likewise CONST_MODE_NUNITS.
9503         * machmode.def (ADJUST_NUNITS): Document.
9504         * genmodes.c (mode_data::need_nunits_adj): New field.
9505         (blank_mode): Update accordingly.
9506         (adj_nunits): New variable.
9507         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9508         parameter.
9509         (emit_mode_size_inline): Set need_bytesize_adj for all modes
9510         listed in adj_nunits.
9511         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9512         listed in adj_nunits.  Don't emit case statements for such modes.
9513         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9514         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
9515         nothing if adj_nunits is nonnull.
9516         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9517         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9518         (emit_mode_fbit): Update use of print_maybe_const_decl.
9519         (emit_move_size): Likewise.  Treat the array as non-const
9520         if adj_nunits.
9521         (emit_mode_adjustments): Handle adj_nunits.
9522
9523 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9524
9525         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9526         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9527         (VECTOR_MODES): Use it.
9528         (make_vector_modes): Take the prefix as an argument.
9529
9530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9531             Alan Hayward  <alan.hayward@arm.com>
9532             David Sherwood  <david.sherwood@arm.com>
9533
9534         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9535         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9536         for MODE_VECTOR_BOOL.
9537         * machmode.def (VECTOR_BOOL_MODE): Document.
9538         * genmodes.c (VECTOR_BOOL_MODE): New macro.
9539         (make_vector_bool_mode): New function.
9540         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9541         MODE_VECTOR_BOOL.
9542         * lto-streamer-in.c (lto_input_mode_table): Likewise.
9543         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9544         Likewise.
9545         * stor-layout.c (int_mode_for_mode): Likewise.
9546         * tree.c (build_vector_type_for_mode): Likewise.
9547         * varasm.c (output_constant_pool_2): Likewise.
9548         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9549         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
9550         for MODE_VECTOR_BOOL.
9551         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9552         of mode class checks.
9553         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9554         instead of a list of mode class checks.
9555         (expand_vector_scalar_condition): Likewise.
9556         (type_for_widest_vector_mode): Handle BImode as an inner mode.
9557
9558 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9559             Alan Hayward  <alan.hayward@arm.com>
9560             David Sherwood  <david.sherwood@arm.com>
9561
9562         * machmode.h (mode_size): Change from unsigned short to
9563         poly_uint16_pod.
9564         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9565         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9566         or if measurement_type is not polynomial.
9567         (fixed_size_mode::includes_p): Check for constant-sized modes.
9568         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9569         return a poly_uint16 rather than an unsigned short.
9570         (emit_mode_size): Change the type of mode_size from unsigned short
9571         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
9572         (emit_mode_adjustments): Cope with polynomial vector sizes.
9573         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9574         for GET_MODE_SIZE.
9575         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9576         for GET_MODE_SIZE.
9577         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9578         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9579         * caller-save.c (setup_save_areas): Likewise.
9580         (replace_reg_with_saved_mem): Likewise.
9581         * calls.c (emit_library_call_value_1): Likewise.
9582         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9583         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9584         (gen_lowpart_for_combine): Likewise.
9585         * convert.c (convert_to_integer_1): Likewise.
9586         * cse.c (equiv_constant, cse_insn): Likewise.
9587         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9588         (cselib_subst_to_values): Likewise.
9589         * dce.c (word_dce_process_block): Likewise.
9590         * df-problems.c (df_word_lr_mark_ref): Likewise.
9591         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9592         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9593         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9594         (rtl_for_decl_location): Likewise.
9595         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9596         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9597         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9598         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9599         (expand_expr_real_1): Likewise.
9600         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9601         (pad_below): Likewise.
9602         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9603         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9604         * ira.c (get_subreg_tracking_sizes): Likewise.
9605         * ira-build.c (ira_create_allocno_objects): Likewise.
9606         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9607         (ira_sort_regnos_for_alter_reg): Likewise.
9608         * ira-costs.c (record_operand_costs): Likewise.
9609         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9610         (resolve_simple_move): Likewise.
9611         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9612         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9613         (lra_constraints): Likewise.
9614         (CONST_POOL_OK_P): Reject variable-sized modes.
9615         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9616         (add_pseudo_to_slot, lra_spill): Likewise.
9617         * omp-low.c (omp_clause_aligned_alignment): Likewise.
9618         * optabs-query.c (get_best_extraction_insn): Likewise.
9619         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9620         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9621         (expand_mult_highpart, valid_multiword_target_p): Likewise.
9622         * recog.c (offsettable_address_addr_space_p): Likewise.
9623         * regcprop.c (maybe_mode_change): Likewise.
9624         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9625         * regrename.c (build_def_use): Likewise.
9626         * regstat.c (dump_reg_info): Likewise.
9627         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9628         (find_reloads, find_reloads_subreg_address): Likewise.
9629         * reload1.c (eliminate_regs_1): Likewise.
9630         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9631         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9632         (simplify_binary_operation_1, simplify_subreg): Likewise.
9633         * targhooks.c (default_function_arg_padding): Likewise.
9634         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9635         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9636         (verify_gimple_assign_ternary): Likewise.
9637         * tree-inline.c (estimate_move_cost): Likewise.
9638         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9639         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9640         (get_address_cost_ainc): Likewise.
9641         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9642         (vect_supportable_dr_alignment): Likewise.
9643         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9644         (vectorizable_reduction): Likewise.
9645         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9646         (vectorizable_operation, vectorizable_load): Likewise.
9647         * tree.c (build_same_sized_truth_vector_type): Likewise.
9648         * valtrack.c (cleanup_auto_inc_dec): Likewise.
9649         * var-tracking.c (emit_note_insn_var_location): Likewise.
9650         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9651         (ADDR_VEC_ALIGN): Likewise.
9652
9653 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9654             Alan Hayward  <alan.hayward@arm.com>
9655             David Sherwood  <david.sherwood@arm.com>
9656
9657         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9658         unsigned short.
9659         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9660         or if measurement_type is polynomial.
9661         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9662         * combine.c (make_extraction): Likewise.
9663         * dse.c (find_shift_sequence): Likewise.
9664         * dwarf2out.c (mem_loc_descriptor): Likewise.
9665         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9666         (extract_bit_field, extract_low_bits): Likewise.
9667         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9668         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9669         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9670         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9671         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9672         * reload.c (find_reloads): Likewise.
9673         * reload1.c (alter_reg): Likewise.
9674         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9675         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9676         * tree-if-conv.c (predicate_mem_writes): Likewise.
9677         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9678         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9679         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9680         * valtrack.c (dead_debug_insert_temp): Likewise.
9681         * varasm.c (mergeable_constant_section): Likewise.
9682         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9683
9684 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9685             Alan Hayward  <alan.hayward@arm.com>
9686             David Sherwood  <david.sherwood@arm.com>
9687
9688         * expr.c (expand_assignment): Cope with polynomial mode sizes
9689         when assigning to a CONCAT.
9690
9691 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9692             Alan Hayward  <alan.hayward@arm.com>
9693             David Sherwood  <david.sherwood@arm.com>
9694
9695         * machmode.h (mode_precision): Change from unsigned short to
9696         poly_uint16_pod.
9697         (mode_to_precision): Return a poly_uint16 rather than an unsigned
9698         short.
9699         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9700         or if measurement_type is not polynomial.
9701         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
9702         in which the mode is already known to be a scalar_int_mode.
9703         * genmodes.c (emit_mode_precision): Change the type of mode_precision
9704         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
9705         initializer.
9706         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9707         for GET_MODE_PRECISION.
9708         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9709         for GET_MODE_PRECISION.
9710         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9711         as polynomial.
9712         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9713         (expand_field_assignment, make_extraction): Likewise.
9714         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9715         (get_last_value): Likewise.
9716         * convert.c (convert_to_integer_1): Likewise.
9717         * cse.c (cse_insn): Likewise.
9718         * expr.c (expand_expr_real_1): Likewise.
9719         * lra-constraints.c (simplify_operand_subreg): Likewise.
9720         * optabs-query.c (can_atomic_load_p): Likewise.
9721         * optabs.c (expand_atomic_load): Likewise.
9722         (expand_atomic_store): Likewise.
9723         * ree.c (combine_reaching_defs): Likewise.
9724         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9725         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9726         * tree.h (type_has_mode_precision_p): Likewise.
9727         * ubsan.c (instrument_si_overflow): Likewise.
9728
9729 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9730             Alan Hayward  <alan.hayward@arm.com>
9731             David Sherwood  <david.sherwood@arm.com>
9732
9733         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9734         polynomial numbers of units.
9735         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9736         (valid_vector_subparts_p): New function.
9737         (build_vector_type): Remove temporary shim and take the number
9738         of units as a poly_uint64 rather than an int.
9739         (build_opaque_vector_type): Take the number of units as a
9740         poly_uint64 rather than an int.
9741         * tree.c (build_vector_from_ctor): Handle polynomial
9742         TYPE_VECTOR_SUBPARTS.
9743         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9744         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9745         (build_vector_from_val): If the number of units is variable,
9746         use build_vec_duplicate_cst for constant operands and
9747         VEC_DUPLICATE_EXPR otherwise.
9748         (make_vector_type): Remove temporary is_constant ().
9749         (build_vector_type, build_opaque_vector_type): Take the number of
9750         units as a poly_uint64 rather than an int.
9751         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9752         VECTOR_CST_NELTS.
9753         * cfgexpand.c (expand_debug_expr): Likewise.
9754         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9755         (store_constructor, expand_expr_real_1): Likewise.
9756         (const_scalar_mask_from_tree): Likewise.
9757         * fold-const-call.c (fold_const_reduction): Likewise.
9758         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9759         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9760         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9761         (fold_relational_const): Likewise.
9762         (native_interpret_vector): Likewise.  Change the size from an
9763         int to an unsigned int.
9764         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9765         TYPE_VECTOR_SUBPARTS.
9766         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9767         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9768         duplicating a non-constant operand into a variable-length vector.
9769         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9770         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9771         * ipa-icf.c (sem_variable::equals): Likewise.
9772         * match.pd: Likewise.
9773         * omp-simd-clone.c (simd_clone_subparts): Likewise.
9774         * print-tree.c (print_node): Likewise.
9775         * stor-layout.c (layout_type): Likewise.
9776         * targhooks.c (default_builtin_vectorization_cost): Likewise.
9777         * tree-cfg.c (verify_gimple_comparison): Likewise.
9778         (verify_gimple_assign_binary): Likewise.
9779         (verify_gimple_assign_ternary): Likewise.
9780         (verify_gimple_assign_single): Likewise.
9781         * tree-pretty-print.c (dump_generic_node): Likewise.
9782         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9783         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9784         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9785         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9786         (vect_shift_permute_load_chain): Likewise.
9787         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9788         (expand_vector_condition, optimize_vector_constructor): Likewise.
9789         (lower_vec_perm, get_compute_type): Likewise.
9790         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9791         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9792         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9793         (vect_recog_mask_conversion_pattern): Likewise.
9794         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9795         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9796         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9797         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9798         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9799         (vectorizable_shift, vectorizable_operation, vectorizable_store)
9800         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9801         (supportable_widening_operation): Likewise.
9802         (supportable_narrowing_operation): Likewise.
9803         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9804         Likewise.
9805         * varasm.c (output_constant): Likewise.
9806
9807 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9808             Alan Hayward  <alan.hayward@arm.com>
9809             David Sherwood  <david.sherwood@arm.com>
9810
9811         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9812         so that both the length == 3 and length != 3 cases set up their
9813         own permute vectors.  Add comments explaining why we know the
9814         number of elements is constant.
9815         (vect_permute_load_chain): Likewise.
9816
9817 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9818             Alan Hayward  <alan.hayward@arm.com>
9819             David Sherwood  <david.sherwood@arm.com>
9820
9821         * machmode.h (mode_nunits): Change from unsigned char to
9822         poly_uint16_pod.
9823         (ONLY_FIXED_SIZE_MODES): New macro.
9824         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9825         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9826         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9827         New typedefs.
9828         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9829         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9830         or if measurement_type is not polynomial.
9831         * genmodes.c (ZERO_COEFFS): New macro.
9832         (emit_mode_nunits_inline): Make mode_nunits_inline return a
9833         poly_uint16.
9834         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9835         Use ZERO_COEFFS when emitting initializers.
9836         * data-streamer.h (bp_pack_poly_value): New function.
9837         (bp_unpack_poly_value): Likewise.
9838         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9839         for GET_MODE_NUNITS.
9840         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9841         for GET_MODE_NUNITS.
9842         * tree.c (make_vector_type): Remove temporary shim and make
9843         the real function take the number of units as a poly_uint64
9844         rather than an int.
9845         (build_vector_type_for_mode): Handle polynomial nunits.
9846         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9847         * emit-rtl.c (const_vec_series_p_1): Likewise.
9848         (gen_rtx_CONST_VECTOR): Likewise.
9849         * fold-const.c (test_vec_duplicate_folding): Likewise.
9850         * genrecog.c (validate_pattern): Likewise.
9851         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9852         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9853         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9854         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9855         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9856         * rtlanal.c (subreg_get_info): Likewise.
9857         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9858         (vect_grouped_load_supported): Likewise.
9859         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9860         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9861         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9862         (simplify_const_unary_operation, simplify_binary_operation_1)
9863         (simplify_const_binary_operation, simplify_ternary_operation)
9864         (test_vector_ops_duplicate, test_vector_ops): Likewise.
9865         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9866         instead of CONST_VECTOR_NUNITS.
9867         * varasm.c (output_constant_pool_2): Likewise.
9868         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9869         explicit-encoded elements in the XVEC for variable-length vectors.
9870
9871 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9872
9873         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9874
9875 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9876             Alan Hayward  <alan.hayward@arm.com>
9877             David Sherwood  <david.sherwood@arm.com>
9878
9879         * coretypes.h (fixed_size_mode): Declare.
9880         (fixed_size_mode_pod): New typedef.
9881         * builtins.h (target_builtins::x_apply_args_mode)
9882         (target_builtins::x_apply_result_mode): Change type to
9883         fixed_size_mode_pod.
9884         * builtins.c (apply_args_size, apply_result_size, result_vector)
9885         (expand_builtin_apply_args_1, expand_builtin_apply)
9886         (expand_builtin_return): Update accordingly.
9887
9888 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9889
9890         * cse.c (hash_rtx_cb): Hash only the encoded elements.
9891         * cselib.c (cselib_hash_rtx): Likewise.
9892         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9893         CONST_VECTOR encoding.
9894
9895 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
9896             Jeff Law  <law@redhat.com>
9897
9898         PR target/83641
9899         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9900         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9901         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9902         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9903
9904         PR target/83641
9905         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9906         explicitly probe *sp in a noreturn function if there were any callee
9907         register saves or frame pointer is needed.
9908
9909 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
9910
9911         PR debug/83621
9912         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9913         BLKmode for ternary, binary or unary expressions.
9914
9915         PR debug/83645
9916         * var-tracking.c (delete_vta_debug_insn): New inline function.
9917         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9918         insns from get_insns () to NULL instead of each bb separately.
9919         Use delete_vta_debug_insn.  No longer static.
9920         (vt_debug_insns_local, variable_tracking_main_1): Adjust
9921         delete_vta_debug_insns callers.
9922         * rtl.h (delete_vta_debug_insns): Declare.
9923         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9924         instead of variable_tracking_main.
9925
9926 2018-01-03  Martin Sebor  <msebor@redhat.com>
9927
9928         PR tree-optimization/83603
9929         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9930         arguments past the endof the argument list in functions declared
9931         without a prototype.
9932         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9933         Avoid checking when arguments are null.
9934
9935 2018-01-03  Martin Sebor  <msebor@redhat.com>
9936
9937         PR c/83559
9938         * doc/extend.texi (attribute const): Fix a typo.
9939         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9940         issuing -Wsuggest-attribute for void functions.
9941
9942 2018-01-03  Martin Sebor  <msebor@redhat.com>
9943
9944         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9945         offset_int::from instead of wide_int::to_shwi.
9946         (maybe_diag_overlap): Remove assertion.
9947         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9948         * gimple-ssa-sprintf.c (format_directive): Same.
9949         (parse_directive): Same.
9950         (sprintf_dom_walker::compute_format_length): Same.
9951         (try_substitute_return_value): Same.
9952
9953 2018-01-03  Jeff Law  <law@redhat.com>
9954
9955         PR middle-end/83654
9956         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9957         non-constant residual for zero at runtime and avoid probing in
9958         that case.  Reorganize code for trailing problem to mirror handling
9959         of the residual.
9960
9961 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
9962
9963         PR tree-optimization/83501
9964         * tree-ssa-strlen.c (get_string_cst): New.
9965         (handle_char_store): Call get_string_cst.
9966
9967 2018-01-03  Martin Liska  <mliska@suse.cz>
9968
9969         PR tree-optimization/83593
9970         * tree-ssa-strlen.c: Include tree-cfg.h.
9971         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9972         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9973         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9974         to false.
9975         (strlen_dom_walker::before_dom_children): Call
9976         gimple_purge_dead_eh_edges. Dump tranformation with details
9977         dump flags.
9978         (strlen_dom_walker::before_dom_children): Update call by adding
9979         new argument cleanup_eh.
9980         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9981
9982 2018-01-03  Martin Liska  <mliska@suse.cz>
9983
9984         PR ipa/83549
9985         * cif-code.def (VARIADIC_THUNK): New enum value.
9986         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9987         thunks.
9988
9989 2018-01-03  Jan Beulich  <jbeulich@suse.com>
9990
9991         * sse.md (mov<mode>_internal): Tighten condition for when to use
9992         vmovdqu<ssescalarsize> for TI and OI modes.
9993
9994 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
9995
9996         Update copyright years.
9997
9998 2018-01-03  Martin Liska  <mliska@suse.cz>
9999
10000         PR ipa/83594
10001         * ipa-visibility.c (function_and_variable_visibility): Skip
10002         functions with noipa attribure.
10003
10004 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
10005
10006         * gcc.c (process_command): Update copyright notice dates.
10007         * gcov-dump.c (print_version): Ditto.
10008         * gcov.c (print_version): Ditto.
10009         * gcov-tool.c (print_version): Ditto.
10010         * gengtype.c (create_file): Ditto.
10011         * doc/cpp.texi: Bump @copying's copyright year.
10012         * doc/cppinternals.texi: Ditto.
10013         * doc/gcc.texi: Ditto.
10014         * doc/gccint.texi: Ditto.
10015         * doc/gcov.texi: Ditto.
10016         * doc/install.texi: Ditto.
10017         * doc/invoke.texi: Ditto.
10018
10019 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10020
10021         * vector-builder.h (vector_builder::m_full_nelts): Change from
10022         unsigned int to poly_uint64.
10023         (vector_builder::full_nelts): Update prototype accordingly.
10024         (vector_builder::new_vector): Likewise.
10025         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10026         (vector_builder::operator ==): Likewise.
10027         (vector_builder::finalize): Likewise.
10028         * int-vector-builder.h (int_vector_builder::int_vector_builder):
10029         Take the number of elements as a poly_uint64 rather than an
10030         unsigned int.
10031         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10032         from unsigned int to poly_uint64.
10033         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10034         (vec_perm_indices::new_vector): Likewise.
10035         (vec_perm_indices::length): Likewise.
10036         (vec_perm_indices::nelts_per_input): Likewise.
10037         (vec_perm_indices::input_nelts): Likewise.
10038         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10039         number of elements per input as a poly_uint64 rather than an
10040         unsigned int.  Use the original encoding for variable-length
10041         vectors, rather than clamping each individual element.
10042         For the second and subsequent elements in each pattern,
10043         clamp the step and base before clamping their sum.
10044         (vec_perm_indices::series_p): Handle polynomial element counts.
10045         (vec_perm_indices::all_in_range_p): Likewise.
10046         (vec_perm_indices_to_tree): Likewise.
10047         (vec_perm_indices_to_rtx): Likewise.
10048         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10049         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10050         (tree_vector_builder::new_binary_operation): Handle polynomial
10051         element counts.  Return false if we need to know the number
10052         of elements at compile time.
10053         * fold-const.c (fold_vec_perm): Punt if the number of elements
10054         isn't known at compile time.
10055
10056 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10057
10058         * vec-perm-indices.h (vec_perm_builder): Change element type
10059         from HOST_WIDE_INT to poly_int64.
10060         (vec_perm_indices::element_type): Update accordingly.
10061         (vec_perm_indices::clamp): Handle polynomial element_types.
10062         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10063         (vec_perm_indices::all_in_range_p): Likewise.
10064         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10065         than shwi trees.
10066         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10067         polynomial vec_perm_indices element types.
10068         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10069         * fold-const.c (fold_vec_perm): Likewise.
10070         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10071         * tree-vect-generic.c (lower_vec_perm): Likewise.
10072         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10073         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10074         element type to HOST_WIDE_INT.
10075
10076 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10077             Alan Hayward  <alan.hayward@arm.com>
10078             David Sherwood  <david.sherwood@arm.com>
10079
10080         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10081         rather than an int.  Use plus_constant.
10082         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10083         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10084
10085 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10086             Alan Hayward  <alan.hayward@arm.com>
10087             David Sherwood  <david.sherwood@arm.com>
10088
10089         * calls.c (emit_call_1, expand_call): Change struct_value_size from
10090         a HOST_WIDE_INT to a poly_int64.
10091
10092 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10093             Alan Hayward  <alan.hayward@arm.com>
10094             David Sherwood  <david.sherwood@arm.com>
10095
10096         * calls.c (load_register_parameters): Cope with polynomial
10097         mode sizes.  Require a constant size for BLKmode parameters
10098         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
10099         forces a parameter to be padded at the lsb end in order to
10100         fill a complete number of words, require the parameter size
10101         to be ordered wrt UNITS_PER_WORD.
10102
10103 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10104             Alan Hayward  <alan.hayward@arm.com>
10105             David Sherwood  <david.sherwood@arm.com>
10106
10107         * reload1.c (spill_stack_slot_width): Change element type
10108         from unsigned int to poly_uint64_pod.
10109         (alter_reg): Treat mode sizes as polynomial.
10110
10111 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10112             Alan Hayward  <alan.hayward@arm.com>
10113             David Sherwood  <david.sherwood@arm.com>
10114
10115         * reload.c (complex_word_subreg_p): New function.
10116         (reload_inner_reg_of_subreg, push_reload): Use it.
10117
10118 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10119             Alan Hayward  <alan.hayward@arm.com>
10120             David Sherwood  <david.sherwood@arm.com>
10121
10122         * lra-constraints.c (process_alt_operands): Reject matched
10123         operands whose sizes aren't ordered.
10124         (match_reload): Refer to this check here.
10125
10126 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10127             Alan Hayward  <alan.hayward@arm.com>
10128             David Sherwood  <david.sherwood@arm.com>
10129
10130         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10131         that the mode size is in the set {1, 2, 4, 8, 16}.
10132
10133 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10134             Alan Hayward  <alan.hayward@arm.com>
10135             David Sherwood  <david.sherwood@arm.com>
10136
10137         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10138         Use plus_constant instead of gen_rtx_PLUS.
10139
10140 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10141             Alan Hayward  <alan.hayward@arm.com>
10142             David Sherwood  <david.sherwood@arm.com>
10143
10144         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10145         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10146         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10147         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10148         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10149         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10150         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10151         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10152         * config/i386/i386.c (ix86_push_rounding): ...this new function.
10153         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10154         a poly_int64.
10155         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10156         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10157         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10158         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10159         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10160         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10161         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10162         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10163         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10164         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10165         function.
10166         * expr.c (emit_move_resolve_push): Treat the input and result
10167         of PUSH_ROUNDING as a poly_int64.
10168         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10169         (emit_push_insn): Likewise.
10170         * lra-eliminations.c (mark_not_eliminable): Likewise.
10171         * recog.c (push_operand): Likewise.
10172         * reload1.c (elimination_effects): Likewise.
10173         * rtlanal.c (nonzero_bits1): Likewise.
10174         * calls.c (store_one_arg): Likewise.  Require the padding to be
10175         known at compile time.
10176
10177 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10178             Alan Hayward  <alan.hayward@arm.com>
10179             David Sherwood  <david.sherwood@arm.com>
10180
10181         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10182         Use plus_constant instead of gen_rtx_PLUS.
10183
10184 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10185             Alan Hayward  <alan.hayward@arm.com>
10186             David Sherwood  <david.sherwood@arm.com>
10187
10188         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10189         rather than an int.
10190
10191 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10192             Alan Hayward  <alan.hayward@arm.com>
10193             David Sherwood  <david.sherwood@arm.com>
10194
10195         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10196         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10197         via stack temporaries.  Treat the mode size as polynomial too.
10198
10199 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10200             Alan Hayward  <alan.hayward@arm.com>
10201             David Sherwood  <david.sherwood@arm.com>
10202
10203         * expr.c (expand_expr_real_2): When handling conversions involving
10204         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10205         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
10206         as a poly_uint64 too.
10207
10208 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10209             Alan Hayward  <alan.hayward@arm.com>
10210             David Sherwood  <david.sherwood@arm.com>
10211
10212         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10213
10214 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10215             Alan Hayward  <alan.hayward@arm.com>
10216             David Sherwood  <david.sherwood@arm.com>
10217
10218         * combine.c (can_change_dest_mode): Handle polynomial
10219         REGMODE_NATURAL_SIZE.
10220         * expmed.c (store_bit_field_1): Likewise.
10221         * expr.c (store_constructor): Likewise.
10222         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10223         and polynomial REGMODE_NATURAL_SIZE.
10224         (gen_lowpart_common): Likewise.
10225         * reginfo.c (record_subregs_of_mode): Likewise.
10226         * rtlanal.c (read_modify_subreg_p): Likewise.
10227
10228 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10229             Alan Hayward  <alan.hayward@arm.com>
10230             David Sherwood  <david.sherwood@arm.com>
10231
10232         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10233         numbers of elements.
10234
10235 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10236             Alan Hayward  <alan.hayward@arm.com>
10237             David Sherwood  <david.sherwood@arm.com>
10238
10239         * match.pd: Cope with polynomial numbers of vector elements.
10240
10241 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10242             Alan Hayward  <alan.hayward@arm.com>
10243             David Sherwood  <david.sherwood@arm.com>
10244
10245         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10246         in a POINTER_PLUS_EXPR.
10247
10248 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10249             Alan Hayward  <alan.hayward@arm.com>
10250             David Sherwood  <david.sherwood@arm.com>
10251
10252         * omp-simd-clone.c (simd_clone_subparts): New function.
10253         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10254         (ipa_simd_modify_function_body): Likewise.
10255
10256 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10257             Alan Hayward  <alan.hayward@arm.com>
10258             David Sherwood  <david.sherwood@arm.com>
10259
10260         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10261         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10262         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10263         (expand_vector_condition, vector_element): Likewise.
10264         (subparts_gt): New function.
10265         (get_compute_type): Use subparts_gt.
10266         (count_type_subparts): Delete.
10267         (expand_vector_operations_1): Use subparts_gt instead of
10268         count_type_subparts.
10269
10270 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10271             Alan Hayward  <alan.hayward@arm.com>
10272             David Sherwood  <david.sherwood@arm.com>
10273
10274         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10275         (vect_compile_time_alias): ...this new function.  Do the calculation
10276         on poly_ints rather than trees.
10277         (vect_prune_runtime_alias_test_list): Update call accordingly.
10278
10279 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10280             Alan Hayward  <alan.hayward@arm.com>
10281             David Sherwood  <david.sherwood@arm.com>
10282
10283         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10284         numbers of units.
10285         (vect_schedule_slp_instance): Likewise.
10286
10287 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10288             Alan Hayward  <alan.hayward@arm.com>
10289             David Sherwood  <david.sherwood@arm.com>
10290
10291         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10292         constant and extern definitions for variable-length vectors.
10293         (vect_get_constant_vectors): Note that the number of units
10294         is known to be constant.
10295
10296 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10297             Alan Hayward  <alan.hayward@arm.com>
10298             David Sherwood  <david.sherwood@arm.com>
10299
10300         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10301         of units as polynomial.  Choose between WIDE and NARROW based
10302         on multiple_p.
10303
10304 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10305             Alan Hayward  <alan.hayward@arm.com>
10306             David Sherwood  <david.sherwood@arm.com>
10307
10308         * tree-vect-stmts.c (simd_clone_subparts): New function.
10309         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10310
10311 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10312             Alan Hayward  <alan.hayward@arm.com>
10313             David Sherwood  <david.sherwood@arm.com>
10314
10315         * tree-vect-stmts.c (vectorizable_call): Treat the number of
10316         vectors as polynomial.  Use build_index_vector for
10317         IFN_GOMP_SIMD_LANE.
10318
10319 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10320             Alan Hayward  <alan.hayward@arm.com>
10321             David Sherwood  <david.sherwood@arm.com>
10322
10323         * tree-vect-stmts.c (get_load_store_type): Treat the number of
10324         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10325         for variable-length vectors.
10326         (vectorizable_mask_load_store): Treat the number of units as
10327         polynomial, asserting that it is constant if the condition has
10328         already been enforced.
10329         (vectorizable_store, vectorizable_load): Likewise.
10330
10331 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10332             Alan Hayward  <alan.hayward@arm.com>
10333             David Sherwood  <david.sherwood@arm.com>
10334
10335         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10336         of units as polynomial.  Punt if we can't tell at compile time
10337         which vector contains the final result.
10338
10339 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10340             Alan Hayward  <alan.hayward@arm.com>
10341             David Sherwood  <david.sherwood@arm.com>
10342
10343         * tree-vect-loop.c (vectorizable_induction): Treat the number
10344         of units as polynomial.  Punt on SLP inductions.  Use an integer
10345         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
10346         cast of such a series for variable-length floating-point
10347         reductions.
10348
10349 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10350             Alan Hayward  <alan.hayward@arm.com>
10351             David Sherwood  <david.sherwood@arm.com>
10352
10353         * tree.h (build_index_vector): Declare.
10354         * tree.c (build_index_vector): New function.
10355         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10356         of units as polynomial, forcibly converting it to a constant if
10357         vectorizable_reduction has already enforced the condition.
10358         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
10359         to create a {1,2,3,...} vector.
10360         (vectorizable_reduction): Treat the number of units as polynomial.
10361         Choose vectype_in based on the largest scalar element size rather
10362         than the smallest number of units.  Enforce the restrictions
10363         relied on above.
10364
10365 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10366             Alan Hayward  <alan.hayward@arm.com>
10367             David Sherwood  <david.sherwood@arm.com>
10368
10369         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10370         number of units as polynomial.
10371
10372 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10373             Alan Hayward  <alan.hayward@arm.com>
10374             David Sherwood  <david.sherwood@arm.com>
10375
10376         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10377         * target.def (autovectorize_vector_sizes): Return the vector sizes
10378         by pointer, using vector_sizes rather than a bitmask.
10379         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10380         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10381         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10382         Likewise.
10383         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10384         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10385         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10386         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10387         * omp-general.c (omp_max_vf): Likewise.
10388         * omp-low.c (omp_clause_aligned_alignment): Likewise.
10389         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10390         * tree-vect-loop.c (vect_analyze_loop): Likewise.
10391         * tree-vect-slp.c (vect_slp_bb): Likewise.
10392         * doc/tm.texi: Regenerate.
10393         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10394         to a poly_uint64.
10395         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10396         the vector size as a poly_uint64 rather than an unsigned int.
10397         (current_vector_size): Change from an unsigned int to a poly_uint64.
10398         (get_vectype_for_scalar_type): Update accordingly.
10399         * tree.h (build_truth_vector_type): Take the size and number of
10400         units as a poly_uint64 rather than an unsigned int.
10401         (build_vector_type): Add a temporary overload that takes
10402         the number of units as a poly_uint64 rather than an unsigned int.
10403         * tree.c (make_vector_type): Likewise.
10404         (build_truth_vector_type): Take the number of units as a poly_uint64
10405         rather than an unsigned int.
10406
10407 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10408             Alan Hayward  <alan.hayward@arm.com>
10409             David Sherwood  <david.sherwood@arm.com>
10410
10411         * target.def (get_mask_mode): Take the number of units and length
10412         as poly_uint64s rather than unsigned ints.
10413         * targhooks.h (default_get_mask_mode): Update accordingly.
10414         * targhooks.c (default_get_mask_mode): Likewise.
10415         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10416         * doc/tm.texi: Regenerate.
10417
10418 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10419             Alan Hayward  <alan.hayward@arm.com>
10420             David Sherwood  <david.sherwood@arm.com>
10421
10422         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10423         * omp-general.c (omp_max_vf): Likewise.
10424         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10425         (expand_omp_simd): Handle polynomial safelen.
10426         * omp-low.c (omplow_simd_context): Add a default constructor.
10427         (omplow_simd_context::max_vf): Change from int to poly_uint64.
10428         (lower_rec_simd_input_clauses): Update accordingly.
10429         (lower_rec_input_clauses): Likewise.
10430
10431 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10432             Alan Hayward  <alan.hayward@arm.com>
10433             David Sherwood  <david.sherwood@arm.com>
10434
10435         * tree-vectorizer.h (vect_nunits_for_cost): New function.
10436         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10437         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10438         (vect_analyze_slp_cost): Likewise.
10439         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10440         (vect_model_load_cost): Likewise.
10441
10442 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10443             Alan Hayward  <alan.hayward@arm.com>
10444             David Sherwood  <david.sherwood@arm.com>
10445
10446         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10447         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10448         from an unsigned int * to a poly_uint64_pod *.
10449         (calculate_unrolling_factor): New function.
10450         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
10451
10452 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10453             Alan Hayward  <alan.hayward@arm.com>
10454             David Sherwood  <david.sherwood@arm.com>
10455
10456         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10457         from an unsigned int to a poly_uint64.
10458         (_loop_vec_info::slp_unrolling_factor): Likewise.
10459         (_loop_vec_info::vectorization_factor): Change from an int
10460         to a poly_uint64.
10461         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10462         (vect_get_num_vectors): New function.
10463         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10464         (vect_get_num_copies): Use vect_get_num_vectors.
10465         (vect_analyze_data_ref_dependences): Change max_vf from an int *
10466         to an unsigned int *.
10467         (vect_analyze_data_refs): Change min_vf from an int * to a
10468         poly_uint64 *.
10469         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10470         than an unsigned HOST_WIDE_INT.
10471         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10472         (vect_analyze_data_ref_dependence): Change max_vf from an int *
10473         to an unsigned int *.
10474         (vect_analyze_data_ref_dependences): Likewise.
10475         (vect_compute_data_ref_alignment): Handle polynomial vf.
10476         (vect_enhance_data_refs_alignment): Likewise.
10477         (vect_prune_runtime_alias_test_list): Likewise.
10478         (vect_shift_permute_load_chain): Likewise.
10479         (vect_supportable_dr_alignment): Likewise.
10480         (dependence_distance_ge_vf): Take the vectorization factor as a
10481         poly_uint64 rather than an unsigned HOST_WIDE_INT.
10482         (vect_analyze_data_refs): Change min_vf from an int * to a
10483         poly_uint64 *.
10484         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10485         vfm1 as a poly_uint64 rather than an int.  Make the same change
10486         for the returned bound_scalar.
10487         (vect_gen_vector_loop_niters): Handle polynomial vf.
10488         (vect_do_peeling): Likewise.  Update call to
10489         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10490         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10491         be constant.
10492         * tree-vect-loop.c (vect_determine_vectorization_factor)
10493         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10494         (vect_get_known_peeling_cost): Likewise.
10495         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10496         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10497         (vect_transform_loop): Likewise.  Use the lowest possible VF when
10498         updating the upper bounds of the loop.
10499         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
10500         rather than an int.
10501         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10502         polynomial unroll factors.
10503         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10504         (vect_make_slp_decision): Likewise.
10505         (vect_supported_load_permutation_p): Likewise, and polynomial
10506         vf too.
10507         (vect_analyze_slp_cost): Handle polynomial vf.
10508         (vect_slp_analyze_node_operations): Likewise.
10509         (vect_slp_analyze_bb_1): Likewise.
10510         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10511         than an unsigned HOST_WIDE_INT.
10512         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10513         (vectorizable_load): Handle polynomial vf.
10514         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10515         a poly_uint64.
10516         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10517
10518 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10519             Alan Hayward  <alan.hayward@arm.com>
10520             David Sherwood  <david.sherwood@arm.com>
10521
10522         * match.pd: Handle bit operations involving three constants
10523         and try to fold one pair.
10524
10525 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10526
10527         * tree-vect-loop-manip.c: Include gimple-fold.h.
10528         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10529         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
10530         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10531         Add a path that uses a step of VF instead of 1, but disable it
10532         for now.
10533         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10534         and niters_no_overflow parameters.  Update calls to
10535         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10536         Create a new SSA name if the latter choses to use a ste other
10537         than zero, and return it via niters_vector_mult_vf_var.
10538         * tree-vect-loop.c (vect_transform_loop): Update calls to
10539         vect_do_peeling, vect_gen_vector_loop_niters and
10540         slpeel_make_loop_iterate_ntimes.
10541         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10542         (vect_gen_vector_loop_niters): Update declarations after above changes.
10543
10544 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
10545
10546         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10547         128-bit round to integer instructions.
10548         (ceil<mode>2): Likewise.
10549         (btrunc<mode>2): Likewise.
10550         (round<mode>2): Likewise.
10551
10552 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
10553
10554         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10555         unaligned VSX load/store on P8/P9.
10556         (expand_block_clear): Allow the use of unaligned VSX
10557         load/store on P8/P9.
10558
10559 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10560
10561         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10562         New function.
10563         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10564         swap associated with both a load and a store.
10565
10566 2018-01-02  Andrew Waterman  <andrew@sifive.com>
10567
10568         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10569         * config/riscv/riscv.md (clear_cache): Use it.
10570
10571 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
10572
10573         * web.c: Remove out-of-date comment.
10574
10575 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10576
10577         * expr.c (fixup_args_size_notes): Check that any existing
10578         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10579         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10580         (emit_single_push_insn): ...here.
10581
10582 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10583
10584         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10585         (const_vector_encoded_nelts): New function.
10586         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10587         (const_vector_int_elt, const_vector_elt): Declare.
10588         * emit-rtl.c (const_vector_int_elt_1): New function.
10589         (const_vector_elt): Likewise.
10590         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10591         of CONST_VECTOR_ELT.
10592
10593 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10594
10595         * expr.c: Include rtx-vector-builder.h.
10596         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10597         directly on the tree encoding.
10598         (const_vector_from_tree): Likewise.
10599         * optabs.c: Include rtx-vector-builder.h.
10600         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10601         sequence of "u" values.
10602         * vec-perm-indices.c: Include rtx-vector-builder.h.
10603         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10604         directly on the vec_perm_indices encoding.
10605
10606 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10607
10608         * doc/rtl.texi (const_vector): Describe new encoding scheme.
10609         * Makefile.in (OBJS): Add rtx-vector-builder.o.
10610         * rtx-vector-builder.h: New file.
10611         * rtx-vector-builder.c: Likewise.
10612         * rtl.h (rtx_def::u2): Add a const_vector field.
10613         (CONST_VECTOR_NPATTERNS): New macro.
10614         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10615         (CONST_VECTOR_DUPLICATE_P): Likewise.
10616         (CONST_VECTOR_STEPPED_P): Likewise.
10617         (CONST_VECTOR_ENCODED_ELT): Likewise.
10618         (const_vec_duplicate_p): Check for a duplicated vector encoding.
10619         (unwrap_const_vec_duplicate): Likewise.
10620         (const_vec_series_p): Check for a non-duplicated vector encoding.
10621         Say that the function only returns true for integer vectors.
10622         * emit-rtl.c: Include rtx-vector-builder.h.
10623         (gen_const_vec_duplicate_1): Delete.
10624         (gen_const_vector): Call gen_const_vec_duplicate instead of
10625         gen_const_vec_duplicate_1.
10626         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10627         (gen_const_vec_duplicate): Use rtx_vector_builder.
10628         (gen_const_vec_series): Likewise.
10629         (gen_rtx_CONST_VECTOR): Likewise.
10630         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10631         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10632         Build a new vector rather than modifying a CONST_VECTOR in-place.
10633         (handle_special_swappables): Update call accordingly.
10634         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10635         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10636         Build a new vector rather than modifying a CONST_VECTOR in-place.
10637         (handle_special_swappables): Update call accordingly.
10638
10639 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10640
10641         * simplify-rtx.c (simplify_const_binary_operation): Use
10642         CONST_VECTOR_ELT instead of XVECEXP.
10643
10644 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10645
10646         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10647         the selector elements to be different from the data elements
10648         if the selector is a VECTOR_CST.
10649         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10650         ssizetype for the selector.
10651
10652 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10653
10654         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10655         before testing each element individually.
10656         * tree-vect-generic.c (lower_vec_perm): Likewise.
10657
10658 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10659
10660         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10661         * selftest-run-tests.c (selftest::run_tests): Call it.
10662         * vector-builder.h (vector_builder::operator ==): New function.
10663         (vector_builder::operator !=): Likewise.
10664         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10665         (vec_perm_indices::all_from_input_p): New function.
10666         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10667         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10668         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10669         instead of reading the VECTOR_CST directly.  Detect whether both
10670         vector inputs are the same before constructing the vec_perm_indices,
10671         and update the number of inputs argument accordingly.  Use the
10672         utility functions added above.  Only construct sel2 if we need to.
10673
10674 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10675
10676         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10677         the broadcast of the low byte.
10678         (expand_mult_highpart): Use an explicit encoding for the permutes.
10679         * optabs-query.c (can_mult_highpart_p): Likewise.
10680         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10681         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10682         (vectorizable_bswap): Likewise.
10683         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10684         explicit encoding for the power-of-2 permutes.
10685         (vect_permute_store_chain): Likewise.
10686         (vect_grouped_load_supported): Likewise.
10687         (vect_permute_load_chain): Likewise.
10688
10689 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10690
10691         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10692         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10693         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10694         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10695         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10696         (vect_gen_perm_mask_any): Likewise.
10697
10698 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10699
10700         * int-vector-builder.h: New file.
10701         * vec-perm-indices.h: Include int-vector-builder.h.
10702         (vec_perm_indices): Redefine as an int_vector_builder.
10703         (auto_vec_perm_indices): Delete.
10704         (vec_perm_builder): Redefine as a stand-alone class.
10705         (vec_perm_indices::vec_perm_indices): New function.
10706         (vec_perm_indices::clamp): Likewise.
10707         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10708         (vec_perm_indices::new_vector): New function.
10709         (vec_perm_indices::new_expanded_vector): Update for new
10710         vec_perm_indices class.
10711         (vec_perm_indices::rotate_inputs): New function.
10712         (vec_perm_indices::all_in_range_p): Operate directly on the
10713         encoded form, without computing elided elements.
10714         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10715         encoding.  Update for new vec_perm_indices class.
10716         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10717         the given vec_perm_builder.
10718         (expand_vec_perm_var): Update vec_perm_builder constructor.
10719         (expand_mult_highpart): Use vec_perm_builder instead of
10720         auto_vec_perm_indices.
10721         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10722         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
10723         or double series encoding as appropriate.
10724         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10725         vec_perm_indices instead of auto_vec_perm_indices.
10726         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10727         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10728         (vect_permute_store_chain): Likewise.
10729         (vect_grouped_load_supported): Likewise.
10730         (vect_permute_load_chain): Likewise.
10731         (vect_shift_permute_load_chain): Likewise.
10732         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10733         (vect_transform_slp_perm_load): Likewise.
10734         (vect_schedule_slp_instance): Likewise.
10735         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10736         (vectorizable_mask_load_store): Likewise.
10737         (vectorizable_bswap): Likewise.
10738         (vectorizable_store): Likewise.
10739         (vectorizable_load): Likewise.
10740         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10741         vec_perm_indices instead of auto_vec_perm_indices.  Use
10742         tree_to_vec_perm_builder to read the vector from a tree.
10743         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10744         vec_perm_builder instead of a vec_perm_indices.
10745         (have_whole_vector_shift): Use vec_perm_builder and
10746         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
10747         truncation to calc_vec_perm_mask_for_shift.
10748         (vect_create_epilog_for_reduction): Likewise.
10749         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10750         from auto_vec_perm_indices to vec_perm_indices.
10751         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10752         instead of changing individual elements.
10753         (aarch64_vectorize_vec_perm_const): Use new_vector to install
10754         the vector in d.perm.
10755         * config/arm/arm.c (expand_vec_perm_d::perm): Change
10756         from auto_vec_perm_indices to vec_perm_indices.
10757         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10758         instead of changing individual elements.
10759         (arm_vectorize_vec_perm_const): Use new_vector to install
10760         the vector in d.perm.
10761         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10762         Update vec_perm_builder constructor.
10763         (rs6000_expand_interleave): Likewise.
10764         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10765         (rs6000_expand_interleave): Likewise.
10766
10767 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10768
10769         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10770         to qimode could truncate the indices.
10771         * optabs.c (expand_vec_perm_var): Likewise.
10772
10773 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10774
10775         * Makefile.in (OBJS): Add vec-perm-indices.o.
10776         * vec-perm-indices.h: New file.
10777         * vec-perm-indices.c: Likewise.
10778         * target.h (vec_perm_indices): Replace with a forward class
10779         declaration.
10780         (auto_vec_perm_indices): Move to vec-perm-indices.h.
10781         * optabs.h: Include vec-perm-indices.h.
10782         (expand_vec_perm): Delete.
10783         (selector_fits_mode_p, expand_vec_perm_var): Declare.
10784         (expand_vec_perm_const): Declare.
10785         * target.def (vec_perm_const_ok): Replace with...
10786         (vec_perm_const): ...this new hook.
10787         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10788         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10789         * doc/tm.texi: Regenerate.
10790         * optabs.def (vec_perm_const): Delete.
10791         * doc/md.texi (vec_perm_const): Likewise.
10792         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10793         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10794         expand_vec_perm for constant permutation vectors.  Assert that
10795         the mode of variable permutation vectors is the integer equivalent
10796         of the mode that is being permuted.
10797         * optabs-query.h (selector_fits_mode_p): Declare.
10798         * optabs-query.c: Include vec-perm-indices.h.
10799         (selector_fits_mode_p): New function.
10800         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10801         is defined, instead of checking whether the vec_perm_const_optab
10802         exists.  Use targetm.vectorize.vec_perm_const instead of
10803         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
10804         fit in the vector mode before using a variable permute.
10805         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10806         vec_perm_indices instead of an rtx.
10807         (expand_vec_perm): Replace with...
10808         (expand_vec_perm_const): ...this new function.  Take the selector
10809         as a vec_perm_indices rather than an rtx.  Also take the mode of
10810         the selector.  Update call to shift_amt_for_vec_perm_mask.
10811         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10812         Use vec_perm_indices::new_expanded_vector to expand the original
10813         selector into bytes.  Check whether the indices fit in the vector
10814         mode before using a variable permute.
10815         (expand_vec_perm_var): Make global.
10816         (expand_mult_highpart): Use expand_vec_perm_const.
10817         * fold-const.c: Includes vec-perm-indices.h.
10818         * tree-ssa-forwprop.c: Likewise.
10819         * tree-vect-data-refs.c: Likewise.
10820         * tree-vect-generic.c: Likewise.
10821         * tree-vect-loop.c: Likewise.
10822         * tree-vect-slp.c: Likewise.
10823         * tree-vect-stmts.c: Likewise.
10824         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10825         Delete.
10826         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10827         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10828         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10829         (aarch64_vectorize_vec_perm_const): ...this new function.
10830         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10831         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10832         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10833         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10834         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10835         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10836         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10837         into...
10838         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
10839         check for NEON modes.
10840         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10841         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10842         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10843         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10844         into...
10845         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
10846         the old VEC_PERM_CONST conditions.
10847         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10848         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10849         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10850         (ia64_vectorize_vec_perm_const_ok): Merge into...
10851         (ia64_vectorize_vec_perm_const): ...this new function.
10852         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10853         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10854         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10855         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10856         * config/mips/mips.c (mips_expand_vec_perm_const)
10857         (mips_vectorize_vec_perm_const_ok): Merge into...
10858         (mips_vectorize_vec_perm_const): ...this new function.
10859         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10860         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10861         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10862         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10863         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10864         (rs6000_expand_vec_perm_const): Delete.
10865         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10866         Delete.
10867         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10868         (altivec_expand_vec_perm_const_le): Take each operand individually.
10869         Operate on constant selectors rather than rtxes.
10870         (altivec_expand_vec_perm_const): Likewise.  Update call to
10871         altivec_expand_vec_perm_const_le.
10872         (rs6000_expand_vec_perm_const): Delete.
10873         (rs6000_vectorize_vec_perm_const_ok): Delete.
10874         (rs6000_vectorize_vec_perm_const): New function.
10875         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10876         an element count and rtx array.
10877         (rs6000_expand_extract_even): Update call accordingly.
10878         (rs6000_expand_interleave): Likewise.
10879         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10880         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10881         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10882         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10883         (rs6000_expand_vec_perm_const): Delete.
10884         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10885         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10886         (altivec_expand_vec_perm_const_le): Take each operand individually.
10887         Operate on constant selectors rather than rtxes.
10888         (altivec_expand_vec_perm_const): Likewise.  Update call to
10889         altivec_expand_vec_perm_const_le.
10890         (rs6000_expand_vec_perm_const): Delete.
10891         (rs6000_vectorize_vec_perm_const_ok): Delete.
10892         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
10893         reference to the SPE evmerge intructions.
10894         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10895         an element count and rtx array.
10896         (rs6000_expand_extract_even): Update call accordingly.
10897         (rs6000_expand_interleave): Likewise.
10898         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10899         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10900         new function.
10901         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10902
10903 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10904
10905         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10906         vector mode and that that mode matches the mode of the data
10907         being permuted.
10908         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10909         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
10910         directly using expand_vec_perm_1 when forcing selectors into
10911         registers.
10912         (expand_vec_perm_var): New function, split out from expand_vec_perm.
10913
10914 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10915
10916         * optabs-query.h (can_vec_perm_p): Delete.
10917         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10918         * optabs-query.c (can_vec_perm_p): Split into...
10919         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10920         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10921         particular selector is valid.
10922         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10923         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10924         (vect_grouped_load_supported): Likewise.
10925         (vect_shift_permute_load_chain): Likewise.
10926         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10927         (vect_transform_slp_perm_load): Likewise.
10928         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10929         (vectorizable_bswap): Likewise.
10930         (vect_gen_perm_mask_checked): Likewise.
10931         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
10932         implementations of variable permutation vectors into account
10933         when deciding which selector to use.
10934         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10935         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10936         with a false third argument.
10937         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10938         to test whether the constant selector is valid and can_vec_perm_var_p
10939         to test whether a variable selector is valid.
10940
10941 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10942
10943         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10944         * optabs-query.c (can_vec_perm_p): Likewise.
10945         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10946         instead of vec_perm_indices.
10947         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10948         (vect_gen_perm_mask_checked): Likewise,
10949         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10950         (vect_gen_perm_mask_checked): Likewise,
10951
10952 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
10953
10954         * optabs-query.h (qimode_for_vec_perm): Declare.
10955         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10956         (qimode_for_vec_perm): ...this new function.
10957         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10958
10959 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
10960
10961         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10962         does not have a conditional at the top.
10963
10964 2018-01-02  Richard Biener  <rguenther@suse.de>
10965
10966         * ipa-inline.c (big_speedup_p): Fix expression.
10967
10968 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
10969
10970         PR target/81616
10971         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10972         for generic 4->6.
10973
10974 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
10975
10976         PR target/81616
10977         Generic tuning.
10978         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10979         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10980         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10981         cond_taken_branch_cost 3->4.
10982
10983 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
10984
10985         PR tree-optimization/83581
10986         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10987         TODO_cleanup_cfg if any changes have been made.
10988
10989         PR middle-end/83608
10990         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10991         convert_modes if target mode has the right side, but different mode
10992         class.
10993
10994         PR middle-end/83609
10995         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10996         last argument when extracting from CONCAT.  If either from_real or
10997         from_imag is NULL, use expansion through memory.  If result is not
10998         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10999         the parts directly to inner mode, if even that fails, use expansion
11000         through memory.
11001
11002         PR middle-end/83623
11003         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11004         check for bswap in mode rather than HImode and use that in expand_unop
11005         too.
11006 \f
11007 Copyright (C) 2018 Free Software Foundation, Inc.
11008
11009 Copying and distribution of this file, with or without modification,
11010 are permitted in any medium without royalty provided the copyright
11011 notice and this notice are preserved.